274 files changed, 19614 insertions, 2501 deletions

diff --git a/.ci/scripts/common/pre-upload.sh b/.ci/scripts/common/pre-upload.sh
index 705362a3c..3583f9840 100644
--- a/.ci/scripts/common/pre-upload.sh
+++ b/.ci/scripts/common/pre-upload.sh
@@ -5,6 +5,6 @@
 
 GITDATE="`git show -s --date=short --format='%ad' | sed 's/-//g'`"
 GITREV="`git show -s --format='%h'`"
-ARTIFACTS_DIR="artifacts"
+ARTIFACTS_DIR="$PWD/artifacts"
 
 mkdir -p "${ARTIFACTS_DIR}/"
diff --git a/.ci/scripts/linux/docker.sh b/.ci/scripts/linux/docker.sh
index 4b0193565..7bba01d62 100755
--- a/.ci/scripts/linux/docker.sh
+++ b/.ci/scripts/linux/docker.sh
@@ -11,7 +11,7 @@ ccache -s
 mkdir build || true && cd build
 cmake .. \
       -DBoost_USE_STATIC_LIBS=ON \
-      -DCMAKE_BUILD_TYPE=Release \
+      -DCMAKE_BUILD_TYPE=RelWithDebInfo \
       -DCMAKE_CXX_FLAGS="-march=x86-64-v2" \
       -DCMAKE_CXX_COMPILER=/usr/lib/ccache/g++ \
       -DCMAKE_C_COMPILER=/usr/lib/ccache/gcc \
@@ -32,6 +32,19 @@ ccache -s
 
 ctest -VV -C Release
 
+# Separate debug symbols from specified executables
+for EXE in yuzu; do
+    EXE_PATH="bin/$EXE"
+    # Copy debug symbols out
+    objcopy --only-keep-debug $EXE_PATH $EXE_PATH.debug
+    # Add debug link and strip debug symbols
+    objcopy -g --add-gnu-debuglink=$EXE_PATH.debug $EXE_PATH $EXE_PATH.out
+    # Overwrite original with stripped copy
+    mv $EXE_PATH.out $EXE_PATH
+done
+# Strip debug symbols from all executables
+find bin/ -type f -not -regex '.*.debug' -exec strip -g {} ';'
+
 DESTDIR="$PWD/AppDir" ninja install
 rm -vf AppDir/usr/bin/yuzu-cmd AppDir/usr/bin/yuzu-tester
 
diff --git a/.ci/scripts/linux/upload.sh b/.ci/scripts/linux/upload.sh
index e0f336427..fbb2d9c1b 100755
--- a/.ci/scripts/linux/upload.sh
+++ b/.ci/scripts/linux/upload.sh
@@ -59,4 +59,9 @@ if [ "${RELEASE_NAME}" = "mainline" ] || [ "${RELEASE_NAME}" = "early-access" ];
     cp "build/${APPIMAGE_NAME}" "${DIR_NAME}/yuzu-${RELEASE_NAME}.AppImage"
 fi
 
+# Copy debug symbols to artifacts
+cd build/bin
+tar $COMPRESSION_FLAGS "${ARTIFACTS_DIR}/${REV_NAME}-debug.tar.xz" *.debug
+cd -
+
 . .ci/scripts/common/post-upload.sh
diff --git a/.codespellrc b/.codespellrc
index 944194a25..3336d31fe 100644
--- a/.codespellrc
+++ b/.codespellrc
@@ -3,4 +3,4 @@
 
 [codespell]
 skip = ./.git,./build,./dist,./Doxyfile,./externals,./LICENSES,./src/android/app/src/main/res
-ignore-words-list = aci,allright,ba,canonicalizations,deques,froms,hda,inout,lod,masia,nam,nax,nd,optin,pullrequests,pullrequest,te,transfered,unstall,uscaled,zink
+ignore-words-list = aci,allright,ba,canonicalizations,deques,froms,hda,inout,lod,masia,nam,nax,nd,optin,pullrequests,pullrequest,te,transfered,unstall,uscaled,vas,zink
diff --git a/.reuse/dep5 b/.reuse/dep5
index 31178fc4c..d682fbdba 100644
--- a/.reuse/dep5
+++ b/.reuse/dep5
@@ -147,3 +147,7 @@ License: GPL-3.0-or-later
 Files: src/android/gradle/wrapper/*
 Copyright: 2023 yuzu Emulator Project
 License: GPL-3.0-or-later
+
+Files: externals/stb/*
+Copyright: Sean Barrett
+License: MIT
diff --git a/CMakeLists.txt b/CMakeLists.txt
index ed757eb0a..9c35e0946 100644
--- a/CMakeLists.txt
+++ b/CMakeLists.txt
@@ -11,7 +11,6 @@ list(APPEND CMAKE_MODULE_PATH "${CMAKE_CURRENT_SOURCE_DIR}/externals/cmake-modul
 include(DownloadExternals)
 include(CMakeDependentOption)
 include(CTest)
-include(FetchContent)
 
 # Set bundled sdl2/qt as dependent options.
 # OFF by default, but if ENABLE_SDL2 and MSVC are true then ON
@@ -99,47 +98,8 @@ if (ANDROID AND YUZU_DOWNLOAD_ANDROID_VVL)
             DESTINATION "${vvl_lib_path}")
 endif()
 
-# On Android, fetch and compile libcxx before doing anything else
 if (ANDROID)
     set(CMAKE_SKIP_INSTALL_RULES ON)
-    set(LLVM_VERSION "15.0.6")
-
-    # Note: even though libcxx and libcxxabi have separate releases on the project page,
-    # the separated releases cannot be compiled. Only in-tree builds work. Therefore we
-    # must fetch the source release for the entire llvm tree.
-    FetchContent_Declare(llvm
-        URL "https://github.com/llvm/llvm-project/releases/download/llvmorg-${LLVM_VERSION}/llvm-project-${LLVM_VERSION}.src.tar.xz"
-        URL_HASH SHA256=9d53ad04dc60cb7b30e810faf64c5ab8157dadef46c8766f67f286238256ff92
-        TLS_VERIFY TRUE
-    )
-    FetchContent_MakeAvailable(llvm)
-
-    # libcxx has support for most of the range library, but it's gated behind a flag:
-    add_compile_definitions(_LIBCPP_ENABLE_EXPERIMENTAL)
-
-    # Disable standard header inclusion
-    set(ANDROID_STL "none")
-
-    # libcxxabi
-    set(LIBCXXABI_INCLUDE_TESTS OFF)
-    set(LIBCXXABI_ENABLE_SHARED FALSE)
-    set(LIBCXXABI_ENABLE_STATIC TRUE)
-    set(LIBCXXABI_LIBCXX_INCLUDES "${LIBCXX_TARGET_INCLUDE_DIRECTORY}" CACHE STRING "" FORCE)
-    add_subdirectory("${llvm_SOURCE_DIR}/libcxxabi" "${llvm_BINARY_DIR}/libcxxabi")
-    link_libraries(cxxabi_static)
-
-    # libcxx
-    set(LIBCXX_ABI_NAMESPACE "__ndk1" CACHE STRING "" FORCE)
-    set(LIBCXX_CXX_ABI "libcxxabi")
-    set(LIBCXX_INCLUDE_TESTS OFF)
-    set(LIBCXX_INCLUDE_BENCHMARKS OFF)
-    set(LIBCXX_INCLUDE_DOCS OFF)
-    set(LIBCXX_ENABLE_SHARED FALSE)
-    set(LIBCXX_ENABLE_STATIC TRUE)
-    set(LIBCXX_ENABLE_ASSERTIONS FALSE)
-    add_subdirectory("${llvm_SOURCE_DIR}/libcxx" "${llvm_BINARY_DIR}/libcxx")
-    set_target_properties(cxx-headers PROPERTIES INTERFACE_COMPILE_OPTIONS "-isystem${CMAKE_BINARY_DIR}/${LIBCXX_INSTALL_INCLUDE_DIR}")
-    link_libraries(cxx_static cxx-headers)
 endif()
 
 if (YUZU_USE_BUNDLED_VCPKG)
@@ -326,11 +286,12 @@ find_package(Boost 1.79.0 REQUIRED context)
 find_package(enet 1.3 MODULE)
 find_package(fmt 9 REQUIRED)
 find_package(inih 52 MODULE COMPONENTS INIReader)
-find_package(LLVM 17 MODULE COMPONENTS Demangle)
+find_package(LLVM 17.0.2 MODULE COMPONENTS Demangle)
 find_package(lz4 REQUIRED)
 find_package(nlohmann_json 3.8 REQUIRED)
 find_package(Opus 1.3 MODULE)
 find_package(RenderDoc MODULE)
+find_package(stb MODULE)
 find_package(VulkanMemoryAllocator CONFIG)
 find_package(ZLIB 1.2 REQUIRED)
 find_package(zstd 1.5 REQUIRED)
@@ -397,6 +358,9 @@ function(set_yuzu_qt_components)
     if (ENABLE_QT_TRANSLATION)
         list(APPEND YUZU_QT_COMPONENTS2 LinguistTools)
     endif()
+    if (USE_DISCORD_PRESENCE)
+        list(APPEND YUZU_QT_COMPONENTS2 Network)
+    endif()
     set(YUZU_QT_COMPONENTS ${YUZU_QT_COMPONENTS2} PARENT_SCOPE)
 endfunction(set_yuzu_qt_components)
 
diff --git a/CMakeModules/Findstb.cmake b/CMakeModules/Findstb.cmake
new file mode 100644
index 000000000..bff998580
--- /dev/null
+++ b/CMakeModules/Findstb.cmake
@@ -0,0 +1,31 @@
+# SPDX-FileCopyrightText: 2023 Alexandre Bouvier <contact@amb.tf>
+#
+# SPDX-License-Identifier: GPL-3.0-or-later
+
+find_path(stb_image_INCLUDE_DIR stb_image.h PATH_SUFFIXES stb)
+find_path(stb_image_resize_INCLUDE_DIR stb_image_resize.h PATH_SUFFIXES stb)
+find_path(stb_image_write_INCLUDE_DIR stb_image_write.h PATH_SUFFIXES stb)
+
+include(FindPackageHandleStandardArgs)
+find_package_handle_standard_args(stb
+    REQUIRED_VARS
+        stb_image_INCLUDE_DIR
+        stb_image_resize_INCLUDE_DIR
+        stb_image_write_INCLUDE_DIR
+)
+
+if (stb_FOUND AND NOT TARGET stb::headers)
+    add_library(stb::headers INTERFACE IMPORTED)
+    set_property(TARGET stb::headers PROPERTY
+        INTERFACE_INCLUDE_DIRECTORIES
+            "${stb_image_INCLUDE_DIR}"
+            "${stb_image_resize_INCLUDE_DIR}"
+            "${stb_image_write_INCLUDE_DIR}"
+    )
+endif()
+
+mark_as_advanced(
+    stb_image_INCLUDE_DIR
+    stb_image_resize_INCLUDE_DIR
+    stb_image_write_INCLUDE_DIR
+)
diff --git a/dist/org.yuzu_emu.yuzu.desktop b/dist/org.yuzu_emu.yuzu.desktop
index 008422863..51e191a8e 100644
--- a/dist/org.yuzu_emu.yuzu.desktop
+++ b/dist/org.yuzu_emu.yuzu.desktop
@@ -13,3 +13,4 @@ Exec=yuzu %f
 Categories=Game;Emulator;Qt;
 MimeType=application/x-nx-nro;application/x-nx-nso;application/x-nx-nsp;application/x-nx-xci;
 Keywords=Nintendo;Switch;
+StartupWMClass=yuzu
diff --git a/dist/qt_themes/default/style.qss b/dist/qt_themes/default/style.qss
index 79960ee0c..921950c6c 100644
--- a/dist/qt_themes/default/style.qss
+++ b/dist/qt_themes/default/style.qss
@@ -120,6 +120,10 @@ QWidget#connectedControllers {
     background: transparent;
 }
 
+QWidget#closeButtons {
+    background: transparent;
+}
+
 QWidget#playersSupported,
 QWidget#controllersSupported,
 QWidget#controllerSupported1,
diff --git a/dist/qt_themes/qdarkstyle/style.qss b/dist/qt_themes/qdarkstyle/style.qss
index 63a636ae6..328ac942f 100644
--- a/dist/qt_themes/qdarkstyle/style.qss
+++ b/dist/qt_themes/qdarkstyle/style.qss
@@ -1380,6 +1380,10 @@ QWidget#connectedControllers {
     background: transparent;
 }
 
+QWidget#closeButtons {
+    background: transparent;
+}
+
 QWidget#playersSupported,
 QWidget#controllersSupported,
 QWidget#controllerSupported1,
diff --git a/dist/qt_themes/qdarkstyle_midnight_blue/style.qss b/dist/qt_themes/qdarkstyle_midnight_blue/style.qss
index 0c53115f6..eb0889b13 100644
--- a/dist/qt_themes/qdarkstyle_midnight_blue/style.qss
+++ b/dist/qt_themes/qdarkstyle_midnight_blue/style.qss
@@ -2305,6 +2305,10 @@ QWidget#connectedControllers {
   background: transparent;
 }
 
+QWidget#closeButtons {
+    background: transparent;
+}
+
 QWidget#playersSupported,
 QWidget#controllersSupported,
 QWidget#controllerSupported1,
diff --git a/externals/CMakeLists.txt b/externals/CMakeLists.txt
index c2009f34b..61baabb03 100644
--- a/externals/CMakeLists.txt
+++ b/externals/CMakeLists.txt
@@ -171,6 +171,10 @@ endif()
 add_library(stb stb/stb_dxt.cpp)
 target_include_directories(stb PUBLIC ./stb)
 
+if (NOT TARGET stb::headers)
+    add_library(stb::headers ALIAS stb)
+endif()
+
 add_library(bc_decoder bc_decoder/bc_decoder.cpp)
 target_include_directories(bc_decoder PUBLIC ./bc_decoder)
 
diff --git a/externals/nx_tzdb/CMakeLists.txt b/externals/nx_tzdb/CMakeLists.txt
index 593786250..0fad24642 100644
--- a/externals/nx_tzdb/CMakeLists.txt
+++ b/externals/nx_tzdb/CMakeLists.txt
@@ -27,7 +27,7 @@ if (CMAKE_SYSTEM_NAME STREQUAL "Windows" OR ANDROID)
     set(CAN_BUILD_NX_TZDB false)
 endif()
 
-set(NX_TZDB_VERSION "220816")
+set(NX_TZDB_VERSION "221202")
 set(NX_TZDB_ARCHIVE "${CMAKE_CURRENT_BINARY_DIR}/${NX_TZDB_VERSION}.zip")
 
 set(NX_TZDB_ROMFS_DIR "${CMAKE_CURRENT_BINARY_DIR}/nx_tzdb")
diff --git a/externals/nx_tzdb/tzdb_to_nx b/externals/nx_tzdb/tzdb_to_nx
-Subproject 212afa2394a74226dcf1b7996a570aae17debb6
+Subproject 0d17dd066d91f954a4c89d46dcb067eead6b1e4
diff --git a/externals/stb/stb_image.h b/externals/stb/stb_image.h
new file mode 100644
index 000000000..5e807a0a6
--- /dev/null
+++ b/externals/stb/stb_image.h
@@ -0,0 +1,7987 @@
+/* stb_image - v2.28 - public domain image loader - http://nothings.org/stb
+                                  no warranty implied; use at your own risk
+
+   Do this:
+      #define STB_IMAGE_IMPLEMENTATION
+   before you include this file in *one* C or C++ file to create the implementation.
+
+   // i.e. it should look like this:
+   #include ...
+   #include ...
+   #include ...
+   #define STB_IMAGE_IMPLEMENTATION
+   #include "stb_image.h"
+
+   You can #define STBI_ASSERT(x) before the #include to avoid using assert.h.
+   And #define STBI_MALLOC, STBI_REALLOC, and STBI_FREE to avoid using malloc,realloc,free
+
+
+   QUICK NOTES:
+      Primarily of interest to game developers and other people who can
+          avoid problematic images and only need the trivial interface
+
+      JPEG baseline & progressive (12 bpc/arithmetic not supported, same as stock IJG lib)
+      PNG 1/2/4/8/16-bit-per-channel
+
+      TGA (not sure what subset, if a subset)
+      BMP non-1bpp, non-RLE
+      PSD (composited view only, no extra channels, 8/16 bit-per-channel)
+
+      GIF (*comp always reports as 4-channel)
+      HDR (radiance rgbE format)
+      PIC (Softimage PIC)
+      PNM (PPM and PGM binary only)
+
+      Animated GIF still needs a proper API, but here's one way to do it:
+          http://gist.github.com/urraka/685d9a6340b26b830d49
+
+      - decode from memory or through FILE (define STBI_NO_STDIO to remove code)
+      - decode from arbitrary I/O callbacks
+      - SIMD acceleration on x86/x64 (SSE2) and ARM (NEON)
+
+   Full documentation under "DOCUMENTATION" below.
+
+
+LICENSE
+
+  See end of file for license information.
+
+RECENT REVISION HISTORY:
+
+      2.28  (2023-01-29) many error fixes, security errors, just tons of stuff
+      2.27  (2021-07-11) document stbi_info better, 16-bit PNM support, bug fixes
+      2.26  (2020-07-13) many minor fixes
+      2.25  (2020-02-02) fix warnings
+      2.24  (2020-02-02) fix warnings; thread-local failure_reason and flip_vertically
+      2.23  (2019-08-11) fix clang static analysis warning
+      2.22  (2019-03-04) gif fixes, fix warnings
+      2.21  (2019-02-25) fix typo in comment
+      2.20  (2019-02-07) support utf8 filenames in Windows; fix warnings and platform ifdefs
+      2.19  (2018-02-11) fix warning
+      2.18  (2018-01-30) fix warnings
+      2.17  (2018-01-29) bugfix, 1-bit BMP, 16-bitness query, fix warnings
+      2.16  (2017-07-23) all functions have 16-bit variants; optimizations; bugfixes
+      2.15  (2017-03-18) fix png-1,2,4; all Imagenet JPGs; no runtime SSE detection on GCC
+      2.14  (2017-03-03) remove deprecated STBI_JPEG_OLD; fixes for Imagenet JPGs
+      2.13  (2016-12-04) experimental 16-bit API, only for PNG so far; fixes
+      2.12  (2016-04-02) fix typo in 2.11 PSD fix that caused crashes
+      2.11  (2016-04-02) 16-bit PNGS; enable SSE2 in non-gcc x64
+                         RGB-format JPEG; remove white matting in PSD;
+                         allocate large structures on the stack;
+                         correct channel count for PNG & BMP
+      2.10  (2016-01-22) avoid warning introduced in 2.09
+      2.09  (2016-01-16) 16-bit TGA; comments in PNM files; STBI_REALLOC_SIZED
+
+   See end of file for full revision history.
+
+
+ ============================    Contributors    =========================
+
+ Image formats                          Extensions, features
+    Sean Barrett (jpeg, png, bmp)          Jetro Lauha (stbi_info)
+    Nicolas Schulz (hdr, psd)              Martin "SpartanJ" Golini (stbi_info)
+    Jonathan Dummer (tga)                  James "moose2000" Brown (iPhone PNG)
+    Jean-Marc Lienher (gif)                Ben "Disch" Wenger (io callbacks)
+    Tom Seddon (pic)                       Omar Cornut (1/2/4-bit PNG)
+    Thatcher Ulrich (psd)                  Nicolas Guillemot (vertical flip)
+    Ken Miller (pgm, ppm)                  Richard Mitton (16-bit PSD)
+    github:urraka (animated gif)           Junggon Kim (PNM comments)
+    Christopher Forseth (animated gif)     Daniel Gibson (16-bit TGA)
+                                           socks-the-fox (16-bit PNG)
+                                           Jeremy Sawicki (handle all ImageNet JPGs)
+ Optimizations & bugfixes                  Mikhail Morozov (1-bit BMP)
+    Fabian "ryg" Giesen                    Anael Seghezzi (is-16-bit query)
+    Arseny Kapoulkine                      Simon Breuss (16-bit PNM)
+    John-Mark Allen
+    Carmelo J Fdez-Aguera
+
+ Bug & warning fixes
+    Marc LeBlanc            David Woo          Guillaume George     Martins Mozeiko
+    Christpher Lloyd        Jerry Jansson      Joseph Thomson       Blazej Dariusz Roszkowski
+    Phil Jordan                                Dave Moore           Roy Eltham
+    Hayaki Saito            Nathan Reed        Won Chun
+    Luke Graham             Johan Duparc       Nick Verigakis       the Horde3D community
+    Thomas Ruf              Ronny Chevalier                         github:rlyeh
+    Janez Zemva             John Bartholomew   Michal Cichon        github:romigrou
+    Jonathan Blow           Ken Hamada         Tero Hanninen        github:svdijk
+    Eugene Golushkov        Laurent Gomila     Cort Stratton        github:snagar
+    Aruelien Pocheville     Sergio Gonzalez    Thibault Reuille     github:Zelex
+    Cass Everitt            Ryamond Barbiero                        github:grim210
+    Paul Du Bois            Engin Manap        Aldo Culquicondor    github:sammyhw
+    Philipp Wiesemann       Dale Weiler        Oriol Ferrer Mesia   github:phprus
+    Josh Tobin              Neil Bickford      Matthew Gregan       github:poppolopoppo
+    Julian Raschke          Gregory Mullen     Christian Floisand   github:darealshinji
+    Baldur Karlsson         Kevin Schmidt      JR Smith             github:Michaelangel007
+                            Brad Weinberger    Matvey Cherevko      github:mosra
+    Luca Sas                Alexander Veselov  Zack Middleton       [reserved]
+    Ryan C. Gordon          [reserved]                              [reserved]
+                     DO NOT ADD YOUR NAME HERE
+
+                     Jacko Dirks
+
+  To add your name to the credits, pick a random blank space in the middle and fill it.
+  80% of merge conflicts on stb PRs are due to people adding their name at the end
+  of the credits.
+*/
+
+#ifndef STBI_INCLUDE_STB_IMAGE_H
+#define STBI_INCLUDE_STB_IMAGE_H
+
+// DOCUMENTATION
+//
+// Limitations:
+//    - no 12-bit-per-channel JPEG
+//    - no JPEGs with arithmetic coding
+//    - GIF always returns *comp=4
+//
+// Basic usage (see HDR discussion below for HDR usage):
+//    int x,y,n;
+//    unsigned char *data = stbi_load(filename, &x, &y, &n, 0);
+//    // ... process data if not NULL ...
+//    // ... x = width, y = height, n = # 8-bit components per pixel ...
+//    // ... replace '0' with '1'..'4' to force that many components per pixel
+//    // ... but 'n' will always be the number that it would have been if you said 0
+//    stbi_image_free(data);
+//
+// Standard parameters:
+//    int *x                 -- outputs image width in pixels
+//    int *y                 -- outputs image height in pixels
+//    int *channels_in_file  -- outputs # of image components in image file
+//    int desired_channels   -- if non-zero, # of image components requested in result
+//
+// The return value from an image loader is an 'unsigned char *' which points
+// to the pixel data, or NULL on an allocation failure or if the image is
+// corrupt or invalid. The pixel data consists of *y scanlines of *x pixels,
+// with each pixel consisting of N interleaved 8-bit components; the first
+// pixel pointed to is top-left-most in the image. There is no padding between
+// image scanlines or between pixels, regardless of format. The number of
+// components N is 'desired_channels' if desired_channels is non-zero, or
+// *channels_in_file otherwise. If desired_channels is non-zero,
+// *channels_in_file has the number of components that _would_ have been
+// output otherwise. E.g. if you set desired_channels to 4, you will always
+// get RGBA output, but you can check *channels_in_file to see if it's trivially
+// opaque because e.g. there were only 3 channels in the source image.
+//
+// An output image with N components has the following components interleaved
+// in this order in each pixel:
+//
+//     N=#comp     components
+//       1           grey
+//       2           grey, alpha
+//       3           red, green, blue
+//       4           red, green, blue, alpha
+//
+// If image loading fails for any reason, the return value will be NULL,
+// and *x, *y, *channels_in_file will be unchanged. The function
+// stbi_failure_reason() can be queried for an extremely brief, end-user
+// unfriendly explanation of why the load failed. Define STBI_NO_FAILURE_STRINGS
+// to avoid compiling these strings at all, and STBI_FAILURE_USERMSG to get slightly
+// more user-friendly ones.
+//
+// Paletted PNG, BMP, GIF, and PIC images are automatically depalettized.
+//
+// To query the width, height and component count of an image without having to
+// decode the full file, you can use the stbi_info family of functions:
+//
+//   int x,y,n,ok;
+//   ok = stbi_info(filename, &x, &y, &n);
+//   // returns ok=1 and sets x, y, n if image is a supported format,
+//   // 0 otherwise.
+//
+// Note that stb_image pervasively uses ints in its public API for sizes,
+// including sizes of memory buffers. This is now part of the API and thus
+// hard to change without causing breakage. As a result, the various image
+// loaders all have certain limits on image size; these differ somewhat
+// by format but generally boil down to either just under 2GB or just under
+// 1GB. When the decoded image would be larger than this, stb_image decoding
+// will fail.
+//
+// Additionally, stb_image will reject image files that have any of their
+// dimensions set to a larger value than the configurable STBI_MAX_DIMENSIONS,
+// which defaults to 2**24 = 16777216 pixels. Due to the above memory limit,
+// the only way to have an image with such dimensions load correctly
+// is for it to have a rather extreme aspect ratio. Either way, the
+// assumption here is that such larger images are likely to be malformed
+// or malicious. If you do need to load an image with individual dimensions
+// larger than that, and it still fits in the overall size limit, you can
+// #define STBI_MAX_DIMENSIONS on your own to be something larger.
+//
+// ===========================================================================
+//
+// UNICODE:
+//
+//   If compiling for Windows and you wish to use Unicode filenames, compile
+//   with
+//       #define STBI_WINDOWS_UTF8
+//   and pass utf8-encoded filenames. Call stbi_convert_wchar_to_utf8 to convert
+//   Windows wchar_t filenames to utf8.
+//
+// ===========================================================================
+//
+// Philosophy
+//
+// stb libraries are designed with the following priorities:
+//
+//    1. easy to use
+//    2. easy to maintain
+//    3. good performance
+//
+// Sometimes I let "good performance" creep up in priority over "easy to maintain",
+// and for best performance I may provide less-easy-to-use APIs that give higher
+// performance, in addition to the easy-to-use ones. Nevertheless, it's important
+// to keep in mind that from the standpoint of you, a client of this library,
+// all you care about is #1 and #3, and stb libraries DO NOT emphasize #3 above all.
+//
+// Some secondary priorities arise directly from the first two, some of which
+// provide more explicit reasons why performance can't be emphasized.
+//
+//    - Portable ("ease of use")
+//    - Small source code footprint ("easy to maintain")
+//    - No dependencies ("ease of use")
+//
+// ===========================================================================
+//
+// I/O callbacks
+//
+// I/O callbacks allow you to read from arbitrary sources, like packaged
+// files or some other source. Data read from callbacks are processed
+// through a small internal buffer (currently 128 bytes) to try to reduce
+// overhead.
+//
+// The three functions you must define are "read" (reads some bytes of data),
+// "skip" (skips some bytes of data), "eof" (reports if the stream is at the end).
+//
+// ===========================================================================
+//
+// SIMD support
+//
+// The JPEG decoder will try to automatically use SIMD kernels on x86 when
+// supported by the compiler. For ARM Neon support, you must explicitly
+// request it.
+//
+// (The old do-it-yourself SIMD API is no longer supported in the current
+// code.)
+//
+// On x86, SSE2 will automatically be used when available based on a run-time
+// test; if not, the generic C versions are used as a fall-back. On ARM targets,
+// the typical path is to have separate builds for NEON and non-NEON devices
+// (at least this is true for iOS and Android). Therefore, the NEON support is
+// toggled by a build flag: define STBI_NEON to get NEON loops.
+//
+// If for some reason you do not want to use any of SIMD code, or if
+// you have issues compiling it, you can disable it entirely by
+// defining STBI_NO_SIMD.
+//
+// ===========================================================================
+//
+// HDR image support   (disable by defining STBI_NO_HDR)
+//
+// stb_image supports loading HDR images in general, and currently the Radiance
+// .HDR file format specifically. You can still load any file through the existing
+// interface; if you attempt to load an HDR file, it will be automatically remapped
+// to LDR, assuming gamma 2.2 and an arbitrary scale factor defaulting to 1;
+// both of these constants can be reconfigured through this interface:
+//
+//     stbi_hdr_to_ldr_gamma(2.2f);
+//     stbi_hdr_to_ldr_scale(1.0f);
+//
+// (note, do not use _inverse_ constants; stbi_image will invert them
+// appropriately).
+//
+// Additionally, there is a new, parallel interface for loading files as
+// (linear) floats to preserve the full dynamic range:
+//
+//    float *data = stbi_loadf(filename, &x, &y, &n, 0);
+//
+// If you load LDR images through this interface, those images will
+// be promoted to floating point values, run through the inverse of
+// constants corresponding to the above:
+//
+//     stbi_ldr_to_hdr_scale(1.0f);
+//     stbi_ldr_to_hdr_gamma(2.2f);
+//
+// Finally, given a filename (or an open file or memory block--see header
+// file for details) containing image data, you can query for the "most
+// appropriate" interface to use (that is, whether the image is HDR or
+// not), using:
+//
+//     stbi_is_hdr(char *filename);
+//
+// ===========================================================================
+//
+// iPhone PNG support:
+//
+// We optionally support converting iPhone-formatted PNGs (which store
+// premultiplied BGRA) back to RGB, even though they're internally encoded
+// differently. To enable this conversion, call
+// stbi_convert_iphone_png_to_rgb(1).
+//
+// Call stbi_set_unpremultiply_on_load(1) as well to force a divide per
+// pixel to remove any premultiplied alpha *only* if the image file explicitly
+// says there's premultiplied data (currently only happens in iPhone images,
+// and only if iPhone convert-to-rgb processing is on).
+//
+// ===========================================================================
+//
+// ADDITIONAL CONFIGURATION
+//
+//  - You can suppress implementation of any of the decoders to reduce
+//    your code footprint by #defining one or more of the following
+//    symbols before creating the implementation.
+//
+//        STBI_NO_JPEG
+//        STBI_NO_PNG
+//        STBI_NO_BMP
+//        STBI_NO_PSD
+//        STBI_NO_TGA
+//        STBI_NO_GIF
+//        STBI_NO_HDR
+//        STBI_NO_PIC
+//        STBI_NO_PNM   (.ppm and .pgm)
+//
+//  - You can request *only* certain decoders and suppress all other ones
+//    (this will be more forward-compatible, as addition of new decoders
+//    doesn't require you to disable them explicitly):
+//
+//        STBI_ONLY_JPEG
+//        STBI_ONLY_PNG
+//        STBI_ONLY_BMP
+//        STBI_ONLY_PSD
+//        STBI_ONLY_TGA
+//        STBI_ONLY_GIF
+//        STBI_ONLY_HDR
+//        STBI_ONLY_PIC
+//        STBI_ONLY_PNM   (.ppm and .pgm)
+//
+//   - If you use STBI_NO_PNG (or _ONLY_ without PNG), and you still
+//     want the zlib decoder to be available, #define STBI_SUPPORT_ZLIB
+//
+//  - If you define STBI_MAX_DIMENSIONS, stb_image will reject images greater
+//    than that size (in either width or height) without further processing.
+//    This is to let programs in the wild set an upper bound to prevent
+//    denial-of-service attacks on untrusted data, as one could generate a
+//    valid image of gigantic dimensions and force stb_image to allocate a
+//    huge block of memory and spend disproportionate time decoding it. By
+//    default this is set to (1 << 24), which is 16777216, but that's still
+//    very big.
+
+#ifndef STBI_NO_STDIO
+#include <stdio.h>
+#endif // STBI_NO_STDIO
+
+#define STBI_VERSION 1
+
+enum
+{
+   STBI_default = 0, // only used for desired_channels
+
+   STBI_grey       = 1,
+   STBI_grey_alpha = 2,
+   STBI_rgb        = 3,
+   STBI_rgb_alpha  = 4
+};
+
+#include <stdlib.h>
+typedef unsigned char stbi_uc;
+typedef unsigned short stbi_us;
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+#ifndef STBIDEF
+#ifdef STB_IMAGE_STATIC
+#define STBIDEF static
+#else
+#define STBIDEF extern
+#endif
+#endif
+
+//////////////////////////////////////////////////////////////////////////////
+//
+// PRIMARY API - works on images of any type
+//
+
+//
+// load image by filename, open file, or memory buffer
+//
+
+typedef struct
+{
+   int      (*read)  (void *user,char *data,int size);   // fill 'data' with 'size' bytes.  return number of bytes actually read
+   void     (*skip)  (void *user,int n);                 // skip the next 'n' bytes, or 'unget' the last -n bytes if negative
+   int      (*eof)   (void *user);                       // returns nonzero if we are at end of file/data
+} stbi_io_callbacks;
+
+////////////////////////////////////
+//
+// 8-bits-per-channel interface
+//
+
+STBIDEF stbi_uc *stbi_load_from_memory   (stbi_uc           const *buffer, int len   , int *x, int *y, int *channels_in_file, int desired_channels);
+STBIDEF stbi_uc *stbi_load_from_callbacks(stbi_io_callbacks const *clbk  , void *user, int *x, int *y, int *channels_in_file, int desired_channels);
+
+#ifndef STBI_NO_STDIO
+STBIDEF stbi_uc *stbi_load            (char const *filename, int *x, int *y, int *channels_in_file, int desired_channels);
+STBIDEF stbi_uc *stbi_load_from_file  (FILE *f, int *x, int *y, int *channels_in_file, int desired_channels);
+// for stbi_load_from_file, file pointer is left pointing immediately after image
+#endif
+
+#ifndef STBI_NO_GIF
+STBIDEF stbi_uc *stbi_load_gif_from_memory(stbi_uc const *buffer, int len, int **delays, int *x, int *y, int *z, int *comp, int req_comp);
+#endif
+
+#ifdef STBI_WINDOWS_UTF8
+STBIDEF int stbi_convert_wchar_to_utf8(char *buffer, size_t bufferlen, const wchar_t* input);
+#endif
+
+////////////////////////////////////
+//
+// 16-bits-per-channel interface
+//
+
+STBIDEF stbi_us *stbi_load_16_from_memory   (stbi_uc const *buffer, int len, int *x, int *y, int *channels_in_file, int desired_channels);
+STBIDEF stbi_us *stbi_load_16_from_callbacks(stbi_io_callbacks const *clbk, void *user, int *x, int *y, int *channels_in_file, int desired_channels);
+
+#ifndef STBI_NO_STDIO
+STBIDEF stbi_us *stbi_load_16          (char const *filename, int *x, int *y, int *channels_in_file, int desired_channels);
+STBIDEF stbi_us *stbi_load_from_file_16(FILE *f, int *x, int *y, int *channels_in_file, int desired_channels);
+#endif
+
+////////////////////////////////////
+//
+// float-per-channel interface
+//
+#ifndef STBI_NO_LINEAR
+   STBIDEF float *stbi_loadf_from_memory     (stbi_uc const *buffer, int len, int *x, int *y, int *channels_in_file, int desired_channels);
+   STBIDEF float *stbi_loadf_from_callbacks  (stbi_io_callbacks const *clbk, void *user, int *x, int *y,  int *channels_in_file, int desired_channels);
+
+   #ifndef STBI_NO_STDIO
+   STBIDEF float *stbi_loadf            (char const *filename, int *x, int *y, int *channels_in_file, int desired_channels);
+   STBIDEF float *stbi_loadf_from_file  (FILE *f, int *x, int *y, int *channels_in_file, int desired_channels);
+   #endif
+#endif
+
+#ifndef STBI_NO_HDR
+   STBIDEF void   stbi_hdr_to_ldr_gamma(float gamma);
+   STBIDEF void   stbi_hdr_to_ldr_scale(float scale);
+#endif // STBI_NO_HDR
+
+#ifndef STBI_NO_LINEAR
+   STBIDEF void   stbi_ldr_to_hdr_gamma(float gamma);
+   STBIDEF void   stbi_ldr_to_hdr_scale(float scale);
+#endif // STBI_NO_LINEAR
+
+// stbi_is_hdr is always defined, but always returns false if STBI_NO_HDR
+STBIDEF int    stbi_is_hdr_from_callbacks(stbi_io_callbacks const *clbk, void *user);
+STBIDEF int    stbi_is_hdr_from_memory(stbi_uc const *buffer, int len);
+#ifndef STBI_NO_STDIO
+STBIDEF int      stbi_is_hdr          (char const *filename);
+STBIDEF int      stbi_is_hdr_from_file(FILE *f);
+#endif // STBI_NO_STDIO
+
+
+// get a VERY brief reason for failure
+// on most compilers (and ALL modern mainstream compilers) this is threadsafe
+STBIDEF const char *stbi_failure_reason  (void);
+
+// free the loaded image -- this is just free()
+STBIDEF void     stbi_image_free      (void *retval_from_stbi_load);
+
+// get image dimensions & components without fully decoding
+STBIDEF int      stbi_info_from_memory(stbi_uc const *buffer, int len, int *x, int *y, int *comp);
+STBIDEF int      stbi_info_from_callbacks(stbi_io_callbacks const *clbk, void *user, int *x, int *y, int *comp);
+STBIDEF int      stbi_is_16_bit_from_memory(stbi_uc const *buffer, int len);
+STBIDEF int      stbi_is_16_bit_from_callbacks(stbi_io_callbacks const *clbk, void *user);
+
+#ifndef STBI_NO_STDIO
+STBIDEF int      stbi_info               (char const *filename,     int *x, int *y, int *comp);
+STBIDEF int      stbi_info_from_file     (FILE *f,                  int *x, int *y, int *comp);
+STBIDEF int      stbi_is_16_bit          (char const *filename);
+STBIDEF int      stbi_is_16_bit_from_file(FILE *f);
+#endif
+
+
+
+// for image formats that explicitly notate that they have premultiplied alpha,
+// we just return the colors as stored in the file. set this flag to force
+// unpremultiplication. results are undefined if the unpremultiply overflow.
+STBIDEF void stbi_set_unpremultiply_on_load(int flag_true_if_should_unpremultiply);
+
+// indicate whether we should process iphone images back to canonical format,
+// or just pass them through "as-is"
+STBIDEF void stbi_convert_iphone_png_to_rgb(int flag_true_if_should_convert);
+
+// flip the image vertically, so the first pixel in the output array is the bottom left
+STBIDEF void stbi_set_flip_vertically_on_load(int flag_true_if_should_flip);
+
+// as above, but only applies to images loaded on the thread that calls the function
+// this function is only available if your compiler supports thread-local variables;
+// calling it will fail to link if your compiler doesn't
+STBIDEF void stbi_set_unpremultiply_on_load_thread(int flag_true_if_should_unpremultiply);
+STBIDEF void stbi_convert_iphone_png_to_rgb_thread(int flag_true_if_should_convert);
+STBIDEF void stbi_set_flip_vertically_on_load_thread(int flag_true_if_should_flip);
+
+// ZLIB client - used by PNG, available for other purposes
+
+STBIDEF char *stbi_zlib_decode_malloc_guesssize(const char *buffer, int len, int initial_size, int *outlen);
+STBIDEF char *stbi_zlib_decode_malloc_guesssize_headerflag(const char *buffer, int len, int initial_size, int *outlen, int parse_header);
+STBIDEF char *stbi_zlib_decode_malloc(const char *buffer, int len, int *outlen);
+STBIDEF int   stbi_zlib_decode_buffer(char *obuffer, int olen, const char *ibuffer, int ilen);
+
+STBIDEF char *stbi_zlib_decode_noheader_malloc(const char *buffer, int len, int *outlen);
+STBIDEF int   stbi_zlib_decode_noheader_buffer(char *obuffer, int olen, const char *ibuffer, int ilen);
+
+
+#ifdef __cplusplus
+}
+#endif
+
+//
+//
+////   end header file   /////////////////////////////////////////////////////
+#endif // STBI_INCLUDE_STB_IMAGE_H
+
+#ifdef STB_IMAGE_IMPLEMENTATION
+
+#if defined(STBI_ONLY_JPEG) || defined(STBI_ONLY_PNG) || defined(STBI_ONLY_BMP) \
+  || defined(STBI_ONLY_TGA) || defined(STBI_ONLY_GIF) || defined(STBI_ONLY_PSD) \
+  || defined(STBI_ONLY_HDR) || defined(STBI_ONLY_PIC) || defined(STBI_ONLY_PNM) \
+  || defined(STBI_ONLY_ZLIB)
+   #ifndef STBI_ONLY_JPEG
+   #define STBI_NO_JPEG
+   #endif
+   #ifndef STBI_ONLY_PNG
+   #define STBI_NO_PNG
+   #endif
+   #ifndef STBI_ONLY_BMP
+   #define STBI_NO_BMP
+   #endif
+   #ifndef STBI_ONLY_PSD
+   #define STBI_NO_PSD
+   #endif
+   #ifndef STBI_ONLY_TGA
+   #define STBI_NO_TGA
+   #endif
+   #ifndef STBI_ONLY_GIF
+   #define STBI_NO_GIF
+   #endif
+   #ifndef STBI_ONLY_HDR
+   #define STBI_NO_HDR
+   #endif
+   #ifndef STBI_ONLY_PIC
+   #define STBI_NO_PIC
+   #endif
+   #ifndef STBI_ONLY_PNM
+   #define STBI_NO_PNM
+   #endif
+#endif
+
+#if defined(STBI_NO_PNG) && !defined(STBI_SUPPORT_ZLIB) && !defined(STBI_NO_ZLIB)
+#define STBI_NO_ZLIB
+#endif
+
+
+#include <stdarg.h>
+#include <stddef.h> // ptrdiff_t on osx
+#include <stdlib.h>
+#include <string.h>
+#include <limits.h>
+
+#if !defined(STBI_NO_LINEAR) || !defined(STBI_NO_HDR)
+#include <math.h>  // ldexp, pow
+#endif
+
+#ifndef STBI_NO_STDIO
+#include <stdio.h>
+#endif
+
+#ifndef STBI_ASSERT
+#include <assert.h>
+#define STBI_ASSERT(x) assert(x)
+#endif
+
+#ifdef __cplusplus
+#define STBI_EXTERN extern "C"
+#else
+#define STBI_EXTERN extern
+#endif
+
+
+#ifndef _MSC_VER
+   #ifdef __cplusplus
+   #define stbi_inline inline
+   #else
+   #define stbi_inline
+   #endif
+#else
+   #define stbi_inline __forceinline
+#endif
+
+#ifndef STBI_NO_THREAD_LOCALS
+   #if defined(__cplusplus) &&  __cplusplus >= 201103L
+      #define STBI_THREAD_LOCAL       thread_local
+   #elif defined(__GNUC__) && __GNUC__ < 5
+      #define STBI_THREAD_LOCAL       __thread
+   #elif defined(_MSC_VER)
+      #define STBI_THREAD_LOCAL       __declspec(thread)
+   #elif defined (__STDC_VERSION__) && __STDC_VERSION__ >= 201112L && !defined(__STDC_NO_THREADS__)
+      #define STBI_THREAD_LOCAL       _Thread_local
+   #endif
+
+   #ifndef STBI_THREAD_LOCAL
+      #if defined(__GNUC__)
+        #define STBI_THREAD_LOCAL       __thread
+      #endif
+   #endif
+#endif
+
+#if defined(_MSC_VER) || defined(__SYMBIAN32__)
+typedef unsigned short stbi__uint16;
+typedef   signed short stbi__int16;
+typedef unsigned int   stbi__uint32;
+typedef   signed int   stbi__int32;
+#else
+#include <stdint.h>
+typedef uint16_t stbi__uint16;
+typedef int16_t  stbi__int16;
+typedef uint32_t stbi__uint32;
+typedef int32_t  stbi__int32;
+#endif
+
+// should produce compiler error if size is wrong
+typedef unsigned char validate_uint32[sizeof(stbi__uint32)==4 ? 1 : -1];
+
+#ifdef _MSC_VER
+#define STBI_NOTUSED(v)  (void)(v)
+#else
+#define STBI_NOTUSED(v)  (void)sizeof(v)
+#endif
+
+#ifdef _MSC_VER
+#define STBI_HAS_LROTL
+#endif
+
+#ifdef STBI_HAS_LROTL
+   #define stbi_lrot(x,y)  _lrotl(x,y)
+#else
+   #define stbi_lrot(x,y)  (((x) << (y)) | ((x) >> (-(y) & 31)))
+#endif
+
+#if defined(STBI_MALLOC) && defined(STBI_FREE) && (defined(STBI_REALLOC) || defined(STBI_REALLOC_SIZED))
+// ok
+#elif !defined(STBI_MALLOC) && !defined(STBI_FREE) && !defined(STBI_REALLOC) && !defined(STBI_REALLOC_SIZED)
+// ok
+#else
+#error "Must define all or none of STBI_MALLOC, STBI_FREE, and STBI_REALLOC (or STBI_REALLOC_SIZED)."
+#endif
+
+#ifndef STBI_MALLOC
+#define STBI_MALLOC(sz)           malloc(sz)
+#define STBI_REALLOC(p,newsz)     realloc(p,newsz)
+#define STBI_FREE(p)              free(p)
+#endif
+
+#ifndef STBI_REALLOC_SIZED
+#define STBI_REALLOC_SIZED(p,oldsz,newsz) STBI_REALLOC(p,newsz)
+#endif
+
+// x86/x64 detection
+#if defined(__x86_64__) || defined(_M_X64)
+#define STBI__X64_TARGET
+#elif defined(__i386) || defined(_M_IX86)
+#define STBI__X86_TARGET
+#endif
+
+#if defined(__GNUC__) && defined(STBI__X86_TARGET) && !defined(__SSE2__) && !defined(STBI_NO_SIMD)
+// gcc doesn't support sse2 intrinsics unless you compile with -msse2,
+// which in turn means it gets to use SSE2 everywhere. This is unfortunate,
+// but previous attempts to provide the SSE2 functions with runtime
+// detection caused numerous issues. The way architecture extensions are
+// exposed in GCC/Clang is, sadly, not really suited for one-file libs.
+// New behavior: if compiled with -msse2, we use SSE2 without any
+// detection; if not, we don't use it at all.
+#define STBI_NO_SIMD
+#endif
+
+#if defined(__MINGW32__) && defined(STBI__X86_TARGET) && !defined(STBI_MINGW_ENABLE_SSE2) && !defined(STBI_NO_SIMD)
+// Note that __MINGW32__ doesn't actually mean 32-bit, so we have to avoid STBI__X64_TARGET
+//
+// 32-bit MinGW wants ESP to be 16-byte aligned, but this is not in the
+// Windows ABI and VC++ as well as Windows DLLs don't maintain that invariant.
+// As a result, enabling SSE2 on 32-bit MinGW is dangerous when not
+// simultaneously enabling "-mstackrealign".
+//
+// See https://github.com/nothings/stb/issues/81 for more information.
+//
+// So default to no SSE2 on 32-bit MinGW. If you've read this far and added
+// -mstackrealign to your build settings, feel free to #define STBI_MINGW_ENABLE_SSE2.
+#define STBI_NO_SIMD
+#endif
+
+#if !defined(STBI_NO_SIMD) && (defined(STBI__X86_TARGET) || defined(STBI__X64_TARGET))
+#define STBI_SSE2
+#include <emmintrin.h>
+
+#ifdef _MSC_VER
+
+#if _MSC_VER >= 1400  // not VC6
+#include <intrin.h> // __cpuid
+static int stbi__cpuid3(void)
+{
+   int info[4];
+   __cpuid(info,1);
+   return info[3];
+}
+#else
+static int stbi__cpuid3(void)
+{
+   int res;
+   __asm {
+      mov  eax,1
+      cpuid
+      mov  res,edx
+   }
+   return res;
+}
+#endif
+
+#define STBI_SIMD_ALIGN(type, name) __declspec(align(16)) type name
+
+#if !defined(STBI_NO_JPEG) && defined(STBI_SSE2)
+static int stbi__sse2_available(void)
+{
+   int info3 = stbi__cpuid3();
+   return ((info3 >> 26) & 1) != 0;
+}
+#endif
+
+#else // assume GCC-style if not VC++
+#define STBI_SIMD_ALIGN(type, name) type name __attribute__((aligned(16)))
+
+#if !defined(STBI_NO_JPEG) && defined(STBI_SSE2)
+static int stbi__sse2_available(void)
+{
+   // If we're even attempting to compile this on GCC/Clang, that means
+   // -msse2 is on, which means the compiler is allowed to use SSE2
+   // instructions at will, and so are we.
+   return 1;
+}
+#endif
+
+#endif
+#endif
+
+// ARM NEON
+#if defined(STBI_NO_SIMD) && defined(STBI_NEON)
+#undef STBI_NEON
+#endif
+
+#ifdef STBI_NEON
+#include <arm_neon.h>
+#ifdef _MSC_VER
+#define STBI_SIMD_ALIGN(type, name) __declspec(align(16)) type name
+#else
+#define STBI_SIMD_ALIGN(type, name) type name __attribute__((aligned(16)))
+#endif
+#endif
+
+#ifndef STBI_SIMD_ALIGN
+#define STBI_SIMD_ALIGN(type, name) type name
+#endif
+
+#ifndef STBI_MAX_DIMENSIONS
+#define STBI_MAX_DIMENSIONS (1 << 24)
+#endif
+
+///////////////////////////////////////////////
+//
+//  stbi__context struct and start_xxx functions
+
+// stbi__context structure is our basic context used by all images, so it
+// contains all the IO context, plus some basic image information
+typedef struct
+{
+   stbi__uint32 img_x, img_y;
+   int img_n, img_out_n;
+
+   stbi_io_callbacks io;
+   void *io_user_data;
+
+   int read_from_callbacks;
+   int buflen;
+   stbi_uc buffer_start[128];
+   int callback_already_read;
+
+   stbi_uc *img_buffer, *img_buffer_end;
+   stbi_uc *img_buffer_original, *img_buffer_original_end;
+} stbi__context;
+
+
+static void stbi__refill_buffer(stbi__context *s);
+
+// initialize a memory-decode context
+static void stbi__start_mem(stbi__context *s, stbi_uc const *buffer, int len)
+{
+   s->io.read = NULL;
+   s->read_from_callbacks = 0;
+   s->callback_already_read = 0;
+   s->img_buffer = s->img_buffer_original = (stbi_uc *) buffer;
+   s->img_buffer_end = s->img_buffer_original_end = (stbi_uc *) buffer+len;
+}
+
+// initialize a callback-based context
+static void stbi__start_callbacks(stbi__context *s, stbi_io_callbacks *c, void *user)
+{
+   s->io = *c;
+   s->io_user_data = user;
+   s->buflen = sizeof(s->buffer_start);
+   s->read_from_callbacks = 1;
+   s->callback_already_read = 0;
+   s->img_buffer = s->img_buffer_original = s->buffer_start;
+   stbi__refill_buffer(s);
+   s->img_buffer_original_end = s->img_buffer_end;
+}
+
+#ifndef STBI_NO_STDIO
+
+static int stbi__stdio_read(void *user, char *data, int size)
+{
+   return (int) fread(data,1,size,(FILE*) user);
+}
+
+static void stbi__stdio_skip(void *user, int n)
+{
+   int ch;
+   fseek((FILE*) user, n, SEEK_CUR);
+   ch = fgetc((FILE*) user);  /* have to read a byte to reset feof()'s flag */
+   if (ch != EOF) {
+      ungetc(ch, (FILE *) user);  /* push byte back onto stream if valid. */
+   }
+}
+
+static int stbi__stdio_eof(void *user)
+{
+   return feof((FILE*) user) || ferror((FILE *) user);
+}
+
+static stbi_io_callbacks stbi__stdio_callbacks =
+{
+   stbi__stdio_read,
+   stbi__stdio_skip,
+   stbi__stdio_eof,
+};
+
+static void stbi__start_file(stbi__context *s, FILE *f)
+{
+   stbi__start_callbacks(s, &stbi__stdio_callbacks, (void *) f);
+}
+
+//static void stop_file(stbi__context *s) { }
+
+#endif // !STBI_NO_STDIO
+
+static void stbi__rewind(stbi__context *s)
+{
+   // conceptually rewind SHOULD rewind to the beginning of the stream,
+   // but we just rewind to the beginning of the initial buffer, because
+   // we only use it after doing 'test', which only ever looks at at most 92 bytes
+   s->img_buffer = s->img_buffer_original;
+   s->img_buffer_end = s->img_buffer_original_end;
+}
+
+enum
+{
+   STBI_ORDER_RGB,
+   STBI_ORDER_BGR
+};
+
+typedef struct
+{
+   int bits_per_channel;
+   int num_channels;
+   int channel_order;
+} stbi__result_info;
+
+#ifndef STBI_NO_JPEG
+static int      stbi__jpeg_test(stbi__context *s);
+static void    *stbi__jpeg_load(stbi__context *s, int *x, int *y, int *comp, int req_comp, stbi__result_info *ri);
+static int      stbi__jpeg_info(stbi__context *s, int *x, int *y, int *comp);
+#endif
+
+#ifndef STBI_NO_PNG
+static int      stbi__png_test(stbi__context *s);
+static void    *stbi__png_load(stbi__context *s, int *x, int *y, int *comp, int req_comp, stbi__result_info *ri);
+static int      stbi__png_info(stbi__context *s, int *x, int *y, int *comp);
+static int      stbi__png_is16(stbi__context *s);
+#endif
+
+#ifndef STBI_NO_BMP
+static int      stbi__bmp_test(stbi__context *s);
+static void    *stbi__bmp_load(stbi__context *s, int *x, int *y, int *comp, int req_comp, stbi__result_info *ri);
+static int      stbi__bmp_info(stbi__context *s, int *x, int *y, int *comp);
+#endif
+
+#ifndef STBI_NO_TGA
+static int      stbi__tga_test(stbi__context *s);
+static void    *stbi__tga_load(stbi__context *s, int *x, int *y, int *comp, int req_comp, stbi__result_info *ri);
+static int      stbi__tga_info(stbi__context *s, int *x, int *y, int *comp);
+#endif
+
+#ifndef STBI_NO_PSD
+static int      stbi__psd_test(stbi__context *s);
+static void    *stbi__psd_load(stbi__context *s, int *x, int *y, int *comp, int req_comp, stbi__result_info *ri, int bpc);
+static int      stbi__psd_info(stbi__context *s, int *x, int *y, int *comp);
+static int      stbi__psd_is16(stbi__context *s);
+#endif
+
+#ifndef STBI_NO_HDR
+static int      stbi__hdr_test(stbi__context *s);
+static float   *stbi__hdr_load(stbi__context *s, int *x, int *y, int *comp, int req_comp, stbi__result_info *ri);
+static int      stbi__hdr_info(stbi__context *s, int *x, int *y, int *comp);
+#endif
+
+#ifndef STBI_NO_PIC
+static int      stbi__pic_test(stbi__context *s);
+static void    *stbi__pic_load(stbi__context *s, int *x, int *y, int *comp, int req_comp, stbi__result_info *ri);
+static int      stbi__pic_info(stbi__context *s, int *x, int *y, int *comp);
+#endif
+
+#ifndef STBI_NO_GIF
+static int      stbi__gif_test(stbi__context *s);
+static void    *stbi__gif_load(stbi__context *s, int *x, int *y, int *comp, int req_comp, stbi__result_info *ri);
+static void    *stbi__load_gif_main(stbi__context *s, int **delays, int *x, int *y, int *z, int *comp, int req_comp);
+static int      stbi__gif_info(stbi__context *s, int *x, int *y, int *comp);
+#endif
+
+#ifndef STBI_NO_PNM
+static int      stbi__pnm_test(stbi__context *s);
+static void    *stbi__pnm_load(stbi__context *s, int *x, int *y, int *comp, int req_comp, stbi__result_info *ri);
+static int      stbi__pnm_info(stbi__context *s, int *x, int *y, int *comp);
+static int      stbi__pnm_is16(stbi__context *s);
+#endif
+
+static
+#ifdef STBI_THREAD_LOCAL
+STBI_THREAD_LOCAL
+#endif
+const char *stbi__g_failure_reason;
+
+STBIDEF const char *stbi_failure_reason(void)
+{
+   return stbi__g_failure_reason;
+}
+
+#ifndef STBI_NO_FAILURE_STRINGS
+static int stbi__err(const char *str)
+{
+   stbi__g_failure_reason = str;
+   return 0;
+}
+#endif
+
+static void *stbi__malloc(size_t size)
+{
+    return STBI_MALLOC(size);
+}
+
+// stb_image uses ints pervasively, including for offset calculations.
+// therefore the largest decoded image size we can support with the
+// current code, even on 64-bit targets, is INT_MAX. this is not a
+// significant limitation for the intended use case.
+//
+// we do, however, need to make sure our size calculations don't
+// overflow. hence a few helper functions for size calculations that
+// multiply integers together, making sure that they're non-negative
+// and no overflow occurs.
+
+// return 1 if the sum is valid, 0 on overflow.
+// negative terms are considered invalid.
+static int stbi__addsizes_valid(int a, int b)
+{
+   if (b < 0) return 0;
+   // now 0 <= b <= INT_MAX, hence also
+   // 0 <= INT_MAX - b <= INTMAX.
+   // And "a + b <= INT_MAX" (which might overflow) is the
+   // same as a <= INT_MAX - b (no overflow)
+   return a <= INT_MAX - b;
+}
+
+// returns 1 if the product is valid, 0 on overflow.
+// negative factors are considered invalid.
+static int stbi__mul2sizes_valid(int a, int b)
+{
+   if (a < 0 || b < 0) return 0;
+   if (b == 0) return 1; // mul-by-0 is always safe
+   // portable way to check for no overflows in a*b
+   return a <= INT_MAX/b;
+}
+
+#if !defined(STBI_NO_JPEG) || !defined(STBI_NO_PNG) || !defined(STBI_NO_TGA) || !defined(STBI_NO_HDR)
+// returns 1 if "a*b + add" has no negative terms/factors and doesn't overflow
+static int stbi__mad2sizes_valid(int a, int b, int add)
+{
+   return stbi__mul2sizes_valid(a, b) && stbi__addsizes_valid(a*b, add);
+}
+#endif
+
+// returns 1 if "a*b*c + add" has no negative terms/factors and doesn't overflow
+static int stbi__mad3sizes_valid(int a, int b, int c, int add)
+{
+   return stbi__mul2sizes_valid(a, b) && stbi__mul2sizes_valid(a*b, c) &&
+      stbi__addsizes_valid(a*b*c, add);
+}
+
+// returns 1 if "a*b*c*d + add" has no negative terms/factors and doesn't overflow
+#if !defined(STBI_NO_LINEAR) || !defined(STBI_NO_HDR) || !defined(STBI_NO_PNM)
+static int stbi__mad4sizes_valid(int a, int b, int c, int d, int add)
+{
+   return stbi__mul2sizes_valid(a, b) && stbi__mul2sizes_valid(a*b, c) &&
+      stbi__mul2sizes_valid(a*b*c, d) && stbi__addsizes_valid(a*b*c*d, add);
+}
+#endif
+
+#if !defined(STBI_NO_JPEG) || !defined(STBI_NO_PNG) || !defined(STBI_NO_TGA) || !defined(STBI_NO_HDR)
+// mallocs with size overflow checking
+static void *stbi__malloc_mad2(int a, int b, int add)
+{
+   if (!stbi__mad2sizes_valid(a, b, add)) return NULL;
+   return stbi__malloc(a*b + add);
+}
+#endif
+
+static void *stbi__malloc_mad3(int a, int b, int c, int add)
+{
+   if (!stbi__mad3sizes_valid(a, b, c, add)) return NULL;
+   return stbi__malloc(a*b*c + add);
+}
+
+#if !defined(STBI_NO_LINEAR) || !defined(STBI_NO_HDR) || !defined(STBI_NO_PNM)
+static void *stbi__malloc_mad4(int a, int b, int c, int d, int add)
+{
+   if (!stbi__mad4sizes_valid(a, b, c, d, add)) return NULL;
+   return stbi__malloc(a*b*c*d + add);
+}
+#endif
+
+// returns 1 if the sum of two signed ints is valid (between -2^31 and 2^31-1 inclusive), 0 on overflow.
+static int stbi__addints_valid(int a, int b)
+{
+   if ((a >= 0) != (b >= 0)) return 1; // a and b have different signs, so no overflow
+   if (a < 0 && b < 0) return a >= INT_MIN - b; // same as a + b >= INT_MIN; INT_MIN - b cannot overflow since b < 0.
+   return a <= INT_MAX - b;
+}
+
+// returns 1 if the product of two signed shorts is valid, 0 on overflow.
+static int stbi__mul2shorts_valid(short a, short b)
+{
+   if (b == 0 || b == -1) return 1; // multiplication by 0 is always 0; check for -1 so SHRT_MIN/b doesn't overflow
+   if ((a >= 0) == (b >= 0)) return a <= SHRT_MAX/b; // product is positive, so similar to mul2sizes_valid
+   if (b < 0) return a <= SHRT_MIN / b; // same as a * b >= SHRT_MIN
+   return a >= SHRT_MIN / b;
+}
+
+// stbi__err - error
+// stbi__errpf - error returning pointer to float
+// stbi__errpuc - error returning pointer to unsigned char
+
+#ifdef STBI_NO_FAILURE_STRINGS
+   #define stbi__err(x,y)  0
+#elif defined(STBI_FAILURE_USERMSG)
+   #define stbi__err(x,y)  stbi__err(y)
+#else
+   #define stbi__err(x,y)  stbi__err(x)
+#endif
+
+#define stbi__errpf(x,y)   ((float *)(size_t) (stbi__err(x,y)?NULL:NULL))
+#define stbi__errpuc(x,y)  ((unsigned char *)(size_t) (stbi__err(x,y)?NULL:NULL))
+
+STBIDEF void stbi_image_free(void *retval_from_stbi_load)
+{
+   STBI_FREE(retval_from_stbi_load);
+}
+
+#ifndef STBI_NO_LINEAR
+static float   *stbi__ldr_to_hdr(stbi_uc *data, int x, int y, int comp);
+#endif
+
+#ifndef STBI_NO_HDR
+static stbi_uc *stbi__hdr_to_ldr(float   *data, int x, int y, int comp);
+#endif
+
+static int stbi__vertically_flip_on_load_global = 0;
+
+STBIDEF void stbi_set_flip_vertically_on_load(int flag_true_if_should_flip)
+{
+   stbi__vertically_flip_on_load_global = flag_true_if_should_flip;
+}
+
+#ifndef STBI_THREAD_LOCAL
+#define stbi__vertically_flip_on_load  stbi__vertically_flip_on_load_global
+#else
+static STBI_THREAD_LOCAL int stbi__vertically_flip_on_load_local, stbi__vertically_flip_on_load_set;
+
+STBIDEF void stbi_set_flip_vertically_on_load_thread(int flag_true_if_should_flip)
+{
+   stbi__vertically_flip_on_load_local = flag_true_if_should_flip;
+   stbi__vertically_flip_on_load_set = 1;
+}
+
+#define stbi__vertically_flip_on_load  (stbi__vertically_flip_on_load_set       \
+                                         ? stbi__vertically_flip_on_load_local  \
+                                         : stbi__vertically_flip_on_load_global)
+#endif // STBI_THREAD_LOCAL
+
+static void *stbi__load_main(stbi__context *s, int *x, int *y, int *comp, int req_comp, stbi__result_info *ri, int bpc)
+{
+   memset(ri, 0, sizeof(*ri)); // make sure it's initialized if we add new fields
+   ri->bits_per_channel = 8; // default is 8 so most paths don't have to be changed
+   ri->channel_order = STBI_ORDER_RGB; // all current input & output are this, but this is here so we can add BGR order
+   ri->num_channels = 0;
+
+   // test the formats with a very explicit header first (at least a FOURCC
+   // or distinctive magic number first)
+   #ifndef STBI_NO_PNG
+   if (stbi__png_test(s))  return stbi__png_load(s,x,y,comp,req_comp, ri);
+   #endif
+   #ifndef STBI_NO_BMP
+   if (stbi__bmp_test(s))  return stbi__bmp_load(s,x,y,comp,req_comp, ri);
+   #endif
+   #ifndef STBI_NO_GIF
+   if (stbi__gif_test(s))  return stbi__gif_load(s,x,y,comp,req_comp, ri);
+   #endif
+   #ifndef STBI_NO_PSD
+   if (stbi__psd_test(s))  return stbi__psd_load(s,x,y,comp,req_comp, ri, bpc);
+   #else
+   STBI_NOTUSED(bpc);
+   #endif
+   #ifndef STBI_NO_PIC
+   if (stbi__pic_test(s))  return stbi__pic_load(s,x,y,comp,req_comp, ri);
+   #endif
+
+   // then the formats that can end up attempting to load with just 1 or 2
+   // bytes matching expectations; these are prone to false positives, so
+   // try them later
+   #ifndef STBI_NO_JPEG
+   if (stbi__jpeg_test(s)) return stbi__jpeg_load(s,x,y,comp,req_comp, ri);
+   #endif
+   #ifndef STBI_NO_PNM
+   if (stbi__pnm_test(s))  return stbi__pnm_load(s,x,y,comp,req_comp, ri);
+   #endif
+
+   #ifndef STBI_NO_HDR
+   if (stbi__hdr_test(s)) {
+      float *hdr = stbi__hdr_load(s, x,y,comp,req_comp, ri);
+      return stbi__hdr_to_ldr(hdr, *x, *y, req_comp ? req_comp : *comp);
+   }
+   #endif
+
+   #ifndef STBI_NO_TGA
+   // test tga last because it's a crappy test!
+   if (stbi__tga_test(s))
+      return stbi__tga_load(s,x,y,comp,req_comp, ri);
+   #endif
+
+   return stbi__errpuc("unknown image type", "Image not of any known type, or corrupt");
+}
+
+static stbi_uc *stbi__convert_16_to_8(stbi__uint16 *orig, int w, int h, int channels)
+{
+   int i;
+   int img_len = w * h * channels;
+   stbi_uc *reduced;
+
+   reduced = (stbi_uc *) stbi__malloc(img_len);
+   if (reduced == NULL) return stbi__errpuc("outofmem", "Out of memory");
+
+   for (i = 0; i < img_len; ++i)
+      reduced[i] = (stbi_uc)((orig[i] >> 8) & 0xFF); // top half of each byte is sufficient approx of 16->8 bit scaling
+
+   STBI_FREE(orig);
+   return reduced;
+}
+
+static stbi__uint16 *stbi__convert_8_to_16(stbi_uc *orig, int w, int h, int channels)
+{
+   int i;
+   int img_len = w * h * channels;
+   stbi__uint16 *enlarged;
+
+   enlarged = (stbi__uint16 *) stbi__malloc(img_len*2);
+   if (enlarged == NULL) return (stbi__uint16 *) stbi__errpuc("outofmem", "Out of memory");
+
+   for (i = 0; i < img_len; ++i)
+      enlarged[i] = (stbi__uint16)((orig[i] << 8) + orig[i]); // replicate to high and low byte, maps 0->0, 255->0xffff
+
+   STBI_FREE(orig);
+   return enlarged;
+}
+
+static void stbi__vertical_flip(void *image, int w, int h, int bytes_per_pixel)
+{
+   int row;
+   size_t bytes_per_row = (size_t)w * bytes_per_pixel;
+   stbi_uc temp[2048];
+   stbi_uc *bytes = (stbi_uc *)image;
+
+   for (row = 0; row < (h>>1); row++) {
+      stbi_uc *row0 = bytes + row*bytes_per_row;
+      stbi_uc *row1 = bytes + (h - row - 1)*bytes_per_row;
+      // swap row0 with row1
+      size_t bytes_left = bytes_per_row;
+      while (bytes_left) {
+         size_t bytes_copy = (bytes_left < sizeof(temp)) ? bytes_left : sizeof(temp);
+         memcpy(temp, row0, bytes_copy);
+         memcpy(row0, row1, bytes_copy);
+         memcpy(row1, temp, bytes_copy);
+         row0 += bytes_copy;
+         row1 += bytes_copy;
+         bytes_left -= bytes_copy;
+      }
+   }
+}
+
+#ifndef STBI_NO_GIF
+static void stbi__vertical_flip_slices(void *image, int w, int h, int z, int bytes_per_pixel)
+{
+   int slice;
+   int slice_size = w * h * bytes_per_pixel;
+
+   stbi_uc *bytes = (stbi_uc *)image;
+   for (slice = 0; slice < z; ++slice) {
+      stbi__vertical_flip(bytes, w, h, bytes_per_pixel);
+      bytes += slice_size;
+   }
+}
+#endif
+
+static unsigned char *stbi__load_and_postprocess_8bit(stbi__context *s, int *x, int *y, int *comp, int req_comp)
+{
+   stbi__result_info ri;
+   void *result = stbi__load_main(s, x, y, comp, req_comp, &ri, 8);
+
+   if (result == NULL)
+      return NULL;
+
+   // it is the responsibility of the loaders to make sure we get either 8 or 16 bit.
+   STBI_ASSERT(ri.bits_per_channel == 8 || ri.bits_per_channel == 16);
+
+   if (ri.bits_per_channel != 8) {
+      result = stbi__convert_16_to_8((stbi__uint16 *) result, *x, *y, req_comp == 0 ? *comp : req_comp);
+      ri.bits_per_channel = 8;
+   }
+
+   // @TODO: move stbi__convert_format to here
+
+   if (stbi__vertically_flip_on_load) {
+      int channels = req_comp ? req_comp : *comp;
+      stbi__vertical_flip(result, *x, *y, channels * sizeof(stbi_uc));
+   }
+
+   return (unsigned char *) result;
+}
+
+static stbi__uint16 *stbi__load_and_postprocess_16bit(stbi__context *s, int *x, int *y, int *comp, int req_comp)
+{
+   stbi__result_info ri;
+   void *result = stbi__load_main(s, x, y, comp, req_comp, &ri, 16);
+
+   if (result == NULL)
+      return NULL;
+
+   // it is the responsibility of the loaders to make sure we get either 8 or 16 bit.
+   STBI_ASSERT(ri.bits_per_channel == 8 || ri.bits_per_channel == 16);
+
+   if (ri.bits_per_channel != 16) {
+      result = stbi__convert_8_to_16((stbi_uc *) result, *x, *y, req_comp == 0 ? *comp : req_comp);
+      ri.bits_per_channel = 16;
+   }
+
+   // @TODO: move stbi__convert_format16 to here
+   // @TODO: special case RGB-to-Y (and RGBA-to-YA) for 8-bit-to-16-bit case to keep more precision
+
+   if (stbi__vertically_flip_on_load) {
+      int channels = req_comp ? req_comp : *comp;
+      stbi__vertical_flip(result, *x, *y, channels * sizeof(stbi__uint16));
+   }
+
+   return (stbi__uint16 *) result;
+}
+
+#if !defined(STBI_NO_HDR) && !defined(STBI_NO_LINEAR)
+static void stbi__float_postprocess(float *result, int *x, int *y, int *comp, int req_comp)
+{
+   if (stbi__vertically_flip_on_load && result != NULL) {
+      int channels = req_comp ? req_comp : *comp;
+      stbi__vertical_flip(result, *x, *y, channels * sizeof(float));
+   }
+}
+#endif
+
+#ifndef STBI_NO_STDIO
+
+#if defined(_WIN32) && defined(STBI_WINDOWS_UTF8)
+STBI_EXTERN __declspec(dllimport) int __stdcall MultiByteToWideChar(unsigned int cp, unsigned long flags, const char *str, int cbmb, wchar_t *widestr, int cchwide);
+STBI_EXTERN __declspec(dllimport) int __stdcall WideCharToMultiByte(unsigned int cp, unsigned long flags, const wchar_t *widestr, int cchwide, char *str, int cbmb, const char *defchar, int *used_default);
+#endif
+
+#if defined(_WIN32) && defined(STBI_WINDOWS_UTF8)
+STBIDEF int stbi_convert_wchar_to_utf8(char *buffer, size_t bufferlen, const wchar_t* input)
+{
+        return WideCharToMultiByte(65001 /* UTF8 */, 0, input, -1, buffer, (int) bufferlen, NULL, NULL);
+}
+#endif
+
+static FILE *stbi__fopen(char const *filename, char const *mode)
+{
+   FILE *f;
+#if defined(_WIN32) && defined(STBI_WINDOWS_UTF8)
+   wchar_t wMode[64];
+   wchar_t wFilename[1024];
+        if (0 == MultiByteToWideChar(65001 /* UTF8 */, 0, filename, -1, wFilename, sizeof(wFilename)/sizeof(*wFilename)))
+      return 0;
+
+        if (0 == MultiByteToWideChar(65001 /* UTF8 */, 0, mode, -1, wMode, sizeof(wMode)/sizeof(*wMode)))
+      return 0;
+
+#if defined(_MSC_VER) && _MSC_VER >= 1400
+        if (0 != _wfopen_s(&f, wFilename, wMode))
+                f = 0;
+#else
+   f = _wfopen(wFilename, wMode);
+#endif
+
+#elif defined(_MSC_VER) && _MSC_VER >= 1400
+   if (0 != fopen_s(&f, filename, mode))
+      f=0;
+#else
+   f = fopen(filename, mode);
+#endif
+   return f;
+}
+
+
+STBIDEF stbi_uc *stbi_load(char const *filename, int *x, int *y, int *comp, int req_comp)
+{
+   FILE *f = stbi__fopen(filename, "rb");
+   unsigned char *result;
+   if (!f) return stbi__errpuc("can't fopen", "Unable to open file");
+   result = stbi_load_from_file(f,x,y,comp,req_comp);
+   fclose(f);
+   return result;
+}
+
+STBIDEF stbi_uc *stbi_load_from_file(FILE *f, int *x, int *y, int *comp, int req_comp)
+{
+   unsigned char *result;
+   stbi__context s;
+   stbi__start_file(&s,f);
+   result = stbi__load_and_postprocess_8bit(&s,x,y,comp,req_comp);
+   if (result) {
+      // need to 'unget' all the characters in the IO buffer
+      fseek(f, - (int) (s.img_buffer_end - s.img_buffer), SEEK_CUR);
+   }
+   return result;
+}
+
+STBIDEF stbi__uint16 *stbi_load_from_file_16(FILE *f, int *x, int *y, int *comp, int req_comp)
+{
+   stbi__uint16 *result;
+   stbi__context s;
+   stbi__start_file(&s,f);
+   result = stbi__load_and_postprocess_16bit(&s,x,y,comp,req_comp);
+   if (result) {
+      // need to 'unget' all the characters in the IO buffer
+      fseek(f, - (int) (s.img_buffer_end - s.img_buffer), SEEK_CUR);
+   }
+   return result;
+}
+
+STBIDEF stbi_us *stbi_load_16(char const *filename, int *x, int *y, int *comp, int req_comp)
+{
+   FILE *f = stbi__fopen(filename, "rb");
+   stbi__uint16 *result;
+   if (!f) return (stbi_us *) stbi__errpuc("can't fopen", "Unable to open file");
+   result = stbi_load_from_file_16(f,x,y,comp,req_comp);
+   fclose(f);
+   return result;
+}
+
+
+#endif //!STBI_NO_STDIO
+
+STBIDEF stbi_us *stbi_load_16_from_memory(stbi_uc const *buffer, int len, int *x, int *y, int *channels_in_file, int desired_channels)
+{
+   stbi__context s;
+   stbi__start_mem(&s,buffer,len);
+   return stbi__load_and_postprocess_16bit(&s,x,y,channels_in_file,desired_channels);
+}
+
+STBIDEF stbi_us *stbi_load_16_from_callbacks(stbi_io_callbacks const *clbk, void *user, int *x, int *y, int *channels_in_file, int desired_channels)
+{
+   stbi__context s;
+   stbi__start_callbacks(&s, (stbi_io_callbacks *)clbk, user);
+   return stbi__load_and_postprocess_16bit(&s,x,y,channels_in_file,desired_channels);
+}
+
+STBIDEF stbi_uc *stbi_load_from_memory(stbi_uc const *buffer, int len, int *x, int *y, int *comp, int req_comp)
+{
+   stbi__context s;
+   stbi__start_mem(&s,buffer,len);
+   return stbi__load_and_postprocess_8bit(&s,x,y,comp,req_comp);
+}
+
+STBIDEF stbi_uc *stbi_load_from_callbacks(stbi_io_callbacks const *clbk, void *user, int *x, int *y, int *comp, int req_comp)
+{
+   stbi__context s;
+   stbi__start_callbacks(&s, (stbi_io_callbacks *) clbk, user);
+   return stbi__load_and_postprocess_8bit(&s,x,y,comp,req_comp);
+}
+
+#ifndef STBI_NO_GIF
+STBIDEF stbi_uc *stbi_load_gif_from_memory(stbi_uc const *buffer, int len, int **delays, int *x, int *y, int *z, int *comp, int req_comp)
+{
+   unsigned char *result;
+   stbi__context s;
+   stbi__start_mem(&s,buffer,len);
+
+   result = (unsigned char*) stbi__load_gif_main(&s, delays, x, y, z, comp, req_comp);
+   if (stbi__vertically_flip_on_load) {
+      stbi__vertical_flip_slices( result, *x, *y, *z, *comp );
+   }
+
+   return result;
+}
+#endif
+
+#ifndef STBI_NO_LINEAR
+static float *stbi__loadf_main(stbi__context *s, int *x, int *y, int *comp, int req_comp)
+{
+   unsigned char *data;
+   #ifndef STBI_NO_HDR
+   if (stbi__hdr_test(s)) {
+      stbi__result_info ri;
+      float *hdr_data = stbi__hdr_load(s,x,y,comp,req_comp, &ri);
+      if (hdr_data)
+         stbi__float_postprocess(hdr_data,x,y,comp,req_comp);
+      return hdr_data;
+   }
+   #endif
+   data = stbi__load_and_postprocess_8bit(s, x, y, comp, req_comp);
+   if (data)
+      return stbi__ldr_to_hdr(data, *x, *y, req_comp ? req_comp : *comp);
+   return stbi__errpf("unknown image type", "Image not of any known type, or corrupt");
+}
+
+STBIDEF float *stbi_loadf_from_memory(stbi_uc const *buffer, int len, int *x, int *y, int *comp, int req_comp)
+{
+   stbi__context s;
+   stbi__start_mem(&s,buffer,len);
+   return stbi__loadf_main(&s,x,y,comp,req_comp);
+}
+
+STBIDEF float *stbi_loadf_from_callbacks(stbi_io_callbacks const *clbk, void *user, int *x, int *y, int *comp, int req_comp)
+{
+   stbi__context s;
+   stbi__start_callbacks(&s, (stbi_io_callbacks *) clbk, user);
+   return stbi__loadf_main(&s,x,y,comp,req_comp);
+}
+
+#ifndef STBI_NO_STDIO
+STBIDEF float *stbi_loadf(char const *filename, int *x, int *y, int *comp, int req_comp)
+{
+   float *result;
+   FILE *f = stbi__fopen(filename, "rb");
+   if (!f) return stbi__errpf("can't fopen", "Unable to open file");
+   result = stbi_loadf_from_file(f,x,y,comp,req_comp);
+   fclose(f);
+   return result;
+}
+
+STBIDEF float *stbi_loadf_from_file(FILE *f, int *x, int *y, int *comp, int req_comp)
+{
+   stbi__context s;
+   stbi__start_file(&s,f);
+   return stbi__loadf_main(&s,x,y,comp,req_comp);
+}
+#endif // !STBI_NO_STDIO
+
+#endif // !STBI_NO_LINEAR
+
+// these is-hdr-or-not is defined independent of whether STBI_NO_LINEAR is
+// defined, for API simplicity; if STBI_NO_LINEAR is defined, it always
+// reports false!
+
+STBIDEF int stbi_is_hdr_from_memory(stbi_uc const *buffer, int len)
+{
+   #ifndef STBI_NO_HDR
+   stbi__context s;
+   stbi__start_mem(&s,buffer,len);
+   return stbi__hdr_test(&s);
+   #else
+   STBI_NOTUSED(buffer);
+   STBI_NOTUSED(len);
+   return 0;
+   #endif
+}
+
+#ifndef STBI_NO_STDIO
+STBIDEF int      stbi_is_hdr          (char const *filename)
+{
+   FILE *f = stbi__fopen(filename, "rb");
+   int result=0;
+   if (f) {
+      result = stbi_is_hdr_from_file(f);
+      fclose(f);
+   }
+   return result;
+}
+
+STBIDEF int stbi_is_hdr_from_file(FILE *f)
+{
+   #ifndef STBI_NO_HDR
+   long pos = ftell(f);
+   int res;
+   stbi__context s;
+   stbi__start_file(&s,f);
+   res = stbi__hdr_test(&s);
+   fseek(f, pos, SEEK_SET);
+   return res;
+   #else
+   STBI_NOTUSED(f);
+   return 0;
+   #endif
+}
+#endif // !STBI_NO_STDIO
+
+STBIDEF int      stbi_is_hdr_from_callbacks(stbi_io_callbacks const *clbk, void *user)
+{
+   #ifndef STBI_NO_HDR
+   stbi__context s;
+   stbi__start_callbacks(&s, (stbi_io_callbacks *) clbk, user);
+   return stbi__hdr_test(&s);
+   #else
+   STBI_NOTUSED(clbk);
+   STBI_NOTUSED(user);
+   return 0;
+   #endif
+}
+
+#ifndef STBI_NO_LINEAR
+static float stbi__l2h_gamma=2.2f, stbi__l2h_scale=1.0f;
+
+STBIDEF void   stbi_ldr_to_hdr_gamma(float gamma) { stbi__l2h_gamma = gamma; }
+STBIDEF void   stbi_ldr_to_hdr_scale(float scale) { stbi__l2h_scale = scale; }
+#endif
+
+static float stbi__h2l_gamma_i=1.0f/2.2f, stbi__h2l_scale_i=1.0f;
+
+STBIDEF void   stbi_hdr_to_ldr_gamma(float gamma) { stbi__h2l_gamma_i = 1/gamma; }
+STBIDEF void   stbi_hdr_to_ldr_scale(float scale) { stbi__h2l_scale_i = 1/scale; }
+
+
+//////////////////////////////////////////////////////////////////////////////
+//
+// Common code used by all image loaders
+//
+
+enum
+{
+   STBI__SCAN_load=0,
+   STBI__SCAN_type,
+   STBI__SCAN_header
+};
+
+static void stbi__refill_buffer(stbi__context *s)
+{
+   int n = (s->io.read)(s->io_user_data,(char*)s->buffer_start,s->buflen);
+   s->callback_already_read += (int) (s->img_buffer - s->img_buffer_original);
+   if (n == 0) {
+      // at end of file, treat same as if from memory, but need to handle case
+      // where s->img_buffer isn't pointing to safe memory, e.g. 0-byte file
+      s->read_from_callbacks = 0;
+      s->img_buffer = s->buffer_start;
+      s->img_buffer_end = s->buffer_start+1;
+      *s->img_buffer = 0;
+   } else {
+      s->img_buffer = s->buffer_start;
+      s->img_buffer_end = s->buffer_start + n;
+   }
+}
+
+stbi_inline static stbi_uc stbi__get8(stbi__context *s)
+{
+   if (s->img_buffer < s->img_buffer_end)
+      return *s->img_buffer++;
+   if (s->read_from_callbacks) {
+      stbi__refill_buffer(s);
+      return *s->img_buffer++;
+   }
+   return 0;
+}
+
+#if defined(STBI_NO_JPEG) && defined(STBI_NO_HDR) && defined(STBI_NO_PIC) && defined(STBI_NO_PNM)
+// nothing
+#else
+stbi_inline static int stbi__at_eof(stbi__context *s)
+{
+   if (s->io.read) {
+      if (!(s->io.eof)(s->io_user_data)) return 0;
+      // if feof() is true, check if buffer = end
+      // special case: we've only got the special 0 character at the end
+      if (s->read_from_callbacks == 0) return 1;
+   }
+
+   return s->img_buffer >= s->img_buffer_end;
+}
+#endif
+
+#if defined(STBI_NO_JPEG) && defined(STBI_NO_PNG) && defined(STBI_NO_BMP) && defined(STBI_NO_PSD) && defined(STBI_NO_TGA) && defined(STBI_NO_GIF) && defined(STBI_NO_PIC)
+// nothing
+#else
+static void stbi__skip(stbi__context *s, int n)
+{
+   if (n == 0) return;  // already there!
+   if (n < 0) {
+      s->img_buffer = s->img_buffer_end;
+      return;
+   }
+   if (s->io.read) {
+      int blen = (int) (s->img_buffer_end - s->img_buffer);
+      if (blen < n) {
+         s->img_buffer = s->img_buffer_end;
+         (s->io.skip)(s->io_user_data, n - blen);
+         return;
+      }
+   }
+   s->img_buffer += n;
+}
+#endif
+
+#if defined(STBI_NO_PNG) && defined(STBI_NO_TGA) && defined(STBI_NO_HDR) && defined(STBI_NO_PNM)
+// nothing
+#else
+static int stbi__getn(stbi__context *s, stbi_uc *buffer, int n)
+{
+   if (s->io.read) {
+      int blen = (int) (s->img_buffer_end - s->img_buffer);
+      if (blen < n) {
+         int res, count;
+
+         memcpy(buffer, s->img_buffer, blen);
+
+         count = (s->io.read)(s->io_user_data, (char*) buffer + blen, n - blen);
+         res = (count == (n-blen));
+         s->img_buffer = s->img_buffer_end;
+         return res;
+      }
+   }
+
+   if (s->img_buffer+n <= s->img_buffer_end) {
+      memcpy(buffer, s->img_buffer, n);
+      s->img_buffer += n;
+      return 1;
+   } else
+      return 0;
+}
+#endif
+
+#if defined(STBI_NO_JPEG) && defined(STBI_NO_PNG) && defined(STBI_NO_PSD) && defined(STBI_NO_PIC)
+// nothing
+#else
+static int stbi__get16be(stbi__context *s)
+{
+   int z = stbi__get8(s);
+   return (z << 8) + stbi__get8(s);
+}
+#endif
+
+#if defined(STBI_NO_PNG) && defined(STBI_NO_PSD) && defined(STBI_NO_PIC)
+// nothing
+#else
+static stbi__uint32 stbi__get32be(stbi__context *s)
+{
+   stbi__uint32 z = stbi__get16be(s);
+   return (z << 16) + stbi__get16be(s);
+}
+#endif
+
+#if defined(STBI_NO_BMP) && defined(STBI_NO_TGA) && defined(STBI_NO_GIF)
+// nothing
+#else
+static int stbi__get16le(stbi__context *s)
+{
+   int z = stbi__get8(s);
+   return z + (stbi__get8(s) << 8);
+}
+#endif
+
+#ifndef STBI_NO_BMP
+static stbi__uint32 stbi__get32le(stbi__context *s)
+{
+   stbi__uint32 z = stbi__get16le(s);
+   z += (stbi__uint32)stbi__get16le(s) << 16;
+   return z;
+}
+#endif
+
+#define STBI__BYTECAST(x)  ((stbi_uc) ((x) & 255))  // truncate int to byte without warnings
+
+#if defined(STBI_NO_JPEG) && defined(STBI_NO_PNG) && defined(STBI_NO_BMP) && defined(STBI_NO_PSD) && defined(STBI_NO_TGA) && defined(STBI_NO_GIF) && defined(STBI_NO_PIC) && defined(STBI_NO_PNM)
+// nothing
+#else
+//////////////////////////////////////////////////////////////////////////////
+//
+//  generic converter from built-in img_n to req_comp
+//    individual types do this automatically as much as possible (e.g. jpeg
+//    does all cases internally since it needs to colorspace convert anyway,
+//    and it never has alpha, so very few cases ). png can automatically
+//    interleave an alpha=255 channel, but falls back to this for other cases
+//
+//  assume data buffer is malloced, so malloc a new one and free that one
+//  only failure mode is malloc failing
+
+static stbi_uc stbi__compute_y(int r, int g, int b)
+{
+   return (stbi_uc) (((r*77) + (g*150) +  (29*b)) >> 8);
+}
+#endif
+
+#if defined(STBI_NO_PNG) && defined(STBI_NO_BMP) && defined(STBI_NO_PSD) && defined(STBI_NO_TGA) && defined(STBI_NO_GIF) && defined(STBI_NO_PIC) && defined(STBI_NO_PNM)
+// nothing
+#else
+static unsigned char *stbi__convert_format(unsigned char *data, int img_n, int req_comp, unsigned int x, unsigned int y)
+{
+   int i,j;
+   unsigned char *good;
+
+   if (req_comp == img_n) return data;
+   STBI_ASSERT(req_comp >= 1 && req_comp <= 4);
+
+   good = (unsigned char *) stbi__malloc_mad3(req_comp, x, y, 0);
+   if (good == NULL) {
+      STBI_FREE(data);
+      return stbi__errpuc("outofmem", "Out of memory");
+   }
+
+   for (j=0; j < (int) y; ++j) {
+      unsigned char *src  = data + j * x * img_n   ;
+      unsigned char *dest = good + j * x * req_comp;
+
+      #define STBI__COMBO(a,b)  ((a)*8+(b))
+      #define STBI__CASE(a,b)   case STBI__COMBO(a,b): for(i=x-1; i >= 0; --i, src += a, dest += b)
+      // convert source image with img_n components to one with req_comp components;
+      // avoid switch per pixel, so use switch per scanline and massive macros
+      switch (STBI__COMBO(img_n, req_comp)) {
+         STBI__CASE(1,2) { dest[0]=src[0]; dest[1]=255;                                     } break;
+         STBI__CASE(1,3) { dest[0]=dest[1]=dest[2]=src[0];                                  } break;
+         STBI__CASE(1,4) { dest[0]=dest[1]=dest[2]=src[0]; dest[3]=255;                     } break;
+         STBI__CASE(2,1) { dest[0]=src[0];                                                  } break;
+         STBI__CASE(2,3) { dest[0]=dest[1]=dest[2]=src[0];                                  } break;
+         STBI__CASE(2,4) { dest[0]=dest[1]=dest[2]=src[0]; dest[3]=src[1];                  } break;
+         STBI__CASE(3,4) { dest[0]=src[0];dest[1]=src[1];dest[2]=src[2];dest[3]=255;        } break;
+         STBI__CASE(3,1) { dest[0]=stbi__compute_y(src[0],src[1],src[2]);                   } break;
+         STBI__CASE(3,2) { dest[0]=stbi__compute_y(src[0],src[1],src[2]); dest[1] = 255;    } break;
+         STBI__CASE(4,1) { dest[0]=stbi__compute_y(src[0],src[1],src[2]);                   } break;
+         STBI__CASE(4,2) { dest[0]=stbi__compute_y(src[0],src[1],src[2]); dest[1] = src[3]; } break;
+         STBI__CASE(4,3) { dest[0]=src[0];dest[1]=src[1];dest[2]=src[2];                    } break;
+         default: STBI_ASSERT(0); STBI_FREE(data); STBI_FREE(good); return stbi__errpuc("unsupported", "Unsupported format conversion");
+      }
+      #undef STBI__CASE
+   }
+
+   STBI_FREE(data);
+   return good;
+}
+#endif
+
+#if defined(STBI_NO_PNG) && defined(STBI_NO_PSD)
+// nothing
+#else
+static stbi__uint16 stbi__compute_y_16(int r, int g, int b)
+{
+   return (stbi__uint16) (((r*77) + (g*150) +  (29*b)) >> 8);
+}
+#endif
+
+#if defined(STBI_NO_PNG) && defined(STBI_NO_PSD)
+// nothing
+#else
+static stbi__uint16 *stbi__convert_format16(stbi__uint16 *data, int img_n, int req_comp, unsigned int x, unsigned int y)
+{
+   int i,j;
+   stbi__uint16 *good;
+
+   if (req_comp == img_n) return data;
+   STBI_ASSERT(req_comp >= 1 && req_comp <= 4);
+
+   good = (stbi__uint16 *) stbi__malloc(req_comp * x * y * 2);
+   if (good == NULL) {
+      STBI_FREE(data);
+      return (stbi__uint16 *) stbi__errpuc("outofmem", "Out of memory");
+   }
+
+   for (j=0; j < (int) y; ++j) {
+      stbi__uint16 *src  = data + j * x * img_n   ;
+      stbi__uint16 *dest = good + j * x * req_comp;
+
+      #define STBI__COMBO(a,b)  ((a)*8+(b))
+      #define STBI__CASE(a,b)   case STBI__COMBO(a,b): for(i=x-1; i >= 0; --i, src += a, dest += b)
+      // convert source image with img_n components to one with req_comp components;
+      // avoid switch per pixel, so use switch per scanline and massive macros
+      switch (STBI__COMBO(img_n, req_comp)) {
+         STBI__CASE(1,2) { dest[0]=src[0]; dest[1]=0xffff;                                     } break;
+         STBI__CASE(1,3) { dest[0]=dest[1]=dest[2]=src[0];                                     } break;
+         STBI__CASE(1,4) { dest[0]=dest[1]=dest[2]=src[0]; dest[3]=0xffff;                     } break;
+         STBI__CASE(2,1) { dest[0]=src[0];                                                     } break;
+         STBI__CASE(2,3) { dest[0]=dest[1]=dest[2]=src[0];                                     } break;
+         STBI__CASE(2,4) { dest[0]=dest[1]=dest[2]=src[0]; dest[3]=src[1];                     } break;
+         STBI__CASE(3,4) { dest[0]=src[0];dest[1]=src[1];dest[2]=src[2];dest[3]=0xffff;        } break;
+         STBI__CASE(3,1) { dest[0]=stbi__compute_y_16(src[0],src[1],src[2]);                   } break;
+         STBI__CASE(3,2) { dest[0]=stbi__compute_y_16(src[0],src[1],src[2]); dest[1] = 0xffff; } break;
+         STBI__CASE(4,1) { dest[0]=stbi__compute_y_16(src[0],src[1],src[2]);                   } break;
+         STBI__CASE(4,2) { dest[0]=stbi__compute_y_16(src[0],src[1],src[2]); dest[1] = src[3]; } break;
+         STBI__CASE(4,3) { dest[0]=src[0];dest[1]=src[1];dest[2]=src[2];                       } break;
+         default: STBI_ASSERT(0); STBI_FREE(data); STBI_FREE(good); return (stbi__uint16*) stbi__errpuc("unsupported", "Unsupported format conversion");
+      }
+      #undef STBI__CASE
+   }
+
+   STBI_FREE(data);
+   return good;
+}
+#endif
+
+#ifndef STBI_NO_LINEAR
+static float   *stbi__ldr_to_hdr(stbi_uc *data, int x, int y, int comp)
+{
+   int i,k,n;
+   float *output;
+   if (!data) return NULL;
+   output = (float *) stbi__malloc_mad4(x, y, comp, sizeof(float), 0);
+   if (output == NULL) { STBI_FREE(data); return stbi__errpf("outofmem", "Out of memory"); }
+   // compute number of non-alpha components
+   if (comp & 1) n = comp; else n = comp-1;
+   for (i=0; i < x*y; ++i) {
+      for (k=0; k < n; ++k) {
+         output[i*comp + k] = (float) (pow(data[i*comp+k]/255.0f, stbi__l2h_gamma) * stbi__l2h_scale);
+      }
+   }
+   if (n < comp) {
+      for (i=0; i < x*y; ++i) {
+         output[i*comp + n] = data[i*comp + n]/255.0f;
+      }
+   }
+   STBI_FREE(data);
+   return output;
+}
+#endif
+
+#ifndef STBI_NO_HDR
+#define stbi__float2int(x)   ((int) (x))
+static stbi_uc *stbi__hdr_to_ldr(float   *data, int x, int y, int comp)
+{
+   int i,k,n;
+   stbi_uc *output;
+   if (!data) return NULL;
+   output = (stbi_uc *) stbi__malloc_mad3(x, y, comp, 0);
+   if (output == NULL) { STBI_FREE(data); return stbi__errpuc("outofmem", "Out of memory"); }
+   // compute number of non-alpha components
+   if (comp & 1) n = comp; else n = comp-1;
+   for (i=0; i < x*y; ++i) {
+      for (k=0; k < n; ++k) {
+         float z = (float) pow(data[i*comp+k]*stbi__h2l_scale_i, stbi__h2l_gamma_i) * 255 + 0.5f;
+         if (z < 0) z = 0;
+         if (z > 255) z = 255;
+         output[i*comp + k] = (stbi_uc) stbi__float2int(z);
+      }
+      if (k < comp) {
+         float z = data[i*comp+k] * 255 + 0.5f;
+         if (z < 0) z = 0;
+         if (z > 255) z = 255;
+         output[i*comp + k] = (stbi_uc) stbi__float2int(z);
+      }
+   }
+   STBI_FREE(data);
+   return output;
+}
+#endif
+
+//////////////////////////////////////////////////////////////////////////////
+//
+//  "baseline" JPEG/JFIF decoder
+//
+//    simple implementation
+//      - doesn't support delayed output of y-dimension
+//      - simple interface (only one output format: 8-bit interleaved RGB)
+//      - doesn't try to recover corrupt jpegs
+//      - doesn't allow partial loading, loading multiple at once
+//      - still fast on x86 (copying globals into locals doesn't help x86)
+//      - allocates lots of intermediate memory (full size of all components)
+//        - non-interleaved case requires this anyway
+//        - allows good upsampling (see next)
+//    high-quality
+//      - upsampled channels are bilinearly interpolated, even across blocks
+//      - quality integer IDCT derived from IJG's 'slow'
+//    performance
+//      - fast huffman; reasonable integer IDCT
+//      - some SIMD kernels for common paths on targets with SSE2/NEON
+//      - uses a lot of intermediate memory, could cache poorly
+
+#ifndef STBI_NO_JPEG
+
+// huffman decoding acceleration
+#define FAST_BITS   9  // larger handles more cases; smaller stomps less cache
+
+typedef struct
+{
+   stbi_uc  fast[1 << FAST_BITS];
+   // weirdly, repacking this into AoS is a 10% speed loss, instead of a win
+   stbi__uint16 code[256];
+   stbi_uc  values[256];
+   stbi_uc  size[257];
+   unsigned int maxcode[18];
+   int    delta[17];   // old 'firstsymbol' - old 'firstcode'
+} stbi__huffman;
+
+typedef struct
+{
+   stbi__context *s;
+   stbi__huffman huff_dc[4];
+   stbi__huffman huff_ac[4];
+   stbi__uint16 dequant[4][64];
+   stbi__int16 fast_ac[4][1 << FAST_BITS];
+
+// sizes for components, interleaved MCUs
+   int img_h_max, img_v_max;
+   int img_mcu_x, img_mcu_y;
+   int img_mcu_w, img_mcu_h;
+
+// definition of jpeg image component
+   struct
+   {
+      int id;
+      int h,v;
+      int tq;
+      int hd,ha;
+      int dc_pred;
+
+      int x,y,w2,h2;
+      stbi_uc *data;
+      void *raw_data, *raw_coeff;
+      stbi_uc *linebuf;
+      short   *coeff;   // progressive only
+      int      coeff_w, coeff_h; // number of 8x8 coefficient blocks
+   } img_comp[4];
+
+   stbi__uint32   code_buffer; // jpeg entropy-coded buffer
+   int            code_bits;   // number of valid bits
+   unsigned char  marker;      // marker seen while filling entropy buffer
+   int            nomore;      // flag if we saw a marker so must stop
+
+   int            progressive;
+   int            spec_start;
+   int            spec_end;
+   int            succ_high;
+   int            succ_low;
+   int            eob_run;
+   int            jfif;
+   int            app14_color_transform; // Adobe APP14 tag
+   int            rgb;
+
+   int scan_n, order[4];
+   int restart_interval, todo;
+
+// kernels
+   void (*idct_block_kernel)(stbi_uc *out, int out_stride, short data[64]);
+   void (*YCbCr_to_RGB_kernel)(stbi_uc *out, const stbi_uc *y, const stbi_uc *pcb, const stbi_uc *pcr, int count, int step);
+   stbi_uc *(*resample_row_hv_2_kernel)(stbi_uc *out, stbi_uc *in_near, stbi_uc *in_far, int w, int hs);
+} stbi__jpeg;
+
+static int stbi__build_huffman(stbi__huffman *h, int *count)
+{
+   int i,j,k=0;
+   unsigned int code;
+   // build size list for each symbol (from JPEG spec)
+   for (i=0; i < 16; ++i) {
+      for (j=0; j < count[i]; ++j) {
+         h->size[k++] = (stbi_uc) (i+1);
+         if(k >= 257) return stbi__err("bad size list","Corrupt JPEG");
+      }
+   }
+   h->size[k] = 0;
+
+   // compute actual symbols (from jpeg spec)
+   code = 0;
+   k = 0;
+   for(j=1; j <= 16; ++j) {
+      // compute delta to add to code to compute symbol id
+      h->delta[j] = k - code;
+      if (h->size[k] == j) {
+         while (h->size[k] == j)
+            h->code[k++] = (stbi__uint16) (code++);
+         if (code-1 >= (1u << j)) return stbi__err("bad code lengths","Corrupt JPEG");
+      }
+      // compute largest code + 1 for this size, preshifted as needed later
+      h->maxcode[j] = code << (16-j);
+      code <<= 1;
+   }
+   h->maxcode[j] = 0xffffffff;
+
+   // build non-spec acceleration table; 255 is flag for not-accelerated
+   memset(h->fast, 255, 1 << FAST_BITS);
+   for (i=0; i < k; ++i) {
+      int s = h->size[i];
+      if (s <= FAST_BITS) {
+         int c = h->code[i] << (FAST_BITS-s);
+         int m = 1 << (FAST_BITS-s);
+         for (j=0; j < m; ++j) {
+            h->fast[c+j] = (stbi_uc) i;
+         }
+      }
+   }
+   return 1;
+}
+
+// build a table that decodes both magnitude and value of small ACs in
+// one go.
+static void stbi__build_fast_ac(stbi__int16 *fast_ac, stbi__huffman *h)
+{
+   int i;
+   for (i=0; i < (1 << FAST_BITS); ++i) {
+      stbi_uc fast = h->fast[i];
+      fast_ac[i] = 0;
+      if (fast < 255) {
+         int rs = h->values[fast];
+         int run = (rs >> 4) & 15;
+         int magbits = rs & 15;
+         int len = h->size[fast];
+
+         if (magbits && len + magbits <= FAST_BITS) {
+            // magnitude code followed by receive_extend code
+            int k = ((i << len) & ((1 << FAST_BITS) - 1)) >> (FAST_BITS - magbits);
+            int m = 1 << (magbits - 1);
+            if (k < m) k += (~0U << magbits) + 1;
+            // if the result is small enough, we can fit it in fast_ac table
+            if (k >= -128 && k <= 127)
+               fast_ac[i] = (stbi__int16) ((k * 256) + (run * 16) + (len + magbits));
+         }
+      }
+   }
+}
+
+static void stbi__grow_buffer_unsafe(stbi__jpeg *j)
+{
+   do {
+      unsigned int b = j->nomore ? 0 : stbi__get8(j->s);
+      if (b == 0xff) {
+         int c = stbi__get8(j->s);
+         while (c == 0xff) c = stbi__get8(j->s); // consume fill bytes
+         if (c != 0) {
+            j->marker = (unsigned char) c;
+            j->nomore = 1;
+            return;
+         }
+      }
+      j->code_buffer |= b << (24 - j->code_bits);
+      j->code_bits += 8;
+   } while (j->code_bits <= 24);
+}
+
+// (1 << n) - 1
+static const stbi__uint32 stbi__bmask[17]={0,1,3,7,15,31,63,127,255,511,1023,2047,4095,8191,16383,32767,65535};
+
+// decode a jpeg huffman value from the bitstream
+stbi_inline static int stbi__jpeg_huff_decode(stbi__jpeg *j, stbi__huffman *h)
+{
+   unsigned int temp;
+   int c,k;
+
+   if (j->code_bits < 16) stbi__grow_buffer_unsafe(j);
+
+   // look at the top FAST_BITS and determine what symbol ID it is,
+   // if the code is <= FAST_BITS
+   c = (j->code_buffer >> (32 - FAST_BITS)) & ((1 << FAST_BITS)-1);
+   k = h->fast[c];
+   if (k < 255) {
+      int s = h->size[k];
+      if (s > j->code_bits)
+         return -1;
+      j->code_buffer <<= s;
+      j->code_bits -= s;
+      return h->values[k];
+   }
+
+   // naive test is to shift the code_buffer down so k bits are
+   // valid, then test against maxcode. To speed this up, we've
+   // preshifted maxcode left so that it has (16-k) 0s at the
+   // end; in other words, regardless of the number of bits, it
+   // wants to be compared against something shifted to have 16;
+   // that way we don't need to shift inside the loop.
+   temp = j->code_buffer >> 16;
+   for (k=FAST_BITS+1 ; ; ++k)
+      if (temp < h->maxcode[k])
+         break;
+   if (k == 17) {
+      // error! code not found
+      j->code_bits -= 16;
+      return -1;
+   }
+
+   if (k > j->code_bits)
+      return -1;
+
+   // convert the huffman code to the symbol id
+   c = ((j->code_buffer >> (32 - k)) & stbi__bmask[k]) + h->delta[k];
+   if(c < 0 || c >= 256) // symbol id out of bounds!
+       return -1;
+   STBI_ASSERT((((j->code_buffer) >> (32 - h->size[c])) & stbi__bmask[h->size[c]]) == h->code[c]);
+
+   // convert the id to a symbol
+   j->code_bits -= k;
+   j->code_buffer <<= k;
+   return h->values[c];
+}
+
+// bias[n] = (-1<<n) + 1
+static const int stbi__jbias[16] = {0,-1,-3,-7,-15,-31,-63,-127,-255,-511,-1023,-2047,-4095,-8191,-16383,-32767};
+
+// combined JPEG 'receive' and JPEG 'extend', since baseline
+// always extends everything it receives.
+stbi_inline static int stbi__extend_receive(stbi__jpeg *j, int n)
+{
+   unsigned int k;
+   int sgn;
+   if (j->code_bits < n) stbi__grow_buffer_unsafe(j);
+   if (j->code_bits < n) return 0; // ran out of bits from stream, return 0s intead of continuing
+
+   sgn = j->code_buffer >> 31; // sign bit always in MSB; 0 if MSB clear (positive), 1 if MSB set (negative)
+   k = stbi_lrot(j->code_buffer, n);
+   j->code_buffer = k & ~stbi__bmask[n];
+   k &= stbi__bmask[n];
+   j->code_bits -= n;
+   return k + (stbi__jbias[n] & (sgn - 1));
+}
+
+// get some unsigned bits
+stbi_inline static int stbi__jpeg_get_bits(stbi__jpeg *j, int n)
+{
+   unsigned int k;
+   if (j->code_bits < n) stbi__grow_buffer_unsafe(j);
+   if (j->code_bits < n) return 0; // ran out of bits from stream, return 0s intead of continuing
+   k = stbi_lrot(j->code_buffer, n);
+   j->code_buffer = k & ~stbi__bmask[n];
+   k &= stbi__bmask[n];
+   j->code_bits -= n;
+   return k;
+}
+
+stbi_inline static int stbi__jpeg_get_bit(stbi__jpeg *j)
+{
+   unsigned int k;
+   if (j->code_bits < 1) stbi__grow_buffer_unsafe(j);
+   if (j->code_bits < 1) return 0; // ran out of bits from stream, return 0s intead of continuing
+   k = j->code_buffer;
+   j->code_buffer <<= 1;
+   --j->code_bits;
+   return k & 0x80000000;
+}
+
+// given a value that's at position X in the zigzag stream,
+// where does it appear in the 8x8 matrix coded as row-major?
+static const stbi_uc stbi__jpeg_dezigzag[64+15] =
+{
+    0,  1,  8, 16,  9,  2,  3, 10,
+   17, 24, 32, 25, 18, 11,  4,  5,
+   12, 19, 26, 33, 40, 48, 41, 34,
+   27, 20, 13,  6,  7, 14, 21, 28,
+   35, 42, 49, 56, 57, 50, 43, 36,
+   29, 22, 15, 23, 30, 37, 44, 51,
+   58, 59, 52, 45, 38, 31, 39, 46,
+   53, 60, 61, 54, 47, 55, 62, 63,
+   // let corrupt input sample past end
+   63, 63, 63, 63, 63, 63, 63, 63,
+   63, 63, 63, 63, 63, 63, 63
+};
+
+// decode one 64-entry block--
+static int stbi__jpeg_decode_block(stbi__jpeg *j, short data[64], stbi__huffman *hdc, stbi__huffman *hac, stbi__int16 *fac, int b, stbi__uint16 *dequant)
+{
+   int diff,dc,k;
+   int t;
+
+   if (j->code_bits < 16) stbi__grow_buffer_unsafe(j);
+   t = stbi__jpeg_huff_decode(j, hdc);
+   if (t < 0 || t > 15) return stbi__err("bad huffman code","Corrupt JPEG");
+
+   // 0 all the ac values now so we can do it 32-bits at a time
+   memset(data,0,64*sizeof(data[0]));
+
+   diff = t ? stbi__extend_receive(j, t) : 0;
+   if (!stbi__addints_valid(j->img_comp[b].dc_pred, diff)) return stbi__err("bad delta","Corrupt JPEG");
+   dc = j->img_comp[b].dc_pred + diff;
+   j->img_comp[b].dc_pred = dc;
+   if (!stbi__mul2shorts_valid(dc, dequant[0])) return stbi__err("can't merge dc and ac", "Corrupt JPEG");
+   data[0] = (short) (dc * dequant[0]);
+
+   // decode AC components, see JPEG spec
+   k = 1;
+   do {
+      unsigned int zig;
+      int c,r,s;
+      if (j->code_bits < 16) stbi__grow_buffer_unsafe(j);
+      c = (j->code_buffer >> (32 - FAST_BITS)) & ((1 << FAST_BITS)-1);
+      r = fac[c];
+      if (r) { // fast-AC path
+         k += (r >> 4) & 15; // run
+         s = r & 15; // combined length
+         if (s > j->code_bits) return stbi__err("bad huffman code", "Combined length longer than code bits available");
+         j->code_buffer <<= s;
+         j->code_bits -= s;
+         // decode into unzigzag'd location
+         zig = stbi__jpeg_dezigzag[k++];
+         data[zig] = (short) ((r >> 8) * dequant[zig]);
+      } else {
+         int rs = stbi__jpeg_huff_decode(j, hac);
+         if (rs < 0) return stbi__err("bad huffman code","Corrupt JPEG");
+         s = rs & 15;
+         r = rs >> 4;
+         if (s == 0) {
+            if (rs != 0xf0) break; // end block
+            k += 16;
+         } else {
+            k += r;
+            // decode into unzigzag'd location
+            zig = stbi__jpeg_dezigzag[k++];
+            data[zig] = (short) (stbi__extend_receive(j,s) * dequant[zig]);
+         }
+      }
+   } while (k < 64);
+   return 1;
+}
+
+static int stbi__jpeg_decode_block_prog_dc(stbi__jpeg *j, short data[64], stbi__huffman *hdc, int b)
+{
+   int diff,dc;
+   int t;
+   if (j->spec_end != 0) return stbi__err("can't merge dc and ac", "Corrupt JPEG");
+
+   if (j->code_bits < 16) stbi__grow_buffer_unsafe(j);
+
+   if (j->succ_high == 0) {
+      // first scan for DC coefficient, must be first
+      memset(data,0,64*sizeof(data[0])); // 0 all the ac values now
+      t = stbi__jpeg_huff_decode(j, hdc);
+      if (t < 0 || t > 15) return stbi__err("can't merge dc and ac", "Corrupt JPEG");
+      diff = t ? stbi__extend_receive(j, t) : 0;
+
+      if (!stbi__addints_valid(j->img_comp[b].dc_pred, diff)) return stbi__err("bad delta", "Corrupt JPEG");
+      dc = j->img_comp[b].dc_pred + diff;
+      j->img_comp[b].dc_pred = dc;
+      if (!stbi__mul2shorts_valid(dc, 1 << j->succ_low)) return stbi__err("can't merge dc and ac", "Corrupt JPEG");
+      data[0] = (short) (dc * (1 << j->succ_low));
+   } else {
+      // refinement scan for DC coefficient
+      if (stbi__jpeg_get_bit(j))
+         data[0] += (short) (1 << j->succ_low);
+   }
+   return 1;
+}
+
+// @OPTIMIZE: store non-zigzagged during the decode passes,
+// and only de-zigzag when dequantizing
+static int stbi__jpeg_decode_block_prog_ac(stbi__jpeg *j, short data[64], stbi__huffman *hac, stbi__int16 *fac)
+{
+   int k;
+   if (j->spec_start == 0) return stbi__err("can't merge dc and ac", "Corrupt JPEG");
+
+   if (j->succ_high == 0) {
+      int shift = j->succ_low;
+
+      if (j->eob_run) {
+         --j->eob_run;
+         return 1;
+      }
+
+      k = j->spec_start;
+      do {
+         unsigned int zig;
+         int c,r,s;
+         if (j->code_bits < 16) stbi__grow_buffer_unsafe(j);
+         c = (j->code_buffer >> (32 - FAST_BITS)) & ((1 << FAST_BITS)-1);
+         r = fac[c];
+         if (r) { // fast-AC path
+            k += (r >> 4) & 15; // run
+            s = r & 15; // combined length
+            if (s > j->code_bits) return stbi__err("bad huffman code", "Combined length longer than code bits available");
+            j->code_buffer <<= s;
+            j->code_bits -= s;
+            zig = stbi__jpeg_dezigzag[k++];
+            data[zig] = (short) ((r >> 8) * (1 << shift));
+         } else {
+            int rs = stbi__jpeg_huff_decode(j, hac);
+            if (rs < 0) return stbi__err("bad huffman code","Corrupt JPEG");
+            s = rs & 15;
+            r = rs >> 4;
+            if (s == 0) {
+               if (r < 15) {
+                  j->eob_run = (1 << r);
+                  if (r)
+                     j->eob_run += stbi__jpeg_get_bits(j, r);
+                  --j->eob_run;
+                  break;
+               }
+               k += 16;
+            } else {
+               k += r;
+               zig = stbi__jpeg_dezigzag[k++];
+               data[zig] = (short) (stbi__extend_receive(j,s) * (1 << shift));
+            }
+         }
+      } while (k <= j->spec_end);
+   } else {
+      // refinement scan for these AC coefficients
+
+      short bit = (short) (1 << j->succ_low);
+
+      if (j->eob_run) {
+         --j->eob_run;
+         for (k = j->spec_start; k <= j->spec_end; ++k) {
+            short *p = &data[stbi__jpeg_dezigzag[k]];
+            if (*p != 0)
+               if (stbi__jpeg_get_bit(j))
+                  if ((*p & bit)==0) {
+                     if (*p > 0)
+                        *p += bit;
+                     else
+                        *p -= bit;
+                  }
+         }
+      } else {
+         k = j->spec_start;
+         do {
+            int r,s;
+            int rs = stbi__jpeg_huff_decode(j, hac); // @OPTIMIZE see if we can use the fast path here, advance-by-r is so slow, eh
+            if (rs < 0) return stbi__err("bad huffman code","Corrupt JPEG");
+            s = rs & 15;
+            r = rs >> 4;
+            if (s == 0) {
+               if (r < 15) {
+                  j->eob_run = (1 << r) - 1;
+                  if (r)
+                     j->eob_run += stbi__jpeg_get_bits(j, r);
+                  r = 64; // force end of block
+               } else {
+                  // r=15 s=0 should write 16 0s, so we just do
+                  // a run of 15 0s and then write s (which is 0),
+                  // so we don't have to do anything special here
+               }
+            } else {
+               if (s != 1) return stbi__err("bad huffman code", "Corrupt JPEG");
+               // sign bit
+               if (stbi__jpeg_get_bit(j))
+                  s = bit;
+               else
+                  s = -bit;
+            }
+
+            // advance by r
+            while (k <= j->spec_end) {
+               short *p = &data[stbi__jpeg_dezigzag[k++]];
+               if (*p != 0) {
+                  if (stbi__jpeg_get_bit(j))
+                     if ((*p & bit)==0) {
+                        if (*p > 0)
+                           *p += bit;
+                        else
+                           *p -= bit;
+                     }
+               } else {
+                  if (r == 0) {
+                     *p = (short) s;
+                     break;
+                  }
+                  --r;
+               }
+            }
+         } while (k <= j->spec_end);
+      }
+   }
+   return 1;
+}
+
+// take a -128..127 value and stbi__clamp it and convert to 0..255
+stbi_inline static stbi_uc stbi__clamp(int x)
+{
+   // trick to use a single test to catch both cases
+   if ((unsigned int) x > 255) {
+      if (x < 0) return 0;
+      if (x > 255) return 255;
+   }
+   return (stbi_uc) x;
+}
+
+#define stbi__f2f(x)  ((int) (((x) * 4096 + 0.5)))
+#define stbi__fsh(x)  ((x) * 4096)
+
+// derived from jidctint -- DCT_ISLOW
+#define STBI__IDCT_1D(s0,s1,s2,s3,s4,s5,s6,s7) \
+   int t0,t1,t2,t3,p1,p2,p3,p4,p5,x0,x1,x2,x3; \
+   p2 = s2;                                    \
+   p3 = s6;                                    \
+   p1 = (p2+p3) * stbi__f2f(0.5411961f);       \
+   t2 = p1 + p3*stbi__f2f(-1.847759065f);      \
+   t3 = p1 + p2*stbi__f2f( 0.765366865f);      \
+   p2 = s0;                                    \
+   p3 = s4;                                    \
+   t0 = stbi__fsh(p2+p3);                      \
+   t1 = stbi__fsh(p2-p3);                      \
+   x0 = t0+t3;                                 \
+   x3 = t0-t3;                                 \
+   x1 = t1+t2;                                 \
+   x2 = t1-t2;                                 \
+   t0 = s7;                                    \
+   t1 = s5;                                    \
+   t2 = s3;                                    \
+   t3 = s1;                                    \
+   p3 = t0+t2;                                 \
+   p4 = t1+t3;                                 \
+   p1 = t0+t3;                                 \
+   p2 = t1+t2;                                 \
+   p5 = (p3+p4)*stbi__f2f( 1.175875602f);      \
+   t0 = t0*stbi__f2f( 0.298631336f);           \
+   t1 = t1*stbi__f2f( 2.053119869f);           \
+   t2 = t2*stbi__f2f( 3.072711026f);           \
+   t3 = t3*stbi__f2f( 1.501321110f);           \
+   p1 = p5 + p1*stbi__f2f(-0.899976223f);      \
+   p2 = p5 + p2*stbi__f2f(-2.562915447f);      \
+   p3 = p3*stbi__f2f(-1.961570560f);           \
+   p4 = p4*stbi__f2f(-0.390180644f);           \
+   t3 += p1+p4;                                \
+   t2 += p2+p3;                                \
+   t1 += p2+p4;                                \
+   t0 += p1+p3;
+
+static void stbi__idct_block(stbi_uc *out, int out_stride, short data[64])
+{
+   int i,val[64],*v=val;
+   stbi_uc *o;
+   short *d = data;
+
+   // columns
+   for (i=0; i < 8; ++i,++d, ++v) {
+      // if all zeroes, shortcut -- this avoids dequantizing 0s and IDCTing
+      if (d[ 8]==0 && d[16]==0 && d[24]==0 && d[32]==0
+           && d[40]==0 && d[48]==0 && d[56]==0) {
+         //    no shortcut                 0     seconds
+         //    (1|2|3|4|5|6|7)==0          0     seconds
+         //    all separate               -0.047 seconds
+         //    1 && 2|3 && 4|5 && 6|7:    -0.047 seconds
+         int dcterm = d[0]*4;
+         v[0] = v[8] = v[16] = v[24] = v[32] = v[40] = v[48] = v[56] = dcterm;
+      } else {
+         STBI__IDCT_1D(d[ 0],d[ 8],d[16],d[24],d[32],d[40],d[48],d[56])
+         // constants scaled things up by 1<<12; let's bring them back
+         // down, but keep 2 extra bits of precision
+         x0 += 512; x1 += 512; x2 += 512; x3 += 512;
+         v[ 0] = (x0+t3) >> 10;
+         v[56] = (x0-t3) >> 10;
+         v[ 8] = (x1+t2) >> 10;
+         v[48] = (x1-t2) >> 10;
+         v[16] = (x2+t1) >> 10;
+         v[40] = (x2-t1) >> 10;
+         v[24] = (x3+t0) >> 10;
+         v[32] = (x3-t0) >> 10;
+      }
+   }
+
+   for (i=0, v=val, o=out; i < 8; ++i,v+=8,o+=out_stride) {
+      // no fast case since the first 1D IDCT spread components out
+      STBI__IDCT_1D(v[0],v[1],v[2],v[3],v[4],v[5],v[6],v[7])
+      // constants scaled things up by 1<<12, plus we had 1<<2 from first
+      // loop, plus horizontal and vertical each scale by sqrt(8) so together
+      // we've got an extra 1<<3, so 1<<17 total we need to remove.
+      // so we want to round that, which means adding 0.5 * 1<<17,
+      // aka 65536. Also, we'll end up with -128 to 127 that we want
+      // to encode as 0..255 by adding 128, so we'll add that before the shift
+      x0 += 65536 + (128<<17);
+      x1 += 65536 + (128<<17);
+      x2 += 65536 + (128<<17);
+      x3 += 65536 + (128<<17);
+      // tried computing the shifts into temps, or'ing the temps to see
+      // if any were out of range, but that was slower
+      o[0] = stbi__clamp((x0+t3) >> 17);
+      o[7] = stbi__clamp((x0-t3) >> 17);
+      o[1] = stbi__clamp((x1+t2) >> 17);
+      o[6] = stbi__clamp((x1-t2) >> 17);
+      o[2] = stbi__clamp((x2+t1) >> 17);
+      o[5] = stbi__clamp((x2-t1) >> 17);
+      o[3] = stbi__clamp((x3+t0) >> 17);
+      o[4] = stbi__clamp((x3-t0) >> 17);
+   }
+}
+
+#ifdef STBI_SSE2
+// sse2 integer IDCT. not the fastest possible implementation but it
+// produces bit-identical results to the generic C version so it's
+// fully "transparent".
+static void stbi__idct_simd(stbi_uc *out, int out_stride, short data[64])
+{
+   // This is constructed to match our regular (generic) integer IDCT exactly.
+   __m128i row0, row1, row2, row3, row4, row5, row6, row7;
+   __m128i tmp;
+
+   // dot product constant: even elems=x, odd elems=y
+   #define dct_const(x,y)  _mm_setr_epi16((x),(y),(x),(y),(x),(y),(x),(y))
+
+   // out(0) = c0[even]*x + c0[odd]*y   (c0, x, y 16-bit, out 32-bit)
+   // out(1) = c1[even]*x + c1[odd]*y
+   #define dct_rot(out0,out1, x,y,c0,c1) \
+      __m128i c0##lo = _mm_unpacklo_epi16((x),(y)); \
+      __m128i c0##hi = _mm_unpackhi_epi16((x),(y)); \
+      __m128i out0##_l = _mm_madd_epi16(c0##lo, c0); \
+      __m128i out0##_h = _mm_madd_epi16(c0##hi, c0); \
+      __m128i out1##_l = _mm_madd_epi16(c0##lo, c1); \
+      __m128i out1##_h = _mm_madd_epi16(c0##hi, c1)
+
+   // out = in << 12  (in 16-bit, out 32-bit)
+   #define dct_widen(out, in) \
+      __m128i out##_l = _mm_srai_epi32(_mm_unpacklo_epi16(_mm_setzero_si128(), (in)), 4); \
+      __m128i out##_h = _mm_srai_epi32(_mm_unpackhi_epi16(_mm_setzero_si128(), (in)), 4)
+
+   // wide add
+   #define dct_wadd(out, a, b) \
+      __m128i out##_l = _mm_add_epi32(a##_l, b##_l); \
+      __m128i out##_h = _mm_add_epi32(a##_h, b##_h)
+
+   // wide sub
+   #define dct_wsub(out, a, b) \
+      __m128i out##_l = _mm_sub_epi32(a##_l, b##_l); \
+      __m128i out##_h = _mm_sub_epi32(a##_h, b##_h)
+
+   // butterfly a/b, add bias, then shift by "s" and pack
+   #define dct_bfly32o(out0, out1, a,b,bias,s) \
+      { \
+         __m128i abiased_l = _mm_add_epi32(a##_l, bias); \
+         __m128i abiased_h = _mm_add_epi32(a##_h, bias); \
+         dct_wadd(sum, abiased, b); \
+         dct_wsub(dif, abiased, b); \
+         out0 = _mm_packs_epi32(_mm_srai_epi32(sum_l, s), _mm_srai_epi32(sum_h, s)); \
+         out1 = _mm_packs_epi32(_mm_srai_epi32(dif_l, s), _mm_srai_epi32(dif_h, s)); \
+      }
+
+   // 8-bit interleave step (for transposes)
+   #define dct_interleave8(a, b) \
+      tmp = a; \
+      a = _mm_unpacklo_epi8(a, b); \
+      b = _mm_unpackhi_epi8(tmp, b)
+
+   // 16-bit interleave step (for transposes)
+   #define dct_interleave16(a, b) \
+      tmp = a; \
+      a = _mm_unpacklo_epi16(a, b); \
+      b = _mm_unpackhi_epi16(tmp, b)
+
+   #define dct_pass(bias,shift) \
+      { \
+         /* even part */ \
+         dct_rot(t2e,t3e, row2,row6, rot0_0,rot0_1); \
+         __m128i sum04 = _mm_add_epi16(row0, row4); \
+         __m128i dif04 = _mm_sub_epi16(row0, row4); \
+         dct_widen(t0e, sum04); \
+         dct_widen(t1e, dif04); \
+         dct_wadd(x0, t0e, t3e); \
+         dct_wsub(x3, t0e, t3e); \
+         dct_wadd(x1, t1e, t2e); \
+         dct_wsub(x2, t1e, t2e); \
+         /* odd part */ \
+         dct_rot(y0o,y2o, row7,row3, rot2_0,rot2_1); \
+         dct_rot(y1o,y3o, row5,row1, rot3_0,rot3_1); \
+         __m128i sum17 = _mm_add_epi16(row1, row7); \
+         __m128i sum35 = _mm_add_epi16(row3, row5); \
+         dct_rot(y4o,y5o, sum17,sum35, rot1_0,rot1_1); \
+         dct_wadd(x4, y0o, y4o); \
+         dct_wadd(x5, y1o, y5o); \
+         dct_wadd(x6, y2o, y5o); \
+         dct_wadd(x7, y3o, y4o); \
+         dct_bfly32o(row0,row7, x0,x7,bias,shift); \
+         dct_bfly32o(row1,row6, x1,x6,bias,shift); \
+         dct_bfly32o(row2,row5, x2,x5,bias,shift); \
+         dct_bfly32o(row3,row4, x3,x4,bias,shift); \
+      }
+
+   __m128i rot0_0 = dct_const(stbi__f2f(0.5411961f), stbi__f2f(0.5411961f) + stbi__f2f(-1.847759065f));
+   __m128i rot0_1 = dct_const(stbi__f2f(0.5411961f) + stbi__f2f( 0.765366865f), stbi__f2f(0.5411961f));
+   __m128i rot1_0 = dct_const(stbi__f2f(1.175875602f) + stbi__f2f(-0.899976223f), stbi__f2f(1.175875602f));
+   __m128i rot1_1 = dct_const(stbi__f2f(1.175875602f), stbi__f2f(1.175875602f) + stbi__f2f(-2.562915447f));
+   __m128i rot2_0 = dct_const(stbi__f2f(-1.961570560f) + stbi__f2f( 0.298631336f), stbi__f2f(-1.961570560f));
+   __m128i rot2_1 = dct_const(stbi__f2f(-1.961570560f), stbi__f2f(-1.961570560f) + stbi__f2f( 3.072711026f));
+   __m128i rot3_0 = dct_const(stbi__f2f(-0.390180644f) + stbi__f2f( 2.053119869f), stbi__f2f(-0.390180644f));
+   __m128i rot3_1 = dct_const(stbi__f2f(-0.390180644f), stbi__f2f(-0.390180644f) + stbi__f2f( 1.501321110f));
+
+   // rounding biases in column/row passes, see stbi__idct_block for explanation.
+   __m128i bias_0 = _mm_set1_epi32(512);
+   __m128i bias_1 = _mm_set1_epi32(65536 + (128<<17));
+
+   // load
+   row0 = _mm_load_si128((const __m128i *) (data + 0*8));
+   row1 = _mm_load_si128((const __m128i *) (data + 1*8));
+   row2 = _mm_load_si128((const __m128i *) (data + 2*8));
+   row3 = _mm_load_si128((const __m128i *) (data + 3*8));
+   row4 = _mm_load_si128((const __m128i *) (data + 4*8));
+   row5 = _mm_load_si128((const __m128i *) (data + 5*8));
+   row6 = _mm_load_si128((const __m128i *) (data + 6*8));
+   row7 = _mm_load_si128((const __m128i *) (data + 7*8));
+
+   // column pass
+   dct_pass(bias_0, 10);
+
+   {
+      // 16bit 8x8 transpose pass 1
+      dct_interleave16(row0, row4);
+      dct_interleave16(row1, row5);
+      dct_interleave16(row2, row6);
+      dct_interleave16(row3, row7);
+
+      // transpose pass 2
+      dct_interleave16(row0, row2);
+      dct_interleave16(row1, row3);
+      dct_interleave16(row4, row6);
+      dct_interleave16(row5, row7);
+
+      // transpose pass 3
+      dct_interleave16(row0, row1);
+      dct_interleave16(row2, row3);
+      dct_interleave16(row4, row5);
+      dct_interleave16(row6, row7);
+   }
+
+   // row pass
+   dct_pass(bias_1, 17);
+
+   {
+      // pack
+      __m128i p0 = _mm_packus_epi16(row0, row1); // a0a1a2a3...a7b0b1b2b3...b7
+      __m128i p1 = _mm_packus_epi16(row2, row3);
+      __m128i p2 = _mm_packus_epi16(row4, row5);
+      __m128i p3 = _mm_packus_epi16(row6, row7);
+
+      // 8bit 8x8 transpose pass 1
+      dct_interleave8(p0, p2); // a0e0a1e1...
+      dct_interleave8(p1, p3); // c0g0c1g1...
+
+      // transpose pass 2
+      dct_interleave8(p0, p1); // a0c0e0g0...
+      dct_interleave8(p2, p3); // b0d0f0h0...
+
+      // transpose pass 3
+      dct_interleave8(p0, p2); // a0b0c0d0...
+      dct_interleave8(p1, p3); // a4b4c4d4...
+
+      // store
+      _mm_storel_epi64((__m128i *) out, p0); out += out_stride;
+      _mm_storel_epi64((__m128i *) out, _mm_shuffle_epi32(p0, 0x4e)); out += out_stride;
+      _mm_storel_epi64((__m128i *) out, p2); out += out_stride;
+      _mm_storel_epi64((__m128i *) out, _mm_shuffle_epi32(p2, 0x4e)); out += out_stride;
+      _mm_storel_epi64((__m128i *) out, p1); out += out_stride;
+      _mm_storel_epi64((__m128i *) out, _mm_shuffle_epi32(p1, 0x4e)); out += out_stride;
+      _mm_storel_epi64((__m128i *) out, p3); out += out_stride;
+      _mm_storel_epi64((__m128i *) out, _mm_shuffle_epi32(p3, 0x4e));
+   }
+
+#undef dct_const
+#undef dct_rot
+#undef dct_widen
+#undef dct_wadd
+#undef dct_wsub
+#undef dct_bfly32o
+#undef dct_interleave8
+#undef dct_interleave16
+#undef dct_pass
+}
+
+#endif // STBI_SSE2
+
+#ifdef STBI_NEON
+
+// NEON integer IDCT. should produce bit-identical
+// results to the generic C version.
+static void stbi__idct_simd(stbi_uc *out, int out_stride, short data[64])
+{
+   int16x8_t row0, row1, row2, row3, row4, row5, row6, row7;
+
+   int16x4_t rot0_0 = vdup_n_s16(stbi__f2f(0.5411961f));
+   int16x4_t rot0_1 = vdup_n_s16(stbi__f2f(-1.847759065f));
+   int16x4_t rot0_2 = vdup_n_s16(stbi__f2f( 0.765366865f));
+   int16x4_t rot1_0 = vdup_n_s16(stbi__f2f( 1.175875602f));
+   int16x4_t rot1_1 = vdup_n_s16(stbi__f2f(-0.899976223f));
+   int16x4_t rot1_2 = vdup_n_s16(stbi__f2f(-2.562915447f));
+   int16x4_t rot2_0 = vdup_n_s16(stbi__f2f(-1.961570560f));
+   int16x4_t rot2_1 = vdup_n_s16(stbi__f2f(-0.390180644f));
+   int16x4_t rot3_0 = vdup_n_s16(stbi__f2f( 0.298631336f));
+   int16x4_t rot3_1 = vdup_n_s16(stbi__f2f( 2.053119869f));
+   int16x4_t rot3_2 = vdup_n_s16(stbi__f2f( 3.072711026f));
+   int16x4_t rot3_3 = vdup_n_s16(stbi__f2f( 1.501321110f));
+
+#define dct_long_mul(out, inq, coeff) \
+   int32x4_t out##_l = vmull_s16(vget_low_s16(inq), coeff); \
+   int32x4_t out##_h = vmull_s16(vget_high_s16(inq), coeff)
+
+#define dct_long_mac(out, acc, inq, coeff) \
+   int32x4_t out##_l = vmlal_s16(acc##_l, vget_low_s16(inq), coeff); \
+   int32x4_t out##_h = vmlal_s16(acc##_h, vget_high_s16(inq), coeff)
+
+#define dct_widen(out, inq) \
+   int32x4_t out##_l = vshll_n_s16(vget_low_s16(inq), 12); \
+   int32x4_t out##_h = vshll_n_s16(vget_high_s16(inq), 12)
+
+// wide add
+#define dct_wadd(out, a, b) \
+   int32x4_t out##_l = vaddq_s32(a##_l, b##_l); \
+   int32x4_t out##_h = vaddq_s32(a##_h, b##_h)
+
+// wide sub
+#define dct_wsub(out, a, b) \
+   int32x4_t out##_l = vsubq_s32(a##_l, b##_l); \
+   int32x4_t out##_h = vsubq_s32(a##_h, b##_h)
+
+// butterfly a/b, then shift using "shiftop" by "s" and pack
+#define dct_bfly32o(out0,out1, a,b,shiftop,s) \
+   { \
+      dct_wadd(sum, a, b); \
+      dct_wsub(dif, a, b); \
+      out0 = vcombine_s16(shiftop(sum_l, s), shiftop(sum_h, s)); \
+      out1 = vcombine_s16(shiftop(dif_l, s), shiftop(dif_h, s)); \
+   }
+
+#define dct_pass(shiftop, shift) \
+   { \
+      /* even part */ \
+      int16x8_t sum26 = vaddq_s16(row2, row6); \
+      dct_long_mul(p1e, sum26, rot0_0); \
+      dct_long_mac(t2e, p1e, row6, rot0_1); \
+      dct_long_mac(t3e, p1e, row2, rot0_2); \
+      int16x8_t sum04 = vaddq_s16(row0, row4); \
+      int16x8_t dif04 = vsubq_s16(row0, row4); \
+      dct_widen(t0e, sum04); \
+      dct_widen(t1e, dif04); \
+      dct_wadd(x0, t0e, t3e); \
+      dct_wsub(x3, t0e, t3e); \
+      dct_wadd(x1, t1e, t2e); \
+      dct_wsub(x2, t1e, t2e); \
+      /* odd part */ \
+      int16x8_t sum15 = vaddq_s16(row1, row5); \
+      int16x8_t sum17 = vaddq_s16(row1, row7); \
+      int16x8_t sum35 = vaddq_s16(row3, row5); \
+      int16x8_t sum37 = vaddq_s16(row3, row7); \
+      int16x8_t sumodd = vaddq_s16(sum17, sum35); \
+      dct_long_mul(p5o, sumodd, rot1_0); \
+      dct_long_mac(p1o, p5o, sum17, rot1_1); \
+      dct_long_mac(p2o, p5o, sum35, rot1_2); \
+      dct_long_mul(p3o, sum37, rot2_0); \
+      dct_long_mul(p4o, sum15, rot2_1); \
+      dct_wadd(sump13o, p1o, p3o); \
+      dct_wadd(sump24o, p2o, p4o); \
+      dct_wadd(sump23o, p2o, p3o); \
+      dct_wadd(sump14o, p1o, p4o); \
+      dct_long_mac(x4, sump13o, row7, rot3_0); \
+      dct_long_mac(x5, sump24o, row5, rot3_1); \
+      dct_long_mac(x6, sump23o, row3, rot3_2); \
+      dct_long_mac(x7, sump14o, row1, rot3_3); \
+      dct_bfly32o(row0,row7, x0,x7,shiftop,shift); \
+      dct_bfly32o(row1,row6, x1,x6,shiftop,shift); \
+      dct_bfly32o(row2,row5, x2,x5,shiftop,shift); \
+      dct_bfly32o(row3,row4, x3,x4,shiftop,shift); \
+   }
+
+   // load
+   row0 = vld1q_s16(data + 0*8);
+   row1 = vld1q_s16(data + 1*8);
+   row2 = vld1q_s16(data + 2*8);
+   row3 = vld1q_s16(data + 3*8);
+   row4 = vld1q_s16(data + 4*8);
+   row5 = vld1q_s16(data + 5*8);
+   row6 = vld1q_s16(data + 6*8);
+   row7 = vld1q_s16(data + 7*8);
+
+   // add DC bias
+   row0 = vaddq_s16(row0, vsetq_lane_s16(1024, vdupq_n_s16(0), 0));
+
+   // column pass
+   dct_pass(vrshrn_n_s32, 10);
+
+   // 16bit 8x8 transpose
+   {
+// these three map to a single VTRN.16, VTRN.32, and VSWP, respectively.
+// whether compilers actually get this is another story, sadly.
+#define dct_trn16(x, y) { int16x8x2_t t = vtrnq_s16(x, y); x = t.val[0]; y = t.val[1]; }
+#define dct_trn32(x, y) { int32x4x2_t t = vtrnq_s32(vreinterpretq_s32_s16(x), vreinterpretq_s32_s16(y)); x = vreinterpretq_s16_s32(t.val[0]); y = vreinterpretq_s16_s32(t.val[1]); }
+#define dct_trn64(x, y) { int16x8_t x0 = x; int16x8_t y0 = y; x = vcombine_s16(vget_low_s16(x0), vget_low_s16(y0)); y = vcombine_s16(vget_high_s16(x0), vget_high_s16(y0)); }
+
+      // pass 1
+      dct_trn16(row0, row1); // a0b0a2b2a4b4a6b6
+      dct_trn16(row2, row3);
+      dct_trn16(row4, row5);
+      dct_trn16(row6, row7);
+
+      // pass 2
+      dct_trn32(row0, row2); // a0b0c0d0a4b4c4d4
+      dct_trn32(row1, row3);
+      dct_trn32(row4, row6);
+      dct_trn32(row5, row7);
+
+      // pass 3
+      dct_trn64(row0, row4); // a0b0c0d0e0f0g0h0
+      dct_trn64(row1, row5);
+      dct_trn64(row2, row6);
+      dct_trn64(row3, row7);
+
+#undef dct_trn16
+#undef dct_trn32
+#undef dct_trn64
+   }
+
+   // row pass
+   // vrshrn_n_s32 only supports shifts up to 16, we need
+   // 17. so do a non-rounding shift of 16 first then follow
+   // up with a rounding shift by 1.
+   dct_pass(vshrn_n_s32, 16);
+
+   {
+      // pack and round
+      uint8x8_t p0 = vqrshrun_n_s16(row0, 1);
+      uint8x8_t p1 = vqrshrun_n_s16(row1, 1);
+      uint8x8_t p2 = vqrshrun_n_s16(row2, 1);
+      uint8x8_t p3 = vqrshrun_n_s16(row3, 1);
+      uint8x8_t p4 = vqrshrun_n_s16(row4, 1);
+      uint8x8_t p5 = vqrshrun_n_s16(row5, 1);
+      uint8x8_t p6 = vqrshrun_n_s16(row6, 1);
+      uint8x8_t p7 = vqrshrun_n_s16(row7, 1);
+
+      // again, these can translate into one instruction, but often don't.
+#define dct_trn8_8(x, y) { uint8x8x2_t t = vtrn_u8(x, y); x = t.val[0]; y = t.val[1]; }
+#define dct_trn8_16(x, y) { uint16x4x2_t t = vtrn_u16(vreinterpret_u16_u8(x), vreinterpret_u16_u8(y)); x = vreinterpret_u8_u16(t.val[0]); y = vreinterpret_u8_u16(t.val[1]); }
+#define dct_trn8_32(x, y) { uint32x2x2_t t = vtrn_u32(vreinterpret_u32_u8(x), vreinterpret_u32_u8(y)); x = vreinterpret_u8_u32(t.val[0]); y = vreinterpret_u8_u32(t.val[1]); }
+
+      // sadly can't use interleaved stores here since we only write
+      // 8 bytes to each scan line!
+
+      // 8x8 8-bit transpose pass 1
+      dct_trn8_8(p0, p1);
+      dct_trn8_8(p2, p3);
+      dct_trn8_8(p4, p5);
+      dct_trn8_8(p6, p7);
+
+      // pass 2
+      dct_trn8_16(p0, p2);
+      dct_trn8_16(p1, p3);
+      dct_trn8_16(p4, p6);
+      dct_trn8_16(p5, p7);
+
+      // pass 3
+      dct_trn8_32(p0, p4);
+      dct_trn8_32(p1, p5);
+      dct_trn8_32(p2, p6);
+      dct_trn8_32(p3, p7);
+
+      // store
+      vst1_u8(out, p0); out += out_stride;
+      vst1_u8(out, p1); out += out_stride;
+      vst1_u8(out, p2); out += out_stride;
+      vst1_u8(out, p3); out += out_stride;
+      vst1_u8(out, p4); out += out_stride;
+      vst1_u8(out, p5); out += out_stride;
+      vst1_u8(out, p6); out += out_stride;
+      vst1_u8(out, p7);
+
+#undef dct_trn8_8
+#undef dct_trn8_16
+#undef dct_trn8_32
+   }
+
+#undef dct_long_mul
+#undef dct_long_mac
+#undef dct_widen
+#undef dct_wadd
+#undef dct_wsub
+#undef dct_bfly32o
+#undef dct_pass
+}
+
+#endif // STBI_NEON
+
+#define STBI__MARKER_none  0xff
+// if there's a pending marker from the entropy stream, return that
+// otherwise, fetch from the stream and get a marker. if there's no
+// marker, return 0xff, which is never a valid marker value
+static stbi_uc stbi__get_marker(stbi__jpeg *j)
+{
+   stbi_uc x;
+   if (j->marker != STBI__MARKER_none) { x = j->marker; j->marker = STBI__MARKER_none; return x; }
+   x = stbi__get8(j->s);
+   if (x != 0xff) return STBI__MARKER_none;
+   while (x == 0xff)
+      x = stbi__get8(j->s); // consume repeated 0xff fill bytes
+   return x;
+}
+
+// in each scan, we'll have scan_n components, and the order
+// of the components is specified by order[]
+#define STBI__RESTART(x)     ((x) >= 0xd0 && (x) <= 0xd7)
+
+// after a restart interval, stbi__jpeg_reset the entropy decoder and
+// the dc prediction
+static void stbi__jpeg_reset(stbi__jpeg *j)
+{
+   j->code_bits = 0;
+   j->code_buffer = 0;
+   j->nomore = 0;
+   j->img_comp[0].dc_pred = j->img_comp[1].dc_pred = j->img_comp[2].dc_pred = j->img_comp[3].dc_pred = 0;
+   j->marker = STBI__MARKER_none;
+   j->todo = j->restart_interval ? j->restart_interval : 0x7fffffff;
+   j->eob_run = 0;
+   // no more than 1<<31 MCUs if no restart_interal? that's plenty safe,
+   // since we don't even allow 1<<30 pixels
+}
+
+static int stbi__parse_entropy_coded_data(stbi__jpeg *z)
+{
+   stbi__jpeg_reset(z);
+   if (!z->progressive) {
+      if (z->scan_n == 1) {
+         int i,j;
+         STBI_SIMD_ALIGN(short, data[64]);
+         int n = z->order[0];
+         // non-interleaved data, we just need to process one block at a time,
+         // in trivial scanline order
+         // number of blocks to do just depends on how many actual "pixels" this
+         // component has, independent of interleaved MCU blocking and such
+         int w = (z->img_comp[n].x+7) >> 3;
+         int h = (z->img_comp[n].y+7) >> 3;
+         for (j=0; j < h; ++j) {
+            for (i=0; i < w; ++i) {
+               int ha = z->img_comp[n].ha;
+               if (!stbi__jpeg_decode_block(z, data, z->huff_dc+z->img_comp[n].hd, z->huff_ac+ha, z->fast_ac[ha], n, z->dequant[z->img_comp[n].tq])) return 0;
+               z->idct_block_kernel(z->img_comp[n].data+z->img_comp[n].w2*j*8+i*8, z->img_comp[n].w2, data);
+               // every data block is an MCU, so countdown the restart interval
+               if (--z->todo <= 0) {
+                  if (z->code_bits < 24) stbi__grow_buffer_unsafe(z);
+                  // if it's NOT a restart, then just bail, so we get corrupt data
+                  // rather than no data
+                  if (!STBI__RESTART(z->marker)) return 1;
+                  stbi__jpeg_reset(z);
+               }
+            }
+         }
+         return 1;
+      } else { // interleaved
+         int i,j,k,x,y;
+         STBI_SIMD_ALIGN(short, data[64]);
+         for (j=0; j < z->img_mcu_y; ++j) {
+            for (i=0; i < z->img_mcu_x; ++i) {
+               // scan an interleaved mcu... process scan_n components in order
+               for (k=0; k < z->scan_n; ++k) {
+                  int n = z->order[k];
+                  // scan out an mcu's worth of this component; that's just determined
+                  // by the basic H and V specified for the component
+                  for (y=0; y < z->img_comp[n].v; ++y) {
+                     for (x=0; x < z->img_comp[n].h; ++x) {
+                        int x2 = (i*z->img_comp[n].h + x)*8;
+                        int y2 = (j*z->img_comp[n].v + y)*8;
+                        int ha = z->img_comp[n].ha;
+                        if (!stbi__jpeg_decode_block(z, data, z->huff_dc+z->img_comp[n].hd, z->huff_ac+ha, z->fast_ac[ha], n, z->dequant[z->img_comp[n].tq])) return 0;
+                        z->idct_block_kernel(z->img_comp[n].data+z->img_comp[n].w2*y2+x2, z->img_comp[n].w2, data);
+                     }
+                  }
+               }
+               // after all interleaved components, that's an interleaved MCU,
+               // so now count down the restart interval
+               if (--z->todo <= 0) {
+                  if (z->code_bits < 24) stbi__grow_buffer_unsafe(z);
+                  if (!STBI__RESTART(z->marker)) return 1;
+                  stbi__jpeg_reset(z);
+               }
+            }
+         }
+         return 1;
+      }
+   } else {
+      if (z->scan_n == 1) {
+         int i,j;
+         int n = z->order[0];
+         // non-interleaved data, we just need to process one block at a time,
+         // in trivial scanline order
+         // number of blocks to do just depends on how many actual "pixels" this
+         // component has, independent of interleaved MCU blocking and such
+         int w = (z->img_comp[n].x+7) >> 3;
+         int h = (z->img_comp[n].y+7) >> 3;
+         for (j=0; j < h; ++j) {
+            for (i=0; i < w; ++i) {
+               short *data = z->img_comp[n].coeff + 64 * (i + j * z->img_comp[n].coeff_w);
+               if (z->spec_start == 0) {
+                  if (!stbi__jpeg_decode_block_prog_dc(z, data, &z->huff_dc[z->img_comp[n].hd], n))
+                     return 0;
+               } else {
+                  int ha = z->img_comp[n].ha;
+                  if (!stbi__jpeg_decode_block_prog_ac(z, data, &z->huff_ac[ha], z->fast_ac[ha]))
+                     return 0;
+               }
+               // every data block is an MCU, so countdown the restart interval
+               if (--z->todo <= 0) {
+                  if (z->code_bits < 24) stbi__grow_buffer_unsafe(z);
+                  if (!STBI__RESTART(z->marker)) return 1;
+                  stbi__jpeg_reset(z);
+               }
+            }
+         }
+         return 1;
+      } else { // interleaved
+         int i,j,k,x,y;
+         for (j=0; j < z->img_mcu_y; ++j) {
+            for (i=0; i < z->img_mcu_x; ++i) {
+               // scan an interleaved mcu... process scan_n components in order
+               for (k=0; k < z->scan_n; ++k) {
+                  int n = z->order[k];
+                  // scan out an mcu's worth of this component; that's just determined
+                  // by the basic H and V specified for the component
+                  for (y=0; y < z->img_comp[n].v; ++y) {
+                     for (x=0; x < z->img_comp[n].h; ++x) {
+                        int x2 = (i*z->img_comp[n].h + x);
+                        int y2 = (j*z->img_comp[n].v + y);
+                        short *data = z->img_comp[n].coeff + 64 * (x2 + y2 * z->img_comp[n].coeff_w);
+                        if (!stbi__jpeg_decode_block_prog_dc(z, data, &z->huff_dc[z->img_comp[n].hd], n))
+                           return 0;
+                     }
+                  }
+               }
+               // after all interleaved components, that's an interleaved MCU,
+               // so now count down the restart interval
+               if (--z->todo <= 0) {
+                  if (z->code_bits < 24) stbi__grow_buffer_unsafe(z);
+                  if (!STBI__RESTART(z->marker)) return 1;
+                  stbi__jpeg_reset(z);
+               }
+            }
+         }
+         return 1;
+      }
+   }
+}
+
+static void stbi__jpeg_dequantize(short *data, stbi__uint16 *dequant)
+{
+   int i;
+   for (i=0; i < 64; ++i)
+      data[i] *= dequant[i];
+}
+
+static void stbi__jpeg_finish(stbi__jpeg *z)
+{
+   if (z->progressive) {
+      // dequantize and idct the data
+      int i,j,n;
+      for (n=0; n < z->s->img_n; ++n) {
+         int w = (z->img_comp[n].x+7) >> 3;
+         int h = (z->img_comp[n].y+7) >> 3;
+         for (j=0; j < h; ++j) {
+            for (i=0; i < w; ++i) {
+               short *data = z->img_comp[n].coeff + 64 * (i + j * z->img_comp[n].coeff_w);
+               stbi__jpeg_dequantize(data, z->dequant[z->img_comp[n].tq]);
+               z->idct_block_kernel(z->img_comp[n].data+z->img_comp[n].w2*j*8+i*8, z->img_comp[n].w2, data);
+            }
+         }
+      }
+   }
+}
+
+static int stbi__process_marker(stbi__jpeg *z, int m)
+{
+   int L;
+   switch (m) {
+      case STBI__MARKER_none: // no marker found
+         return stbi__err("expected marker","Corrupt JPEG");
+
+      case 0xDD: // DRI - specify restart interval
+         if (stbi__get16be(z->s) != 4) return stbi__err("bad DRI len","Corrupt JPEG");
+         z->restart_interval = stbi__get16be(z->s);
+         return 1;
+
+      case 0xDB: // DQT - define quantization table
+         L = stbi__get16be(z->s)-2;
+         while (L > 0) {
+            int q = stbi__get8(z->s);
+            int p = q >> 4, sixteen = (p != 0);
+            int t = q & 15,i;
+            if (p != 0 && p != 1) return stbi__err("bad DQT type","Corrupt JPEG");
+            if (t > 3) return stbi__err("bad DQT table","Corrupt JPEG");
+
+            for (i=0; i < 64; ++i)
+               z->dequant[t][stbi__jpeg_dezigzag[i]] = (stbi__uint16)(sixteen ? stbi__get16be(z->s) : stbi__get8(z->s));
+            L -= (sixteen ? 129 : 65);
+         }
+         return L==0;
+
+      case 0xC4: // DHT - define huffman table
+         L = stbi__get16be(z->s)-2;
+         while (L > 0) {
+            stbi_uc *v;
+            int sizes[16],i,n=0;
+            int q = stbi__get8(z->s);
+            int tc = q >> 4;
+            int th = q & 15;
+            if (tc > 1 || th > 3) return stbi__err("bad DHT header","Corrupt JPEG");
+            for (i=0; i < 16; ++i) {
+               sizes[i] = stbi__get8(z->s);
+               n += sizes[i];
+            }
+            if(n > 256) return stbi__err("bad DHT header","Corrupt JPEG"); // Loop over i < n would write past end of values!
+            L -= 17;
+            if (tc == 0) {
+               if (!stbi__build_huffman(z->huff_dc+th, sizes)) return 0;
+               v = z->huff_dc[th].values;
+            } else {
+               if (!stbi__build_huffman(z->huff_ac+th, sizes)) return 0;
+               v = z->huff_ac[th].values;
+            }
+            for (i=0; i < n; ++i)
+               v[i] = stbi__get8(z->s);
+            if (tc != 0)
+               stbi__build_fast_ac(z->fast_ac[th], z->huff_ac + th);
+            L -= n;
+         }
+         return L==0;
+   }
+
+   // check for comment block or APP blocks
+   if ((m >= 0xE0 && m <= 0xEF) || m == 0xFE) {
+      L = stbi__get16be(z->s);
+      if (L < 2) {
+         if (m == 0xFE)
+            return stbi__err("bad COM len","Corrupt JPEG");
+         else
+            return stbi__err("bad APP len","Corrupt JPEG");
+      }
+      L -= 2;
+
+      if (m == 0xE0 && L >= 5) { // JFIF APP0 segment
+         static const unsigned char tag[5] = {'J','F','I','F','\0'};
+         int ok = 1;
+         int i;
+         for (i=0; i < 5; ++i)
+            if (stbi__get8(z->s) != tag[i])
+               ok = 0;
+         L -= 5;
+         if (ok)
+            z->jfif = 1;
+      } else if (m == 0xEE && L >= 12) { // Adobe APP14 segment
+         static const unsigned char tag[6] = {'A','d','o','b','e','\0'};
+         int ok = 1;
+         int i;
+         for (i=0; i < 6; ++i)
+            if (stbi__get8(z->s) != tag[i])
+               ok = 0;
+         L -= 6;
+         if (ok) {
+            stbi__get8(z->s); // version
+            stbi__get16be(z->s); // flags0
+            stbi__get16be(z->s); // flags1
+            z->app14_color_transform = stbi__get8(z->s); // color transform
+            L -= 6;
+         }
+      }
+
+      stbi__skip(z->s, L);
+      return 1;
+   }
+
+   return stbi__err("unknown marker","Corrupt JPEG");
+}
+
+// after we see SOS
+static int stbi__process_scan_header(stbi__jpeg *z)
+{
+   int i;
+   int Ls = stbi__get16be(z->s);
+   z->scan_n = stbi__get8(z->s);
+   if (z->scan_n < 1 || z->scan_n > 4 || z->scan_n > (int) z->s->img_n) return stbi__err("bad SOS component count","Corrupt JPEG");
+   if (Ls != 6+2*z->scan_n) return stbi__err("bad SOS len","Corrupt JPEG");
+   for (i=0; i < z->scan_n; ++i) {
+      int id = stbi__get8(z->s), which;
+      int q = stbi__get8(z->s);
+      for (which = 0; which < z->s->img_n; ++which)
+         if (z->img_comp[which].id == id)
+            break;
+      if (which == z->s->img_n) return 0; // no match
+      z->img_comp[which].hd = q >> 4;   if (z->img_comp[which].hd > 3) return stbi__err("bad DC huff","Corrupt JPEG");
+      z->img_comp[which].ha = q & 15;   if (z->img_comp[which].ha > 3) return stbi__err("bad AC huff","Corrupt JPEG");
+      z->order[i] = which;
+   }
+
+   {
+      int aa;
+      z->spec_start = stbi__get8(z->s);
+      z->spec_end   = stbi__get8(z->s); // should be 63, but might be 0
+      aa = stbi__get8(z->s);
+      z->succ_high = (aa >> 4);
+      z->succ_low  = (aa & 15);
+      if (z->progressive) {
+         if (z->spec_start > 63 || z->spec_end > 63  || z->spec_start > z->spec_end || z->succ_high > 13 || z->succ_low > 13)
+            return stbi__err("bad SOS", "Corrupt JPEG");
+      } else {
+         if (z->spec_start != 0) return stbi__err("bad SOS","Corrupt JPEG");
+         if (z->succ_high != 0 || z->succ_low != 0) return stbi__err("bad SOS","Corrupt JPEG");
+         z->spec_end = 63;
+      }
+   }
+
+   return 1;
+}
+
+static int stbi__free_jpeg_components(stbi__jpeg *z, int ncomp, int why)
+{
+   int i;
+   for (i=0; i < ncomp; ++i) {
+      if (z->img_comp[i].raw_data) {
+         STBI_FREE(z->img_comp[i].raw_data);
+         z->img_comp[i].raw_data = NULL;
+         z->img_comp[i].data = NULL;
+      }
+      if (z->img_comp[i].raw_coeff) {
+         STBI_FREE(z->img_comp[i].raw_coeff);
+         z->img_comp[i].raw_coeff = 0;
+         z->img_comp[i].coeff = 0;
+      }
+      if (z->img_comp[i].linebuf) {
+         STBI_FREE(z->img_comp[i].linebuf);
+         z->img_comp[i].linebuf = NULL;
+      }
+   }
+   return why;
+}
+
+static int stbi__process_frame_header(stbi__jpeg *z, int scan)
+{
+   stbi__context *s = z->s;
+   int Lf,p,i,q, h_max=1,v_max=1,c;
+   Lf = stbi__get16be(s);         if (Lf < 11) return stbi__err("bad SOF len","Corrupt JPEG"); // JPEG
+   p  = stbi__get8(s);            if (p != 8) return stbi__err("only 8-bit","JPEG format not supported: 8-bit only"); // JPEG baseline
+   s->img_y = stbi__get16be(s);   if (s->img_y == 0) return stbi__err("no header height", "JPEG format not supported: delayed height"); // Legal, but we don't handle it--but neither does IJG
+   s->img_x = stbi__get16be(s);   if (s->img_x == 0) return stbi__err("0 width","Corrupt JPEG"); // JPEG requires
+   if (s->img_y > STBI_MAX_DIMENSIONS) return stbi__err("too large","Very large image (corrupt?)");
+   if (s->img_x > STBI_MAX_DIMENSIONS) return stbi__err("too large","Very large image (corrupt?)");
+   c = stbi__get8(s);
+   if (c != 3 && c != 1 && c != 4) return stbi__err("bad component count","Corrupt JPEG");
+   s->img_n = c;
+   for (i=0; i < c; ++i) {
+      z->img_comp[i].data = NULL;
+      z->img_comp[i].linebuf = NULL;
+   }
+
+   if (Lf != 8+3*s->img_n) return stbi__err("bad SOF len","Corrupt JPEG");
+
+   z->rgb = 0;
+   for (i=0; i < s->img_n; ++i) {
+      static const unsigned char rgb[3] = { 'R', 'G', 'B' };
+      z->img_comp[i].id = stbi__get8(s);
+      if (s->img_n == 3 && z->img_comp[i].id == rgb[i])
+         ++z->rgb;
+      q = stbi__get8(s);
+      z->img_comp[i].h = (q >> 4);  if (!z->img_comp[i].h || z->img_comp[i].h > 4) return stbi__err("bad H","Corrupt JPEG");
+      z->img_comp[i].v = q & 15;    if (!z->img_comp[i].v || z->img_comp[i].v > 4) return stbi__err("bad V","Corrupt JPEG");
+      z->img_comp[i].tq = stbi__get8(s);  if (z->img_comp[i].tq > 3) return stbi__err("bad TQ","Corrupt JPEG");
+   }
+
+   if (scan != STBI__SCAN_load) return 1;
+
+   if (!stbi__mad3sizes_valid(s->img_x, s->img_y, s->img_n, 0)) return stbi__err("too large", "Image too large to decode");
+
+   for (i=0; i < s->img_n; ++i) {
+      if (z->img_comp[i].h > h_max) h_max = z->img_comp[i].h;
+      if (z->img_comp[i].v > v_max) v_max = z->img_comp[i].v;
+   }
+
+   // check that plane subsampling factors are integer ratios; our resamplers can't deal with fractional ratios
+   // and I've never seen a non-corrupted JPEG file actually use them
+   for (i=0; i < s->img_n; ++i) {
+      if (h_max % z->img_comp[i].h != 0) return stbi__err("bad H","Corrupt JPEG");
+      if (v_max % z->img_comp[i].v != 0) return stbi__err("bad V","Corrupt JPEG");
+   }
+
+   // compute interleaved mcu info
+   z->img_h_max = h_max;
+   z->img_v_max = v_max;
+   z->img_mcu_w = h_max * 8;
+   z->img_mcu_h = v_max * 8;
+   // these sizes can't be more than 17 bits
+   z->img_mcu_x = (s->img_x + z->img_mcu_w-1) / z->img_mcu_w;
+   z->img_mcu_y = (s->img_y + z->img_mcu_h-1) / z->img_mcu_h;
+
+   for (i=0; i < s->img_n; ++i) {
+      // number of effective pixels (e.g. for non-interleaved MCU)
+      z->img_comp[i].x = (s->img_x * z->img_comp[i].h + h_max-1) / h_max;
+      z->img_comp[i].y = (s->img_y * z->img_comp[i].v + v_max-1) / v_max;
+      // to simplify generation, we'll allocate enough memory to decode
+      // the bogus oversized data from using interleaved MCUs and their
+      // big blocks (e.g. a 16x16 iMCU on an image of width 33); we won't
+      // discard the extra data until colorspace conversion
+      //
+      // img_mcu_x, img_mcu_y: <=17 bits; comp[i].h and .v are <=4 (checked earlier)
+      // so these muls can't overflow with 32-bit ints (which we require)
+      z->img_comp[i].w2 = z->img_mcu_x * z->img_comp[i].h * 8;
+      z->img_comp[i].h2 = z->img_mcu_y * z->img_comp[i].v * 8;
+      z->img_comp[i].coeff = 0;
+      z->img_comp[i].raw_coeff = 0;
+      z->img_comp[i].linebuf = NULL;
+      z->img_comp[i].raw_data = stbi__malloc_mad2(z->img_comp[i].w2, z->img_comp[i].h2, 15);
+      if (z->img_comp[i].raw_data == NULL)
+         return stbi__free_jpeg_components(z, i+1, stbi__err("outofmem", "Out of memory"));
+      // align blocks for idct using mmx/sse
+      z->img_comp[i].data = (stbi_uc*) (((size_t) z->img_comp[i].raw_data + 15) & ~15);
+      if (z->progressive) {
+         // w2, h2 are multiples of 8 (see above)
+         z->img_comp[i].coeff_w = z->img_comp[i].w2 / 8;
+         z->img_comp[i].coeff_h = z->img_comp[i].h2 / 8;
+         z->img_comp[i].raw_coeff = stbi__malloc_mad3(z->img_comp[i].w2, z->img_comp[i].h2, sizeof(short), 15);
+         if (z->img_comp[i].raw_coeff == NULL)
+            return stbi__free_jpeg_components(z, i+1, stbi__err("outofmem", "Out of memory"));
+         z->img_comp[i].coeff = (short*) (((size_t) z->img_comp[i].raw_coeff + 15) & ~15);
+      }
+   }
+
+   return 1;
+}
+
+// use comparisons since in some cases we handle more than one case (e.g. SOF)
+#define stbi__DNL(x)         ((x) == 0xdc)
+#define stbi__SOI(x)         ((x) == 0xd8)
+#define stbi__EOI(x)         ((x) == 0xd9)
+#define stbi__SOF(x)         ((x) == 0xc0 || (x) == 0xc1 || (x) == 0xc2)
+#define stbi__SOS(x)         ((x) == 0xda)
+
+#define stbi__SOF_progressive(x)   ((x) == 0xc2)
+
+static int stbi__decode_jpeg_header(stbi__jpeg *z, int scan)
+{
+   int m;
+   z->jfif = 0;
+   z->app14_color_transform = -1; // valid values are 0,1,2
+   z->marker = STBI__MARKER_none; // initialize cached marker to empty
+   m = stbi__get_marker(z);
+   if (!stbi__SOI(m)) return stbi__err("no SOI","Corrupt JPEG");
+   if (scan == STBI__SCAN_type) return 1;
+   m = stbi__get_marker(z);
+   while (!stbi__SOF(m)) {
+      if (!stbi__process_marker(z,m)) return 0;
+      m = stbi__get_marker(z);
+      while (m == STBI__MARKER_none) {
+         // some files have extra padding after their blocks, so ok, we'll scan
+         if (stbi__at_eof(z->s)) return stbi__err("no SOF", "Corrupt JPEG");
+         m = stbi__get_marker(z);
+      }
+   }
+   z->progressive = stbi__SOF_progressive(m);
+   if (!stbi__process_frame_header(z, scan)) return 0;
+   return 1;
+}
+
+static int stbi__skip_jpeg_junk_at_end(stbi__jpeg *j)
+{
+   // some JPEGs have junk at end, skip over it but if we find what looks
+   // like a valid marker, resume there
+   while (!stbi__at_eof(j->s)) {
+      int x = stbi__get8(j->s);
+      while (x == 255) { // might be a marker
+         if (stbi__at_eof(j->s)) return STBI__MARKER_none;
+         x = stbi__get8(j->s);
+         if (x != 0x00 && x != 0xff) {
+            // not a stuffed zero or lead-in to another marker, looks
+            // like an actual marker, return it
+            return x;
+         }
+         // stuffed zero has x=0 now which ends the loop, meaning we go
+         // back to regular scan loop.
+         // repeated 0xff keeps trying to read the next byte of the marker.
+      }
+   }
+   return STBI__MARKER_none;
+}
+
+// decode image to YCbCr format
+static int stbi__decode_jpeg_image(stbi__jpeg *j)
+{
+   int m;
+   for (m = 0; m < 4; m++) {
+      j->img_comp[m].raw_data = NULL;
+      j->img_comp[m].raw_coeff = NULL;
+   }
+   j->restart_interval = 0;
+   if (!stbi__decode_jpeg_header(j, STBI__SCAN_load)) return 0;
+   m = stbi__get_marker(j);
+   while (!stbi__EOI(m)) {
+      if (stbi__SOS(m)) {
+         if (!stbi__process_scan_header(j)) return 0;
+         if (!stbi__parse_entropy_coded_data(j)) return 0;
+         if (j->marker == STBI__MARKER_none ) {
+         j->marker = stbi__skip_jpeg_junk_at_end(j);
+            // if we reach eof without hitting a marker, stbi__get_marker() below will fail and we'll eventually return 0
+         }
+         m = stbi__get_marker(j);
+         if (STBI__RESTART(m))
+            m = stbi__get_marker(j);
+      } else if (stbi__DNL(m)) {
+         int Ld = stbi__get16be(j->s);
+         stbi__uint32 NL = stbi__get16be(j->s);
+         if (Ld != 4) return stbi__err("bad DNL len", "Corrupt JPEG");
+         if (NL != j->s->img_y) return stbi__err("bad DNL height", "Corrupt JPEG");
+         m = stbi__get_marker(j);
+      } else {
+         if (!stbi__process_marker(j, m)) return 1;
+         m = stbi__get_marker(j);
+      }
+   }
+   if (j->progressive)
+      stbi__jpeg_finish(j);
+   return 1;
+}
+
+// static jfif-centered resampling (across block boundaries)
+
+typedef stbi_uc *(*resample_row_func)(stbi_uc *out, stbi_uc *in0, stbi_uc *in1,
+                                    int w, int hs);
+
+#define stbi__div4(x) ((stbi_uc) ((x) >> 2))
+
+static stbi_uc *resample_row_1(stbi_uc *out, stbi_uc *in_near, stbi_uc *in_far, int w, int hs)
+{
+   STBI_NOTUSED(out);
+   STBI_NOTUSED(in_far);
+   STBI_NOTUSED(w);
+   STBI_NOTUSED(hs);
+   return in_near;
+}
+
+static stbi_uc* stbi__resample_row_v_2(stbi_uc *out, stbi_uc *in_near, stbi_uc *in_far, int w, int hs)
+{
+   // need to generate two samples vertically for every one in input
+   int i;
+   STBI_NOTUSED(hs);
+   for (i=0; i < w; ++i)
+      out[i] = stbi__div4(3*in_near[i] + in_far[i] + 2);
+   return out;
+}
+
+static stbi_uc*  stbi__resample_row_h_2(stbi_uc *out, stbi_uc *in_near, stbi_uc *in_far, int w, int hs)
+{
+   // need to generate two samples horizontally for every one in input
+   int i;
+   stbi_uc *input = in_near;
+
+   if (w == 1) {
+      // if only one sample, can't do any interpolation
+      out[0] = out[1] = input[0];
+      return out;
+   }
+
+   out[0] = input[0];
+   out[1] = stbi__div4(input[0]*3 + input[1] + 2);
+   for (i=1; i < w-1; ++i) {
+      int n = 3*input[i]+2;
+      out[i*2+0] = stbi__div4(n+input[i-1]);
+      out[i*2+1] = stbi__div4(n+input[i+1]);
+   }
+   out[i*2+0] = stbi__div4(input[w-2]*3 + input[w-1] + 2);
+   out[i*2+1] = input[w-1];
+
+   STBI_NOTUSED(in_far);
+   STBI_NOTUSED(hs);
+
+   return out;
+}
+
+#define stbi__div16(x) ((stbi_uc) ((x) >> 4))
+
+static stbi_uc *stbi__resample_row_hv_2(stbi_uc *out, stbi_uc *in_near, stbi_uc *in_far, int w, int hs)
+{
+   // need to generate 2x2 samples for every one in input
+   int i,t0,t1;
+   if (w == 1) {
+      out[0] = out[1] = stbi__div4(3*in_near[0] + in_far[0] + 2);
+      return out;
+   }
+
+   t1 = 3*in_near[0] + in_far[0];
+   out[0] = stbi__div4(t1+2);
+   for (i=1; i < w; ++i) {
+      t0 = t1;
+      t1 = 3*in_near[i]+in_far[i];
+      out[i*2-1] = stbi__div16(3*t0 + t1 + 8);
+      out[i*2  ] = stbi__div16(3*t1 + t0 + 8);
+   }
+   out[w*2-1] = stbi__div4(t1+2);
+
+   STBI_NOTUSED(hs);
+
+   return out;
+}
+
+#if defined(STBI_SSE2) || defined(STBI_NEON)
+static stbi_uc *stbi__resample_row_hv_2_simd(stbi_uc *out, stbi_uc *in_near, stbi_uc *in_far, int w, int hs)
+{
+   // need to generate 2x2 samples for every one in input
+   int i=0,t0,t1;
+
+   if (w == 1) {
+      out[0] = out[1] = stbi__div4(3*in_near[0] + in_far[0] + 2);
+      return out;
+   }
+
+   t1 = 3*in_near[0] + in_far[0];
+   // process groups of 8 pixels for as long as we can.
+   // note we can't handle the last pixel in a row in this loop
+   // because we need to handle the filter boundary conditions.
+   for (; i < ((w-1) & ~7); i += 8) {
+#if defined(STBI_SSE2)
+      // load and perform the vertical filtering pass
+      // this uses 3*x + y = 4*x + (y - x)
+      __m128i zero  = _mm_setzero_si128();
+      __m128i farb  = _mm_loadl_epi64((__m128i *) (in_far + i));
+      __m128i nearb = _mm_loadl_epi64((__m128i *) (in_near + i));
+      __m128i farw  = _mm_unpacklo_epi8(farb, zero);
+      __m128i nearw = _mm_unpacklo_epi8(nearb, zero);
+      __m128i diff  = _mm_sub_epi16(farw, nearw);
+      __m128i nears = _mm_slli_epi16(nearw, 2);
+      __m128i curr  = _mm_add_epi16(nears, diff); // current row
+
+      // horizontal filter works the same based on shifted vers of current
+      // row. "prev" is current row shifted right by 1 pixel; we need to
+      // insert the previous pixel value (from t1).
+      // "next" is current row shifted left by 1 pixel, with first pixel
+      // of next block of 8 pixels added in.
+      __m128i prv0 = _mm_slli_si128(curr, 2);
+      __m128i nxt0 = _mm_srli_si128(curr, 2);
+      __m128i prev = _mm_insert_epi16(prv0, t1, 0);
+      __m128i next = _mm_insert_epi16(nxt0, 3*in_near[i+8] + in_far[i+8], 7);
+
+      // horizontal filter, polyphase implementation since it's convenient:
+      // even pixels = 3*cur + prev = cur*4 + (prev - cur)
+      // odd  pixels = 3*cur + next = cur*4 + (next - cur)
+      // note the shared term.
+      __m128i bias  = _mm_set1_epi16(8);
+      __m128i curs = _mm_slli_epi16(curr, 2);
+      __m128i prvd = _mm_sub_epi16(prev, curr);
+      __m128i nxtd = _mm_sub_epi16(next, curr);
+      __m128i curb = _mm_add_epi16(curs, bias);
+      __m128i even = _mm_add_epi16(prvd, curb);
+      __m128i odd  = _mm_add_epi16(nxtd, curb);
+
+      // interleave even and odd pixels, then undo scaling.
+      __m128i int0 = _mm_unpacklo_epi16(even, odd);
+      __m128i int1 = _mm_unpackhi_epi16(even, odd);
+      __m128i de0  = _mm_srli_epi16(int0, 4);
+      __m128i de1  = _mm_srli_epi16(int1, 4);
+
+      // pack and write output
+      __m128i outv = _mm_packus_epi16(de0, de1);
+      _mm_storeu_si128((__m128i *) (out + i*2), outv);
+#elif defined(STBI_NEON)
+      // load and perform the vertical filtering pass
+      // this uses 3*x + y = 4*x + (y - x)
+      uint8x8_t farb  = vld1_u8(in_far + i);
+      uint8x8_t nearb = vld1_u8(in_near + i);
+      int16x8_t diff  = vreinterpretq_s16_u16(vsubl_u8(farb, nearb));
+      int16x8_t nears = vreinterpretq_s16_u16(vshll_n_u8(nearb, 2));
+      int16x8_t curr  = vaddq_s16(nears, diff); // current row
+
+      // horizontal filter works the same based on shifted vers of current
+      // row. "prev" is current row shifted right by 1 pixel; we need to
+      // insert the previous pixel value (from t1).
+      // "next" is current row shifted left by 1 pixel, with first pixel
+      // of next block of 8 pixels added in.
+      int16x8_t prv0 = vextq_s16(curr, curr, 7);
+      int16x8_t nxt0 = vextq_s16(curr, curr, 1);
+      int16x8_t prev = vsetq_lane_s16(t1, prv0, 0);
+      int16x8_t next = vsetq_lane_s16(3*in_near[i+8] + in_far[i+8], nxt0, 7);
+
+      // horizontal filter, polyphase implementation since it's convenient:
+      // even pixels = 3*cur + prev = cur*4 + (prev - cur)
+      // odd  pixels = 3*cur + next = cur*4 + (next - cur)
+      // note the shared term.
+      int16x8_t curs = vshlq_n_s16(curr, 2);
+      int16x8_t prvd = vsubq_s16(prev, curr);
+      int16x8_t nxtd = vsubq_s16(next, curr);
+      int16x8_t even = vaddq_s16(curs, prvd);
+      int16x8_t odd  = vaddq_s16(curs, nxtd);
+
+      // undo scaling and round, then store with even/odd phases interleaved
+      uint8x8x2_t o;
+      o.val[0] = vqrshrun_n_s16(even, 4);
+      o.val[1] = vqrshrun_n_s16(odd,  4);
+      vst2_u8(out + i*2, o);
+#endif
+
+      // "previous" value for next iter
+      t1 = 3*in_near[i+7] + in_far[i+7];
+   }
+
+   t0 = t1;
+   t1 = 3*in_near[i] + in_far[i];
+   out[i*2] = stbi__div16(3*t1 + t0 + 8);
+
+   for (++i; i < w; ++i) {
+      t0 = t1;
+      t1 = 3*in_near[i]+in_far[i];
+      out[i*2-1] = stbi__div16(3*t0 + t1 + 8);
+      out[i*2  ] = stbi__div16(3*t1 + t0 + 8);
+   }
+   out[w*2-1] = stbi__div4(t1+2);
+
+   STBI_NOTUSED(hs);
+
+   return out;
+}
+#endif
+
+static stbi_uc *stbi__resample_row_generic(stbi_uc *out, stbi_uc *in_near, stbi_uc *in_far, int w, int hs)
+{
+   // resample with nearest-neighbor
+   int i,j;
+   STBI_NOTUSED(in_far);
+   for (i=0; i < w; ++i)
+      for (j=0; j < hs; ++j)
+         out[i*hs+j] = in_near[i];
+   return out;
+}
+
+// this is a reduced-precision calculation of YCbCr-to-RGB introduced
+// to make sure the code produces the same results in both SIMD and scalar
+#define stbi__float2fixed(x)  (((int) ((x) * 4096.0f + 0.5f)) << 8)
+static void stbi__YCbCr_to_RGB_row(stbi_uc *out, const stbi_uc *y, const stbi_uc *pcb, const stbi_uc *pcr, int count, int step)
+{
+   int i;
+   for (i=0; i < count; ++i) {
+      int y_fixed = (y[i] << 20) + (1<<19); // rounding
+      int r,g,b;
+      int cr = pcr[i] - 128;
+      int cb = pcb[i] - 128;
+      r = y_fixed +  cr* stbi__float2fixed(1.40200f);
+      g = y_fixed + (cr*-stbi__float2fixed(0.71414f)) + ((cb*-stbi__float2fixed(0.34414f)) & 0xffff0000);
+      b = y_fixed                                     +   cb* stbi__float2fixed(1.77200f);
+      r >>= 20;
+      g >>= 20;
+      b >>= 20;
+      if ((unsigned) r > 255) { if (r < 0) r = 0; else r = 255; }
+      if ((unsigned) g > 255) { if (g < 0) g = 0; else g = 255; }
+      if ((unsigned) b > 255) { if (b < 0) b = 0; else b = 255; }
+      out[0] = (stbi_uc)r;
+      out[1] = (stbi_uc)g;
+      out[2] = (stbi_uc)b;
+      out[3] = 255;
+      out += step;
+   }
+}
+
+#if defined(STBI_SSE2) || defined(STBI_NEON)
+static void stbi__YCbCr_to_RGB_simd(stbi_uc *out, stbi_uc const *y, stbi_uc const *pcb, stbi_uc const *pcr, int count, int step)
+{
+   int i = 0;
+
+#ifdef STBI_SSE2
+   // step == 3 is pretty ugly on the final interleave, and i'm not convinced
+   // it's useful in practice (you wouldn't use it for textures, for example).
+   // so just accelerate step == 4 case.
+   if (step == 4) {
+      // this is a fairly straightforward implementation and not super-optimized.
+      __m128i signflip  = _mm_set1_epi8(-0x80);
+      __m128i cr_const0 = _mm_set1_epi16(   (short) ( 1.40200f*4096.0f+0.5f));
+      __m128i cr_const1 = _mm_set1_epi16( - (short) ( 0.71414f*4096.0f+0.5f));
+      __m128i cb_const0 = _mm_set1_epi16( - (short) ( 0.34414f*4096.0f+0.5f));
+      __m128i cb_const1 = _mm_set1_epi16(   (short) ( 1.77200f*4096.0f+0.5f));
+      __m128i y_bias = _mm_set1_epi8((char) (unsigned char) 128);
+      __m128i xw = _mm_set1_epi16(255); // alpha channel
+
+      for (; i+7 < count; i += 8) {
+         // load
+         __m128i y_bytes = _mm_loadl_epi64((__m128i *) (y+i));
+         __m128i cr_bytes = _mm_loadl_epi64((__m128i *) (pcr+i));
+         __m128i cb_bytes = _mm_loadl_epi64((__m128i *) (pcb+i));
+         __m128i cr_biased = _mm_xor_si128(cr_bytes, signflip); // -128
+         __m128i cb_biased = _mm_xor_si128(cb_bytes, signflip); // -128
+
+         // unpack to short (and left-shift cr, cb by 8)
+         __m128i yw  = _mm_unpacklo_epi8(y_bias, y_bytes);
+         __m128i crw = _mm_unpacklo_epi8(_mm_setzero_si128(), cr_biased);
+         __m128i cbw = _mm_unpacklo_epi8(_mm_setzero_si128(), cb_biased);
+
+         // color transform
+         __m128i yws = _mm_srli_epi16(yw, 4);
+         __m128i cr0 = _mm_mulhi_epi16(cr_const0, crw);
+         __m128i cb0 = _mm_mulhi_epi16(cb_const0, cbw);
+         __m128i cb1 = _mm_mulhi_epi16(cbw, cb_const1);
+         __m128i cr1 = _mm_mulhi_epi16(crw, cr_const1);
+         __m128i rws = _mm_add_epi16(cr0, yws);
+         __m128i gwt = _mm_add_epi16(cb0, yws);
+         __m128i bws = _mm_add_epi16(yws, cb1);
+         __m128i gws = _mm_add_epi16(gwt, cr1);
+
+         // descale
+         __m128i rw = _mm_srai_epi16(rws, 4);
+         __m128i bw = _mm_srai_epi16(bws, 4);
+         __m128i gw = _mm_srai_epi16(gws, 4);
+
+         // back to byte, set up for transpose
+         __m128i brb = _mm_packus_epi16(rw, bw);
+         __m128i gxb = _mm_packus_epi16(gw, xw);
+
+         // transpose to interleave channels
+         __m128i t0 = _mm_unpacklo_epi8(brb, gxb);
+         __m128i t1 = _mm_unpackhi_epi8(brb, gxb);
+         __m128i o0 = _mm_unpacklo_epi16(t0, t1);
+         __m128i o1 = _mm_unpackhi_epi16(t0, t1);
+
+         // store
+         _mm_storeu_si128((__m128i *) (out + 0), o0);
+         _mm_storeu_si128((__m128i *) (out + 16), o1);
+         out += 32;
+      }
+   }
+#endif
+
+#ifdef STBI_NEON
+   // in this version, step=3 support would be easy to add. but is there demand?
+   if (step == 4) {
+      // this is a fairly straightforward implementation and not super-optimized.
+      uint8x8_t signflip = vdup_n_u8(0x80);
+      int16x8_t cr_const0 = vdupq_n_s16(   (short) ( 1.40200f*4096.0f+0.5f));
+      int16x8_t cr_const1 = vdupq_n_s16( - (short) ( 0.71414f*4096.0f+0.5f));
+      int16x8_t cb_const0 = vdupq_n_s16( - (short) ( 0.34414f*4096.0f+0.5f));
+      int16x8_t cb_const1 = vdupq_n_s16(   (short) ( 1.77200f*4096.0f+0.5f));
+
+      for (; i+7 < count; i += 8) {
+         // load
+         uint8x8_t y_bytes  = vld1_u8(y + i);
+         uint8x8_t cr_bytes = vld1_u8(pcr + i);
+         uint8x8_t cb_bytes = vld1_u8(pcb + i);
+         int8x8_t cr_biased = vreinterpret_s8_u8(vsub_u8(cr_bytes, signflip));
+         int8x8_t cb_biased = vreinterpret_s8_u8(vsub_u8(cb_bytes, signflip));
+
+         // expand to s16
+         int16x8_t yws = vreinterpretq_s16_u16(vshll_n_u8(y_bytes, 4));
+         int16x8_t crw = vshll_n_s8(cr_biased, 7);
+         int16x8_t cbw = vshll_n_s8(cb_biased, 7);
+
+         // color transform
+         int16x8_t cr0 = vqdmulhq_s16(crw, cr_const0);
+         int16x8_t cb0 = vqdmulhq_s16(cbw, cb_const0);
+         int16x8_t cr1 = vqdmulhq_s16(crw, cr_const1);
+         int16x8_t cb1 = vqdmulhq_s16(cbw, cb_const1);
+         int16x8_t rws = vaddq_s16(yws, cr0);
+         int16x8_t gws = vaddq_s16(vaddq_s16(yws, cb0), cr1);
+         int16x8_t bws = vaddq_s16(yws, cb1);
+
+         // undo scaling, round, convert to byte
+         uint8x8x4_t o;
+         o.val[0] = vqrshrun_n_s16(rws, 4);
+         o.val[1] = vqrshrun_n_s16(gws, 4);
+         o.val[2] = vqrshrun_n_s16(bws, 4);
+         o.val[3] = vdup_n_u8(255);
+
+         // store, interleaving r/g/b/a
+         vst4_u8(out, o);
+         out += 8*4;
+      }
+   }
+#endif
+
+   for (; i < count; ++i) {
+      int y_fixed = (y[i] << 20) + (1<<19); // rounding
+      int r,g,b;
+      int cr = pcr[i] - 128;
+      int cb = pcb[i] - 128;
+      r = y_fixed + cr* stbi__float2fixed(1.40200f);
+      g = y_fixed + cr*-stbi__float2fixed(0.71414f) + ((cb*-stbi__float2fixed(0.34414f)) & 0xffff0000);
+      b = y_fixed                                   +   cb* stbi__float2fixed(1.77200f);
+      r >>= 20;
+      g >>= 20;
+      b >>= 20;
+      if ((unsigned) r > 255) { if (r < 0) r = 0; else r = 255; }
+      if ((unsigned) g > 255) { if (g < 0) g = 0; else g = 255; }
+      if ((unsigned) b > 255) { if (b < 0) b = 0; else b = 255; }
+      out[0] = (stbi_uc)r;
+      out[1] = (stbi_uc)g;
+      out[2] = (stbi_uc)b;
+      out[3] = 255;
+      out += step;
+   }
+}
+#endif
+
+// set up the kernels
+static void stbi__setup_jpeg(stbi__jpeg *j)
+{
+   j->idct_block_kernel = stbi__idct_block;
+   j->YCbCr_to_RGB_kernel = stbi__YCbCr_to_RGB_row;
+   j->resample_row_hv_2_kernel = stbi__resample_row_hv_2;
+
+#ifdef STBI_SSE2
+   if (stbi__sse2_available()) {
+      j->idct_block_kernel = stbi__idct_simd;
+      j->YCbCr_to_RGB_kernel = stbi__YCbCr_to_RGB_simd;
+      j->resample_row_hv_2_kernel = stbi__resample_row_hv_2_simd;
+   }
+#endif
+
+#ifdef STBI_NEON
+   j->idct_block_kernel = stbi__idct_simd;
+   j->YCbCr_to_RGB_kernel = stbi__YCbCr_to_RGB_simd;
+   j->resample_row_hv_2_kernel = stbi__resample_row_hv_2_simd;
+#endif
+}
+
+// clean up the temporary component buffers
+static void stbi__cleanup_jpeg(stbi__jpeg *j)
+{
+   stbi__free_jpeg_components(j, j->s->img_n, 0);
+}
+
+typedef struct
+{
+   resample_row_func resample;
+   stbi_uc *line0,*line1;
+   int hs,vs;   // expansion factor in each axis
+   int w_lores; // horizontal pixels pre-expansion
+   int ystep;   // how far through vertical expansion we are
+   int ypos;    // which pre-expansion row we're on
+} stbi__resample;
+
+// fast 0..255 * 0..255 => 0..255 rounded multiplication
+static stbi_uc stbi__blinn_8x8(stbi_uc x, stbi_uc y)
+{
+   unsigned int t = x*y + 128;
+   return (stbi_uc) ((t + (t >>8)) >> 8);
+}
+
+static stbi_uc *load_jpeg_image(stbi__jpeg *z, int *out_x, int *out_y, int *comp, int req_comp)
+{
+   int n, decode_n, is_rgb;
+   z->s->img_n = 0; // make stbi__cleanup_jpeg safe
+
+   // validate req_comp
+   if (req_comp < 0 || req_comp > 4) return stbi__errpuc("bad req_comp", "Internal error");
+
+   // load a jpeg image from whichever source, but leave in YCbCr format
+   if (!stbi__decode_jpeg_image(z)) { stbi__cleanup_jpeg(z); return NULL; }
+
+   // determine actual number of components to generate
+   n = req_comp ? req_comp : z->s->img_n >= 3 ? 3 : 1;
+
+   is_rgb = z->s->img_n == 3 && (z->rgb == 3 || (z->app14_color_transform == 0 && !z->jfif));
+
+   if (z->s->img_n == 3 && n < 3 && !is_rgb)
+      decode_n = 1;
+   else
+      decode_n = z->s->img_n;
+
+   // nothing to do if no components requested; check this now to avoid
+   // accessing uninitialized coutput[0] later
+   if (decode_n <= 0) { stbi__cleanup_jpeg(z); return NULL; }
+
+   // resample and color-convert
+   {
+      int k;
+      unsigned int i,j;
+      stbi_uc *output;
+      stbi_uc *coutput[4] = { NULL, NULL, NULL, NULL };
+
+      stbi__resample res_comp[4];
+
+      for (k=0; k < decode_n; ++k) {
+         stbi__resample *r = &res_comp[k];
+
+         // allocate line buffer big enough for upsampling off the edges
+         // with upsample factor of 4
+         z->img_comp[k].linebuf = (stbi_uc *) stbi__malloc(z->s->img_x + 3);
+         if (!z->img_comp[k].linebuf) { stbi__cleanup_jpeg(z); return stbi__errpuc("outofmem", "Out of memory"); }
+
+         r->hs      = z->img_h_max / z->img_comp[k].h;
+         r->vs      = z->img_v_max / z->img_comp[k].v;
+         r->ystep   = r->vs >> 1;
+         r->w_lores = (z->s->img_x + r->hs-1) / r->hs;
+         r->ypos    = 0;
+         r->line0   = r->line1 = z->img_comp[k].data;
+
+         if      (r->hs == 1 && r->vs == 1) r->resample = resample_row_1;
+         else if (r->hs == 1 && r->vs == 2) r->resample = stbi__resample_row_v_2;
+         else if (r->hs == 2 && r->vs == 1) r->resample = stbi__resample_row_h_2;
+         else if (r->hs == 2 && r->vs == 2) r->resample = z->resample_row_hv_2_kernel;
+         else                               r->resample = stbi__resample_row_generic;
+      }
+
+      // can't error after this so, this is safe
+      output = (stbi_uc *) stbi__malloc_mad3(n, z->s->img_x, z->s->img_y, 1);
+      if (!output) { stbi__cleanup_jpeg(z); return stbi__errpuc("outofmem", "Out of memory"); }
+
+      // now go ahead and resample
+      for (j=0; j < z->s->img_y; ++j) {
+         stbi_uc *out = output + n * z->s->img_x * j;
+         for (k=0; k < decode_n; ++k) {
+            stbi__resample *r = &res_comp[k];
+            int y_bot = r->ystep >= (r->vs >> 1);
+            coutput[k] = r->resample(z->img_comp[k].linebuf,
+                                     y_bot ? r->line1 : r->line0,
+                                     y_bot ? r->line0 : r->line1,
+                                     r->w_lores, r->hs);
+            if (++r->ystep >= r->vs) {
+               r->ystep = 0;
+               r->line0 = r->line1;
+               if (++r->ypos < z->img_comp[k].y)
+                  r->line1 += z->img_comp[k].w2;
+            }
+         }
+         if (n >= 3) {
+            stbi_uc *y = coutput[0];
+            if (z->s->img_n == 3) {
+               if (is_rgb) {
+                  for (i=0; i < z->s->img_x; ++i) {
+                     out[0] = y[i];
+                     out[1] = coutput[1][i];
+                     out[2] = coutput[2][i];
+                     out[3] = 255;
+                     out += n;
+                  }
+               } else {
+                  z->YCbCr_to_RGB_kernel(out, y, coutput[1], coutput[2], z->s->img_x, n);
+               }
+            } else if (z->s->img_n == 4) {
+               if (z->app14_color_transform == 0) { // CMYK
+                  for (i=0; i < z->s->img_x; ++i) {
+                     stbi_uc m = coutput[3][i];
+                     out[0] = stbi__blinn_8x8(coutput[0][i], m);
+                     out[1] = stbi__blinn_8x8(coutput[1][i], m);
+                     out[2] = stbi__blinn_8x8(coutput[2][i], m);
+                     out[3] = 255;
+                     out += n;
+                  }
+               } else if (z->app14_color_transform == 2) { // YCCK
+                  z->YCbCr_to_RGB_kernel(out, y, coutput[1], coutput[2], z->s->img_x, n);
+                  for (i=0; i < z->s->img_x; ++i) {
+                     stbi_uc m = coutput[3][i];
+                     out[0] = stbi__blinn_8x8(255 - out[0], m);
+                     out[1] = stbi__blinn_8x8(255 - out[1], m);
+                     out[2] = stbi__blinn_8x8(255 - out[2], m);
+                     out += n;
+                  }
+               } else { // YCbCr + alpha?  Ignore the fourth channel for now
+                  z->YCbCr_to_RGB_kernel(out, y, coutput[1], coutput[2], z->s->img_x, n);
+               }
+            } else
+               for (i=0; i < z->s->img_x; ++i) {
+                  out[0] = out[1] = out[2] = y[i];
+                  out[3] = 255; // not used if n==3
+                  out += n;
+               }
+         } else {
+            if (is_rgb) {
+               if (n == 1)
+                  for (i=0; i < z->s->img_x; ++i)
+                     *out++ = stbi__compute_y(coutput[0][i], coutput[1][i], coutput[2][i]);
+               else {
+                  for (i=0; i < z->s->img_x; ++i, out += 2) {
+                     out[0] = stbi__compute_y(coutput[0][i], coutput[1][i], coutput[2][i]);
+                     out[1] = 255;
+                  }
+               }
+            } else if (z->s->img_n == 4 && z->app14_color_transform == 0) {
+               for (i=0; i < z->s->img_x; ++i) {
+                  stbi_uc m = coutput[3][i];
+                  stbi_uc r = stbi__blinn_8x8(coutput[0][i], m);
+                  stbi_uc g = stbi__blinn_8x8(coutput[1][i], m);
+                  stbi_uc b = stbi__blinn_8x8(coutput[2][i], m);
+                  out[0] = stbi__compute_y(r, g, b);
+                  out[1] = 255;
+                  out += n;
+               }
+            } else if (z->s->img_n == 4 && z->app14_color_transform == 2) {
+               for (i=0; i < z->s->img_x; ++i) {
+                  out[0] = stbi__blinn_8x8(255 - coutput[0][i], coutput[3][i]);
+                  out[1] = 255;
+                  out += n;
+               }
+            } else {
+               stbi_uc *y = coutput[0];
+               if (n == 1)
+                  for (i=0; i < z->s->img_x; ++i) out[i] = y[i];
+               else
+                  for (i=0; i < z->s->img_x; ++i) { *out++ = y[i]; *out++ = 255; }
+            }
+         }
+      }
+      stbi__cleanup_jpeg(z);
+      *out_x = z->s->img_x;
+      *out_y = z->s->img_y;
+      if (comp) *comp = z->s->img_n >= 3 ? 3 : 1; // report original components, not output
+      return output;
+   }
+}
+
+static void *stbi__jpeg_load(stbi__context *s, int *x, int *y, int *comp, int req_comp, stbi__result_info *ri)
+{
+   unsigned char* result;
+   stbi__jpeg* j = (stbi__jpeg*) stbi__malloc(sizeof(stbi__jpeg));
+   if (!j) return stbi__errpuc("outofmem", "Out of memory");
+   memset(j, 0, sizeof(stbi__jpeg));
+   STBI_NOTUSED(ri);
+   j->s = s;
+   stbi__setup_jpeg(j);
+   result = load_jpeg_image(j, x,y,comp,req_comp);
+   STBI_FREE(j);
+   return result;
+}
+
+static int stbi__jpeg_test(stbi__context *s)
+{
+   int r;
+   stbi__jpeg* j = (stbi__jpeg*)stbi__malloc(sizeof(stbi__jpeg));
+   if (!j) return stbi__err("outofmem", "Out of memory");
+   memset(j, 0, sizeof(stbi__jpeg));
+   j->s = s;
+   stbi__setup_jpeg(j);
+   r = stbi__decode_jpeg_header(j, STBI__SCAN_type);
+   stbi__rewind(s);
+   STBI_FREE(j);
+   return r;
+}
+
+static int stbi__jpeg_info_raw(stbi__jpeg *j, int *x, int *y, int *comp)
+{
+   if (!stbi__decode_jpeg_header(j, STBI__SCAN_header)) {
+      stbi__rewind( j->s );
+      return 0;
+   }
+   if (x) *x = j->s->img_x;
+   if (y) *y = j->s->img_y;
+   if (comp) *comp = j->s->img_n >= 3 ? 3 : 1;
+   return 1;
+}
+
+static int stbi__jpeg_info(stbi__context *s, int *x, int *y, int *comp)
+{
+   int result;
+   stbi__jpeg* j = (stbi__jpeg*) (stbi__malloc(sizeof(stbi__jpeg)));
+   if (!j) return stbi__err("outofmem", "Out of memory");
+   memset(j, 0, sizeof(stbi__jpeg));
+   j->s = s;
+   result = stbi__jpeg_info_raw(j, x, y, comp);
+   STBI_FREE(j);
+   return result;
+}
+#endif
+
+// public domain zlib decode    v0.2  Sean Barrett 2006-11-18
+//    simple implementation
+//      - all input must be provided in an upfront buffer
+//      - all output is written to a single output buffer (can malloc/realloc)
+//    performance
+//      - fast huffman
+
+#ifndef STBI_NO_ZLIB
+
+// fast-way is faster to check than jpeg huffman, but slow way is slower
+#define STBI__ZFAST_BITS  9 // accelerate all cases in default tables
+#define STBI__ZFAST_MASK  ((1 << STBI__ZFAST_BITS) - 1)
+#define STBI__ZNSYMS 288 // number of symbols in literal/length alphabet
+
+// zlib-style huffman encoding
+// (jpegs packs from left, zlib from right, so can't share code)
+typedef struct
+{
+   stbi__uint16 fast[1 << STBI__ZFAST_BITS];
+   stbi__uint16 firstcode[16];
+   int maxcode[17];
+   stbi__uint16 firstsymbol[16];
+   stbi_uc  size[STBI__ZNSYMS];
+   stbi__uint16 value[STBI__ZNSYMS];
+} stbi__zhuffman;
+
+stbi_inline static int stbi__bitreverse16(int n)
+{
+  n = ((n & 0xAAAA) >>  1) | ((n & 0x5555) << 1);
+  n = ((n & 0xCCCC) >>  2) | ((n & 0x3333) << 2);
+  n = ((n & 0xF0F0) >>  4) | ((n & 0x0F0F) << 4);
+  n = ((n & 0xFF00) >>  8) | ((n & 0x00FF) << 8);
+  return n;
+}
+
+stbi_inline static int stbi__bit_reverse(int v, int bits)
+{
+   STBI_ASSERT(bits <= 16);
+   // to bit reverse n bits, reverse 16 and shift
+   // e.g. 11 bits, bit reverse and shift away 5
+   return stbi__bitreverse16(v) >> (16-bits);
+}
+
+static int stbi__zbuild_huffman(stbi__zhuffman *z, const stbi_uc *sizelist, int num)
+{
+   int i,k=0;
+   int code, next_code[16], sizes[17];
+
+   // DEFLATE spec for generating codes
+   memset(sizes, 0, sizeof(sizes));
+   memset(z->fast, 0, sizeof(z->fast));
+   for (i=0; i < num; ++i)
+      ++sizes[sizelist[i]];
+   sizes[0] = 0;
+   for (i=1; i < 16; ++i)
+      if (sizes[i] > (1 << i))
+         return stbi__err("bad sizes", "Corrupt PNG");
+   code = 0;
+   for (i=1; i < 16; ++i) {
+      next_code[i] = code;
+      z->firstcode[i] = (stbi__uint16) code;
+      z->firstsymbol[i] = (stbi__uint16) k;
+      code = (code + sizes[i]);
+      if (sizes[i])
+         if (code-1 >= (1 << i)) return stbi__err("bad codelengths","Corrupt PNG");
+      z->maxcode[i] = code << (16-i); // preshift for inner loop
+      code <<= 1;
+      k += sizes[i];
+   }
+   z->maxcode[16] = 0x10000; // sentinel
+   for (i=0; i < num; ++i) {
+      int s = sizelist[i];
+      if (s) {
+         int c = next_code[s] - z->firstcode[s] + z->firstsymbol[s];
+         stbi__uint16 fastv = (stbi__uint16) ((s << 9) | i);
+         z->size [c] = (stbi_uc     ) s;
+         z->value[c] = (stbi__uint16) i;
+         if (s <= STBI__ZFAST_BITS) {
+            int j = stbi__bit_reverse(next_code[s],s);
+            while (j < (1 << STBI__ZFAST_BITS)) {
+               z->fast[j] = fastv;
+               j += (1 << s);
+            }
+         }
+         ++next_code[s];
+      }
+   }
+   return 1;
+}
+
+// zlib-from-memory implementation for PNG reading
+//    because PNG allows splitting the zlib stream arbitrarily,
+//    and it's annoying structurally to have PNG call ZLIB call PNG,
+//    we require PNG read all the IDATs and combine them into a single
+//    memory buffer
+
+typedef struct
+{
+   stbi_uc *zbuffer, *zbuffer_end;
+   int num_bits;
+   stbi__uint32 code_buffer;
+
+   char *zout;
+   char *zout_start;
+   char *zout_end;
+   int   z_expandable;
+
+   stbi__zhuffman z_length, z_distance;
+} stbi__zbuf;
+
+stbi_inline static int stbi__zeof(stbi__zbuf *z)
+{
+   return (z->zbuffer >= z->zbuffer_end);
+}
+
+stbi_inline static stbi_uc stbi__zget8(stbi__zbuf *z)
+{
+   return stbi__zeof(z) ? 0 : *z->zbuffer++;
+}
+
+static void stbi__fill_bits(stbi__zbuf *z)
+{
+   do {
+      if (z->code_buffer >= (1U << z->num_bits)) {
+        z->zbuffer = z->zbuffer_end;  /* treat this as EOF so we fail. */
+        return;
+      }
+      z->code_buffer |= (unsigned int) stbi__zget8(z) << z->num_bits;
+      z->num_bits += 8;
+   } while (z->num_bits <= 24);
+}
+
+stbi_inline static unsigned int stbi__zreceive(stbi__zbuf *z, int n)
+{
+   unsigned int k;
+   if (z->num_bits < n) stbi__fill_bits(z);
+   k = z->code_buffer & ((1 << n) - 1);
+   z->code_buffer >>= n;
+   z->num_bits -= n;
+   return k;
+}
+
+static int stbi__zhuffman_decode_slowpath(stbi__zbuf *a, stbi__zhuffman *z)
+{
+   int b,s,k;
+   // not resolved by fast table, so compute it the slow way
+   // use jpeg approach, which requires MSbits at top
+   k = stbi__bit_reverse(a->code_buffer, 16);
+   for (s=STBI__ZFAST_BITS+1; ; ++s)
+      if (k < z->maxcode[s])
+         break;
+   if (s >= 16) return -1; // invalid code!
+   // code size is s, so:
+   b = (k >> (16-s)) - z->firstcode[s] + z->firstsymbol[s];
+   if (b >= STBI__ZNSYMS) return -1; // some data was corrupt somewhere!
+   if (z->size[b] != s) return -1;  // was originally an assert, but report failure instead.
+   a->code_buffer >>= s;
+   a->num_bits -= s;
+   return z->value[b];
+}
+
+stbi_inline static int stbi__zhuffman_decode(stbi__zbuf *a, stbi__zhuffman *z)
+{
+   int b,s;
+   if (a->num_bits < 16) {
+      if (stbi__zeof(a)) {
+         return -1;   /* report error for unexpected end of data. */
+      }
+      stbi__fill_bits(a);
+   }
+   b = z->fast[a->code_buffer & STBI__ZFAST_MASK];
+   if (b) {
+      s = b >> 9;
+      a->code_buffer >>= s;
+      a->num_bits -= s;
+      return b & 511;
+   }
+   return stbi__zhuffman_decode_slowpath(a, z);
+}
+
+static int stbi__zexpand(stbi__zbuf *z, char *zout, int n)  // need to make room for n bytes
+{
+   char *q;
+   unsigned int cur, limit, old_limit;
+   z->zout = zout;
+   if (!z->z_expandable) return stbi__err("output buffer limit","Corrupt PNG");
+   cur   = (unsigned int) (z->zout - z->zout_start);
+   limit = old_limit = (unsigned) (z->zout_end - z->zout_start);
+   if (UINT_MAX - cur < (unsigned) n) return stbi__err("outofmem", "Out of memory");
+   while (cur + n > limit) {
+      if(limit > UINT_MAX / 2) return stbi__err("outofmem", "Out of memory");
+      limit *= 2;
+   }
+   q = (char *) STBI_REALLOC_SIZED(z->zout_start, old_limit, limit);
+   STBI_NOTUSED(old_limit);
+   if (q == NULL) return stbi__err("outofmem", "Out of memory");
+   z->zout_start = q;
+   z->zout       = q + cur;
+   z->zout_end   = q + limit;
+   return 1;
+}
+
+static const int stbi__zlength_base[31] = {
+   3,4,5,6,7,8,9,10,11,13,
+   15,17,19,23,27,31,35,43,51,59,
+   67,83,99,115,131,163,195,227,258,0,0 };
+
+static const int stbi__zlength_extra[31]=
+{ 0,0,0,0,0,0,0,0,1,1,1,1,2,2,2,2,3,3,3,3,4,4,4,4,5,5,5,5,0,0,0 };
+
+static const int stbi__zdist_base[32] = { 1,2,3,4,5,7,9,13,17,25,33,49,65,97,129,193,
+257,385,513,769,1025,1537,2049,3073,4097,6145,8193,12289,16385,24577,0,0};
+
+static const int stbi__zdist_extra[32] =
+{ 0,0,0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7,8,8,9,9,10,10,11,11,12,12,13,13};
+
+static int stbi__parse_huffman_block(stbi__zbuf *a)
+{
+   char *zout = a->zout;
+   for(;;) {
+      int z = stbi__zhuffman_decode(a, &a->z_length);
+      if (z < 256) {
+         if (z < 0) return stbi__err("bad huffman code","Corrupt PNG"); // error in huffman codes
+         if (zout >= a->zout_end) {
+            if (!stbi__zexpand(a, zout, 1)) return 0;
+            zout = a->zout;
+         }
+         *zout++ = (char) z;
+      } else {
+         stbi_uc *p;
+         int len,dist;
+         if (z == 256) {
+            a->zout = zout;
+            return 1;
+         }
+         if (z >= 286) return stbi__err("bad huffman code","Corrupt PNG"); // per DEFLATE, length codes 286 and 287 must not appear in compressed data
+         z -= 257;
+         len = stbi__zlength_base[z];
+         if (stbi__zlength_extra[z]) len += stbi__zreceive(a, stbi__zlength_extra[z]);
+         z = stbi__zhuffman_decode(a, &a->z_distance);
+         if (z < 0 || z >= 30) return stbi__err("bad huffman code","Corrupt PNG"); // per DEFLATE, distance codes 30 and 31 must not appear in compressed data
+         dist = stbi__zdist_base[z];
+         if (stbi__zdist_extra[z]) dist += stbi__zreceive(a, stbi__zdist_extra[z]);
+         if (zout - a->zout_start < dist) return stbi__err("bad dist","Corrupt PNG");
+         if (zout + len > a->zout_end) {
+            if (!stbi__zexpand(a, zout, len)) return 0;
+            zout = a->zout;
+         }
+         p = (stbi_uc *) (zout - dist);
+         if (dist == 1) { // run of one byte; common in images.
+            stbi_uc v = *p;
+            if (len) { do *zout++ = v; while (--len); }
+         } else {
+            if (len) { do *zout++ = *p++; while (--len); }
+         }
+      }
+   }
+}
+
+static int stbi__compute_huffman_codes(stbi__zbuf *a)
+{
+   static const stbi_uc length_dezigzag[19] = { 16,17,18,0,8,7,9,6,10,5,11,4,12,3,13,2,14,1,15 };
+   stbi__zhuffman z_codelength;
+   stbi_uc lencodes[286+32+137];//padding for maximum single op
+   stbi_uc codelength_sizes[19];
+   int i,n;
+
+   int hlit  = stbi__zreceive(a,5) + 257;
+   int hdist = stbi__zreceive(a,5) + 1;
+   int hclen = stbi__zreceive(a,4) + 4;
+   int ntot  = hlit + hdist;
+
+   memset(codelength_sizes, 0, sizeof(codelength_sizes));
+   for (i=0; i < hclen; ++i) {
+      int s = stbi__zreceive(a,3);
+      codelength_sizes[length_dezigzag[i]] = (stbi_uc) s;
+   }
+   if (!stbi__zbuild_huffman(&z_codelength, codelength_sizes, 19)) return 0;
+
+   n = 0;
+   while (n < ntot) {
+      int c = stbi__zhuffman_decode(a, &z_codelength);
+      if (c < 0 || c >= 19) return stbi__err("bad codelengths", "Corrupt PNG");
+      if (c < 16)
+         lencodes[n++] = (stbi_uc) c;
+      else {
+         stbi_uc fill = 0;
+         if (c == 16) {
+            c = stbi__zreceive(a,2)+3;
+            if (n == 0) return stbi__err("bad codelengths", "Corrupt PNG");
+            fill = lencodes[n-1];
+         } else if (c == 17) {
+            c = stbi__zreceive(a,3)+3;
+         } else if (c == 18) {
+            c = stbi__zreceive(a,7)+11;
+         } else {
+            return stbi__err("bad codelengths", "Corrupt PNG");
+         }
+         if (ntot - n < c) return stbi__err("bad codelengths", "Corrupt PNG");
+         memset(lencodes+n, fill, c);
+         n += c;
+      }
+   }
+   if (n != ntot) return stbi__err("bad codelengths","Corrupt PNG");
+   if (!stbi__zbuild_huffman(&a->z_length, lencodes, hlit)) return 0;
+   if (!stbi__zbuild_huffman(&a->z_distance, lencodes+hlit, hdist)) return 0;
+   return 1;
+}
+
+static int stbi__parse_uncompressed_block(stbi__zbuf *a)
+{
+   stbi_uc header[4];
+   int len,nlen,k;
+   if (a->num_bits & 7)
+      stbi__zreceive(a, a->num_bits & 7); // discard
+   // drain the bit-packed data into header
+   k = 0;
+   while (a->num_bits > 0) {
+      header[k++] = (stbi_uc) (a->code_buffer & 255); // suppress MSVC run-time check
+      a->code_buffer >>= 8;
+      a->num_bits -= 8;
+   }
+   if (a->num_bits < 0) return stbi__err("zlib corrupt","Corrupt PNG");
+   // now fill header the normal way
+   while (k < 4)
+      header[k++] = stbi__zget8(a);
+   len  = header[1] * 256 + header[0];
+   nlen = header[3] * 256 + header[2];
+   if (nlen != (len ^ 0xffff)) return stbi__err("zlib corrupt","Corrupt PNG");
+   if (a->zbuffer + len > a->zbuffer_end) return stbi__err("read past buffer","Corrupt PNG");
+   if (a->zout + len > a->zout_end)
+      if (!stbi__zexpand(a, a->zout, len)) return 0;
+   memcpy(a->zout, a->zbuffer, len);
+   a->zbuffer += len;
+   a->zout += len;
+   return 1;
+}
+
+static int stbi__parse_zlib_header(stbi__zbuf *a)
+{
+   int cmf   = stbi__zget8(a);
+   int cm    = cmf & 15;
+   /* int cinfo = cmf >> 4; */
+   int flg   = stbi__zget8(a);
+   if (stbi__zeof(a)) return stbi__err("bad zlib header","Corrupt PNG"); // zlib spec
+   if ((cmf*256+flg) % 31 != 0) return stbi__err("bad zlib header","Corrupt PNG"); // zlib spec
+   if (flg & 32) return stbi__err("no preset dict","Corrupt PNG"); // preset dictionary not allowed in png
+   if (cm != 8) return stbi__err("bad compression","Corrupt PNG"); // DEFLATE required for png
+   // window = 1 << (8 + cinfo)... but who cares, we fully buffer output
+   return 1;
+}
+
+static const stbi_uc stbi__zdefault_length[STBI__ZNSYMS] =
+{
+   8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8, 8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,
+   8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8, 8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,
+   8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8, 8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,
+   8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8, 8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,
+   8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8, 9,9,9,9,9,9,9,9,9,9,9,9,9,9,9,9,
+   9,9,9,9,9,9,9,9,9,9,9,9,9,9,9,9, 9,9,9,9,9,9,9,9,9,9,9,9,9,9,9,9,
+   9,9,9,9,9,9,9,9,9,9,9,9,9,9,9,9, 9,9,9,9,9,9,9,9,9,9,9,9,9,9,9,9,
+   9,9,9,9,9,9,9,9,9,9,9,9,9,9,9,9, 9,9,9,9,9,9,9,9,9,9,9,9,9,9,9,9,
+   7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7, 7,7,7,7,7,7,7,7,8,8,8,8,8,8,8,8
+};
+static const stbi_uc stbi__zdefault_distance[32] =
+{
+   5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5
+};
+/*
+Init algorithm:
+{
+   int i;   // use <= to match clearly with spec
+   for (i=0; i <= 143; ++i)     stbi__zdefault_length[i]   = 8;
+   for (   ; i <= 255; ++i)     stbi__zdefault_length[i]   = 9;
+   for (   ; i <= 279; ++i)     stbi__zdefault_length[i]   = 7;
+   for (   ; i <= 287; ++i)     stbi__zdefault_length[i]   = 8;
+
+   for (i=0; i <=  31; ++i)     stbi__zdefault_distance[i] = 5;
+}
+*/
+
+static int stbi__parse_zlib(stbi__zbuf *a, int parse_header)
+{
+   int final, type;
+   if (parse_header)
+      if (!stbi__parse_zlib_header(a)) return 0;
+   a->num_bits = 0;
+   a->code_buffer = 0;
+   do {
+      final = stbi__zreceive(a,1);
+      type = stbi__zreceive(a,2);
+      if (type == 0) {
+         if (!stbi__parse_uncompressed_block(a)) return 0;
+      } else if (type == 3) {
+         return 0;
+      } else {
+         if (type == 1) {
+            // use fixed code lengths
+            if (!stbi__zbuild_huffman(&a->z_length  , stbi__zdefault_length  , STBI__ZNSYMS)) return 0;
+            if (!stbi__zbuild_huffman(&a->z_distance, stbi__zdefault_distance,  32)) return 0;
+         } else {
+            if (!stbi__compute_huffman_codes(a)) return 0;
+         }
+         if (!stbi__parse_huffman_block(a)) return 0;
+      }
+   } while (!final);
+   return 1;
+}
+
+static int stbi__do_zlib(stbi__zbuf *a, char *obuf, int olen, int exp, int parse_header)
+{
+   a->zout_start = obuf;
+   a->zout       = obuf;
+   a->zout_end   = obuf + olen;
+   a->z_expandable = exp;
+
+   return stbi__parse_zlib(a, parse_header);
+}
+
+STBIDEF char *stbi_zlib_decode_malloc_guesssize(const char *buffer, int len, int initial_size, int *outlen)
+{
+   stbi__zbuf a;
+   char *p = (char *) stbi__malloc(initial_size);
+   if (p == NULL) return NULL;
+   a.zbuffer = (stbi_uc *) buffer;
+   a.zbuffer_end = (stbi_uc *) buffer + len;
+   if (stbi__do_zlib(&a, p, initial_size, 1, 1)) {
+      if (outlen) *outlen = (int) (a.zout - a.zout_start);
+      return a.zout_start;
+   } else {
+      STBI_FREE(a.zout_start);
+      return NULL;
+   }
+}
+
+STBIDEF char *stbi_zlib_decode_malloc(char const *buffer, int len, int *outlen)
+{
+   return stbi_zlib_decode_malloc_guesssize(buffer, len, 16384, outlen);
+}
+
+STBIDEF char *stbi_zlib_decode_malloc_guesssize_headerflag(const char *buffer, int len, int initial_size, int *outlen, int parse_header)
+{
+   stbi__zbuf a;
+   char *p = (char *) stbi__malloc(initial_size);
+   if (p == NULL) return NULL;
+   a.zbuffer = (stbi_uc *) buffer;
+   a.zbuffer_end = (stbi_uc *) buffer + len;
+   if (stbi__do_zlib(&a, p, initial_size, 1, parse_header)) {
+      if (outlen) *outlen = (int) (a.zout - a.zout_start);
+      return a.zout_start;
+   } else {
+      STBI_FREE(a.zout_start);
+      return NULL;
+   }
+}
+
+STBIDEF int stbi_zlib_decode_buffer(char *obuffer, int olen, char const *ibuffer, int ilen)
+{
+   stbi__zbuf a;
+   a.zbuffer = (stbi_uc *) ibuffer;
+   a.zbuffer_end = (stbi_uc *) ibuffer + ilen;
+   if (stbi__do_zlib(&a, obuffer, olen, 0, 1))
+      return (int) (a.zout - a.zout_start);
+   else
+      return -1;
+}
+
+STBIDEF char *stbi_zlib_decode_noheader_malloc(char const *buffer, int len, int *outlen)
+{
+   stbi__zbuf a;
+   char *p = (char *) stbi__malloc(16384);
+   if (p == NULL) return NULL;
+   a.zbuffer = (stbi_uc *) buffer;
+   a.zbuffer_end = (stbi_uc *) buffer+len;
+   if (stbi__do_zlib(&a, p, 16384, 1, 0)) {
+      if (outlen) *outlen = (int) (a.zout - a.zout_start);
+      return a.zout_start;
+   } else {
+      STBI_FREE(a.zout_start);
+      return NULL;
+   }
+}
+
+STBIDEF int stbi_zlib_decode_noheader_buffer(char *obuffer, int olen, const char *ibuffer, int ilen)
+{
+   stbi__zbuf a;
+   a.zbuffer = (stbi_uc *) ibuffer;
+   a.zbuffer_end = (stbi_uc *) ibuffer + ilen;
+   if (stbi__do_zlib(&a, obuffer, olen, 0, 0))
+      return (int) (a.zout - a.zout_start);
+   else
+      return -1;
+}
+#endif
+
+// public domain "baseline" PNG decoder   v0.10  Sean Barrett 2006-11-18
+//    simple implementation
+//      - only 8-bit samples
+//      - no CRC checking
+//      - allocates lots of intermediate memory
+//        - avoids problem of streaming data between subsystems
+//        - avoids explicit window management
+//    performance
+//      - uses stb_zlib, a PD zlib implementation with fast huffman decoding
+
+#ifndef STBI_NO_PNG
+typedef struct
+{
+   stbi__uint32 length;
+   stbi__uint32 type;
+} stbi__pngchunk;
+
+static stbi__pngchunk stbi__get_chunk_header(stbi__context *s)
+{
+   stbi__pngchunk c;
+   c.length = stbi__get32be(s);
+   c.type   = stbi__get32be(s);
+   return c;
+}
+
+static int stbi__check_png_header(stbi__context *s)
+{
+   static const stbi_uc png_sig[8] = { 137,80,78,71,13,10,26,10 };
+   int i;
+   for (i=0; i < 8; ++i)
+      if (stbi__get8(s) != png_sig[i]) return stbi__err("bad png sig","Not a PNG");
+   return 1;
+}
+
+typedef struct
+{
+   stbi__context *s;
+   stbi_uc *idata, *expanded, *out;
+   int depth;
+} stbi__png;
+
+
+enum {
+   STBI__F_none=0,
+   STBI__F_sub=1,
+   STBI__F_up=2,
+   STBI__F_avg=3,
+   STBI__F_paeth=4,
+   // synthetic filters used for first scanline to avoid needing a dummy row of 0s
+   STBI__F_avg_first,
+   STBI__F_paeth_first
+};
+
+static stbi_uc first_row_filter[5] =
+{
+   STBI__F_none,
+   STBI__F_sub,
+   STBI__F_none,
+   STBI__F_avg_first,
+   STBI__F_paeth_first
+};
+
+static int stbi__paeth(int a, int b, int c)
+{
+   int p = a + b - c;
+   int pa = abs(p-a);
+   int pb = abs(p-b);
+   int pc = abs(p-c);
+   if (pa <= pb && pa <= pc) return a;
+   if (pb <= pc) return b;
+   return c;
+}
+
+static const stbi_uc stbi__depth_scale_table[9] = { 0, 0xff, 0x55, 0, 0x11, 0,0,0, 0x01 };
+
+// create the png data from post-deflated data
+static int stbi__create_png_image_raw(stbi__png *a, stbi_uc *raw, stbi__uint32 raw_len, int out_n, stbi__uint32 x, stbi__uint32 y, int depth, int color)
+{
+   int bytes = (depth == 16? 2 : 1);
+   stbi__context *s = a->s;
+   stbi__uint32 i,j,stride = x*out_n*bytes;
+   stbi__uint32 img_len, img_width_bytes;
+   int k;
+   int img_n = s->img_n; // copy it into a local for later
+
+   int output_bytes = out_n*bytes;
+   int filter_bytes = img_n*bytes;
+   int width = x;
+
+   STBI_ASSERT(out_n == s->img_n || out_n == s->img_n+1);
+   a->out = (stbi_uc *) stbi__malloc_mad3(x, y, output_bytes, 0); // extra bytes to write off the end into
+   if (!a->out) return stbi__err("outofmem", "Out of memory");
+
+   if (!stbi__mad3sizes_valid(img_n, x, depth, 7)) return stbi__err("too large", "Corrupt PNG");
+   img_width_bytes = (((img_n * x * depth) + 7) >> 3);
+   img_len = (img_width_bytes + 1) * y;
+
+   // we used to check for exact match between raw_len and img_len on non-interlaced PNGs,
+   // but issue #276 reported a PNG in the wild that had extra data at the end (all zeros),
+   // so just check for raw_len < img_len always.
+   if (raw_len < img_len) return stbi__err("not enough pixels","Corrupt PNG");
+
+   for (j=0; j < y; ++j) {
+      stbi_uc *cur = a->out + stride*j;
+      stbi_uc *prior;
+      int filter = *raw++;
+
+      if (filter > 4)
+         return stbi__err("invalid filter","Corrupt PNG");
+
+      if (depth < 8) {
+         if (img_width_bytes > x) return stbi__err("invalid width","Corrupt PNG");
+         cur += x*out_n - img_width_bytes; // store output to the rightmost img_len bytes, so we can decode in place
+         filter_bytes = 1;
+         width = img_width_bytes;
+      }
+      prior = cur - stride; // bugfix: need to compute this after 'cur +=' computation above
+
+      // if first row, use special filter that doesn't sample previous row
+      if (j == 0) filter = first_row_filter[filter];
+
+      // handle first byte explicitly
+      for (k=0; k < filter_bytes; ++k) {
+         switch (filter) {
+            case STBI__F_none       : cur[k] = raw[k]; break;
+            case STBI__F_sub        : cur[k] = raw[k]; break;
+            case STBI__F_up         : cur[k] = STBI__BYTECAST(raw[k] + prior[k]); break;
+            case STBI__F_avg        : cur[k] = STBI__BYTECAST(raw[k] + (prior[k]>>1)); break;
+            case STBI__F_paeth      : cur[k] = STBI__BYTECAST(raw[k] + stbi__paeth(0,prior[k],0)); break;
+            case STBI__F_avg_first  : cur[k] = raw[k]; break;
+            case STBI__F_paeth_first: cur[k] = raw[k]; break;
+         }
+      }
+
+      if (depth == 8) {
+         if (img_n != out_n)
+            cur[img_n] = 255; // first pixel
+         raw += img_n;
+         cur += out_n;
+         prior += out_n;
+      } else if (depth == 16) {
+         if (img_n != out_n) {
+            cur[filter_bytes]   = 255; // first pixel top byte
+            cur[filter_bytes+1] = 255; // first pixel bottom byte
+         }
+         raw += filter_bytes;
+         cur += output_bytes;
+         prior += output_bytes;
+      } else {
+         raw += 1;
+         cur += 1;
+         prior += 1;
+      }
+
+      // this is a little gross, so that we don't switch per-pixel or per-component
+      if (depth < 8 || img_n == out_n) {
+         int nk = (width - 1)*filter_bytes;
+         #define STBI__CASE(f) \
+             case f:     \
+                for (k=0; k < nk; ++k)
+         switch (filter) {
+            // "none" filter turns into a memcpy here; make that explicit.
+            case STBI__F_none:         memcpy(cur, raw, nk); break;
+            STBI__CASE(STBI__F_sub)          { cur[k] = STBI__BYTECAST(raw[k] + cur[k-filter_bytes]); } break;
+            STBI__CASE(STBI__F_up)           { cur[k] = STBI__BYTECAST(raw[k] + prior[k]); } break;
+            STBI__CASE(STBI__F_avg)          { cur[k] = STBI__BYTECAST(raw[k] + ((prior[k] + cur[k-filter_bytes])>>1)); } break;
+            STBI__CASE(STBI__F_paeth)        { cur[k] = STBI__BYTECAST(raw[k] + stbi__paeth(cur[k-filter_bytes],prior[k],prior[k-filter_bytes])); } break;
+            STBI__CASE(STBI__F_avg_first)    { cur[k] = STBI__BYTECAST(raw[k] + (cur[k-filter_bytes] >> 1)); } break;
+            STBI__CASE(STBI__F_paeth_first)  { cur[k] = STBI__BYTECAST(raw[k] + stbi__paeth(cur[k-filter_bytes],0,0)); } break;
+         }
+         #undef STBI__CASE
+         raw += nk;
+      } else {
+         STBI_ASSERT(img_n+1 == out_n);
+         #define STBI__CASE(f) \
+             case f:     \
+                for (i=x-1; i >= 1; --i, cur[filter_bytes]=255,raw+=filter_bytes,cur+=output_bytes,prior+=output_bytes) \
+                   for (k=0; k < filter_bytes; ++k)
+         switch (filter) {
+            STBI__CASE(STBI__F_none)         { cur[k] = raw[k]; } break;
+            STBI__CASE(STBI__F_sub)          { cur[k] = STBI__BYTECAST(raw[k] + cur[k- output_bytes]); } break;
+            STBI__CASE(STBI__F_up)           { cur[k] = STBI__BYTECAST(raw[k] + prior[k]); } break;
+            STBI__CASE(STBI__F_avg)          { cur[k] = STBI__BYTECAST(raw[k] + ((prior[k] + cur[k- output_bytes])>>1)); } break;
+            STBI__CASE(STBI__F_paeth)        { cur[k] = STBI__BYTECAST(raw[k] + stbi__paeth(cur[k- output_bytes],prior[k],prior[k- output_bytes])); } break;
+            STBI__CASE(STBI__F_avg_first)    { cur[k] = STBI__BYTECAST(raw[k] + (cur[k- output_bytes] >> 1)); } break;
+            STBI__CASE(STBI__F_paeth_first)  { cur[k] = STBI__BYTECAST(raw[k] + stbi__paeth(cur[k- output_bytes],0,0)); } break;
+         }
+         #undef STBI__CASE
+
+         // the loop above sets the high byte of the pixels' alpha, but for
+         // 16 bit png files we also need the low byte set. we'll do that here.
+         if (depth == 16) {
+            cur = a->out + stride*j; // start at the beginning of the row again
+            for (i=0; i < x; ++i,cur+=output_bytes) {
+               cur[filter_bytes+1] = 255;
+            }
+         }
+      }
+   }
+
+   // we make a separate pass to expand bits to pixels; for performance,
+   // this could run two scanlines behind the above code, so it won't
+   // intefere with filtering but will still be in the cache.
+   if (depth < 8) {
+      for (j=0; j < y; ++j) {
+         stbi_uc *cur = a->out + stride*j;
+         stbi_uc *in  = a->out + stride*j + x*out_n - img_width_bytes;
+         // unpack 1/2/4-bit into a 8-bit buffer. allows us to keep the common 8-bit path optimal at minimal cost for 1/2/4-bit
+         // png guarante byte alignment, if width is not multiple of 8/4/2 we'll decode dummy trailing data that will be skipped in the later loop
+         stbi_uc scale = (color == 0) ? stbi__depth_scale_table[depth] : 1; // scale grayscale values to 0..255 range
+
+         // note that the final byte might overshoot and write more data than desired.
+         // we can allocate enough data that this never writes out of memory, but it
+         // could also overwrite the next scanline. can it overwrite non-empty data
+         // on the next scanline? yes, consider 1-pixel-wide scanlines with 1-bit-per-pixel.
+         // so we need to explicitly clamp the final ones
+
+         if (depth == 4) {
+            for (k=x*img_n; k >= 2; k-=2, ++in) {
+               *cur++ = scale * ((*in >> 4)       );
+               *cur++ = scale * ((*in     ) & 0x0f);
+            }
+            if (k > 0) *cur++ = scale * ((*in >> 4)       );
+         } else if (depth == 2) {
+            for (k=x*img_n; k >= 4; k-=4, ++in) {
+               *cur++ = scale * ((*in >> 6)       );
+               *cur++ = scale * ((*in >> 4) & 0x03);
+               *cur++ = scale * ((*in >> 2) & 0x03);
+               *cur++ = scale * ((*in     ) & 0x03);
+            }
+            if (k > 0) *cur++ = scale * ((*in >> 6)       );
+            if (k > 1) *cur++ = scale * ((*in >> 4) & 0x03);
+            if (k > 2) *cur++ = scale * ((*in >> 2) & 0x03);
+         } else if (depth == 1) {
+            for (k=x*img_n; k >= 8; k-=8, ++in) {
+               *cur++ = scale * ((*in >> 7)       );
+               *cur++ = scale * ((*in >> 6) & 0x01);
+               *cur++ = scale * ((*in >> 5) & 0x01);
+               *cur++ = scale * ((*in >> 4) & 0x01);
+               *cur++ = scale * ((*in >> 3) & 0x01);
+               *cur++ = scale * ((*in >> 2) & 0x01);
+               *cur++ = scale * ((*in >> 1) & 0x01);
+               *cur++ = scale * ((*in     ) & 0x01);
+            }
+            if (k > 0) *cur++ = scale * ((*in >> 7)       );
+            if (k > 1) *cur++ = scale * ((*in >> 6) & 0x01);
+            if (k > 2) *cur++ = scale * ((*in >> 5) & 0x01);
+            if (k > 3) *cur++ = scale * ((*in >> 4) & 0x01);
+            if (k > 4) *cur++ = scale * ((*in >> 3) & 0x01);
+            if (k > 5) *cur++ = scale * ((*in >> 2) & 0x01);
+            if (k > 6) *cur++ = scale * ((*in >> 1) & 0x01);
+         }
+         if (img_n != out_n) {
+            int q;
+            // insert alpha = 255
+            cur = a->out + stride*j;
+            if (img_n == 1) {
+               for (q=x-1; q >= 0; --q) {
+                  cur[q*2+1] = 255;
+                  cur[q*2+0] = cur[q];
+               }
+            } else {
+               STBI_ASSERT(img_n == 3);
+               for (q=x-1; q >= 0; --q) {
+                  cur[q*4+3] = 255;
+                  cur[q*4+2] = cur[q*3+2];
+                  cur[q*4+1] = cur[q*3+1];
+                  cur[q*4+0] = cur[q*3+0];
+               }
+            }
+         }
+      }
+   } else if (depth == 16) {
+      // force the image data from big-endian to platform-native.
+      // this is done in a separate pass due to the decoding relying
+      // on the data being untouched, but could probably be done
+      // per-line during decode if care is taken.
+      stbi_uc *cur = a->out;
+      stbi__uint16 *cur16 = (stbi__uint16*)cur;
+
+      for(i=0; i < x*y*out_n; ++i,cur16++,cur+=2) {
+         *cur16 = (cur[0] << 8) | cur[1];
+      }
+   }
+
+   return 1;
+}
+
+static int stbi__create_png_image(stbi__png *a, stbi_uc *image_data, stbi__uint32 image_data_len, int out_n, int depth, int color, int interlaced)
+{
+   int bytes = (depth == 16 ? 2 : 1);
+   int out_bytes = out_n * bytes;
+   stbi_uc *final;
+   int p;
+   if (!interlaced)
+      return stbi__create_png_image_raw(a, image_data, image_data_len, out_n, a->s->img_x, a->s->img_y, depth, color);
+
+   // de-interlacing
+   final = (stbi_uc *) stbi__malloc_mad3(a->s->img_x, a->s->img_y, out_bytes, 0);
+   if (!final) return stbi__err("outofmem", "Out of memory");
+   for (p=0; p < 7; ++p) {
+      int xorig[] = { 0,4,0,2,0,1,0 };
+      int yorig[] = { 0,0,4,0,2,0,1 };
+      int xspc[]  = { 8,8,4,4,2,2,1 };
+      int yspc[]  = { 8,8,8,4,4,2,2 };
+      int i,j,x,y;
+      // pass1_x[4] = 0, pass1_x[5] = 1, pass1_x[12] = 1
+      x = (a->s->img_x - xorig[p] + xspc[p]-1) / xspc[p];
+      y = (a->s->img_y - yorig[p] + yspc[p]-1) / yspc[p];
+      if (x && y) {
+         stbi__uint32 img_len = ((((a->s->img_n * x * depth) + 7) >> 3) + 1) * y;
+         if (!stbi__create_png_image_raw(a, image_data, image_data_len, out_n, x, y, depth, color)) {
+            STBI_FREE(final);
+            return 0;
+         }
+         for (j=0; j < y; ++j) {
+            for (i=0; i < x; ++i) {
+               int out_y = j*yspc[p]+yorig[p];
+               int out_x = i*xspc[p]+xorig[p];
+               memcpy(final + out_y*a->s->img_x*out_bytes + out_x*out_bytes,
+                      a->out + (j*x+i)*out_bytes, out_bytes);
+            }
+         }
+         STBI_FREE(a->out);
+         image_data += img_len;
+         image_data_len -= img_len;
+      }
+   }
+   a->out = final;
+
+   return 1;
+}
+
+static int stbi__compute_transparency(stbi__png *z, stbi_uc tc[3], int out_n)
+{
+   stbi__context *s = z->s;
+   stbi__uint32 i, pixel_count = s->img_x * s->img_y;
+   stbi_uc *p = z->out;
+
+   // compute color-based transparency, assuming we've
+   // already got 255 as the alpha value in the output
+   STBI_ASSERT(out_n == 2 || out_n == 4);
+
+   if (out_n == 2) {
+      for (i=0; i < pixel_count; ++i) {
+         p[1] = (p[0] == tc[0] ? 0 : 255);
+         p += 2;
+      }
+   } else {
+      for (i=0; i < pixel_count; ++i) {
+         if (p[0] == tc[0] && p[1] == tc[1] && p[2] == tc[2])
+            p[3] = 0;
+         p += 4;
+      }
+   }
+   return 1;
+}
+
+static int stbi__compute_transparency16(stbi__png *z, stbi__uint16 tc[3], int out_n)
+{
+   stbi__context *s = z->s;
+   stbi__uint32 i, pixel_count = s->img_x * s->img_y;
+   stbi__uint16 *p = (stbi__uint16*) z->out;
+
+   // compute color-based transparency, assuming we've
+   // already got 65535 as the alpha value in the output
+   STBI_ASSERT(out_n == 2 || out_n == 4);
+
+   if (out_n == 2) {
+      for (i = 0; i < pixel_count; ++i) {
+         p[1] = (p[0] == tc[0] ? 0 : 65535);
+         p += 2;
+      }
+   } else {
+      for (i = 0; i < pixel_count; ++i) {
+         if (p[0] == tc[0] && p[1] == tc[1] && p[2] == tc[2])
+            p[3] = 0;
+         p += 4;
+      }
+   }
+   return 1;
+}
+
+static int stbi__expand_png_palette(stbi__png *a, stbi_uc *palette, int len, int pal_img_n)
+{
+   stbi__uint32 i, pixel_count = a->s->img_x * a->s->img_y;
+   stbi_uc *p, *temp_out, *orig = a->out;
+
+   p = (stbi_uc *) stbi__malloc_mad2(pixel_count, pal_img_n, 0);
+   if (p == NULL) return stbi__err("outofmem", "Out of memory");
+
+   // between here and free(out) below, exitting would leak
+   temp_out = p;
+
+   if (pal_img_n == 3) {
+      for (i=0; i < pixel_count; ++i) {
+         int n = orig[i]*4;
+         p[0] = palette[n  ];
+         p[1] = palette[n+1];
+         p[2] = palette[n+2];
+         p += 3;
+      }
+   } else {
+      for (i=0; i < pixel_count; ++i) {
+         int n = orig[i]*4;
+         p[0] = palette[n  ];
+         p[1] = palette[n+1];
+         p[2] = palette[n+2];
+         p[3] = palette[n+3];
+         p += 4;
+      }
+   }
+   STBI_FREE(a->out);
+   a->out = temp_out;
+
+   STBI_NOTUSED(len);
+
+   return 1;
+}
+
+static int stbi__unpremultiply_on_load_global = 0;
+static int stbi__de_iphone_flag_global = 0;
+
+STBIDEF void stbi_set_unpremultiply_on_load(int flag_true_if_should_unpremultiply)
+{
+   stbi__unpremultiply_on_load_global = flag_true_if_should_unpremultiply;
+}
+
+STBIDEF void stbi_convert_iphone_png_to_rgb(int flag_true_if_should_convert)
+{
+   stbi__de_iphone_flag_global = flag_true_if_should_convert;
+}
+
+#ifndef STBI_THREAD_LOCAL
+#define stbi__unpremultiply_on_load  stbi__unpremultiply_on_load_global
+#define stbi__de_iphone_flag  stbi__de_iphone_flag_global
+#else
+static STBI_THREAD_LOCAL int stbi__unpremultiply_on_load_local, stbi__unpremultiply_on_load_set;
+static STBI_THREAD_LOCAL int stbi__de_iphone_flag_local, stbi__de_iphone_flag_set;
+
+STBIDEF void stbi_set_unpremultiply_on_load_thread(int flag_true_if_should_unpremultiply)
+{
+   stbi__unpremultiply_on_load_local = flag_true_if_should_unpremultiply;
+   stbi__unpremultiply_on_load_set = 1;
+}
+
+STBIDEF void stbi_convert_iphone_png_to_rgb_thread(int flag_true_if_should_convert)
+{
+   stbi__de_iphone_flag_local = flag_true_if_should_convert;
+   stbi__de_iphone_flag_set = 1;
+}
+
+#define stbi__unpremultiply_on_load  (stbi__unpremultiply_on_load_set           \
+                                       ? stbi__unpremultiply_on_load_local      \
+                                       : stbi__unpremultiply_on_load_global)
+#define stbi__de_iphone_flag  (stbi__de_iphone_flag_set                         \
+                                ? stbi__de_iphone_flag_local                    \
+                                : stbi__de_iphone_flag_global)
+#endif // STBI_THREAD_LOCAL
+
+static void stbi__de_iphone(stbi__png *z)
+{
+   stbi__context *s = z->s;
+   stbi__uint32 i, pixel_count = s->img_x * s->img_y;
+   stbi_uc *p = z->out;
+
+   if (s->img_out_n == 3) {  // convert bgr to rgb
+      for (i=0; i < pixel_count; ++i) {
+         stbi_uc t = p[0];
+         p[0] = p[2];
+         p[2] = t;
+         p += 3;
+      }
+   } else {
+      STBI_ASSERT(s->img_out_n == 4);
+      if (stbi__unpremultiply_on_load) {
+         // convert bgr to rgb and unpremultiply
+         for (i=0; i < pixel_count; ++i) {
+            stbi_uc a = p[3];
+            stbi_uc t = p[0];
+            if (a) {
+               stbi_uc half = a / 2;
+               p[0] = (p[2] * 255 + half) / a;
+               p[1] = (p[1] * 255 + half) / a;
+               p[2] = ( t   * 255 + half) / a;
+            } else {
+               p[0] = p[2];
+               p[2] = t;
+            }
+            p += 4;
+         }
+      } else {
+         // convert bgr to rgb
+         for (i=0; i < pixel_count; ++i) {
+            stbi_uc t = p[0];
+            p[0] = p[2];
+            p[2] = t;
+            p += 4;
+         }
+      }
+   }
+}
+
+#define STBI__PNG_TYPE(a,b,c,d)  (((unsigned) (a) << 24) + ((unsigned) (b) << 16) + ((unsigned) (c) << 8) + (unsigned) (d))
+
+static int stbi__parse_png_file(stbi__png *z, int scan, int req_comp)
+{
+   stbi_uc palette[1024], pal_img_n=0;
+   stbi_uc has_trans=0, tc[3]={0};
+   stbi__uint16 tc16[3];
+   stbi__uint32 ioff=0, idata_limit=0, i, pal_len=0;
+   int first=1,k,interlace=0, color=0, is_iphone=0;
+   stbi__context *s = z->s;
+
+   z->expanded = NULL;
+   z->idata = NULL;
+   z->out = NULL;
+
+   if (!stbi__check_png_header(s)) return 0;
+
+   if (scan == STBI__SCAN_type) return 1;
+
+   for (;;) {
+      stbi__pngchunk c = stbi__get_chunk_header(s);
+      switch (c.type) {
+         case STBI__PNG_TYPE('C','g','B','I'):
+            is_iphone = 1;
+            stbi__skip(s, c.length);
+            break;
+         case STBI__PNG_TYPE('I','H','D','R'): {
+            int comp,filter;
+            if (!first) return stbi__err("multiple IHDR","Corrupt PNG");
+            first = 0;
+            if (c.length != 13) return stbi__err("bad IHDR len","Corrupt PNG");
+            s->img_x = stbi__get32be(s);
+            s->img_y = stbi__get32be(s);
+            if (s->img_y > STBI_MAX_DIMENSIONS) return stbi__err("too large","Very large image (corrupt?)");
+            if (s->img_x > STBI_MAX_DIMENSIONS) return stbi__err("too large","Very large image (corrupt?)");
+            z->depth = stbi__get8(s);  if (z->depth != 1 && z->depth != 2 && z->depth != 4 && z->depth != 8 && z->depth != 16)  return stbi__err("1/2/4/8/16-bit only","PNG not supported: 1/2/4/8/16-bit only");
+            color = stbi__get8(s);  if (color > 6)         return stbi__err("bad ctype","Corrupt PNG");
+            if (color == 3 && z->depth == 16)                  return stbi__err("bad ctype","Corrupt PNG");
+            if (color == 3) pal_img_n = 3; else if (color & 1) return stbi__err("bad ctype","Corrupt PNG");
+            comp  = stbi__get8(s);  if (comp) return stbi__err("bad comp method","Corrupt PNG");
+            filter= stbi__get8(s);  if (filter) return stbi__err("bad filter method","Corrupt PNG");
+            interlace = stbi__get8(s); if (interlace>1) return stbi__err("bad interlace method","Corrupt PNG");
+            if (!s->img_x || !s->img_y) return stbi__err("0-pixel image","Corrupt PNG");
+            if (!pal_img_n) {
+               s->img_n = (color & 2 ? 3 : 1) + (color & 4 ? 1 : 0);
+               if ((1 << 30) / s->img_x / s->img_n < s->img_y) return stbi__err("too large", "Image too large to decode");
+            } else {
+               // if paletted, then pal_n is our final components, and
+               // img_n is # components to decompress/filter.
+               s->img_n = 1;
+               if ((1 << 30) / s->img_x / 4 < s->img_y) return stbi__err("too large","Corrupt PNG");
+            }
+            // even with SCAN_header, have to scan to see if we have a tRNS
+            break;
+         }
+
+         case STBI__PNG_TYPE('P','L','T','E'):  {
+            if (first) return stbi__err("first not IHDR", "Corrupt PNG");
+            if (c.length > 256*3) return stbi__err("invalid PLTE","Corrupt PNG");
+            pal_len = c.length / 3;
+            if (pal_len * 3 != c.length) return stbi__err("invalid PLTE","Corrupt PNG");
+            for (i=0; i < pal_len; ++i) {
+               palette[i*4+0] = stbi__get8(s);
+               palette[i*4+1] = stbi__get8(s);
+               palette[i*4+2] = stbi__get8(s);
+               palette[i*4+3] = 255;
+            }
+            break;
+         }
+
+         case STBI__PNG_TYPE('t','R','N','S'): {
+            if (first) return stbi__err("first not IHDR", "Corrupt PNG");
+            if (z->idata) return stbi__err("tRNS after IDAT","Corrupt PNG");
+            if (pal_img_n) {
+               if (scan == STBI__SCAN_header) { s->img_n = 4; return 1; }
+               if (pal_len == 0) return stbi__err("tRNS before PLTE","Corrupt PNG");
+               if (c.length > pal_len) return stbi__err("bad tRNS len","Corrupt PNG");
+               pal_img_n = 4;
+               for (i=0; i < c.length; ++i)
+                  palette[i*4+3] = stbi__get8(s);
+            } else {
+               if (!(s->img_n & 1)) return stbi__err("tRNS with alpha","Corrupt PNG");
+               if (c.length != (stbi__uint32) s->img_n*2) return stbi__err("bad tRNS len","Corrupt PNG");
+               has_trans = 1;
+               // non-paletted with tRNS = constant alpha. if header-scanning, we can stop now.
+               if (scan == STBI__SCAN_header) { ++s->img_n; return 1; }
+               if (z->depth == 16) {
+                  for (k = 0; k < s->img_n; ++k) tc16[k] = (stbi__uint16)stbi__get16be(s); // copy the values as-is
+               } else {
+                  for (k = 0; k < s->img_n; ++k) tc[k] = (stbi_uc)(stbi__get16be(s) & 255) * stbi__depth_scale_table[z->depth]; // non 8-bit images will be larger
+               }
+            }
+            break;
+         }
+
+         case STBI__PNG_TYPE('I','D','A','T'): {
+            if (first) return stbi__err("first not IHDR", "Corrupt PNG");
+            if (pal_img_n && !pal_len) return stbi__err("no PLTE","Corrupt PNG");
+            if (scan == STBI__SCAN_header) {
+               // header scan definitely stops at first IDAT
+               if (pal_img_n)
+                  s->img_n = pal_img_n;
+               return 1;
+            }
+            if (c.length > (1u << 30)) return stbi__err("IDAT size limit", "IDAT section larger than 2^30 bytes");
+            if ((int)(ioff + c.length) < (int)ioff) return 0;
+            if (ioff + c.length > idata_limit) {
+               stbi__uint32 idata_limit_old = idata_limit;
+               stbi_uc *p;
+               if (idata_limit == 0) idata_limit = c.length > 4096 ? c.length : 4096;
+               while (ioff + c.length > idata_limit)
+                  idata_limit *= 2;
+               STBI_NOTUSED(idata_limit_old);
+               p = (stbi_uc *) STBI_REALLOC_SIZED(z->idata, idata_limit_old, idata_limit); if (p == NULL) return stbi__err("outofmem", "Out of memory");
+               z->idata = p;
+            }
+            if (!stbi__getn(s, z->idata+ioff,c.length)) return stbi__err("outofdata","Corrupt PNG");
+            ioff += c.length;
+            break;
+         }
+
+         case STBI__PNG_TYPE('I','E','N','D'): {
+            stbi__uint32 raw_len, bpl;
+            if (first) return stbi__err("first not IHDR", "Corrupt PNG");
+            if (scan != STBI__SCAN_load) return 1;
+            if (z->idata == NULL) return stbi__err("no IDAT","Corrupt PNG");
+            // initial guess for decoded data size to avoid unnecessary reallocs
+            bpl = (s->img_x * z->depth + 7) / 8; // bytes per line, per component
+            raw_len = bpl * s->img_y * s->img_n /* pixels */ + s->img_y /* filter mode per row */;
+            z->expanded = (stbi_uc *) stbi_zlib_decode_malloc_guesssize_headerflag((char *) z->idata, ioff, raw_len, (int *) &raw_len, !is_iphone);
+            if (z->expanded == NULL) return 0; // zlib should set error
+            STBI_FREE(z->idata); z->idata = NULL;
+            if ((req_comp == s->img_n+1 && req_comp != 3 && !pal_img_n) || has_trans)
+               s->img_out_n = s->img_n+1;
+            else
+               s->img_out_n = s->img_n;
+            if (!stbi__create_png_image(z, z->expanded, raw_len, s->img_out_n, z->depth, color, interlace)) return 0;
+            if (has_trans) {
+               if (z->depth == 16) {
+                  if (!stbi__compute_transparency16(z, tc16, s->img_out_n)) return 0;
+               } else {
+                  if (!stbi__compute_transparency(z, tc, s->img_out_n)) return 0;
+               }
+            }
+            if (is_iphone && stbi__de_iphone_flag && s->img_out_n > 2)
+               stbi__de_iphone(z);
+            if (pal_img_n) {
+               // pal_img_n == 3 or 4
+               s->img_n = pal_img_n; // record the actual colors we had
+               s->img_out_n = pal_img_n;
+               if (req_comp >= 3) s->img_out_n = req_comp;
+               if (!stbi__expand_png_palette(z, palette, pal_len, s->img_out_n))
+                  return 0;
+            } else if (has_trans) {
+               // non-paletted image with tRNS -> source image has (constant) alpha
+               ++s->img_n;
+            }
+            STBI_FREE(z->expanded); z->expanded = NULL;
+            // end of PNG chunk, read and skip CRC
+            stbi__get32be(s);
+            return 1;
+         }
+
+         default:
+            // if critical, fail
+            if (first) return stbi__err("first not IHDR", "Corrupt PNG");
+            if ((c.type & (1 << 29)) == 0) {
+               #ifndef STBI_NO_FAILURE_STRINGS
+               // not threadsafe
+               static char invalid_chunk[] = "XXXX PNG chunk not known";
+               invalid_chunk[0] = STBI__BYTECAST(c.type >> 24);
+               invalid_chunk[1] = STBI__BYTECAST(c.type >> 16);
+               invalid_chunk[2] = STBI__BYTECAST(c.type >>  8);
+               invalid_chunk[3] = STBI__BYTECAST(c.type >>  0);
+               #endif
+               return stbi__err(invalid_chunk, "PNG not supported: unknown PNG chunk type");
+            }
+            stbi__skip(s, c.length);
+            break;
+      }
+      // end of PNG chunk, read and skip CRC
+      stbi__get32be(s);
+   }
+}
+
+static void *stbi__do_png(stbi__png *p, int *x, int *y, int *n, int req_comp, stbi__result_info *ri)
+{
+   void *result=NULL;
+   if (req_comp < 0 || req_comp > 4) return stbi__errpuc("bad req_comp", "Internal error");
+   if (stbi__parse_png_file(p, STBI__SCAN_load, req_comp)) {
+      if (p->depth <= 8)
+         ri->bits_per_channel = 8;
+      else if (p->depth == 16)
+         ri->bits_per_channel = 16;
+      else
+         return stbi__errpuc("bad bits_per_channel", "PNG not supported: unsupported color depth");
+      result = p->out;
+      p->out = NULL;
+      if (req_comp && req_comp != p->s->img_out_n) {
+         if (ri->bits_per_channel == 8)
+            result = stbi__convert_format((unsigned char *) result, p->s->img_out_n, req_comp, p->s->img_x, p->s->img_y);
+         else
+            result = stbi__convert_format16((stbi__uint16 *) result, p->s->img_out_n, req_comp, p->s->img_x, p->s->img_y);
+         p->s->img_out_n = req_comp;
+         if (result == NULL) return result;
+      }
+      *x = p->s->img_x;
+      *y = p->s->img_y;
+      if (n) *n = p->s->img_n;
+   }
+   STBI_FREE(p->out);      p->out      = NULL;
+   STBI_FREE(p->expanded); p->expanded = NULL;
+   STBI_FREE(p->idata);    p->idata    = NULL;
+
+   return result;
+}
+
+static void *stbi__png_load(stbi__context *s, int *x, int *y, int *comp, int req_comp, stbi__result_info *ri)
+{
+   stbi__png p;
+   p.s = s;
+   return stbi__do_png(&p, x,y,comp,req_comp, ri);
+}
+
+static int stbi__png_test(stbi__context *s)
+{
+   int r;
+   r = stbi__check_png_header(s);
+   stbi__rewind(s);
+   return r;
+}
+
+static int stbi__png_info_raw(stbi__png *p, int *x, int *y, int *comp)
+{
+   if (!stbi__parse_png_file(p, STBI__SCAN_header, 0)) {
+      stbi__rewind( p->s );
+      return 0;
+   }
+   if (x) *x = p->s->img_x;
+   if (y) *y = p->s->img_y;
+   if (comp) *comp = p->s->img_n;
+   return 1;
+}
+
+static int stbi__png_info(stbi__context *s, int *x, int *y, int *comp)
+{
+   stbi__png p;
+   p.s = s;
+   return stbi__png_info_raw(&p, x, y, comp);
+}
+
+static int stbi__png_is16(stbi__context *s)
+{
+   stbi__png p;
+   p.s = s;
+   if (!stbi__png_info_raw(&p, NULL, NULL, NULL))
+           return 0;
+   if (p.depth != 16) {
+      stbi__rewind(p.s);
+      return 0;
+   }
+   return 1;
+}
+#endif
+
+// Microsoft/Windows BMP image
+
+#ifndef STBI_NO_BMP
+static int stbi__bmp_test_raw(stbi__context *s)
+{
+   int r;
+   int sz;
+   if (stbi__get8(s) != 'B') return 0;
+   if (stbi__get8(s) != 'M') return 0;
+   stbi__get32le(s); // discard filesize
+   stbi__get16le(s); // discard reserved
+   stbi__get16le(s); // discard reserved
+   stbi__get32le(s); // discard data offset
+   sz = stbi__get32le(s);
+   r = (sz == 12 || sz == 40 || sz == 56 || sz == 108 || sz == 124);
+   return r;
+}
+
+static int stbi__bmp_test(stbi__context *s)
+{
+   int r = stbi__bmp_test_raw(s);
+   stbi__rewind(s);
+   return r;
+}
+
+
+// returns 0..31 for the highest set bit
+static int stbi__high_bit(unsigned int z)
+{
+   int n=0;
+   if (z == 0) return -1;
+   if (z >= 0x10000) { n += 16; z >>= 16; }
+   if (z >= 0x00100) { n +=  8; z >>=  8; }
+   if (z >= 0x00010) { n +=  4; z >>=  4; }
+   if (z >= 0x00004) { n +=  2; z >>=  2; }
+   if (z >= 0x00002) { n +=  1;/* >>=  1;*/ }
+   return n;
+}
+
+static int stbi__bitcount(unsigned int a)
+{
+   a = (a & 0x55555555) + ((a >>  1) & 0x55555555); // max 2
+   a = (a & 0x33333333) + ((a >>  2) & 0x33333333); // max 4
+   a = (a + (a >> 4)) & 0x0f0f0f0f; // max 8 per 4, now 8 bits
+   a = (a + (a >> 8)); // max 16 per 8 bits
+   a = (a + (a >> 16)); // max 32 per 8 bits
+   return a & 0xff;
+}
+
+// extract an arbitrarily-aligned N-bit value (N=bits)
+// from v, and then make it 8-bits long and fractionally
+// extend it to full full range.
+static int stbi__shiftsigned(unsigned int v, int shift, int bits)
+{
+   static unsigned int mul_table[9] = {
+      0,
+      0xff/*0b11111111*/, 0x55/*0b01010101*/, 0x49/*0b01001001*/, 0x11/*0b00010001*/,
+      0x21/*0b00100001*/, 0x41/*0b01000001*/, 0x81/*0b10000001*/, 0x01/*0b00000001*/,
+   };
+   static unsigned int shift_table[9] = {
+      0, 0,0,1,0,2,4,6,0,
+   };
+   if (shift < 0)
+      v <<= -shift;
+   else
+      v >>= shift;
+   STBI_ASSERT(v < 256);
+   v >>= (8-bits);
+   STBI_ASSERT(bits >= 0 && bits <= 8);
+   return (int) ((unsigned) v * mul_table[bits]) >> shift_table[bits];
+}
+
+typedef struct
+{
+   int bpp, offset, hsz;
+   unsigned int mr,mg,mb,ma, all_a;
+   int extra_read;
+} stbi__bmp_data;
+
+static int stbi__bmp_set_mask_defaults(stbi__bmp_data *info, int compress)
+{
+   // BI_BITFIELDS specifies masks explicitly, don't override
+   if (compress == 3)
+      return 1;
+
+   if (compress == 0) {
+      if (info->bpp == 16) {
+         info->mr = 31u << 10;
+         info->mg = 31u <<  5;
+         info->mb = 31u <<  0;
+      } else if (info->bpp == 32) {
+         info->mr = 0xffu << 16;
+         info->mg = 0xffu <<  8;
+         info->mb = 0xffu <<  0;
+         info->ma = 0xffu << 24;
+         info->all_a = 0; // if all_a is 0 at end, then we loaded alpha channel but it was all 0
+      } else {
+         // otherwise, use defaults, which is all-0
+         info->mr = info->mg = info->mb = info->ma = 0;
+      }
+      return 1;
+   }
+   return 0; // error
+}
+
+static void *stbi__bmp_parse_header(stbi__context *s, stbi__bmp_data *info)
+{
+   int hsz;
+   if (stbi__get8(s) != 'B' || stbi__get8(s) != 'M') return stbi__errpuc("not BMP", "Corrupt BMP");
+   stbi__get32le(s); // discard filesize
+   stbi__get16le(s); // discard reserved
+   stbi__get16le(s); // discard reserved
+   info->offset = stbi__get32le(s);
+   info->hsz = hsz = stbi__get32le(s);
+   info->mr = info->mg = info->mb = info->ma = 0;
+   info->extra_read = 14;
+
+   if (info->offset < 0) return stbi__errpuc("bad BMP", "bad BMP");
+
+   if (hsz != 12 && hsz != 40 && hsz != 56 && hsz != 108 && hsz != 124) return stbi__errpuc("unknown BMP", "BMP type not supported: unknown");
+   if (hsz == 12) {
+      s->img_x = stbi__get16le(s);
+      s->img_y = stbi__get16le(s);
+   } else {
+      s->img_x = stbi__get32le(s);
+      s->img_y = stbi__get32le(s);
+   }
+   if (stbi__get16le(s) != 1) return stbi__errpuc("bad BMP", "bad BMP");
+   info->bpp = stbi__get16le(s);
+   if (hsz != 12) {
+      int compress = stbi__get32le(s);
+      if (compress == 1 || compress == 2) return stbi__errpuc("BMP RLE", "BMP type not supported: RLE");
+      if (compress >= 4) return stbi__errpuc("BMP JPEG/PNG", "BMP type not supported: unsupported compression"); // this includes PNG/JPEG modes
+      if (compress == 3 && info->bpp != 16 && info->bpp != 32) return stbi__errpuc("bad BMP", "bad BMP"); // bitfields requires 16 or 32 bits/pixel
+      stbi__get32le(s); // discard sizeof
+      stbi__get32le(s); // discard hres
+      stbi__get32le(s); // discard vres
+      stbi__get32le(s); // discard colorsused
+      stbi__get32le(s); // discard max important
+      if (hsz == 40 || hsz == 56) {
+         if (hsz == 56) {
+            stbi__get32le(s);
+            stbi__get32le(s);
+            stbi__get32le(s);
+            stbi__get32le(s);
+         }
+         if (info->bpp == 16 || info->bpp == 32) {
+            if (compress == 0) {
+               stbi__bmp_set_mask_defaults(info, compress);
+            } else if (compress == 3) {
+               info->mr = stbi__get32le(s);
+               info->mg = stbi__get32le(s);
+               info->mb = stbi__get32le(s);
+               info->extra_read += 12;
+               // not documented, but generated by photoshop and handled by mspaint
+               if (info->mr == info->mg && info->mg == info->mb) {
+                  // ?!?!?
+                  return stbi__errpuc("bad BMP", "bad BMP");
+               }
+            } else
+               return stbi__errpuc("bad BMP", "bad BMP");
+         }
+      } else {
+         // V4/V5 header
+         int i;
+         if (hsz != 108 && hsz != 124)
+            return stbi__errpuc("bad BMP", "bad BMP");
+         info->mr = stbi__get32le(s);
+         info->mg = stbi__get32le(s);
+         info->mb = stbi__get32le(s);
+         info->ma = stbi__get32le(s);
+         if (compress != 3) // override mr/mg/mb unless in BI_BITFIELDS mode, as per docs
+            stbi__bmp_set_mask_defaults(info, compress);
+         stbi__get32le(s); // discard color space
+         for (i=0; i < 12; ++i)
+            stbi__get32le(s); // discard color space parameters
+         if (hsz == 124) {
+            stbi__get32le(s); // discard rendering intent
+            stbi__get32le(s); // discard offset of profile data
+            stbi__get32le(s); // discard size of profile data
+            stbi__get32le(s); // discard reserved
+         }
+      }
+   }
+   return (void *) 1;
+}
+
+
+static void *stbi__bmp_load(stbi__context *s, int *x, int *y, int *comp, int req_comp, stbi__result_info *ri)
+{
+   stbi_uc *out;
+   unsigned int mr=0,mg=0,mb=0,ma=0, all_a;
+   stbi_uc pal[256][4];
+   int psize=0,i,j,width;
+   int flip_vertically, pad, target;
+   stbi__bmp_data info;
+   STBI_NOTUSED(ri);
+
+   info.all_a = 255;
+   if (stbi__bmp_parse_header(s, &info) == NULL)
+      return NULL; // error code already set
+
+   flip_vertically = ((int) s->img_y) > 0;
+   s->img_y = abs((int) s->img_y);
+
+   if (s->img_y > STBI_MAX_DIMENSIONS) return stbi__errpuc("too large","Very large image (corrupt?)");
+   if (s->img_x > STBI_MAX_DIMENSIONS) return stbi__errpuc("too large","Very large image (corrupt?)");
+
+   mr = info.mr;
+   mg = info.mg;
+   mb = info.mb;
+   ma = info.ma;
+   all_a = info.all_a;
+
+   if (info.hsz == 12) {
+      if (info.bpp < 24)
+         psize = (info.offset - info.extra_read - 24) / 3;
+   } else {
+      if (info.bpp < 16)
+         psize = (info.offset - info.extra_read - info.hsz) >> 2;
+   }
+   if (psize == 0) {
+      // accept some number of extra bytes after the header, but if the offset points either to before
+      // the header ends or implies a large amount of extra data, reject the file as malformed
+      int bytes_read_so_far = s->callback_already_read + (int)(s->img_buffer - s->img_buffer_original);
+      int header_limit = 1024; // max we actually read is below 256 bytes currently.
+      int extra_data_limit = 256*4; // what ordinarily goes here is a palette; 256 entries*4 bytes is its max size.
+      if (bytes_read_so_far <= 0 || bytes_read_so_far > header_limit) {
+         return stbi__errpuc("bad header", "Corrupt BMP");
+      }
+      // we established that bytes_read_so_far is positive and sensible.
+      // the first half of this test rejects offsets that are either too small positives, or
+      // negative, and guarantees that info.offset >= bytes_read_so_far > 0. this in turn
+      // ensures the number computed in the second half of the test can't overflow.
+      if (info.offset < bytes_read_so_far || info.offset - bytes_read_so_far > extra_data_limit) {
+         return stbi__errpuc("bad offset", "Corrupt BMP");
+      } else {
+         stbi__skip(s, info.offset - bytes_read_so_far);
+      }
+   }
+
+   if (info.bpp == 24 && ma == 0xff000000)
+      s->img_n = 3;
+   else
+      s->img_n = ma ? 4 : 3;
+   if (req_comp && req_comp >= 3) // we can directly decode 3 or 4
+      target = req_comp;
+   else
+      target = s->img_n; // if they want monochrome, we'll post-convert
+
+   // sanity-check size
+   if (!stbi__mad3sizes_valid(target, s->img_x, s->img_y, 0))
+      return stbi__errpuc("too large", "Corrupt BMP");
+
+   out = (stbi_uc *) stbi__malloc_mad3(target, s->img_x, s->img_y, 0);
+   if (!out) return stbi__errpuc("outofmem", "Out of memory");
+   if (info.bpp < 16) {
+      int z=0;
+      if (psize == 0 || psize > 256) { STBI_FREE(out); return stbi__errpuc("invalid", "Corrupt BMP"); }
+      for (i=0; i < psize; ++i) {
+         pal[i][2] = stbi__get8(s);
+         pal[i][1] = stbi__get8(s);
+         pal[i][0] = stbi__get8(s);
+         if (info.hsz != 12) stbi__get8(s);
+         pal[i][3] = 255;
+      }
+      stbi__skip(s, info.offset - info.extra_read - info.hsz - psize * (info.hsz == 12 ? 3 : 4));
+      if (info.bpp == 1) width = (s->img_x + 7) >> 3;
+      else if (info.bpp == 4) width = (s->img_x + 1) >> 1;
+      else if (info.bpp == 8) width = s->img_x;
+      else { STBI_FREE(out); return stbi__errpuc("bad bpp", "Corrupt BMP"); }
+      pad = (-width)&3;
+      if (info.bpp == 1) {
+         for (j=0; j < (int) s->img_y; ++j) {
+            int bit_offset = 7, v = stbi__get8(s);
+            for (i=0; i < (int) s->img_x; ++i) {
+               int color = (v>>bit_offset)&0x1;
+               out[z++] = pal[color][0];
+               out[z++] = pal[color][1];
+               out[z++] = pal[color][2];
+               if (target == 4) out[z++] = 255;
+               if (i+1 == (int) s->img_x) break;
+               if((--bit_offset) < 0) {
+                  bit_offset = 7;
+                  v = stbi__get8(s);
+               }
+            }
+            stbi__skip(s, pad);
+         }
+      } else {
+         for (j=0; j < (int) s->img_y; ++j) {
+            for (i=0; i < (int) s->img_x; i += 2) {
+               int v=stbi__get8(s),v2=0;
+               if (info.bpp == 4) {
+                  v2 = v & 15;
+                  v >>= 4;
+               }
+               out[z++] = pal[v][0];
+               out[z++] = pal[v][1];
+               out[z++] = pal[v][2];
+               if (target == 4) out[z++] = 255;
+               if (i+1 == (int) s->img_x) break;
+               v = (info.bpp == 8) ? stbi__get8(s) : v2;
+               out[z++] = pal[v][0];
+               out[z++] = pal[v][1];
+               out[z++] = pal[v][2];
+               if (target == 4) out[z++] = 255;
+            }
+            stbi__skip(s, pad);
+         }
+      }
+   } else {
+      int rshift=0,gshift=0,bshift=0,ashift=0,rcount=0,gcount=0,bcount=0,acount=0;
+      int z = 0;
+      int easy=0;
+      stbi__skip(s, info.offset - info.extra_read - info.hsz);
+      if (info.bpp == 24) width = 3 * s->img_x;
+      else if (info.bpp == 16) width = 2*s->img_x;
+      else /* bpp = 32 and pad = 0 */ width=0;
+      pad = (-width) & 3;
+      if (info.bpp == 24) {
+         easy = 1;
+      } else if (info.bpp == 32) {
+         if (mb == 0xff && mg == 0xff00 && mr == 0x00ff0000 && ma == 0xff000000)
+            easy = 2;
+      }
+      if (!easy) {
+         if (!mr || !mg || !mb) { STBI_FREE(out); return stbi__errpuc("bad masks", "Corrupt BMP"); }
+         // right shift amt to put high bit in position #7
+         rshift = stbi__high_bit(mr)-7; rcount = stbi__bitcount(mr);
+         gshift = stbi__high_bit(mg)-7; gcount = stbi__bitcount(mg);
+         bshift = stbi__high_bit(mb)-7; bcount = stbi__bitcount(mb);
+         ashift = stbi__high_bit(ma)-7; acount = stbi__bitcount(ma);
+         if (rcount > 8 || gcount > 8 || bcount > 8 || acount > 8) { STBI_FREE(out); return stbi__errpuc("bad masks", "Corrupt BMP"); }
+      }
+      for (j=0; j < (int) s->img_y; ++j) {
+         if (easy) {
+            for (i=0; i < (int) s->img_x; ++i) {
+               unsigned char a;
+               out[z+2] = stbi__get8(s);
+               out[z+1] = stbi__get8(s);
+               out[z+0] = stbi__get8(s);
+               z += 3;
+               a = (easy == 2 ? stbi__get8(s) : 255);
+               all_a |= a;
+               if (target == 4) out[z++] = a;
+            }
+         } else {
+            int bpp = info.bpp;
+            for (i=0; i < (int) s->img_x; ++i) {
+               stbi__uint32 v = (bpp == 16 ? (stbi__uint32) stbi__get16le(s) : stbi__get32le(s));
+               unsigned int a;
+               out[z++] = STBI__BYTECAST(stbi__shiftsigned(v & mr, rshift, rcount));
+               out[z++] = STBI__BYTECAST(stbi__shiftsigned(v & mg, gshift, gcount));
+               out[z++] = STBI__BYTECAST(stbi__shiftsigned(v & mb, bshift, bcount));
+               a = (ma ? stbi__shiftsigned(v & ma, ashift, acount) : 255);
+               all_a |= a;
+               if (target == 4) out[z++] = STBI__BYTECAST(a);
+            }
+         }
+         stbi__skip(s, pad);
+      }
+   }
+
+   // if alpha channel is all 0s, replace with all 255s
+   if (target == 4 && all_a == 0)
+      for (i=4*s->img_x*s->img_y-1; i >= 0; i -= 4)
+         out[i] = 255;
+
+   if (flip_vertically) {
+      stbi_uc t;
+      for (j=0; j < (int) s->img_y>>1; ++j) {
+         stbi_uc *p1 = out +      j     *s->img_x*target;
+         stbi_uc *p2 = out + (s->img_y-1-j)*s->img_x*target;
+         for (i=0; i < (int) s->img_x*target; ++i) {
+            t = p1[i]; p1[i] = p2[i]; p2[i] = t;
+         }
+      }
+   }
+
+   if (req_comp && req_comp != target) {
+      out = stbi__convert_format(out, target, req_comp, s->img_x, s->img_y);
+      if (out == NULL) return out; // stbi__convert_format frees input on failure
+   }
+
+   *x = s->img_x;
+   *y = s->img_y;
+   if (comp) *comp = s->img_n;
+   return out;
+}
+#endif
+
+// Targa Truevision - TGA
+// by Jonathan Dummer
+#ifndef STBI_NO_TGA
+// returns STBI_rgb or whatever, 0 on error
+static int stbi__tga_get_comp(int bits_per_pixel, int is_grey, int* is_rgb16)
+{
+   // only RGB or RGBA (incl. 16bit) or grey allowed
+   if (is_rgb16) *is_rgb16 = 0;
+   switch(bits_per_pixel) {
+      case 8:  return STBI_grey;
+      case 16: if(is_grey) return STBI_grey_alpha;
+               // fallthrough
+      case 15: if(is_rgb16) *is_rgb16 = 1;
+               return STBI_rgb;
+      case 24: // fallthrough
+      case 32: return bits_per_pixel/8;
+      default: return 0;
+   }
+}
+
+static int stbi__tga_info(stbi__context *s, int *x, int *y, int *comp)
+{
+    int tga_w, tga_h, tga_comp, tga_image_type, tga_bits_per_pixel, tga_colormap_bpp;
+    int sz, tga_colormap_type;
+    stbi__get8(s);                   // discard Offset
+    tga_colormap_type = stbi__get8(s); // colormap type
+    if( tga_colormap_type > 1 ) {
+        stbi__rewind(s);
+        return 0;      // only RGB or indexed allowed
+    }
+    tga_image_type = stbi__get8(s); // image type
+    if ( tga_colormap_type == 1 ) { // colormapped (paletted) image
+        if (tga_image_type != 1 && tga_image_type != 9) {
+            stbi__rewind(s);
+            return 0;
+        }
+        stbi__skip(s,4);       // skip index of first colormap entry and number of entries
+        sz = stbi__get8(s);    //   check bits per palette color entry
+        if ( (sz != 8) && (sz != 15) && (sz != 16) && (sz != 24) && (sz != 32) ) {
+            stbi__rewind(s);
+            return 0;
+        }
+        stbi__skip(s,4);       // skip image x and y origin
+        tga_colormap_bpp = sz;
+    } else { // "normal" image w/o colormap - only RGB or grey allowed, +/- RLE
+        if ( (tga_image_type != 2) && (tga_image_type != 3) && (tga_image_type != 10) && (tga_image_type != 11) ) {
+            stbi__rewind(s);
+            return 0; // only RGB or grey allowed, +/- RLE
+        }
+        stbi__skip(s,9); // skip colormap specification and image x/y origin
+        tga_colormap_bpp = 0;
+    }
+    tga_w = stbi__get16le(s);
+    if( tga_w < 1 ) {
+        stbi__rewind(s);
+        return 0;   // test width
+    }
+    tga_h = stbi__get16le(s);
+    if( tga_h < 1 ) {
+        stbi__rewind(s);
+        return 0;   // test height
+    }
+    tga_bits_per_pixel = stbi__get8(s); // bits per pixel
+    stbi__get8(s); // ignore alpha bits
+    if (tga_colormap_bpp != 0) {
+        if((tga_bits_per_pixel != 8) && (tga_bits_per_pixel != 16)) {
+            // when using a colormap, tga_bits_per_pixel is the size of the indexes
+            // I don't think anything but 8 or 16bit indexes makes sense
+            stbi__rewind(s);
+            return 0;
+        }
+        tga_comp = stbi__tga_get_comp(tga_colormap_bpp, 0, NULL);
+    } else {
+        tga_comp = stbi__tga_get_comp(tga_bits_per_pixel, (tga_image_type == 3) || (tga_image_type == 11), NULL);
+    }
+    if(!tga_comp) {
+      stbi__rewind(s);
+      return 0;
+    }
+    if (x) *x = tga_w;
+    if (y) *y = tga_h;
+    if (comp) *comp = tga_comp;
+    return 1;                   // seems to have passed everything
+}
+
+static int stbi__tga_test(stbi__context *s)
+{
+   int res = 0;
+   int sz, tga_color_type;
+   stbi__get8(s);      //   discard Offset
+   tga_color_type = stbi__get8(s);   //   color type
+   if ( tga_color_type > 1 ) goto errorEnd;   //   only RGB or indexed allowed
+   sz = stbi__get8(s);   //   image type
+   if ( tga_color_type == 1 ) { // colormapped (paletted) image
+      if (sz != 1 && sz != 9) goto errorEnd; // colortype 1 demands image type 1 or 9
+      stbi__skip(s,4);       // skip index of first colormap entry and number of entries
+      sz = stbi__get8(s);    //   check bits per palette color entry
+      if ( (sz != 8) && (sz != 15) && (sz != 16) && (sz != 24) && (sz != 32) ) goto errorEnd;
+      stbi__skip(s,4);       // skip image x and y origin
+   } else { // "normal" image w/o colormap
+      if ( (sz != 2) && (sz != 3) && (sz != 10) && (sz != 11) ) goto errorEnd; // only RGB or grey allowed, +/- RLE
+      stbi__skip(s,9); // skip colormap specification and image x/y origin
+   }
+   if ( stbi__get16le(s) < 1 ) goto errorEnd;      //   test width
+   if ( stbi__get16le(s) < 1 ) goto errorEnd;      //   test height
+   sz = stbi__get8(s);   //   bits per pixel
+   if ( (tga_color_type == 1) && (sz != 8) && (sz != 16) ) goto errorEnd; // for colormapped images, bpp is size of an index
+   if ( (sz != 8) && (sz != 15) && (sz != 16) && (sz != 24) && (sz != 32) ) goto errorEnd;
+
+   res = 1; // if we got this far, everything's good and we can return 1 instead of 0
+
+errorEnd:
+   stbi__rewind(s);
+   return res;
+}
+
+// read 16bit value and convert to 24bit RGB
+static void stbi__tga_read_rgb16(stbi__context *s, stbi_uc* out)
+{
+   stbi__uint16 px = (stbi__uint16)stbi__get16le(s);
+   stbi__uint16 fiveBitMask = 31;
+   // we have 3 channels with 5bits each
+   int r = (px >> 10) & fiveBitMask;
+   int g = (px >> 5) & fiveBitMask;
+   int b = px & fiveBitMask;
+   // Note that this saves the data in RGB(A) order, so it doesn't need to be swapped later
+   out[0] = (stbi_uc)((r * 255)/31);
+   out[1] = (stbi_uc)((g * 255)/31);
+   out[2] = (stbi_uc)((b * 255)/31);
+
+   // some people claim that the most significant bit might be used for alpha
+   // (possibly if an alpha-bit is set in the "image descriptor byte")
+   // but that only made 16bit test images completely translucent..
+   // so let's treat all 15 and 16bit TGAs as RGB with no alpha.
+}
+
+static void *stbi__tga_load(stbi__context *s, int *x, int *y, int *comp, int req_comp, stbi__result_info *ri)
+{
+   //   read in the TGA header stuff
+   int tga_offset = stbi__get8(s);
+   int tga_indexed = stbi__get8(s);
+   int tga_image_type = stbi__get8(s);
+   int tga_is_RLE = 0;
+   int tga_palette_start = stbi__get16le(s);
+   int tga_palette_len = stbi__get16le(s);
+   int tga_palette_bits = stbi__get8(s);
+   int tga_x_origin = stbi__get16le(s);
+   int tga_y_origin = stbi__get16le(s);
+   int tga_width = stbi__get16le(s);
+   int tga_height = stbi__get16le(s);
+   int tga_bits_per_pixel = stbi__get8(s);
+   int tga_comp, tga_rgb16=0;
+   int tga_inverted = stbi__get8(s);
+   // int tga_alpha_bits = tga_inverted & 15; // the 4 lowest bits - unused (useless?)
+   //   image data
+   unsigned char *tga_data;
+   unsigned char *tga_palette = NULL;
+   int i, j;
+   unsigned char raw_data[4] = {0};
+   int RLE_count = 0;
+   int RLE_repeating = 0;
+   int read_next_pixel = 1;
+   STBI_NOTUSED(ri);
+   STBI_NOTUSED(tga_x_origin); // @TODO
+   STBI_NOTUSED(tga_y_origin); // @TODO
+
+   if (tga_height > STBI_MAX_DIMENSIONS) return stbi__errpuc("too large","Very large image (corrupt?)");
+   if (tga_width > STBI_MAX_DIMENSIONS) return stbi__errpuc("too large","Very large image (corrupt?)");
+
+   //   do a tiny bit of precessing
+   if ( tga_image_type >= 8 )
+   {
+      tga_image_type -= 8;
+      tga_is_RLE = 1;
+   }
+   tga_inverted = 1 - ((tga_inverted >> 5) & 1);
+
+   //   If I'm paletted, then I'll use the number of bits from the palette
+   if ( tga_indexed ) tga_comp = stbi__tga_get_comp(tga_palette_bits, 0, &tga_rgb16);
+   else tga_comp = stbi__tga_get_comp(tga_bits_per_pixel, (tga_image_type == 3), &tga_rgb16);
+
+   if(!tga_comp) // shouldn't really happen, stbi__tga_test() should have ensured basic consistency
+      return stbi__errpuc("bad format", "Can't find out TGA pixelformat");
+
+   //   tga info
+   *x = tga_width;
+   *y = tga_height;
+   if (comp) *comp = tga_comp;
+
+   if (!stbi__mad3sizes_valid(tga_width, tga_height, tga_comp, 0))
+      return stbi__errpuc("too large", "Corrupt TGA");
+
+   tga_data = (unsigned char*)stbi__malloc_mad3(tga_width, tga_height, tga_comp, 0);
+   if (!tga_data) return stbi__errpuc("outofmem", "Out of memory");
+
+   // skip to the data's starting position (offset usually = 0)
+   stbi__skip(s, tga_offset );
+
+   if ( !tga_indexed && !tga_is_RLE && !tga_rgb16 ) {
+      for (i=0; i < tga_height; ++i) {
+         int row = tga_inverted ? tga_height -i - 1 : i;
+         stbi_uc *tga_row = tga_data + row*tga_width*tga_comp;
+         stbi__getn(s, tga_row, tga_width * tga_comp);
+      }
+   } else  {
+      //   do I need to load a palette?
+      if ( tga_indexed)
+      {
+         if (tga_palette_len == 0) {  /* you have to have at least one entry! */
+            STBI_FREE(tga_data);
+            return stbi__errpuc("bad palette", "Corrupt TGA");
+         }
+
+         //   any data to skip? (offset usually = 0)
+         stbi__skip(s, tga_palette_start );
+         //   load the palette
+         tga_palette = (unsigned char*)stbi__malloc_mad2(tga_palette_len, tga_comp, 0);
+         if (!tga_palette) {
+            STBI_FREE(tga_data);
+            return stbi__errpuc("outofmem", "Out of memory");
+         }
+         if (tga_rgb16) {
+            stbi_uc *pal_entry = tga_palette;
+            STBI_ASSERT(tga_comp == STBI_rgb);
+            for (i=0; i < tga_palette_len; ++i) {
+               stbi__tga_read_rgb16(s, pal_entry);
+               pal_entry += tga_comp;
+            }
+         } else if (!stbi__getn(s, tga_palette, tga_palette_len * tga_comp)) {
+               STBI_FREE(tga_data);
+               STBI_FREE(tga_palette);
+               return stbi__errpuc("bad palette", "Corrupt TGA");
+         }
+      }
+      //   load the data
+      for (i=0; i < tga_width * tga_height; ++i)
+      {
+         //   if I'm in RLE mode, do I need to get a RLE stbi__pngchunk?
+         if ( tga_is_RLE )
+         {
+            if ( RLE_count == 0 )
+            {
+               //   yep, get the next byte as a RLE command
+               int RLE_cmd = stbi__get8(s);
+               RLE_count = 1 + (RLE_cmd & 127);
+               RLE_repeating = RLE_cmd >> 7;
+               read_next_pixel = 1;
+            } else if ( !RLE_repeating )
+            {
+               read_next_pixel = 1;
+            }
+         } else
+         {
+            read_next_pixel = 1;
+         }
+         //   OK, if I need to read a pixel, do it now
+         if ( read_next_pixel )
+         {
+            //   load however much data we did have
+            if ( tga_indexed )
+            {
+               // read in index, then perform the lookup
+               int pal_idx = (tga_bits_per_pixel == 8) ? stbi__get8(s) : stbi__get16le(s);
+               if ( pal_idx >= tga_palette_len ) {
+                  // invalid index
+                  pal_idx = 0;
+               }
+               pal_idx *= tga_comp;
+               for (j = 0; j < tga_comp; ++j) {
+                  raw_data[j] = tga_palette[pal_idx+j];
+               }
+            } else if(tga_rgb16) {
+               STBI_ASSERT(tga_comp == STBI_rgb);
+               stbi__tga_read_rgb16(s, raw_data);
+            } else {
+               //   read in the data raw
+               for (j = 0; j < tga_comp; ++j) {
+                  raw_data[j] = stbi__get8(s);
+               }
+            }
+            //   clear the reading flag for the next pixel
+            read_next_pixel = 0;
+         } // end of reading a pixel
+
+         // copy data
+         for (j = 0; j < tga_comp; ++j)
+           tga_data[i*tga_comp+j] = raw_data[j];
+
+         //   in case we're in RLE mode, keep counting down
+         --RLE_count;
+      }
+      //   do I need to invert the image?
+      if ( tga_inverted )
+      {
+         for (j = 0; j*2 < tga_height; ++j)
+         {
+            int index1 = j * tga_width * tga_comp;
+            int index2 = (tga_height - 1 - j) * tga_width * tga_comp;
+            for (i = tga_width * tga_comp; i > 0; --i)
+            {
+               unsigned char temp = tga_data[index1];
+               tga_data[index1] = tga_data[index2];
+               tga_data[index2] = temp;
+               ++index1;
+               ++index2;
+            }
+         }
+      }
+      //   clear my palette, if I had one
+      if ( tga_palette != NULL )
+      {
+         STBI_FREE( tga_palette );
+      }
+   }
+
+   // swap RGB - if the source data was RGB16, it already is in the right order
+   if (tga_comp >= 3 && !tga_rgb16)
+   {
+      unsigned char* tga_pixel = tga_data;
+      for (i=0; i < tga_width * tga_height; ++i)
+      {
+         unsigned char temp = tga_pixel[0];
+         tga_pixel[0] = tga_pixel[2];
+         tga_pixel[2] = temp;
+         tga_pixel += tga_comp;
+      }
+   }
+
+   // convert to target component count
+   if (req_comp && req_comp != tga_comp)
+      tga_data = stbi__convert_format(tga_data, tga_comp, req_comp, tga_width, tga_height);
+
+   //   the things I do to get rid of an error message, and yet keep
+   //   Microsoft's C compilers happy... [8^(
+   tga_palette_start = tga_palette_len = tga_palette_bits =
+         tga_x_origin = tga_y_origin = 0;
+   STBI_NOTUSED(tga_palette_start);
+   //   OK, done
+   return tga_data;
+}
+#endif
+
+// *************************************************************************************************
+// Photoshop PSD loader -- PD by Thatcher Ulrich, integration by Nicolas Schulz, tweaked by STB
+
+#ifndef STBI_NO_PSD
+static int stbi__psd_test(stbi__context *s)
+{
+   int r = (stbi__get32be(s) == 0x38425053);
+   stbi__rewind(s);
+   return r;
+}
+
+static int stbi__psd_decode_rle(stbi__context *s, stbi_uc *p, int pixelCount)
+{
+   int count, nleft, len;
+
+   count = 0;
+   while ((nleft = pixelCount - count) > 0) {
+      len = stbi__get8(s);
+      if (len == 128) {
+         // No-op.
+      } else if (len < 128) {
+         // Copy next len+1 bytes literally.
+         len++;
+         if (len > nleft) return 0; // corrupt data
+         count += len;
+         while (len) {
+            *p = stbi__get8(s);
+            p += 4;
+            len--;
+         }
+      } else if (len > 128) {
+         stbi_uc   val;
+         // Next -len+1 bytes in the dest are replicated from next source byte.
+         // (Interpret len as a negative 8-bit int.)
+         len = 257 - len;
+         if (len > nleft) return 0; // corrupt data
+         val = stbi__get8(s);
+         count += len;
+         while (len) {
+            *p = val;
+            p += 4;
+            len--;
+         }
+      }
+   }
+
+   return 1;
+}
+
+static void *stbi__psd_load(stbi__context *s, int *x, int *y, int *comp, int req_comp, stbi__result_info *ri, int bpc)
+{
+   int pixelCount;
+   int channelCount, compression;
+   int channel, i;
+   int bitdepth;
+   int w,h;
+   stbi_uc *out;
+   STBI_NOTUSED(ri);
+
+   // Check identifier
+   if (stbi__get32be(s) != 0x38425053)   // "8BPS"
+      return stbi__errpuc("not PSD", "Corrupt PSD image");
+
+   // Check file type version.
+   if (stbi__get16be(s) != 1)
+      return stbi__errpuc("wrong version", "Unsupported version of PSD image");
+
+   // Skip 6 reserved bytes.
+   stbi__skip(s, 6 );
+
+   // Read the number of channels (R, G, B, A, etc).
+   channelCount = stbi__get16be(s);
+   if (channelCount < 0 || channelCount > 16)
+      return stbi__errpuc("wrong channel count", "Unsupported number of channels in PSD image");
+
+   // Read the rows and columns of the image.
+   h = stbi__get32be(s);
+   w = stbi__get32be(s);
+
+   if (h > STBI_MAX_DIMENSIONS) return stbi__errpuc("too large","Very large image (corrupt?)");
+   if (w > STBI_MAX_DIMENSIONS) return stbi__errpuc("too large","Very large image (corrupt?)");
+
+   // Make sure the depth is 8 bits.
+   bitdepth = stbi__get16be(s);
+   if (bitdepth != 8 && bitdepth != 16)
+      return stbi__errpuc("unsupported bit depth", "PSD bit depth is not 8 or 16 bit");
+
+   // Make sure the color mode is RGB.
+   // Valid options are:
+   //   0: Bitmap
+   //   1: Grayscale
+   //   2: Indexed color
+   //   3: RGB color
+   //   4: CMYK color
+   //   7: Multichannel
+   //   8: Duotone
+   //   9: Lab color
+   if (stbi__get16be(s) != 3)
+      return stbi__errpuc("wrong color format", "PSD is not in RGB color format");
+
+   // Skip the Mode Data.  (It's the palette for indexed color; other info for other modes.)
+   stbi__skip(s,stbi__get32be(s) );
+
+   // Skip the image resources.  (resolution, pen tool paths, etc)
+   stbi__skip(s, stbi__get32be(s) );
+
+   // Skip the reserved data.
+   stbi__skip(s, stbi__get32be(s) );
+
+   // Find out if the data is compressed.
+   // Known values:
+   //   0: no compression
+   //   1: RLE compressed
+   compression = stbi__get16be(s);
+   if (compression > 1)
+      return stbi__errpuc("bad compression", "PSD has an unknown compression format");
+
+   // Check size
+   if (!stbi__mad3sizes_valid(4, w, h, 0))
+      return stbi__errpuc("too large", "Corrupt PSD");
+
+   // Create the destination image.
+
+   if (!compression && bitdepth == 16 && bpc == 16) {
+      out = (stbi_uc *) stbi__malloc_mad3(8, w, h, 0);
+      ri->bits_per_channel = 16;
+   } else
+      out = (stbi_uc *) stbi__malloc(4 * w*h);
+
+   if (!out) return stbi__errpuc("outofmem", "Out of memory");
+   pixelCount = w*h;
+
+   // Initialize the data to zero.
+   //memset( out, 0, pixelCount * 4 );
+
+   // Finally, the image data.
+   if (compression) {
+      // RLE as used by .PSD and .TIFF
+      // Loop until you get the number of unpacked bytes you are expecting:
+      //     Read the next source byte into n.
+      //     If n is between 0 and 127 inclusive, copy the next n+1 bytes literally.
+      //     Else if n is between -127 and -1 inclusive, copy the next byte -n+1 times.
+      //     Else if n is 128, noop.
+      // Endloop
+
+      // The RLE-compressed data is preceded by a 2-byte data count for each row in the data,
+      // which we're going to just skip.
+      stbi__skip(s, h * channelCount * 2 );
+
+      // Read the RLE data by channel.
+      for (channel = 0; channel < 4; channel++) {
+         stbi_uc *p;
+
+         p = out+channel;
+         if (channel >= channelCount) {
+            // Fill this channel with default data.
+            for (i = 0; i < pixelCount; i++, p += 4)
+               *p = (channel == 3 ? 255 : 0);
+         } else {
+            // Read the RLE data.
+            if (!stbi__psd_decode_rle(s, p, pixelCount)) {
+               STBI_FREE(out);
+               return stbi__errpuc("corrupt", "bad RLE data");
+            }
+         }
+      }
+
+   } else {
+      // We're at the raw image data.  It's each channel in order (Red, Green, Blue, Alpha, ...)
+      // where each channel consists of an 8-bit (or 16-bit) value for each pixel in the image.
+
+      // Read the data by channel.
+      for (channel = 0; channel < 4; channel++) {
+         if (channel >= channelCount) {
+            // Fill this channel with default data.
+            if (bitdepth == 16 && bpc == 16) {
+               stbi__uint16 *q = ((stbi__uint16 *) out) + channel;
+               stbi__uint16 val = channel == 3 ? 65535 : 0;
+               for (i = 0; i < pixelCount; i++, q += 4)
+                  *q = val;
+            } else {
+               stbi_uc *p = out+channel;
+               stbi_uc val = channel == 3 ? 255 : 0;
+               for (i = 0; i < pixelCount; i++, p += 4)
+                  *p = val;
+            }
+         } else {
+            if (ri->bits_per_channel == 16) {    // output bpc
+               stbi__uint16 *q = ((stbi__uint16 *) out) + channel;
+               for (i = 0; i < pixelCount; i++, q += 4)
+                  *q = (stbi__uint16) stbi__get16be(s);
+            } else {
+               stbi_uc *p = out+channel;
+               if (bitdepth == 16) {  // input bpc
+                  for (i = 0; i < pixelCount; i++, p += 4)
+                     *p = (stbi_uc) (stbi__get16be(s) >> 8);
+               } else {
+                  for (i = 0; i < pixelCount; i++, p += 4)
+                     *p = stbi__get8(s);
+               }
+            }
+         }
+      }
+   }
+
+   // remove weird white matte from PSD
+   if (channelCount >= 4) {
+      if (ri->bits_per_channel == 16) {
+         for (i=0; i < w*h; ++i) {
+            stbi__uint16 *pixel = (stbi__uint16 *) out + 4*i;
+            if (pixel[3] != 0 && pixel[3] != 65535) {
+               float a = pixel[3] / 65535.0f;
+               float ra = 1.0f / a;
+               float inv_a = 65535.0f * (1 - ra);
+               pixel[0] = (stbi__uint16) (pixel[0]*ra + inv_a);
+               pixel[1] = (stbi__uint16) (pixel[1]*ra + inv_a);
+               pixel[2] = (stbi__uint16) (pixel[2]*ra + inv_a);
+            }
+         }
+      } else {
+         for (i=0; i < w*h; ++i) {
+            unsigned char *pixel = out + 4*i;
+            if (pixel[3] != 0 && pixel[3] != 255) {
+               float a = pixel[3] / 255.0f;
+               float ra = 1.0f / a;
+               float inv_a = 255.0f * (1 - ra);
+               pixel[0] = (unsigned char) (pixel[0]*ra + inv_a);
+               pixel[1] = (unsigned char) (pixel[1]*ra + inv_a);
+               pixel[2] = (unsigned char) (pixel[2]*ra + inv_a);
+            }
+         }
+      }
+   }
+
+   // convert to desired output format
+   if (req_comp && req_comp != 4) {
+      if (ri->bits_per_channel == 16)
+         out = (stbi_uc *) stbi__convert_format16((stbi__uint16 *) out, 4, req_comp, w, h);
+      else
+         out = stbi__convert_format(out, 4, req_comp, w, h);
+      if (out == NULL) return out; // stbi__convert_format frees input on failure
+   }
+
+   if (comp) *comp = 4;
+   *y = h;
+   *x = w;
+
+   return out;
+}
+#endif
+
+// *************************************************************************************************
+// Softimage PIC loader
+// by Tom Seddon
+//
+// See http://softimage.wiki.softimage.com/index.php/INFO:_PIC_file_format
+// See http://ozviz.wasp.uwa.edu.au/~pbourke/dataformats/softimagepic/
+
+#ifndef STBI_NO_PIC
+static int stbi__pic_is4(stbi__context *s,const char *str)
+{
+   int i;
+   for (i=0; i<4; ++i)
+      if (stbi__get8(s) != (stbi_uc)str[i])
+         return 0;
+
+   return 1;
+}
+
+static int stbi__pic_test_core(stbi__context *s)
+{
+   int i;
+
+   if (!stbi__pic_is4(s,"\x53\x80\xF6\x34"))
+      return 0;
+
+   for(i=0;i<84;++i)
+      stbi__get8(s);
+
+   if (!stbi__pic_is4(s,"PICT"))
+      return 0;
+
+   return 1;
+}
+
+typedef struct
+{
+   stbi_uc size,type,channel;
+} stbi__pic_packet;
+
+static stbi_uc *stbi__readval(stbi__context *s, int channel, stbi_uc *dest)
+{
+   int mask=0x80, i;
+
+   for (i=0; i<4; ++i, mask>>=1) {
+      if (channel & mask) {
+         if (stbi__at_eof(s)) return stbi__errpuc("bad file","PIC file too short");
+         dest[i]=stbi__get8(s);
+      }
+   }
+
+   return dest;
+}
+
+static void stbi__copyval(int channel,stbi_uc *dest,const stbi_uc *src)
+{
+   int mask=0x80,i;
+
+   for (i=0;i<4; ++i, mask>>=1)
+      if (channel&mask)
+         dest[i]=src[i];
+}
+
+static stbi_uc *stbi__pic_load_core(stbi__context *s,int width,int height,int *comp, stbi_uc *result)
+{
+   int act_comp=0,num_packets=0,y,chained;
+   stbi__pic_packet packets[10];
+
+   // this will (should...) cater for even some bizarre stuff like having data
+    // for the same channel in multiple packets.
+   do {
+      stbi__pic_packet *packet;
+
+      if (num_packets==sizeof(packets)/sizeof(packets[0]))
+         return stbi__errpuc("bad format","too many packets");
+
+      packet = &packets[num_packets++];
+
+      chained = stbi__get8(s);
+      packet->size    = stbi__get8(s);
+      packet->type    = stbi__get8(s);
+      packet->channel = stbi__get8(s);
+
+      act_comp |= packet->channel;
+
+      if (stbi__at_eof(s))          return stbi__errpuc("bad file","file too short (reading packets)");
+      if (packet->size != 8)  return stbi__errpuc("bad format","packet isn't 8bpp");
+   } while (chained);
+
+   *comp = (act_comp & 0x10 ? 4 : 3); // has alpha channel?
+
+   for(y=0; y<height; ++y) {
+      int packet_idx;
+
+      for(packet_idx=0; packet_idx < num_packets; ++packet_idx) {
+         stbi__pic_packet *packet = &packets[packet_idx];
+         stbi_uc *dest = result+y*width*4;
+
+         switch (packet->type) {
+            default:
+               return stbi__errpuc("bad format","packet has bad compression type");
+
+            case 0: {//uncompressed
+               int x;
+
+               for(x=0;x<width;++x, dest+=4)
+                  if (!stbi__readval(s,packet->channel,dest))
+                     return 0;
+               break;
+            }
+
+            case 1://Pure RLE
+               {
+                  int left=width, i;
+
+                  while (left>0) {
+                     stbi_uc count,value[4];
+
+                     count=stbi__get8(s);
+                     if (stbi__at_eof(s))   return stbi__errpuc("bad file","file too short (pure read count)");
+
+                     if (count > left)
+                        count = (stbi_uc) left;
+
+                     if (!stbi__readval(s,packet->channel,value))  return 0;
+
+                     for(i=0; i<count; ++i,dest+=4)
+                        stbi__copyval(packet->channel,dest,value);
+                     left -= count;
+                  }
+               }
+               break;
+
+            case 2: {//Mixed RLE
+               int left=width;
+               while (left>0) {
+                  int count = stbi__get8(s), i;
+                  if (stbi__at_eof(s))  return stbi__errpuc("bad file","file too short (mixed read count)");
+
+                  if (count >= 128) { // Repeated
+                     stbi_uc value[4];
+
+                     if (count==128)
+                        count = stbi__get16be(s);
+                     else
+                        count -= 127;
+                     if (count > left)
+                        return stbi__errpuc("bad file","scanline overrun");
+
+                     if (!stbi__readval(s,packet->channel,value))
+                        return 0;
+
+                     for(i=0;i<count;++i, dest += 4)
+                        stbi__copyval(packet->channel,dest,value);
+                  } else { // Raw
+                     ++count;
+                     if (count>left) return stbi__errpuc("bad file","scanline overrun");
+
+                     for(i=0;i<count;++i, dest+=4)
+                        if (!stbi__readval(s,packet->channel,dest))
+                           return 0;
+                  }
+                  left-=count;
+               }
+               break;
+            }
+         }
+      }
+   }
+
+   return result;
+}
+
+static void *stbi__pic_load(stbi__context *s,int *px,int *py,int *comp,int req_comp, stbi__result_info *ri)
+{
+   stbi_uc *result;
+   int i, x,y, internal_comp;
+   STBI_NOTUSED(ri);
+
+   if (!comp) comp = &internal_comp;
+
+   for (i=0; i<92; ++i)
+      stbi__get8(s);
+
+   x = stbi__get16be(s);
+   y = stbi__get16be(s);
+
+   if (y > STBI_MAX_DIMENSIONS) return stbi__errpuc("too large","Very large image (corrupt?)");
+   if (x > STBI_MAX_DIMENSIONS) return stbi__errpuc("too large","Very large image (corrupt?)");
+
+   if (stbi__at_eof(s))  return stbi__errpuc("bad file","file too short (pic header)");
+   if (!stbi__mad3sizes_valid(x, y, 4, 0)) return stbi__errpuc("too large", "PIC image too large to decode");
+
+   stbi__get32be(s); //skip `ratio'
+   stbi__get16be(s); //skip `fields'
+   stbi__get16be(s); //skip `pad'
+
+   // intermediate buffer is RGBA
+   result = (stbi_uc *) stbi__malloc_mad3(x, y, 4, 0);
+   if (!result) return stbi__errpuc("outofmem", "Out of memory");
+   memset(result, 0xff, x*y*4);
+
+   if (!stbi__pic_load_core(s,x,y,comp, result)) {
+      STBI_FREE(result);
+      result=0;
+   }
+   *px = x;
+   *py = y;
+   if (req_comp == 0) req_comp = *comp;
+   result=stbi__convert_format(result,4,req_comp,x,y);
+
+   return result;
+}
+
+static int stbi__pic_test(stbi__context *s)
+{
+   int r = stbi__pic_test_core(s);
+   stbi__rewind(s);
+   return r;
+}
+#endif
+
+// *************************************************************************************************
+// GIF loader -- public domain by Jean-Marc Lienher -- simplified/shrunk by stb
+
+#ifndef STBI_NO_GIF
+typedef struct
+{
+   stbi__int16 prefix;
+   stbi_uc first;
+   stbi_uc suffix;
+} stbi__gif_lzw;
+
+typedef struct
+{
+   int w,h;
+   stbi_uc *out;                 // output buffer (always 4 components)
+   stbi_uc *background;          // The current "background" as far as a gif is concerned
+   stbi_uc *history;
+   int flags, bgindex, ratio, transparent, eflags;
+   stbi_uc  pal[256][4];
+   stbi_uc lpal[256][4];
+   stbi__gif_lzw codes[8192];
+   stbi_uc *color_table;
+   int parse, step;
+   int lflags;
+   int start_x, start_y;
+   int max_x, max_y;
+   int cur_x, cur_y;
+   int line_size;
+   int delay;
+} stbi__gif;
+
+static int stbi__gif_test_raw(stbi__context *s)
+{
+   int sz;
+   if (stbi__get8(s) != 'G' || stbi__get8(s) != 'I' || stbi__get8(s) != 'F' || stbi__get8(s) != '8') return 0;
+   sz = stbi__get8(s);
+   if (sz != '9' && sz != '7') return 0;
+   if (stbi__get8(s) != 'a') return 0;
+   return 1;
+}
+
+static int stbi__gif_test(stbi__context *s)
+{
+   int r = stbi__gif_test_raw(s);
+   stbi__rewind(s);
+   return r;
+}
+
+static void stbi__gif_parse_colortable(stbi__context *s, stbi_uc pal[256][4], int num_entries, int transp)
+{
+   int i;
+   for (i=0; i < num_entries; ++i) {
+      pal[i][2] = stbi__get8(s);
+      pal[i][1] = stbi__get8(s);
+      pal[i][0] = stbi__get8(s);
+      pal[i][3] = transp == i ? 0 : 255;
+   }
+}
+
+static int stbi__gif_header(stbi__context *s, stbi__gif *g, int *comp, int is_info)
+{
+   stbi_uc version;
+   if (stbi__get8(s) != 'G' || stbi__get8(s) != 'I' || stbi__get8(s) != 'F' || stbi__get8(s) != '8')
+      return stbi__err("not GIF", "Corrupt GIF");
+
+   version = stbi__get8(s);
+   if (version != '7' && version != '9')    return stbi__err("not GIF", "Corrupt GIF");
+   if (stbi__get8(s) != 'a')                return stbi__err("not GIF", "Corrupt GIF");
+
+   stbi__g_failure_reason = "";
+   g->w = stbi__get16le(s);
+   g->h = stbi__get16le(s);
+   g->flags = stbi__get8(s);
+   g->bgindex = stbi__get8(s);
+   g->ratio = stbi__get8(s);
+   g->transparent = -1;
+
+   if (g->w > STBI_MAX_DIMENSIONS) return stbi__err("too large","Very large image (corrupt?)");
+   if (g->h > STBI_MAX_DIMENSIONS) return stbi__err("too large","Very large image (corrupt?)");
+
+   if (comp != 0) *comp = 4;  // can't actually tell whether it's 3 or 4 until we parse the comments
+
+   if (is_info) return 1;
+
+   if (g->flags & 0x80)
+      stbi__gif_parse_colortable(s,g->pal, 2 << (g->flags & 7), -1);
+
+   return 1;
+}
+
+static int stbi__gif_info_raw(stbi__context *s, int *x, int *y, int *comp)
+{
+   stbi__gif* g = (stbi__gif*) stbi__malloc(sizeof(stbi__gif));
+   if (!g) return stbi__err("outofmem", "Out of memory");
+   if (!stbi__gif_header(s, g, comp, 1)) {
+      STBI_FREE(g);
+      stbi__rewind( s );
+      return 0;
+   }
+   if (x) *x = g->w;
+   if (y) *y = g->h;
+   STBI_FREE(g);
+   return 1;
+}
+
+static void stbi__out_gif_code(stbi__gif *g, stbi__uint16 code)
+{
+   stbi_uc *p, *c;
+   int idx;
+
+   // recurse to decode the prefixes, since the linked-list is backwards,
+   // and working backwards through an interleaved image would be nasty
+   if (g->codes[code].prefix >= 0)
+      stbi__out_gif_code(g, g->codes[code].prefix);
+
+   if (g->cur_y >= g->max_y) return;
+
+   idx = g->cur_x + g->cur_y;
+   p = &g->out[idx];
+   g->history[idx / 4] = 1;
+
+   c = &g->color_table[g->codes[code].suffix * 4];
+   if (c[3] > 128) { // don't render transparent pixels;
+      p[0] = c[2];
+      p[1] = c[1];
+      p[2] = c[0];
+      p[3] = c[3];
+   }
+   g->cur_x += 4;
+
+   if (g->cur_x >= g->max_x) {
+      g->cur_x = g->start_x;
+      g->cur_y += g->step;
+
+      while (g->cur_y >= g->max_y && g->parse > 0) {
+         g->step = (1 << g->parse) * g->line_size;
+         g->cur_y = g->start_y + (g->step >> 1);
+         --g->parse;
+      }
+   }
+}
+
+static stbi_uc *stbi__process_gif_raster(stbi__context *s, stbi__gif *g)
+{
+   stbi_uc lzw_cs;
+   stbi__int32 len, init_code;
+   stbi__uint32 first;
+   stbi__int32 codesize, codemask, avail, oldcode, bits, valid_bits, clear;
+   stbi__gif_lzw *p;
+
+   lzw_cs = stbi__get8(s);
+   if (lzw_cs > 12) return NULL;
+   clear = 1 << lzw_cs;
+   first = 1;
+   codesize = lzw_cs + 1;
+   codemask = (1 << codesize) - 1;
+   bits = 0;
+   valid_bits = 0;
+   for (init_code = 0; init_code < clear; init_code++) {
+      g->codes[init_code].prefix = -1;
+      g->codes[init_code].first = (stbi_uc) init_code;
+      g->codes[init_code].suffix = (stbi_uc) init_code;
+   }
+
+   // support no starting clear code
+   avail = clear+2;
+   oldcode = -1;
+
+   len = 0;
+   for(;;) {
+      if (valid_bits < codesize) {
+         if (len == 0) {
+            len = stbi__get8(s); // start new block
+            if (len == 0)
+               return g->out;
+         }
+         --len;
+         bits |= (stbi__int32) stbi__get8(s) << valid_bits;
+         valid_bits += 8;
+      } else {
+         stbi__int32 code = bits & codemask;
+         bits >>= codesize;
+         valid_bits -= codesize;
+         // @OPTIMIZE: is there some way we can accelerate the non-clear path?
+         if (code == clear) {  // clear code
+            codesize = lzw_cs + 1;
+            codemask = (1 << codesize) - 1;
+            avail = clear + 2;
+            oldcode = -1;
+            first = 0;
+         } else if (code == clear + 1) { // end of stream code
+            stbi__skip(s, len);
+            while ((len = stbi__get8(s)) > 0)
+               stbi__skip(s,len);
+            return g->out;
+         } else if (code <= avail) {
+            if (first) {
+               return stbi__errpuc("no clear code", "Corrupt GIF");
+            }
+
+            if (oldcode >= 0) {
+               p = &g->codes[avail++];
+               if (avail > 8192) {
+                  return stbi__errpuc("too many codes", "Corrupt GIF");
+               }
+
+               p->prefix = (stbi__int16) oldcode;
+               p->first = g->codes[oldcode].first;
+               p->suffix = (code == avail) ? p->first : g->codes[code].first;
+            } else if (code == avail)
+               return stbi__errpuc("illegal code in raster", "Corrupt GIF");
+
+            stbi__out_gif_code(g, (stbi__uint16) code);
+
+            if ((avail & codemask) == 0 && avail <= 0x0FFF) {
+               codesize++;
+               codemask = (1 << codesize) - 1;
+            }
+
+            oldcode = code;
+         } else {
+            return stbi__errpuc("illegal code in raster", "Corrupt GIF");
+         }
+      }
+   }
+}
+
+// this function is designed to support animated gifs, although stb_image doesn't support it
+// two back is the image from two frames ago, used for a very specific disposal format
+static stbi_uc *stbi__gif_load_next(stbi__context *s, stbi__gif *g, int *comp, int req_comp, stbi_uc *two_back)
+{
+   int dispose;
+   int first_frame;
+   int pi;
+   int pcount;
+   STBI_NOTUSED(req_comp);
+
+   // on first frame, any non-written pixels get the background colour (non-transparent)
+   first_frame = 0;
+   if (g->out == 0) {
+      if (!stbi__gif_header(s, g, comp,0)) return 0; // stbi__g_failure_reason set by stbi__gif_header
+      if (!stbi__mad3sizes_valid(4, g->w, g->h, 0))
+         return stbi__errpuc("too large", "GIF image is too large");
+      pcount = g->w * g->h;
+      g->out = (stbi_uc *) stbi__malloc(4 * pcount);
+      g->background = (stbi_uc *) stbi__malloc(4 * pcount);
+      g->history = (stbi_uc *) stbi__malloc(pcount);
+      if (!g->out || !g->background || !g->history)
+         return stbi__errpuc("outofmem", "Out of memory");
+
+      // image is treated as "transparent" at the start - ie, nothing overwrites the current background;
+      // background colour is only used for pixels that are not rendered first frame, after that "background"
+      // color refers to the color that was there the previous frame.
+      memset(g->out, 0x00, 4 * pcount);
+      memset(g->background, 0x00, 4 * pcount); // state of the background (starts transparent)
+      memset(g->history, 0x00, pcount);        // pixels that were affected previous frame
+      first_frame = 1;
+   } else {
+      // second frame - how do we dispose of the previous one?
+      dispose = (g->eflags & 0x1C) >> 2;
+      pcount = g->w * g->h;
+
+      if ((dispose == 3) && (two_back == 0)) {
+         dispose = 2; // if I don't have an image to revert back to, default to the old background
+      }
+
+      if (dispose == 3) { // use previous graphic
+         for (pi = 0; pi < pcount; ++pi) {
+            if (g->history[pi]) {
+               memcpy( &g->out[pi * 4], &two_back[pi * 4], 4 );
+            }
+         }
+      } else if (dispose == 2) {
+         // restore what was changed last frame to background before that frame;
+         for (pi = 0; pi < pcount; ++pi) {
+            if (g->history[pi]) {
+               memcpy( &g->out[pi * 4], &g->background[pi * 4], 4 );
+            }
+         }
+      } else {
+         // This is a non-disposal case eithe way, so just
+         // leave the pixels as is, and they will become the new background
+         // 1: do not dispose
+         // 0:  not specified.
+      }
+
+      // background is what out is after the undoing of the previou frame;
+      memcpy( g->background, g->out, 4 * g->w * g->h );
+   }
+
+   // clear my history;
+   memset( g->history, 0x00, g->w * g->h );        // pixels that were affected previous frame
+
+   for (;;) {
+      int tag = stbi__get8(s);
+      switch (tag) {
+         case 0x2C: /* Image Descriptor */
+         {
+            stbi__int32 x, y, w, h;
+            stbi_uc *o;
+
+            x = stbi__get16le(s);
+            y = stbi__get16le(s);
+            w = stbi__get16le(s);
+            h = stbi__get16le(s);
+            if (((x + w) > (g->w)) || ((y + h) > (g->h)))
+               return stbi__errpuc("bad Image Descriptor", "Corrupt GIF");
+
+            g->line_size = g->w * 4;
+            g->start_x = x * 4;
+            g->start_y = y * g->line_size;
+            g->max_x   = g->start_x + w * 4;
+            g->max_y   = g->start_y + h * g->line_size;
+            g->cur_x   = g->start_x;
+            g->cur_y   = g->start_y;
+
+            // if the width of the specified rectangle is 0, that means
+            // we may not see *any* pixels or the image is malformed;
+            // to make sure this is caught, move the current y down to
+            // max_y (which is what out_gif_code checks).
+            if (w == 0)
+               g->cur_y = g->max_y;
+
+            g->lflags = stbi__get8(s);
+
+            if (g->lflags & 0x40) {
+               g->step = 8 * g->line_size; // first interlaced spacing
+               g->parse = 3;
+            } else {
+               g->step = g->line_size;
+               g->parse = 0;
+            }
+
+            if (g->lflags & 0x80) {
+               stbi__gif_parse_colortable(s,g->lpal, 2 << (g->lflags & 7), g->eflags & 0x01 ? g->transparent : -1);
+               g->color_table = (stbi_uc *) g->lpal;
+            } else if (g->flags & 0x80) {
+               g->color_table = (stbi_uc *) g->pal;
+            } else
+               return stbi__errpuc("missing color table", "Corrupt GIF");
+
+            o = stbi__process_gif_raster(s, g);
+            if (!o) return NULL;
+
+            // if this was the first frame,
+            pcount = g->w * g->h;
+            if (first_frame && (g->bgindex > 0)) {
+               // if first frame, any pixel not drawn to gets the background color
+               for (pi = 0; pi < pcount; ++pi) {
+                  if (g->history[pi] == 0) {
+                     g->pal[g->bgindex][3] = 255; // just in case it was made transparent, undo that; It will be reset next frame if need be;
+                     memcpy( &g->out[pi * 4], &g->pal[g->bgindex], 4 );
+                  }
+               }
+            }
+
+            return o;
+         }
+
+         case 0x21: // Comment Extension.
+         {
+            int len;
+            int ext = stbi__get8(s);
+            if (ext == 0xF9) { // Graphic Control Extension.
+               len = stbi__get8(s);
+               if (len == 4) {
+                  g->eflags = stbi__get8(s);
+                  g->delay = 10 * stbi__get16le(s); // delay - 1/100th of a second, saving as 1/1000ths.
+
+                  // unset old transparent
+                  if (g->transparent >= 0) {
+                     g->pal[g->transparent][3] = 255;
+                  }
+                  if (g->eflags & 0x01) {
+                     g->transparent = stbi__get8(s);
+                     if (g->transparent >= 0) {
+                        g->pal[g->transparent][3] = 0;
+                     }
+                  } else {
+                     // don't need transparent
+                     stbi__skip(s, 1);
+                     g->transparent = -1;
+                  }
+               } else {
+                  stbi__skip(s, len);
+                  break;
+               }
+            }
+            while ((len = stbi__get8(s)) != 0) {
+               stbi__skip(s, len);
+            }
+            break;
+         }
+
+         case 0x3B: // gif stream termination code
+            return (stbi_uc *) s; // using '1' causes warning on some compilers
+
+         default:
+            return stbi__errpuc("unknown code", "Corrupt GIF");
+      }
+   }
+}
+
+static void *stbi__load_gif_main_outofmem(stbi__gif *g, stbi_uc *out, int **delays)
+{
+   STBI_FREE(g->out);
+   STBI_FREE(g->history);
+   STBI_FREE(g->background);
+
+   if (out) STBI_FREE(out);
+   if (delays && *delays) STBI_FREE(*delays);
+   return stbi__errpuc("outofmem", "Out of memory");
+}
+
+static void *stbi__load_gif_main(stbi__context *s, int **delays, int *x, int *y, int *z, int *comp, int req_comp)
+{
+   if (stbi__gif_test(s)) {
+      int layers = 0;
+      stbi_uc *u = 0;
+      stbi_uc *out = 0;
+      stbi_uc *two_back = 0;
+      stbi__gif g;
+      int stride;
+      int out_size = 0;
+      int delays_size = 0;
+
+      STBI_NOTUSED(out_size);
+      STBI_NOTUSED(delays_size);
+
+      memset(&g, 0, sizeof(g));
+      if (delays) {
+         *delays = 0;
+      }
+
+      do {
+         u = stbi__gif_load_next(s, &g, comp, req_comp, two_back);
+         if (u == (stbi_uc *) s) u = 0;  // end of animated gif marker
+
+         if (u) {
+            *x = g.w;
+            *y = g.h;
+            ++layers;
+            stride = g.w * g.h * 4;
+
+            if (out) {
+               void *tmp = (stbi_uc*) STBI_REALLOC_SIZED( out, out_size, layers * stride );
+               if (!tmp)
+                  return stbi__load_gif_main_outofmem(&g, out, delays);
+               else {
+                   out = (stbi_uc*) tmp;
+                   out_size = layers * stride;
+               }
+
+               if (delays) {
+                  int *new_delays = (int*) STBI_REALLOC_SIZED( *delays, delays_size, sizeof(int) * layers );
+                  if (!new_delays)
+                     return stbi__load_gif_main_outofmem(&g, out, delays);
+                  *delays = new_delays;
+                  delays_size = layers * sizeof(int);
+               }
+            } else {
+               out = (stbi_uc*)stbi__malloc( layers * stride );
+               if (!out)
+                  return stbi__load_gif_main_outofmem(&g, out, delays);
+               out_size = layers * stride;
+               if (delays) {
+                  *delays = (int*) stbi__malloc( layers * sizeof(int) );
+                  if (!*delays)
+                     return stbi__load_gif_main_outofmem(&g, out, delays);
+                  delays_size = layers * sizeof(int);
+               }
+            }
+            memcpy( out + ((layers - 1) * stride), u, stride );
+            if (layers >= 2) {
+               two_back = out - 2 * stride;
+            }
+
+            if (delays) {
+               (*delays)[layers - 1U] = g.delay;
+            }
+         }
+      } while (u != 0);
+
+      // free temp buffer;
+      STBI_FREE(g.out);
+      STBI_FREE(g.history);
+      STBI_FREE(g.background);
+
+      // do the final conversion after loading everything;
+      if (req_comp && req_comp != 4)
+         out = stbi__convert_format(out, 4, req_comp, layers * g.w, g.h);
+
+      *z = layers;
+      return out;
+   } else {
+      return stbi__errpuc("not GIF", "Image was not as a gif type.");
+   }
+}
+
+static void *stbi__gif_load(stbi__context *s, int *x, int *y, int *comp, int req_comp, stbi__result_info *ri)
+{
+   stbi_uc *u = 0;
+   stbi__gif g;
+   memset(&g, 0, sizeof(g));
+   STBI_NOTUSED(ri);
+
+   u = stbi__gif_load_next(s, &g, comp, req_comp, 0);
+   if (u == (stbi_uc *) s) u = 0;  // end of animated gif marker
+   if (u) {
+      *x = g.w;
+      *y = g.h;
+
+      // moved conversion to after successful load so that the same
+      // can be done for multiple frames.
+      if (req_comp && req_comp != 4)
+         u = stbi__convert_format(u, 4, req_comp, g.w, g.h);
+   } else if (g.out) {
+      // if there was an error and we allocated an image buffer, free it!
+      STBI_FREE(g.out);
+   }
+
+   // free buffers needed for multiple frame loading;
+   STBI_FREE(g.history);
+   STBI_FREE(g.background);
+
+   return u;
+}
+
+static int stbi__gif_info(stbi__context *s, int *x, int *y, int *comp)
+{
+   return stbi__gif_info_raw(s,x,y,comp);
+}
+#endif
+
+// *************************************************************************************************
+// Radiance RGBE HDR loader
+// originally by Nicolas Schulz
+#ifndef STBI_NO_HDR
+static int stbi__hdr_test_core(stbi__context *s, const char *signature)
+{
+   int i;
+   for (i=0; signature[i]; ++i)
+      if (stbi__get8(s) != signature[i])
+          return 0;
+   stbi__rewind(s);
+   return 1;
+}
+
+static int stbi__hdr_test(stbi__context* s)
+{
+   int r = stbi__hdr_test_core(s, "#?RADIANCE\n");
+   stbi__rewind(s);
+   if(!r) {
+       r = stbi__hdr_test_core(s, "#?RGBE\n");
+       stbi__rewind(s);
+   }
+   return r;
+}
+
+#define STBI__HDR_BUFLEN  1024
+static char *stbi__hdr_gettoken(stbi__context *z, char *buffer)
+{
+   int len=0;
+   char c = '\0';
+
+   c = (char) stbi__get8(z);
+
+   while (!stbi__at_eof(z) && c != '\n') {
+      buffer[len++] = c;
+      if (len == STBI__HDR_BUFLEN-1) {
+         // flush to end of line
+         while (!stbi__at_eof(z) && stbi__get8(z) != '\n')
+            ;
+         break;
+      }
+      c = (char) stbi__get8(z);
+   }
+
+   buffer[len] = 0;
+   return buffer;
+}
+
+static void stbi__hdr_convert(float *output, stbi_uc *input, int req_comp)
+{
+   if ( input[3] != 0 ) {
+      float f1;
+      // Exponent
+      f1 = (float) ldexp(1.0f, input[3] - (int)(128 + 8));
+      if (req_comp <= 2)
+         output[0] = (input[0] + input[1] + input[2]) * f1 / 3;
+      else {
+         output[0] = input[0] * f1;
+         output[1] = input[1] * f1;
+         output[2] = input[2] * f1;
+      }
+      if (req_comp == 2) output[1] = 1;
+      if (req_comp == 4) output[3] = 1;
+   } else {
+      switch (req_comp) {
+         case 4: output[3] = 1; /* fallthrough */
+         case 3: output[0] = output[1] = output[2] = 0;
+                 break;
+         case 2: output[1] = 1; /* fallthrough */
+         case 1: output[0] = 0;
+                 break;
+      }
+   }
+}
+
+static float *stbi__hdr_load(stbi__context *s, int *x, int *y, int *comp, int req_comp, stbi__result_info *ri)
+{
+   char buffer[STBI__HDR_BUFLEN];
+   char *token;
+   int valid = 0;
+   int width, height;
+   stbi_uc *scanline;
+   float *hdr_data;
+   int len;
+   unsigned char count, value;
+   int i, j, k, c1,c2, z;
+   const char *headerToken;
+   STBI_NOTUSED(ri);
+
+   // Check identifier
+   headerToken = stbi__hdr_gettoken(s,buffer);
+   if (strcmp(headerToken, "#?RADIANCE") != 0 && strcmp(headerToken, "#?RGBE") != 0)
+      return stbi__errpf("not HDR", "Corrupt HDR image");
+
+   // Parse header
+   for(;;) {
+      token = stbi__hdr_gettoken(s,buffer);
+      if (token[0] == 0) break;
+      if (strcmp(token, "FORMAT=32-bit_rle_rgbe") == 0) valid = 1;
+   }
+
+   if (!valid)    return stbi__errpf("unsupported format", "Unsupported HDR format");
+
+   // Parse width and height
+   // can't use sscanf() if we're not using stdio!
+   token = stbi__hdr_gettoken(s,buffer);
+   if (strncmp(token, "-Y ", 3))  return stbi__errpf("unsupported data layout", "Unsupported HDR format");
+   token += 3;
+   height = (int) strtol(token, &token, 10);
+   while (*token == ' ') ++token;
+   if (strncmp(token, "+X ", 3))  return stbi__errpf("unsupported data layout", "Unsupported HDR format");
+   token += 3;
+   width = (int) strtol(token, NULL, 10);
+
+   if (height > STBI_MAX_DIMENSIONS) return stbi__errpf("too large","Very large image (corrupt?)");
+   if (width > STBI_MAX_DIMENSIONS) return stbi__errpf("too large","Very large image (corrupt?)");
+
+   *x = width;
+   *y = height;
+
+   if (comp) *comp = 3;
+   if (req_comp == 0) req_comp = 3;
+
+   if (!stbi__mad4sizes_valid(width, height, req_comp, sizeof(float), 0))
+      return stbi__errpf("too large", "HDR image is too large");
+
+   // Read data
+   hdr_data = (float *) stbi__malloc_mad4(width, height, req_comp, sizeof(float), 0);
+   if (!hdr_data)
+      return stbi__errpf("outofmem", "Out of memory");
+
+   // Load image data
+   // image data is stored as some number of sca
+   if ( width < 8 || width >= 32768) {
+      // Read flat data
+      for (j=0; j < height; ++j) {
+         for (i=0; i < width; ++i) {
+            stbi_uc rgbe[4];
+           main_decode_loop:
+            stbi__getn(s, rgbe, 4);
+            stbi__hdr_convert(hdr_data + j * width * req_comp + i * req_comp, rgbe, req_comp);
+         }
+      }
+   } else {
+      // Read RLE-encoded data
+      scanline = NULL;
+
+      for (j = 0; j < height; ++j) {
+         c1 = stbi__get8(s);
+         c2 = stbi__get8(s);
+         len = stbi__get8(s);
+         if (c1 != 2 || c2 != 2 || (len & 0x80)) {
+            // not run-length encoded, so we have to actually use THIS data as a decoded
+            // pixel (note this can't be a valid pixel--one of RGB must be >= 128)
+            stbi_uc rgbe[4];
+            rgbe[0] = (stbi_uc) c1;
+            rgbe[1] = (stbi_uc) c2;
+            rgbe[2] = (stbi_uc) len;
+            rgbe[3] = (stbi_uc) stbi__get8(s);
+            stbi__hdr_convert(hdr_data, rgbe, req_comp);
+            i = 1;
+            j = 0;
+            STBI_FREE(scanline);
+            goto main_decode_loop; // yes, this makes no sense
+         }
+         len <<= 8;
+         len |= stbi__get8(s);
+         if (len != width) { STBI_FREE(hdr_data); STBI_FREE(scanline); return stbi__errpf("invalid decoded scanline length", "corrupt HDR"); }
+         if (scanline == NULL) {
+            scanline = (stbi_uc *) stbi__malloc_mad2(width, 4, 0);
+            if (!scanline) {
+               STBI_FREE(hdr_data);
+               return stbi__errpf("outofmem", "Out of memory");
+            }
+         }
+
+         for (k = 0; k < 4; ++k) {
+            int nleft;
+            i = 0;
+            while ((nleft = width - i) > 0) {
+               count = stbi__get8(s);
+               if (count > 128) {
+                  // Run
+                  value = stbi__get8(s);
+                  count -= 128;
+                  if ((count == 0) || (count > nleft)) { STBI_FREE(hdr_data); STBI_FREE(scanline); return stbi__errpf("corrupt", "bad RLE data in HDR"); }
+                  for (z = 0; z < count; ++z)
+                     scanline[i++ * 4 + k] = value;
+               } else {
+                  // Dump
+                  if ((count == 0) || (count > nleft)) { STBI_FREE(hdr_data); STBI_FREE(scanline); return stbi__errpf("corrupt", "bad RLE data in HDR"); }
+                  for (z = 0; z < count; ++z)
+                     scanline[i++ * 4 + k] = stbi__get8(s);
+               }
+            }
+         }
+         for (i=0; i < width; ++i)
+            stbi__hdr_convert(hdr_data+(j*width + i)*req_comp, scanline + i*4, req_comp);
+      }
+      if (scanline)
+         STBI_FREE(scanline);
+   }
+
+   return hdr_data;
+}
+
+static int stbi__hdr_info(stbi__context *s, int *x, int *y, int *comp)
+{
+   char buffer[STBI__HDR_BUFLEN];
+   char *token;
+   int valid = 0;
+   int dummy;
+
+   if (!x) x = &dummy;
+   if (!y) y = &dummy;
+   if (!comp) comp = &dummy;
+
+   if (stbi__hdr_test(s) == 0) {
+       stbi__rewind( s );
+       return 0;
+   }
+
+   for(;;) {
+      token = stbi__hdr_gettoken(s,buffer);
+      if (token[0] == 0) break;
+      if (strcmp(token, "FORMAT=32-bit_rle_rgbe") == 0) valid = 1;
+   }
+
+   if (!valid) {
+       stbi__rewind( s );
+       return 0;
+   }
+   token = stbi__hdr_gettoken(s,buffer);
+   if (strncmp(token, "-Y ", 3)) {
+       stbi__rewind( s );
+       return 0;
+   }
+   token += 3;
+   *y = (int) strtol(token, &token, 10);
+   while (*token == ' ') ++token;
+   if (strncmp(token, "+X ", 3)) {
+       stbi__rewind( s );
+       return 0;
+   }
+   token += 3;
+   *x = (int) strtol(token, NULL, 10);
+   *comp = 3;
+   return 1;
+}
+#endif // STBI_NO_HDR
+
+#ifndef STBI_NO_BMP
+static int stbi__bmp_info(stbi__context *s, int *x, int *y, int *comp)
+{
+   void *p;
+   stbi__bmp_data info;
+
+   info.all_a = 255;
+   p = stbi__bmp_parse_header(s, &info);
+   if (p == NULL) {
+      stbi__rewind( s );
+      return 0;
+   }
+   if (x) *x = s->img_x;
+   if (y) *y = s->img_y;
+   if (comp) {
+      if (info.bpp == 24 && info.ma == 0xff000000)
+         *comp = 3;
+      else
+         *comp = info.ma ? 4 : 3;
+   }
+   return 1;
+}
+#endif
+
+#ifndef STBI_NO_PSD
+static int stbi__psd_info(stbi__context *s, int *x, int *y, int *comp)
+{
+   int channelCount, dummy, depth;
+   if (!x) x = &dummy;
+   if (!y) y = &dummy;
+   if (!comp) comp = &dummy;
+   if (stbi__get32be(s) != 0x38425053) {
+       stbi__rewind( s );
+       return 0;
+   }
+   if (stbi__get16be(s) != 1) {
+       stbi__rewind( s );
+       return 0;
+   }
+   stbi__skip(s, 6);
+   channelCount = stbi__get16be(s);
+   if (channelCount < 0 || channelCount > 16) {
+       stbi__rewind( s );
+       return 0;
+   }
+   *y = stbi__get32be(s);
+   *x = stbi__get32be(s);
+   depth = stbi__get16be(s);
+   if (depth != 8 && depth != 16) {
+       stbi__rewind( s );
+       return 0;
+   }
+   if (stbi__get16be(s) != 3) {
+       stbi__rewind( s );
+       return 0;
+   }
+   *comp = 4;
+   return 1;
+}
+
+static int stbi__psd_is16(stbi__context *s)
+{
+   int channelCount, depth;
+   if (stbi__get32be(s) != 0x38425053) {
+       stbi__rewind( s );
+       return 0;
+   }
+   if (stbi__get16be(s) != 1) {
+       stbi__rewind( s );
+       return 0;
+   }
+   stbi__skip(s, 6);
+   channelCount = stbi__get16be(s);
+   if (channelCount < 0 || channelCount > 16) {
+       stbi__rewind( s );
+       return 0;
+   }
+   STBI_NOTUSED(stbi__get32be(s));
+   STBI_NOTUSED(stbi__get32be(s));
+   depth = stbi__get16be(s);
+   if (depth != 16) {
+       stbi__rewind( s );
+       return 0;
+   }
+   return 1;
+}
+#endif
+
+#ifndef STBI_NO_PIC
+static int stbi__pic_info(stbi__context *s, int *x, int *y, int *comp)
+{
+   int act_comp=0,num_packets=0,chained,dummy;
+   stbi__pic_packet packets[10];
+
+   if (!x) x = &dummy;
+   if (!y) y = &dummy;
+   if (!comp) comp = &dummy;
+
+   if (!stbi__pic_is4(s,"\x53\x80\xF6\x34")) {
+      stbi__rewind(s);
+      return 0;
+   }
+
+   stbi__skip(s, 88);
+
+   *x = stbi__get16be(s);
+   *y = stbi__get16be(s);
+   if (stbi__at_eof(s)) {
+      stbi__rewind( s);
+      return 0;
+   }
+   if ( (*x) != 0 && (1 << 28) / (*x) < (*y)) {
+      stbi__rewind( s );
+      return 0;
+   }
+
+   stbi__skip(s, 8);
+
+   do {
+      stbi__pic_packet *packet;
+
+      if (num_packets==sizeof(packets)/sizeof(packets[0]))
+         return 0;
+
+      packet = &packets[num_packets++];
+      chained = stbi__get8(s);
+      packet->size    = stbi__get8(s);
+      packet->type    = stbi__get8(s);
+      packet->channel = stbi__get8(s);
+      act_comp |= packet->channel;
+
+      if (stbi__at_eof(s)) {
+          stbi__rewind( s );
+          return 0;
+      }
+      if (packet->size != 8) {
+          stbi__rewind( s );
+          return 0;
+      }
+   } while (chained);
+
+   *comp = (act_comp & 0x10 ? 4 : 3);
+
+   return 1;
+}
+#endif
+
+// *************************************************************************************************
+// Portable Gray Map and Portable Pixel Map loader
+// by Ken Miller
+//
+// PGM: http://netpbm.sourceforge.net/doc/pgm.html
+// PPM: http://netpbm.sourceforge.net/doc/ppm.html
+//
+// Known limitations:
+//    Does not support comments in the header section
+//    Does not support ASCII image data (formats P2 and P3)
+
+#ifndef STBI_NO_PNM
+
+static int      stbi__pnm_test(stbi__context *s)
+{
+   char p, t;
+   p = (char) stbi__get8(s);
+   t = (char) stbi__get8(s);
+   if (p != 'P' || (t != '5' && t != '6')) {
+       stbi__rewind( s );
+       return 0;
+   }
+   return 1;
+}
+
+static void *stbi__pnm_load(stbi__context *s, int *x, int *y, int *comp, int req_comp, stbi__result_info *ri)
+{
+   stbi_uc *out;
+   STBI_NOTUSED(ri);
+
+   ri->bits_per_channel = stbi__pnm_info(s, (int *)&s->img_x, (int *)&s->img_y, (int *)&s->img_n);
+   if (ri->bits_per_channel == 0)
+      return 0;
+
+   if (s->img_y > STBI_MAX_DIMENSIONS) return stbi__errpuc("too large","Very large image (corrupt?)");
+   if (s->img_x > STBI_MAX_DIMENSIONS) return stbi__errpuc("too large","Very large image (corrupt?)");
+
+   *x = s->img_x;
+   *y = s->img_y;
+   if (comp) *comp = s->img_n;
+
+   if (!stbi__mad4sizes_valid(s->img_n, s->img_x, s->img_y, ri->bits_per_channel / 8, 0))
+      return stbi__errpuc("too large", "PNM too large");
+
+   out = (stbi_uc *) stbi__malloc_mad4(s->img_n, s->img_x, s->img_y, ri->bits_per_channel / 8, 0);
+   if (!out) return stbi__errpuc("outofmem", "Out of memory");
+   if (!stbi__getn(s, out, s->img_n * s->img_x * s->img_y * (ri->bits_per_channel / 8))) {
+      STBI_FREE(out);
+      return stbi__errpuc("bad PNM", "PNM file truncated");
+   }
+
+   if (req_comp && req_comp != s->img_n) {
+      if (ri->bits_per_channel == 16) {
+         out = (stbi_uc *) stbi__convert_format16((stbi__uint16 *) out, s->img_n, req_comp, s->img_x, s->img_y);
+      } else {
+         out = stbi__convert_format(out, s->img_n, req_comp, s->img_x, s->img_y);
+      }
+      if (out == NULL) return out; // stbi__convert_format frees input on failure
+   }
+   return out;
+}
+
+static int      stbi__pnm_isspace(char c)
+{
+   return c == ' ' || c == '\t' || c == '\n' || c == '\v' || c == '\f' || c == '\r';
+}
+
+static void     stbi__pnm_skip_whitespace(stbi__context *s, char *c)
+{
+   for (;;) {
+      while (!stbi__at_eof(s) && stbi__pnm_isspace(*c))
+         *c = (char) stbi__get8(s);
+
+      if (stbi__at_eof(s) || *c != '#')
+         break;
+
+      while (!stbi__at_eof(s) && *c != '\n' && *c != '\r' )
+         *c = (char) stbi__get8(s);
+   }
+}
+
+static int      stbi__pnm_isdigit(char c)
+{
+   return c >= '0' && c <= '9';
+}
+
+static int      stbi__pnm_getinteger(stbi__context *s, char *c)
+{
+   int value = 0;
+
+   while (!stbi__at_eof(s) && stbi__pnm_isdigit(*c)) {
+      value = value*10 + (*c - '0');
+      *c = (char) stbi__get8(s);
+      if((value > 214748364) || (value == 214748364 && *c > '7'))
+          return stbi__err("integer parse overflow", "Parsing an integer in the PPM header overflowed a 32-bit int");
+   }
+
+   return value;
+}
+
+static int      stbi__pnm_info(stbi__context *s, int *x, int *y, int *comp)
+{
+   int maxv, dummy;
+   char c, p, t;
+
+   if (!x) x = &dummy;
+   if (!y) y = &dummy;
+   if (!comp) comp = &dummy;
+
+   stbi__rewind(s);
+
+   // Get identifier
+   p = (char) stbi__get8(s);
+   t = (char) stbi__get8(s);
+   if (p != 'P' || (t != '5' && t != '6')) {
+       stbi__rewind(s);
+       return 0;
+   }
+
+   *comp = (t == '6') ? 3 : 1;  // '5' is 1-component .pgm; '6' is 3-component .ppm
+
+   c = (char) stbi__get8(s);
+   stbi__pnm_skip_whitespace(s, &c);
+
+   *x = stbi__pnm_getinteger(s, &c); // read width
+   if(*x == 0)
+       return stbi__err("invalid width", "PPM image header had zero or overflowing width");
+   stbi__pnm_skip_whitespace(s, &c);
+
+   *y = stbi__pnm_getinteger(s, &c); // read height
+   if (*y == 0)
+       return stbi__err("invalid width", "PPM image header had zero or overflowing width");
+   stbi__pnm_skip_whitespace(s, &c);
+
+   maxv = stbi__pnm_getinteger(s, &c);  // read max value
+   if (maxv > 65535)
+      return stbi__err("max value > 65535", "PPM image supports only 8-bit and 16-bit images");
+   else if (maxv > 255)
+      return 16;
+   else
+      return 8;
+}
+
+static int stbi__pnm_is16(stbi__context *s)
+{
+   if (stbi__pnm_info(s, NULL, NULL, NULL) == 16)
+           return 1;
+   return 0;
+}
+#endif
+
+static int stbi__info_main(stbi__context *s, int *x, int *y, int *comp)
+{
+   #ifndef STBI_NO_JPEG
+   if (stbi__jpeg_info(s, x, y, comp)) return 1;
+   #endif
+
+   #ifndef STBI_NO_PNG
+   if (stbi__png_info(s, x, y, comp))  return 1;
+   #endif
+
+   #ifndef STBI_NO_GIF
+   if (stbi__gif_info(s, x, y, comp))  return 1;
+   #endif
+
+   #ifndef STBI_NO_BMP
+   if (stbi__bmp_info(s, x, y, comp))  return 1;
+   #endif
+
+   #ifndef STBI_NO_PSD
+   if (stbi__psd_info(s, x, y, comp))  return 1;
+   #endif
+
+   #ifndef STBI_NO_PIC
+   if (stbi__pic_info(s, x, y, comp))  return 1;
+   #endif
+
+   #ifndef STBI_NO_PNM
+   if (stbi__pnm_info(s, x, y, comp))  return 1;
+   #endif
+
+   #ifndef STBI_NO_HDR
+   if (stbi__hdr_info(s, x, y, comp))  return 1;
+   #endif
+
+   // test tga last because it's a crappy test!
+   #ifndef STBI_NO_TGA
+   if (stbi__tga_info(s, x, y, comp))
+       return 1;
+   #endif
+   return stbi__err("unknown image type", "Image not of any known type, or corrupt");
+}
+
+static int stbi__is_16_main(stbi__context *s)
+{
+   #ifndef STBI_NO_PNG
+   if (stbi__png_is16(s))  return 1;
+   #endif
+
+   #ifndef STBI_NO_PSD
+   if (stbi__psd_is16(s))  return 1;
+   #endif
+
+   #ifndef STBI_NO_PNM
+   if (stbi__pnm_is16(s))  return 1;
+   #endif
+   return 0;
+}
+
+#ifndef STBI_NO_STDIO
+STBIDEF int stbi_info(char const *filename, int *x, int *y, int *comp)
+{
+    FILE *f = stbi__fopen(filename, "rb");
+    int result;
+    if (!f) return stbi__err("can't fopen", "Unable to open file");
+    result = stbi_info_from_file(f, x, y, comp);
+    fclose(f);
+    return result;
+}
+
+STBIDEF int stbi_info_from_file(FILE *f, int *x, int *y, int *comp)
+{
+   int r;
+   stbi__context s;
+   long pos = ftell(f);
+   stbi__start_file(&s, f);
+   r = stbi__info_main(&s,x,y,comp);
+   fseek(f,pos,SEEK_SET);
+   return r;
+}
+
+STBIDEF int stbi_is_16_bit(char const *filename)
+{
+    FILE *f = stbi__fopen(filename, "rb");
+    int result;
+    if (!f) return stbi__err("can't fopen", "Unable to open file");
+    result = stbi_is_16_bit_from_file(f);
+    fclose(f);
+    return result;
+}
+
+STBIDEF int stbi_is_16_bit_from_file(FILE *f)
+{
+   int r;
+   stbi__context s;
+   long pos = ftell(f);
+   stbi__start_file(&s, f);
+   r = stbi__is_16_main(&s);
+   fseek(f,pos,SEEK_SET);
+   return r;
+}
+#endif // !STBI_NO_STDIO
+
+STBIDEF int stbi_info_from_memory(stbi_uc const *buffer, int len, int *x, int *y, int *comp)
+{
+   stbi__context s;
+   stbi__start_mem(&s,buffer,len);
+   return stbi__info_main(&s,x,y,comp);
+}
+
+STBIDEF int stbi_info_from_callbacks(stbi_io_callbacks const *c, void *user, int *x, int *y, int *comp)
+{
+   stbi__context s;
+   stbi__start_callbacks(&s, (stbi_io_callbacks *) c, user);
+   return stbi__info_main(&s,x,y,comp);
+}
+
+STBIDEF int stbi_is_16_bit_from_memory(stbi_uc const *buffer, int len)
+{
+   stbi__context s;
+   stbi__start_mem(&s,buffer,len);
+   return stbi__is_16_main(&s);
+}
+
+STBIDEF int stbi_is_16_bit_from_callbacks(stbi_io_callbacks const *c, void *user)
+{
+   stbi__context s;
+   stbi__start_callbacks(&s, (stbi_io_callbacks *) c, user);
+   return stbi__is_16_main(&s);
+}
+
+#endif // STB_IMAGE_IMPLEMENTATION
+
+/*
+   revision history:
+      2.20  (2019-02-07) support utf8 filenames in Windows; fix warnings and platform ifdefs
+      2.19  (2018-02-11) fix warning
+      2.18  (2018-01-30) fix warnings
+      2.17  (2018-01-29) change sbti__shiftsigned to avoid clang -O2 bug
+                         1-bit BMP
+                         *_is_16_bit api
+                         avoid warnings
+      2.16  (2017-07-23) all functions have 16-bit variants;
+                         STBI_NO_STDIO works again;
+                         compilation fixes;
+                         fix rounding in unpremultiply;
+                         optimize vertical flip;
+                         disable raw_len validation;
+                         documentation fixes
+      2.15  (2017-03-18) fix png-1,2,4 bug; now all Imagenet JPGs decode;
+                         warning fixes; disable run-time SSE detection on gcc;
+                         uniform handling of optional "return" values;
+                         thread-safe initialization of zlib tables
+      2.14  (2017-03-03) remove deprecated STBI_JPEG_OLD; fixes for Imagenet JPGs
+      2.13  (2016-11-29) add 16-bit API, only supported for PNG right now
+      2.12  (2016-04-02) fix typo in 2.11 PSD fix that caused crashes
+      2.11  (2016-04-02) allocate large structures on the stack
+                         remove white matting for transparent PSD
+                         fix reported channel count for PNG & BMP
+                         re-enable SSE2 in non-gcc 64-bit
+                         support RGB-formatted JPEG
+                         read 16-bit PNGs (only as 8-bit)
+      2.10  (2016-01-22) avoid warning introduced in 2.09 by STBI_REALLOC_SIZED
+      2.09  (2016-01-16) allow comments in PNM files
+                         16-bit-per-pixel TGA (not bit-per-component)
+                         info() for TGA could break due to .hdr handling
+                         info() for BMP to shares code instead of sloppy parse
+                         can use STBI_REALLOC_SIZED if allocator doesn't support realloc
+                         code cleanup
+      2.08  (2015-09-13) fix to 2.07 cleanup, reading RGB PSD as RGBA
+      2.07  (2015-09-13) fix compiler warnings
+                         partial animated GIF support
+                         limited 16-bpc PSD support
+                         #ifdef unused functions
+                         bug with < 92 byte PIC,PNM,HDR,TGA
+      2.06  (2015-04-19) fix bug where PSD returns wrong '*comp' value
+      2.05  (2015-04-19) fix bug in progressive JPEG handling, fix warning
+      2.04  (2015-04-15) try to re-enable SIMD on MinGW 64-bit
+      2.03  (2015-04-12) extra corruption checking (mmozeiko)
+                         stbi_set_flip_vertically_on_load (nguillemot)
+                         fix NEON support; fix mingw support
+      2.02  (2015-01-19) fix incorrect assert, fix warning
+      2.01  (2015-01-17) fix various warnings; suppress SIMD on gcc 32-bit without -msse2
+      2.00b (2014-12-25) fix STBI_MALLOC in progressive JPEG
+      2.00  (2014-12-25) optimize JPG, including x86 SSE2 & NEON SIMD (ryg)
+                         progressive JPEG (stb)
+                         PGM/PPM support (Ken Miller)
+                         STBI_MALLOC,STBI_REALLOC,STBI_FREE
+                         GIF bugfix -- seemingly never worked
+                         STBI_NO_*, STBI_ONLY_*
+      1.48  (2014-12-14) fix incorrectly-named assert()
+      1.47  (2014-12-14) 1/2/4-bit PNG support, both direct and paletted (Omar Cornut & stb)
+                         optimize PNG (ryg)
+                         fix bug in interlaced PNG with user-specified channel count (stb)
+      1.46  (2014-08-26)
+              fix broken tRNS chunk (colorkey-style transparency) in non-paletted PNG
+      1.45  (2014-08-16)
+              fix MSVC-ARM internal compiler error by wrapping malloc
+      1.44  (2014-08-07)
+              various warning fixes from Ronny Chevalier
+      1.43  (2014-07-15)
+              fix MSVC-only compiler problem in code changed in 1.42
+      1.42  (2014-07-09)
+              don't define _CRT_SECURE_NO_WARNINGS (affects user code)
+              fixes to stbi__cleanup_jpeg path
+              added STBI_ASSERT to avoid requiring assert.h
+      1.41  (2014-06-25)
+              fix search&replace from 1.36 that messed up comments/error messages
+      1.40  (2014-06-22)
+              fix gcc struct-initialization warning
+      1.39  (2014-06-15)
+              fix to TGA optimization when req_comp != number of components in TGA;
+              fix to GIF loading because BMP wasn't rewinding (whoops, no GIFs in my test suite)
+              add support for BMP version 5 (more ignored fields)
+      1.38  (2014-06-06)
+              suppress MSVC warnings on integer casts truncating values
+              fix accidental rename of 'skip' field of I/O
+      1.37  (2014-06-04)
+              remove duplicate typedef
+      1.36  (2014-06-03)
+              convert to header file single-file library
+              if de-iphone isn't set, load iphone images color-swapped instead of returning NULL
+      1.35  (2014-05-27)
+              various warnings
+              fix broken STBI_SIMD path
+              fix bug where stbi_load_from_file no longer left file pointer in correct place
+              fix broken non-easy path for 32-bit BMP (possibly never used)
+              TGA optimization by Arseny Kapoulkine
+      1.34  (unknown)
+              use STBI_NOTUSED in stbi__resample_row_generic(), fix one more leak in tga failure case
+      1.33  (2011-07-14)
+              make stbi_is_hdr work in STBI_NO_HDR (as specified), minor compiler-friendly improvements
+      1.32  (2011-07-13)
+              support for "info" function for all supported filetypes (SpartanJ)
+      1.31  (2011-06-20)
+              a few more leak fixes, bug in PNG handling (SpartanJ)
+      1.30  (2011-06-11)
+              added ability to load files via callbacks to accomidate custom input streams (Ben Wenger)
+              removed deprecated format-specific test/load functions
+              removed support for installable file formats (stbi_loader) -- would have been broken for IO callbacks anyway
+              error cases in bmp and tga give messages and don't leak (Raymond Barbiero, grisha)
+              fix inefficiency in decoding 32-bit BMP (David Woo)
+      1.29  (2010-08-16)
+              various warning fixes from Aurelien Pocheville
+      1.28  (2010-08-01)
+              fix bug in GIF palette transparency (SpartanJ)
+      1.27  (2010-08-01)
+              cast-to-stbi_uc to fix warnings
+      1.26  (2010-07-24)
+              fix bug in file buffering for PNG reported by SpartanJ
+      1.25  (2010-07-17)
+              refix trans_data warning (Won Chun)
+      1.24  (2010-07-12)
+              perf improvements reading from files on platforms with lock-heavy fgetc()
+              minor perf improvements for jpeg
+              deprecated type-specific functions so we'll get feedback if they're needed
+              attempt to fix trans_data warning (Won Chun)
+      1.23    fixed bug in iPhone support
+      1.22  (2010-07-10)
+              removed image *writing* support
+              stbi_info support from Jetro Lauha
+              GIF support from Jean-Marc Lienher
+              iPhone PNG-extensions from James Brown
+              warning-fixes from Nicolas Schulz and Janez Zemva (i.stbi__err. Janez (U+017D)emva)
+      1.21    fix use of 'stbi_uc' in header (reported by jon blow)
+      1.20    added support for Softimage PIC, by Tom Seddon
+      1.19    bug in interlaced PNG corruption check (found by ryg)
+      1.18  (2008-08-02)
+              fix a threading bug (local mutable static)
+      1.17    support interlaced PNG
+      1.16    major bugfix - stbi__convert_format converted one too many pixels
+      1.15    initialize some fields for thread safety
+      1.14    fix threadsafe conversion bug
+              header-file-only version (#define STBI_HEADER_FILE_ONLY before including)
+      1.13    threadsafe
+      1.12    const qualifiers in the API
+      1.11    Support installable IDCT, colorspace conversion routines
+      1.10    Fixes for 64-bit (don't use "unsigned long")
+              optimized upsampling by Fabian "ryg" Giesen
+      1.09    Fix format-conversion for PSD code (bad global variables!)
+      1.08    Thatcher Ulrich's PSD code integrated by Nicolas Schulz
+      1.07    attempt to fix C++ warning/errors again
+      1.06    attempt to fix C++ warning/errors again
+      1.05    fix TGA loading to return correct *comp and use good luminance calc
+      1.04    default float alpha is 1, not 255; use 'void *' for stbi_image_free
+      1.03    bugfixes to STBI_NO_STDIO, STBI_NO_HDR
+      1.02    support for (subset of) HDR files, float interface for preferred access to them
+      1.01    fix bug: possible bug in handling right-side up bmps... not sure
+              fix bug: the stbi__bmp_load() and stbi__tga_load() functions didn't work at all
+      1.00    interface to zlib that skips zlib header
+      0.99    correct handling of alpha in palette
+      0.98    TGA loader by lonesock; dynamically add loaders (untested)
+      0.97    jpeg errors on too large a file; also catch another malloc failure
+      0.96    fix detection of invalid v value - particleman@mollyrocket forum
+      0.95    during header scan, seek to markers in case of padding
+      0.94    STBI_NO_STDIO to disable stdio usage; rename all #defines the same
+      0.93    handle jpegtran output; verbose errors
+      0.92    read 4,8,16,24,32-bit BMP files of several formats
+      0.91    output 24-bit Windows 3.0 BMP files
+      0.90    fix a few more warnings; bump version number to approach 1.0
+      0.61    bugfixes due to Marc LeBlanc, Christopher Lloyd
+      0.60    fix compiling as c++
+      0.59    fix warnings: merge Dave Moore's -Wall fixes
+      0.58    fix bug: zlib uncompressed mode len/nlen was wrong endian
+      0.57    fix bug: jpg last huffman symbol before marker was >9 bits but less than 16 available
+      0.56    fix bug: zlib uncompressed mode len vs. nlen
+      0.55    fix bug: restart_interval not initialized to 0
+      0.54    allow NULL for 'int *comp'
+      0.53    fix bug in png 3->4; speedup png decoding
+      0.52    png handles req_comp=3,4 directly; minor cleanup; jpeg comments
+      0.51    obey req_comp requests, 1-component jpegs return as 1-component,
+              on 'test' only check type, not whether we support this variant
+      0.50  (2006-11-19)
+              first released version
+*/
+
+
+/*
+------------------------------------------------------------------------------
+This software is available under 2 licenses -- choose whichever you prefer.
+------------------------------------------------------------------------------
+ALTERNATIVE A - MIT License
+Copyright (c) 2017 Sean Barrett
+Permission is hereby granted, free of charge, to any person obtaining a copy of
+this software and associated documentation files (the "Software"), to deal in
+the Software without restriction, including without limitation the rights to
+use, copy, modify, merge, publish, distribute, sublicense, and/or sell copies
+of the Software, and to permit persons to whom the Software is furnished to do
+so, subject to the following conditions:
+The above copyright notice and this permission notice shall be included in all
+copies or substantial portions of the Software.
+THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+SOFTWARE.
+------------------------------------------------------------------------------
+ALTERNATIVE B - Public Domain (www.unlicense.org)
+This is free and unencumbered software released into the public domain.
+Anyone is free to copy, modify, publish, use, compile, sell, or distribute this
+software, either in source code form or as a compiled binary, for any purpose,
+commercial or non-commercial, and by any means.
+In jurisdictions that recognize copyright laws, the author or authors of this
+software dedicate any and all copyright interest in the software to the public
+domain. We make this dedication for the benefit of the public at large and to
+the detriment of our heirs and successors. We intend this dedication to be an
+overt act of relinquishment in perpetuity of all present and future rights to
+this software under copyright law.
+THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+AUTHORS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
+ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION
+WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+------------------------------------------------------------------------------
+*/
diff --git a/externals/stb/stb_image_resize.h b/externals/stb/stb_image_resize.h
new file mode 100644
index 000000000..ef9e6fe87
--- /dev/null
+++ b/externals/stb/stb_image_resize.h
@@ -0,0 +1,2634 @@
+/* stb_image_resize - v0.97 - public domain image resizing
+   by Jorge L Rodriguez (@VinoBS) - 2014
+   http://github.com/nothings/stb
+
+   Written with emphasis on usability, portability, and efficiency. (No
+   SIMD or threads, so it be easily outperformed by libs that use those.)
+   Only scaling and translation is supported, no rotations or shears.
+   Easy API downsamples w/Mitchell filter, upsamples w/cubic interpolation.
+
+   COMPILING & LINKING
+      In one C/C++ file that #includes this file, do this:
+         #define STB_IMAGE_RESIZE_IMPLEMENTATION
+      before the #include. That will create the implementation in that file.
+
+   QUICKSTART
+      stbir_resize_uint8(      input_pixels , in_w , in_h , 0,
+                               output_pixels, out_w, out_h, 0, num_channels)
+      stbir_resize_float(...)
+      stbir_resize_uint8_srgb( input_pixels , in_w , in_h , 0,
+                               output_pixels, out_w, out_h, 0,
+                               num_channels , alpha_chan  , 0)
+      stbir_resize_uint8_srgb_edgemode(
+                               input_pixels , in_w , in_h , 0,
+                               output_pixels, out_w, out_h, 0,
+                               num_channels , alpha_chan  , 0, STBIR_EDGE_CLAMP)
+                                                            // WRAP/REFLECT/ZERO
+
+   FULL API
+      See the "header file" section of the source for API documentation.
+
+   ADDITIONAL DOCUMENTATION
+
+      SRGB & FLOATING POINT REPRESENTATION
+         The sRGB functions presume IEEE floating point. If you do not have
+         IEEE floating point, define STBIR_NON_IEEE_FLOAT. This will use
+         a slower implementation.
+
+      MEMORY ALLOCATION
+         The resize functions here perform a single memory allocation using
+         malloc. To control the memory allocation, before the #include that
+         triggers the implementation, do:
+
+            #define STBIR_MALLOC(size,context) ...
+            #define STBIR_FREE(ptr,context)   ...
+
+         Each resize function makes exactly one call to malloc/free, so to use
+         temp memory, store the temp memory in the context and return that.
+
+      ASSERT
+         Define STBIR_ASSERT(boolval) to override assert() and not use assert.h
+
+      OPTIMIZATION
+         Define STBIR_SATURATE_INT to compute clamp values in-range using
+         integer operations instead of float operations. This may be faster
+         on some platforms.
+
+      DEFAULT FILTERS
+         For functions which don't provide explicit control over what filters
+         to use, you can change the compile-time defaults with
+
+            #define STBIR_DEFAULT_FILTER_UPSAMPLE     STBIR_FILTER_something
+            #define STBIR_DEFAULT_FILTER_DOWNSAMPLE   STBIR_FILTER_something
+
+         See stbir_filter in the header-file section for the list of filters.
+
+      NEW FILTERS
+         A number of 1D filter kernels are used. For a list of
+         supported filters see the stbir_filter enum. To add a new filter,
+         write a filter function and add it to stbir__filter_info_table.
+
+      PROGRESS
+         For interactive use with slow resize operations, you can install
+         a progress-report callback:
+
+            #define STBIR_PROGRESS_REPORT(val)   some_func(val)
+
+         The parameter val is a float which goes from 0 to 1 as progress is made.
+
+         For example:
+
+            static void my_progress_report(float progress);
+            #define STBIR_PROGRESS_REPORT(val) my_progress_report(val)
+
+            #define STB_IMAGE_RESIZE_IMPLEMENTATION
+            #include "stb_image_resize.h"
+
+            static void my_progress_report(float progress)
+            {
+               printf("Progress: %f%%\n", progress*100);
+            }
+
+      MAX CHANNELS
+         If your image has more than 64 channels, define STBIR_MAX_CHANNELS
+         to the max you'll have.
+
+      ALPHA CHANNEL
+         Most of the resizing functions provide the ability to control how
+         the alpha channel of an image is processed. The important things
+         to know about this:
+
+         1. The best mathematically-behaved version of alpha to use is
+         called "premultiplied alpha", in which the other color channels
+         have had the alpha value multiplied in. If you use premultiplied
+         alpha, linear filtering (such as image resampling done by this
+         library, or performed in texture units on GPUs) does the "right
+         thing". While premultiplied alpha is standard in the movie CGI
+         industry, it is still uncommon in the videogame/real-time world.
+
+         If you linearly filter non-premultiplied alpha, strange effects
+         occur. (For example, the 50/50 average of 99% transparent bright green
+         and 1% transparent black produces 50% transparent dark green when
+         non-premultiplied, whereas premultiplied it produces 50%
+         transparent near-black. The former introduces green energy
+         that doesn't exist in the source image.)
+
+         2. Artists should not edit premultiplied-alpha images; artists
+         want non-premultiplied alpha images. Thus, art tools generally output
+         non-premultiplied alpha images.
+
+         3. You will get best results in most cases by converting images
+         to premultiplied alpha before processing them mathematically.
+
+         4. If you pass the flag STBIR_FLAG_ALPHA_PREMULTIPLIED, the
+         resizer does not do anything special for the alpha channel;
+         it is resampled identically to other channels. This produces
+         the correct results for premultiplied-alpha images, but produces
+         less-than-ideal results for non-premultiplied-alpha images.
+
+         5. If you do not pass the flag STBIR_FLAG_ALPHA_PREMULTIPLIED,
+         then the resizer weights the contribution of input pixels
+         based on their alpha values, or, equivalently, it multiplies
+         the alpha value into the color channels, resamples, then divides
+         by the resultant alpha value. Input pixels which have alpha=0 do
+         not contribute at all to output pixels unless _all_ of the input
+         pixels affecting that output pixel have alpha=0, in which case
+         the result for that pixel is the same as it would be without
+         STBIR_FLAG_ALPHA_PREMULTIPLIED. However, this is only true for
+         input images in integer formats. For input images in float format,
+         input pixels with alpha=0 have no effect, and output pixels
+         which have alpha=0 will be 0 in all channels. (For float images,
+         you can manually achieve the same result by adding a tiny epsilon
+         value to the alpha channel of every image, and then subtracting
+         or clamping it at the end.)
+
+         6. You can suppress the behavior described in #5 and make
+         all-0-alpha pixels have 0 in all channels by #defining
+         STBIR_NO_ALPHA_EPSILON.
+
+         7. You can separately control whether the alpha channel is
+         interpreted as linear or affected by the colorspace. By default
+         it is linear; you almost never want to apply the colorspace.
+         (For example, graphics hardware does not apply sRGB conversion
+         to the alpha channel.)
+
+   CONTRIBUTORS
+      Jorge L Rodriguez: Implementation
+      Sean Barrett: API design, optimizations
+      Aras Pranckevicius: bugfix
+      Nathan Reed: warning fixes
+
+   REVISIONS
+      0.97 (2020-02-02) fixed warning
+      0.96 (2019-03-04) fixed warnings
+      0.95 (2017-07-23) fixed warnings
+      0.94 (2017-03-18) fixed warnings
+      0.93 (2017-03-03) fixed bug with certain combinations of heights
+      0.92 (2017-01-02) fix integer overflow on large (>2GB) images
+      0.91 (2016-04-02) fix warnings; fix handling of subpixel regions
+      0.90 (2014-09-17) first released version
+
+   LICENSE
+     See end of file for license information.
+
+   TODO
+      Don't decode all of the image data when only processing a partial tile
+      Don't use full-width decode buffers when only processing a partial tile
+      When processing wide images, break processing into tiles so data fits in L1 cache
+      Installable filters?
+      Resize that respects alpha test coverage
+         (Reference code: FloatImage::alphaTestCoverage and FloatImage::scaleAlphaToCoverage:
+         https://code.google.com/p/nvidia-texture-tools/source/browse/trunk/src/nvimage/FloatImage.cpp )
+*/
+
+#ifndef STBIR_INCLUDE_STB_IMAGE_RESIZE_H
+#define STBIR_INCLUDE_STB_IMAGE_RESIZE_H
+
+#ifdef _MSC_VER
+typedef unsigned char  stbir_uint8;
+typedef unsigned short stbir_uint16;
+typedef unsigned int   stbir_uint32;
+#else
+#include <stdint.h>
+typedef uint8_t  stbir_uint8;
+typedef uint16_t stbir_uint16;
+typedef uint32_t stbir_uint32;
+#endif
+
+#ifndef STBIRDEF
+#ifdef STB_IMAGE_RESIZE_STATIC
+#define STBIRDEF static
+#else
+#ifdef __cplusplus
+#define STBIRDEF extern "C"
+#else
+#define STBIRDEF extern
+#endif
+#endif
+#endif
+
+//////////////////////////////////////////////////////////////////////////////
+//
+// Easy-to-use API:
+//
+//     * "input pixels" points to an array of image data with 'num_channels' channels (e.g. RGB=3, RGBA=4)
+//     * input_w is input image width (x-axis), input_h is input image height (y-axis)
+//     * stride is the offset between successive rows of image data in memory, in bytes. you can
+//       specify 0 to mean packed continuously in memory
+//     * alpha channel is treated identically to other channels.
+//     * colorspace is linear or sRGB as specified by function name
+//     * returned result is 1 for success or 0 in case of an error.
+//       #define STBIR_ASSERT() to trigger an assert on parameter validation errors.
+//     * Memory required grows approximately linearly with input and output size, but with
+//       discontinuities at input_w == output_w and input_h == output_h.
+//     * These functions use a "default" resampling filter defined at compile time. To change the filter,
+//       you can change the compile-time defaults by #defining STBIR_DEFAULT_FILTER_UPSAMPLE
+//       and STBIR_DEFAULT_FILTER_DOWNSAMPLE, or you can use the medium-complexity API.
+
+STBIRDEF int stbir_resize_uint8(     const unsigned char *input_pixels , int input_w , int input_h , int input_stride_in_bytes,
+                                           unsigned char *output_pixels, int output_w, int output_h, int output_stride_in_bytes,
+                                     int num_channels);
+
+STBIRDEF int stbir_resize_float(     const float *input_pixels , int input_w , int input_h , int input_stride_in_bytes,
+                                           float *output_pixels, int output_w, int output_h, int output_stride_in_bytes,
+                                     int num_channels);
+
+
+// The following functions interpret image data as gamma-corrected sRGB.
+// Specify STBIR_ALPHA_CHANNEL_NONE if you have no alpha channel,
+// or otherwise provide the index of the alpha channel. Flags value
+// of 0 will probably do the right thing if you're not sure what
+// the flags mean.
+
+#define STBIR_ALPHA_CHANNEL_NONE       -1
+
+// Set this flag if your texture has premultiplied alpha. Otherwise, stbir will
+// use alpha-weighted resampling (effectively premultiplying, resampling,
+// then unpremultiplying).
+#define STBIR_FLAG_ALPHA_PREMULTIPLIED    (1 << 0)
+// The specified alpha channel should be handled as gamma-corrected value even
+// when doing sRGB operations.
+#define STBIR_FLAG_ALPHA_USES_COLORSPACE  (1 << 1)
+
+STBIRDEF int stbir_resize_uint8_srgb(const unsigned char *input_pixels , int input_w , int input_h , int input_stride_in_bytes,
+                                           unsigned char *output_pixels, int output_w, int output_h, int output_stride_in_bytes,
+                                     int num_channels, int alpha_channel, int flags);
+
+
+typedef enum
+{
+    STBIR_EDGE_CLAMP   = 1,
+    STBIR_EDGE_REFLECT = 2,
+    STBIR_EDGE_WRAP    = 3,
+    STBIR_EDGE_ZERO    = 4,
+} stbir_edge;
+
+// This function adds the ability to specify how requests to sample off the edge of the image are handled.
+STBIRDEF int stbir_resize_uint8_srgb_edgemode(const unsigned char *input_pixels , int input_w , int input_h , int input_stride_in_bytes,
+                                                    unsigned char *output_pixels, int output_w, int output_h, int output_stride_in_bytes,
+                                              int num_channels, int alpha_channel, int flags,
+                                              stbir_edge edge_wrap_mode);
+
+//////////////////////////////////////////////////////////////////////////////
+//
+// Medium-complexity API
+//
+// This extends the easy-to-use API as follows:
+//
+//     * Alpha-channel can be processed separately
+//       * If alpha_channel is not STBIR_ALPHA_CHANNEL_NONE
+//         * Alpha channel will not be gamma corrected (unless flags&STBIR_FLAG_GAMMA_CORRECT)
+//         * Filters will be weighted by alpha channel (unless flags&STBIR_FLAG_ALPHA_PREMULTIPLIED)
+//     * Filter can be selected explicitly
+//     * uint16 image type
+//     * sRGB colorspace available for all types
+//     * context parameter for passing to STBIR_MALLOC
+
+typedef enum
+{
+    STBIR_FILTER_DEFAULT      = 0,  // use same filter type that easy-to-use API chooses
+    STBIR_FILTER_BOX          = 1,  // A trapezoid w/1-pixel wide ramps, same result as box for integer scale ratios
+    STBIR_FILTER_TRIANGLE     = 2,  // On upsampling, produces same results as bilinear texture filtering
+    STBIR_FILTER_CUBICBSPLINE = 3,  // The cubic b-spline (aka Mitchell-Netrevalli with B=1,C=0), gaussian-esque
+    STBIR_FILTER_CATMULLROM   = 4,  // An interpolating cubic spline
+    STBIR_FILTER_MITCHELL     = 5,  // Mitchell-Netrevalli filter with B=1/3, C=1/3
+} stbir_filter;
+
+typedef enum
+{
+    STBIR_COLORSPACE_LINEAR,
+    STBIR_COLORSPACE_SRGB,
+
+    STBIR_MAX_COLORSPACES,
+} stbir_colorspace;
+
+// The following functions are all identical except for the type of the image data
+
+STBIRDEF int stbir_resize_uint8_generic( const unsigned char *input_pixels , int input_w , int input_h , int input_stride_in_bytes,
+                                               unsigned char *output_pixels, int output_w, int output_h, int output_stride_in_bytes,
+                                         int num_channels, int alpha_channel, int flags,
+                                         stbir_edge edge_wrap_mode, stbir_filter filter, stbir_colorspace space,
+                                         void *alloc_context);
+
+STBIRDEF int stbir_resize_uint16_generic(const stbir_uint16 *input_pixels  , int input_w , int input_h , int input_stride_in_bytes,
+                                               stbir_uint16 *output_pixels , int output_w, int output_h, int output_stride_in_bytes,
+                                         int num_channels, int alpha_channel, int flags,
+                                         stbir_edge edge_wrap_mode, stbir_filter filter, stbir_colorspace space,
+                                         void *alloc_context);
+
+STBIRDEF int stbir_resize_float_generic( const float *input_pixels         , int input_w , int input_h , int input_stride_in_bytes,
+                                               float *output_pixels        , int output_w, int output_h, int output_stride_in_bytes,
+                                         int num_channels, int alpha_channel, int flags,
+                                         stbir_edge edge_wrap_mode, stbir_filter filter, stbir_colorspace space,
+                                         void *alloc_context);
+
+
+
+//////////////////////////////////////////////////////////////////////////////
+//
+// Full-complexity API
+//
+// This extends the medium API as follows:
+//
+//       * uint32 image type
+//     * not typesafe
+//     * separate filter types for each axis
+//     * separate edge modes for each axis
+//     * can specify scale explicitly for subpixel correctness
+//     * can specify image source tile using texture coordinates
+
+typedef enum
+{
+    STBIR_TYPE_UINT8 ,
+    STBIR_TYPE_UINT16,
+    STBIR_TYPE_UINT32,
+    STBIR_TYPE_FLOAT ,
+
+    STBIR_MAX_TYPES
+} stbir_datatype;
+
+STBIRDEF int stbir_resize(         const void *input_pixels , int input_w , int input_h , int input_stride_in_bytes,
+                                         void *output_pixels, int output_w, int output_h, int output_stride_in_bytes,
+                                   stbir_datatype datatype,
+                                   int num_channels, int alpha_channel, int flags,
+                                   stbir_edge edge_mode_horizontal, stbir_edge edge_mode_vertical,
+                                   stbir_filter filter_horizontal,  stbir_filter filter_vertical,
+                                   stbir_colorspace space, void *alloc_context);
+
+STBIRDEF int stbir_resize_subpixel(const void *input_pixels , int input_w , int input_h , int input_stride_in_bytes,
+                                         void *output_pixels, int output_w, int output_h, int output_stride_in_bytes,
+                                   stbir_datatype datatype,
+                                   int num_channels, int alpha_channel, int flags,
+                                   stbir_edge edge_mode_horizontal, stbir_edge edge_mode_vertical,
+                                   stbir_filter filter_horizontal,  stbir_filter filter_vertical,
+                                   stbir_colorspace space, void *alloc_context,
+                                   float x_scale, float y_scale,
+                                   float x_offset, float y_offset);
+
+STBIRDEF int stbir_resize_region(  const void *input_pixels , int input_w , int input_h , int input_stride_in_bytes,
+                                         void *output_pixels, int output_w, int output_h, int output_stride_in_bytes,
+                                   stbir_datatype datatype,
+                                   int num_channels, int alpha_channel, int flags,
+                                   stbir_edge edge_mode_horizontal, stbir_edge edge_mode_vertical,
+                                   stbir_filter filter_horizontal,  stbir_filter filter_vertical,
+                                   stbir_colorspace space, void *alloc_context,
+                                   float s0, float t0, float s1, float t1);
+// (s0, t0) & (s1, t1) are the top-left and bottom right corner (uv addressing style: [0, 1]x[0, 1]) of a region of the input image to use.
+
+//
+//
+////   end header file   /////////////////////////////////////////////////////
+#endif // STBIR_INCLUDE_STB_IMAGE_RESIZE_H
+
+
+
+
+
+#ifdef STB_IMAGE_RESIZE_IMPLEMENTATION
+
+#ifndef STBIR_ASSERT
+#include <assert.h>
+#define STBIR_ASSERT(x) assert(x)
+#endif
+
+// For memset
+#include <string.h>
+
+#include <math.h>
+
+#ifndef STBIR_MALLOC
+#include <stdlib.h>
+// use comma operator to evaluate c, to avoid "unused parameter" warnings
+#define STBIR_MALLOC(size,c) ((void)(c), malloc(size))
+#define STBIR_FREE(ptr,c)    ((void)(c), free(ptr))
+#endif
+
+#ifndef _MSC_VER
+#ifdef __cplusplus
+#define stbir__inline inline
+#else
+#define stbir__inline
+#endif
+#else
+#define stbir__inline __forceinline
+#endif
+
+
+// should produce compiler error if size is wrong
+typedef unsigned char stbir__validate_uint32[sizeof(stbir_uint32) == 4 ? 1 : -1];
+
+#ifdef _MSC_VER
+#define STBIR__NOTUSED(v)  (void)(v)
+#else
+#define STBIR__NOTUSED(v)  (void)sizeof(v)
+#endif
+
+#define STBIR__ARRAY_SIZE(a) (sizeof((a))/sizeof((a)[0]))
+
+#ifndef STBIR_DEFAULT_FILTER_UPSAMPLE
+#define STBIR_DEFAULT_FILTER_UPSAMPLE    STBIR_FILTER_CATMULLROM
+#endif
+
+#ifndef STBIR_DEFAULT_FILTER_DOWNSAMPLE
+#define STBIR_DEFAULT_FILTER_DOWNSAMPLE  STBIR_FILTER_MITCHELL
+#endif
+
+#ifndef STBIR_PROGRESS_REPORT
+#define STBIR_PROGRESS_REPORT(float_0_to_1)
+#endif
+
+#ifndef STBIR_MAX_CHANNELS
+#define STBIR_MAX_CHANNELS 64
+#endif
+
+#if STBIR_MAX_CHANNELS > 65536
+#error "Too many channels; STBIR_MAX_CHANNELS must be no more than 65536."
+// because we store the indices in 16-bit variables
+#endif
+
+// This value is added to alpha just before premultiplication to avoid
+// zeroing out color values. It is equivalent to 2^-80. If you don't want
+// that behavior (it may interfere if you have floating point images with
+// very small alpha values) then you can define STBIR_NO_ALPHA_EPSILON to
+// disable it.
+#ifndef STBIR_ALPHA_EPSILON
+#define STBIR_ALPHA_EPSILON ((float)1 / (1 << 20) / (1 << 20) / (1 << 20) / (1 << 20))
+#endif
+
+
+
+#ifdef _MSC_VER
+#define STBIR__UNUSED_PARAM(v)  (void)(v)
+#else
+#define STBIR__UNUSED_PARAM(v)  (void)sizeof(v)
+#endif
+
+// must match stbir_datatype
+static unsigned char stbir__type_size[] = {
+    1, // STBIR_TYPE_UINT8
+    2, // STBIR_TYPE_UINT16
+    4, // STBIR_TYPE_UINT32
+    4, // STBIR_TYPE_FLOAT
+};
+
+// Kernel function centered at 0
+typedef float (stbir__kernel_fn)(float x, float scale);
+typedef float (stbir__support_fn)(float scale);
+
+typedef struct
+{
+    stbir__kernel_fn* kernel;
+    stbir__support_fn* support;
+} stbir__filter_info;
+
+// When upsampling, the contributors are which source pixels contribute.
+// When downsampling, the contributors are which destination pixels are contributed to.
+typedef struct
+{
+    int n0; // First contributing pixel
+    int n1; // Last contributing pixel
+} stbir__contributors;
+
+typedef struct
+{
+    const void* input_data;
+    int input_w;
+    int input_h;
+    int input_stride_bytes;
+
+    void* output_data;
+    int output_w;
+    int output_h;
+    int output_stride_bytes;
+
+    float s0, t0, s1, t1;
+
+    float horizontal_shift; // Units: output pixels
+    float vertical_shift;   // Units: output pixels
+    float horizontal_scale;
+    float vertical_scale;
+
+    int channels;
+    int alpha_channel;
+    stbir_uint32 flags;
+    stbir_datatype type;
+    stbir_filter horizontal_filter;
+    stbir_filter vertical_filter;
+    stbir_edge edge_horizontal;
+    stbir_edge edge_vertical;
+    stbir_colorspace colorspace;
+
+    stbir__contributors* horizontal_contributors;
+    float* horizontal_coefficients;
+
+    stbir__contributors* vertical_contributors;
+    float* vertical_coefficients;
+
+    int decode_buffer_pixels;
+    float* decode_buffer;
+
+    float* horizontal_buffer;
+
+    // cache these because ceil/floor are inexplicably showing up in profile
+    int horizontal_coefficient_width;
+    int vertical_coefficient_width;
+    int horizontal_filter_pixel_width;
+    int vertical_filter_pixel_width;
+    int horizontal_filter_pixel_margin;
+    int vertical_filter_pixel_margin;
+    int horizontal_num_contributors;
+    int vertical_num_contributors;
+
+    int ring_buffer_length_bytes;   // The length of an individual entry in the ring buffer. The total number of ring buffers is stbir__get_filter_pixel_width(filter)
+    int ring_buffer_num_entries;    // Total number of entries in the ring buffer.
+    int ring_buffer_first_scanline;
+    int ring_buffer_last_scanline;
+    int ring_buffer_begin_index;    // first_scanline is at this index in the ring buffer
+    float* ring_buffer;
+
+    float* encode_buffer; // A temporary buffer to store floats so we don't lose precision while we do multiply-adds.
+
+    int horizontal_contributors_size;
+    int horizontal_coefficients_size;
+    int vertical_contributors_size;
+    int vertical_coefficients_size;
+    int decode_buffer_size;
+    int horizontal_buffer_size;
+    int ring_buffer_size;
+    int encode_buffer_size;
+} stbir__info;
+
+
+static const float stbir__max_uint8_as_float  = 255.0f;
+static const float stbir__max_uint16_as_float = 65535.0f;
+static const double stbir__max_uint32_as_float = 4294967295.0;
+
+
+static stbir__inline int stbir__min(int a, int b)
+{
+    return a < b ? a : b;
+}
+
+static stbir__inline float stbir__saturate(float x)
+{
+    if (x < 0)
+        return 0;
+
+    if (x > 1)
+        return 1;
+
+    return x;
+}
+
+#ifdef STBIR_SATURATE_INT
+static stbir__inline stbir_uint8 stbir__saturate8(int x)
+{
+    if ((unsigned int) x <= 255)
+        return x;
+
+    if (x < 0)
+        return 0;
+
+    return 255;
+}
+
+static stbir__inline stbir_uint16 stbir__saturate16(int x)
+{
+    if ((unsigned int) x <= 65535)
+        return x;
+
+    if (x < 0)
+        return 0;
+
+    return 65535;
+}
+#endif
+
+static float stbir__srgb_uchar_to_linear_float[256] = {
+    0.000000f, 0.000304f, 0.000607f, 0.000911f, 0.001214f, 0.001518f, 0.001821f, 0.002125f, 0.002428f, 0.002732f, 0.003035f,
+    0.003347f, 0.003677f, 0.004025f, 0.004391f, 0.004777f, 0.005182f, 0.005605f, 0.006049f, 0.006512f, 0.006995f, 0.007499f,
+    0.008023f, 0.008568f, 0.009134f, 0.009721f, 0.010330f, 0.010960f, 0.011612f, 0.012286f, 0.012983f, 0.013702f, 0.014444f,
+    0.015209f, 0.015996f, 0.016807f, 0.017642f, 0.018500f, 0.019382f, 0.020289f, 0.021219f, 0.022174f, 0.023153f, 0.024158f,
+    0.025187f, 0.026241f, 0.027321f, 0.028426f, 0.029557f, 0.030713f, 0.031896f, 0.033105f, 0.034340f, 0.035601f, 0.036889f,
+    0.038204f, 0.039546f, 0.040915f, 0.042311f, 0.043735f, 0.045186f, 0.046665f, 0.048172f, 0.049707f, 0.051269f, 0.052861f,
+    0.054480f, 0.056128f, 0.057805f, 0.059511f, 0.061246f, 0.063010f, 0.064803f, 0.066626f, 0.068478f, 0.070360f, 0.072272f,
+    0.074214f, 0.076185f, 0.078187f, 0.080220f, 0.082283f, 0.084376f, 0.086500f, 0.088656f, 0.090842f, 0.093059f, 0.095307f,
+    0.097587f, 0.099899f, 0.102242f, 0.104616f, 0.107023f, 0.109462f, 0.111932f, 0.114435f, 0.116971f, 0.119538f, 0.122139f,
+    0.124772f, 0.127438f, 0.130136f, 0.132868f, 0.135633f, 0.138432f, 0.141263f, 0.144128f, 0.147027f, 0.149960f, 0.152926f,
+    0.155926f, 0.158961f, 0.162029f, 0.165132f, 0.168269f, 0.171441f, 0.174647f, 0.177888f, 0.181164f, 0.184475f, 0.187821f,
+    0.191202f, 0.194618f, 0.198069f, 0.201556f, 0.205079f, 0.208637f, 0.212231f, 0.215861f, 0.219526f, 0.223228f, 0.226966f,
+    0.230740f, 0.234551f, 0.238398f, 0.242281f, 0.246201f, 0.250158f, 0.254152f, 0.258183f, 0.262251f, 0.266356f, 0.270498f,
+    0.274677f, 0.278894f, 0.283149f, 0.287441f, 0.291771f, 0.296138f, 0.300544f, 0.304987f, 0.309469f, 0.313989f, 0.318547f,
+    0.323143f, 0.327778f, 0.332452f, 0.337164f, 0.341914f, 0.346704f, 0.351533f, 0.356400f, 0.361307f, 0.366253f, 0.371238f,
+    0.376262f, 0.381326f, 0.386430f, 0.391573f, 0.396755f, 0.401978f, 0.407240f, 0.412543f, 0.417885f, 0.423268f, 0.428691f,
+    0.434154f, 0.439657f, 0.445201f, 0.450786f, 0.456411f, 0.462077f, 0.467784f, 0.473532f, 0.479320f, 0.485150f, 0.491021f,
+    0.496933f, 0.502887f, 0.508881f, 0.514918f, 0.520996f, 0.527115f, 0.533276f, 0.539480f, 0.545725f, 0.552011f, 0.558340f,
+    0.564712f, 0.571125f, 0.577581f, 0.584078f, 0.590619f, 0.597202f, 0.603827f, 0.610496f, 0.617207f, 0.623960f, 0.630757f,
+    0.637597f, 0.644480f, 0.651406f, 0.658375f, 0.665387f, 0.672443f, 0.679543f, 0.686685f, 0.693872f, 0.701102f, 0.708376f,
+    0.715694f, 0.723055f, 0.730461f, 0.737911f, 0.745404f, 0.752942f, 0.760525f, 0.768151f, 0.775822f, 0.783538f, 0.791298f,
+    0.799103f, 0.806952f, 0.814847f, 0.822786f, 0.830770f, 0.838799f, 0.846873f, 0.854993f, 0.863157f, 0.871367f, 0.879622f,
+    0.887923f, 0.896269f, 0.904661f, 0.913099f, 0.921582f, 0.930111f, 0.938686f, 0.947307f, 0.955974f, 0.964686f, 0.973445f,
+    0.982251f, 0.991102f, 1.0f
+};
+
+static float stbir__srgb_to_linear(float f)
+{
+    if (f <= 0.04045f)
+        return f / 12.92f;
+    else
+        return (float)pow((f + 0.055f) / 1.055f, 2.4f);
+}
+
+static float stbir__linear_to_srgb(float f)
+{
+    if (f <= 0.0031308f)
+        return f * 12.92f;
+    else
+        return 1.055f * (float)pow(f, 1 / 2.4f) - 0.055f;
+}
+
+#ifndef STBIR_NON_IEEE_FLOAT
+// From https://gist.github.com/rygorous/2203834
+
+typedef union
+{
+    stbir_uint32 u;
+    float f;
+} stbir__FP32;
+
+static const stbir_uint32 fp32_to_srgb8_tab4[104] = {
+    0x0073000d, 0x007a000d, 0x0080000d, 0x0087000d, 0x008d000d, 0x0094000d, 0x009a000d, 0x00a1000d,
+    0x00a7001a, 0x00b4001a, 0x00c1001a, 0x00ce001a, 0x00da001a, 0x00e7001a, 0x00f4001a, 0x0101001a,
+    0x010e0033, 0x01280033, 0x01410033, 0x015b0033, 0x01750033, 0x018f0033, 0x01a80033, 0x01c20033,
+    0x01dc0067, 0x020f0067, 0x02430067, 0x02760067, 0x02aa0067, 0x02dd0067, 0x03110067, 0x03440067,
+    0x037800ce, 0x03df00ce, 0x044600ce, 0x04ad00ce, 0x051400ce, 0x057b00c5, 0x05dd00bc, 0x063b00b5,
+    0x06970158, 0x07420142, 0x07e30130, 0x087b0120, 0x090b0112, 0x09940106, 0x0a1700fc, 0x0a9500f2,
+    0x0b0f01cb, 0x0bf401ae, 0x0ccb0195, 0x0d950180, 0x0e56016e, 0x0f0d015e, 0x0fbc0150, 0x10630143,
+    0x11070264, 0x1238023e, 0x1357021d, 0x14660201, 0x156601e9, 0x165a01d3, 0x174401c0, 0x182401af,
+    0x18fe0331, 0x1a9602fe, 0x1c1502d2, 0x1d7e02ad, 0x1ed4028d, 0x201a0270, 0x21520256, 0x227d0240,
+    0x239f0443, 0x25c003fe, 0x27bf03c4, 0x29a10392, 0x2b6a0367, 0x2d1d0341, 0x2ebe031f, 0x304d0300,
+    0x31d105b0, 0x34a80555, 0x37520507, 0x39d504c5, 0x3c37048b, 0x3e7c0458, 0x40a8042a, 0x42bd0401,
+    0x44c20798, 0x488e071e, 0x4c1c06b6, 0x4f76065d, 0x52a50610, 0x55ac05cc, 0x5892058f, 0x5b590559,
+    0x5e0c0a23, 0x631c0980, 0x67db08f6, 0x6c55087f, 0x70940818, 0x74a007bd, 0x787d076c, 0x7c330723,
+};
+
+static stbir_uint8 stbir__linear_to_srgb_uchar(float in)
+{
+    static const stbir__FP32 almostone = { 0x3f7fffff }; // 1-eps
+    static const stbir__FP32 minval = { (127-13) << 23 };
+    stbir_uint32 tab,bias,scale,t;
+    stbir__FP32 f;
+
+    // Clamp to [2^(-13), 1-eps]; these two values map to 0 and 1, respectively.
+    // The tests are carefully written so that NaNs map to 0, same as in the reference
+    // implementation.
+    if (!(in > minval.f)) // written this way to catch NaNs
+        in = minval.f;
+    if (in > almostone.f)
+        in = almostone.f;
+
+    // Do the table lookup and unpack bias, scale
+    f.f = in;
+    tab = fp32_to_srgb8_tab4[(f.u - minval.u) >> 20];
+    bias = (tab >> 16) << 9;
+    scale = tab & 0xffff;
+
+    // Grab next-highest mantissa bits and perform linear interpolation
+    t = (f.u >> 12) & 0xff;
+    return (unsigned char) ((bias + scale*t) >> 16);
+}
+
+#else
+// sRGB transition values, scaled by 1<<28
+static int stbir__srgb_offset_to_linear_scaled[256] =
+{
+            0,     40738,    122216,    203693,    285170,    366648,    448125,    529603,
+       611080,    692557,    774035,    855852,    942009,   1033024,   1128971,   1229926,
+      1335959,   1447142,   1563542,   1685229,   1812268,   1944725,   2082664,   2226148,
+      2375238,   2529996,   2690481,   2856753,   3028870,   3206888,   3390865,   3580856,
+      3776916,   3979100,   4187460,   4402049,   4622919,   4850123,   5083710,   5323731,
+      5570236,   5823273,   6082892,   6349140,   6622065,   6901714,   7188133,   7481369,
+      7781466,   8088471,   8402427,   8723380,   9051372,   9386448,   9728650,  10078021,
+     10434603,  10798439,  11169569,  11548036,  11933879,  12327139,  12727857,  13136073,
+     13551826,  13975156,  14406100,  14844697,  15290987,  15745007,  16206795,  16676389,
+     17153826,  17639142,  18132374,  18633560,  19142734,  19659934,  20185196,  20718552,
+     21260042,  21809696,  22367554,  22933648,  23508010,  24090680,  24681686,  25281066,
+     25888850,  26505076,  27129772,  27762974,  28404716,  29055026,  29713942,  30381490,
+     31057708,  31742624,  32436272,  33138682,  33849884,  34569912,  35298800,  36036568,
+     36783260,  37538896,  38303512,  39077136,  39859796,  40651528,  41452360,  42262316,
+     43081432,  43909732,  44747252,  45594016,  46450052,  47315392,  48190064,  49074096,
+     49967516,  50870356,  51782636,  52704392,  53635648,  54576432,  55526772,  56486700,
+     57456236,  58435408,  59424248,  60422780,  61431036,  62449032,  63476804,  64514376,
+     65561776,  66619028,  67686160,  68763192,  69850160,  70947088,  72053992,  73170912,
+     74297864,  75434880,  76581976,  77739184,  78906536,  80084040,  81271736,  82469648,
+     83677792,  84896192,  86124888,  87363888,  88613232,  89872928,  91143016,  92423512,
+     93714432,  95015816,  96327688,  97650056,  98982952, 100326408, 101680440, 103045072,
+    104420320, 105806224, 107202800, 108610064, 110028048, 111456776, 112896264, 114346544,
+    115807632, 117279552, 118762328, 120255976, 121760536, 123276016, 124802440, 126339832,
+    127888216, 129447616, 131018048, 132599544, 134192112, 135795792, 137410592, 139036528,
+    140673648, 142321952, 143981456, 145652208, 147334208, 149027488, 150732064, 152447968,
+    154175200, 155913792, 157663776, 159425168, 161197984, 162982240, 164777968, 166585184,
+    168403904, 170234160, 172075968, 173929344, 175794320, 177670896, 179559120, 181458992,
+    183370528, 185293776, 187228736, 189175424, 191133888, 193104112, 195086128, 197079968,
+    199085648, 201103184, 203132592, 205173888, 207227120, 209292272, 211369392, 213458480,
+    215559568, 217672656, 219797792, 221934976, 224084240, 226245600, 228419056, 230604656,
+    232802400, 235012320, 237234432, 239468736, 241715280, 243974080, 246245120, 248528464,
+    250824112, 253132064, 255452368, 257785040, 260130080, 262487520, 264857376, 267239664,
+};
+
+static stbir_uint8 stbir__linear_to_srgb_uchar(float f)
+{
+    int x = (int) (f * (1 << 28)); // has headroom so you don't need to clamp
+    int v = 0;
+    int i;
+
+    // Refine the guess with a short binary search.
+    i = v + 128; if (x >= stbir__srgb_offset_to_linear_scaled[i]) v = i;
+    i = v +  64; if (x >= stbir__srgb_offset_to_linear_scaled[i]) v = i;
+    i = v +  32; if (x >= stbir__srgb_offset_to_linear_scaled[i]) v = i;
+    i = v +  16; if (x >= stbir__srgb_offset_to_linear_scaled[i]) v = i;
+    i = v +   8; if (x >= stbir__srgb_offset_to_linear_scaled[i]) v = i;
+    i = v +   4; if (x >= stbir__srgb_offset_to_linear_scaled[i]) v = i;
+    i = v +   2; if (x >= stbir__srgb_offset_to_linear_scaled[i]) v = i;
+    i = v +   1; if (x >= stbir__srgb_offset_to_linear_scaled[i]) v = i;
+
+    return (stbir_uint8) v;
+}
+#endif
+
+static float stbir__filter_trapezoid(float x, float scale)
+{
+    float halfscale = scale / 2;
+    float t = 0.5f + halfscale;
+    STBIR_ASSERT(scale <= 1);
+
+    x = (float)fabs(x);
+
+    if (x >= t)
+        return 0;
+    else
+    {
+        float r = 0.5f - halfscale;
+        if (x <= r)
+            return 1;
+        else
+            return (t - x) / scale;
+    }
+}
+
+static float stbir__support_trapezoid(float scale)
+{
+    STBIR_ASSERT(scale <= 1);
+    return 0.5f + scale / 2;
+}
+
+static float stbir__filter_triangle(float x, float s)
+{
+    STBIR__UNUSED_PARAM(s);
+
+    x = (float)fabs(x);
+
+    if (x <= 1.0f)
+        return 1 - x;
+    else
+        return 0;
+}
+
+static float stbir__filter_cubic(float x, float s)
+{
+    STBIR__UNUSED_PARAM(s);
+
+    x = (float)fabs(x);
+
+    if (x < 1.0f)
+        return (4 + x*x*(3*x - 6))/6;
+    else if (x < 2.0f)
+        return (8 + x*(-12 + x*(6 - x)))/6;
+
+    return (0.0f);
+}
+
+static float stbir__filter_catmullrom(float x, float s)
+{
+    STBIR__UNUSED_PARAM(s);
+
+    x = (float)fabs(x);
+
+    if (x < 1.0f)
+        return 1 - x*x*(2.5f - 1.5f*x);
+    else if (x < 2.0f)
+        return 2 - x*(4 + x*(0.5f*x - 2.5f));
+
+    return (0.0f);
+}
+
+static float stbir__filter_mitchell(float x, float s)
+{
+    STBIR__UNUSED_PARAM(s);
+
+    x = (float)fabs(x);
+
+    if (x < 1.0f)
+        return (16 + x*x*(21 * x - 36))/18;
+    else if (x < 2.0f)
+        return (32 + x*(-60 + x*(36 - 7*x)))/18;
+
+    return (0.0f);
+}
+
+static float stbir__support_zero(float s)
+{
+    STBIR__UNUSED_PARAM(s);
+    return 0;
+}
+
+static float stbir__support_one(float s)
+{
+    STBIR__UNUSED_PARAM(s);
+    return 1;
+}
+
+static float stbir__support_two(float s)
+{
+    STBIR__UNUSED_PARAM(s);
+    return 2;
+}
+
+static stbir__filter_info stbir__filter_info_table[] = {
+        { NULL,                     stbir__support_zero },
+        { stbir__filter_trapezoid,  stbir__support_trapezoid },
+        { stbir__filter_triangle,   stbir__support_one },
+        { stbir__filter_cubic,      stbir__support_two },
+        { stbir__filter_catmullrom, stbir__support_two },
+        { stbir__filter_mitchell,   stbir__support_two },
+};
+
+stbir__inline static int stbir__use_upsampling(float ratio)
+{
+    return ratio > 1;
+}
+
+stbir__inline static int stbir__use_width_upsampling(stbir__info* stbir_info)
+{
+    return stbir__use_upsampling(stbir_info->horizontal_scale);
+}
+
+stbir__inline static int stbir__use_height_upsampling(stbir__info* stbir_info)
+{
+    return stbir__use_upsampling(stbir_info->vertical_scale);
+}
+
+// This is the maximum number of input samples that can affect an output sample
+// with the given filter
+static int stbir__get_filter_pixel_width(stbir_filter filter, float scale)
+{
+    STBIR_ASSERT(filter != 0);
+    STBIR_ASSERT(filter < STBIR__ARRAY_SIZE(stbir__filter_info_table));
+
+    if (stbir__use_upsampling(scale))
+        return (int)ceil(stbir__filter_info_table[filter].support(1/scale) * 2);
+    else
+        return (int)ceil(stbir__filter_info_table[filter].support(scale) * 2 / scale);
+}
+
+// This is how much to expand buffers to account for filters seeking outside
+// the image boundaries.
+static int stbir__get_filter_pixel_margin(stbir_filter filter, float scale)
+{
+    return stbir__get_filter_pixel_width(filter, scale) / 2;
+}
+
+static int stbir__get_coefficient_width(stbir_filter filter, float scale)
+{
+    if (stbir__use_upsampling(scale))
+        return (int)ceil(stbir__filter_info_table[filter].support(1 / scale) * 2);
+    else
+        return (int)ceil(stbir__filter_info_table[filter].support(scale) * 2);
+}
+
+static int stbir__get_contributors(float scale, stbir_filter filter, int input_size, int output_size)
+{
+    if (stbir__use_upsampling(scale))
+        return output_size;
+    else
+        return (input_size + stbir__get_filter_pixel_margin(filter, scale) * 2);
+}
+
+static int stbir__get_total_horizontal_coefficients(stbir__info* info)
+{
+    return info->horizontal_num_contributors
+         * stbir__get_coefficient_width      (info->horizontal_filter, info->horizontal_scale);
+}
+
+static int stbir__get_total_vertical_coefficients(stbir__info* info)
+{
+    return info->vertical_num_contributors
+         * stbir__get_coefficient_width      (info->vertical_filter, info->vertical_scale);
+}
+
+static stbir__contributors* stbir__get_contributor(stbir__contributors* contributors, int n)
+{
+    return &contributors[n];
+}
+
+// For perf reasons this code is duplicated in stbir__resample_horizontal_upsample/downsample,
+// if you change it here change it there too.
+static float* stbir__get_coefficient(float* coefficients, stbir_filter filter, float scale, int n, int c)
+{
+    int width = stbir__get_coefficient_width(filter, scale);
+    return &coefficients[width*n + c];
+}
+
+static int stbir__edge_wrap_slow(stbir_edge edge, int n, int max)
+{
+    switch (edge)
+    {
+    case STBIR_EDGE_ZERO:
+        return 0; // we'll decode the wrong pixel here, and then overwrite with 0s later
+
+    case STBIR_EDGE_CLAMP:
+        if (n < 0)
+            return 0;
+
+        if (n >= max)
+            return max - 1;
+
+        return n; // NOTREACHED
+
+    case STBIR_EDGE_REFLECT:
+    {
+        if (n < 0)
+        {
+            if (n < max)
+                return -n;
+            else
+                return max - 1;
+        }
+
+        if (n >= max)
+        {
+            int max2 = max * 2;
+            if (n >= max2)
+                return 0;
+            else
+                return max2 - n - 1;
+        }
+
+        return n; // NOTREACHED
+    }
+
+    case STBIR_EDGE_WRAP:
+        if (n >= 0)
+            return (n % max);
+        else
+        {
+            int m = (-n) % max;
+
+            if (m != 0)
+                m = max - m;
+
+            return (m);
+        }
+        // NOTREACHED
+
+    default:
+        STBIR_ASSERT(!"Unimplemented edge type");
+        return 0;
+    }
+}
+
+stbir__inline static int stbir__edge_wrap(stbir_edge edge, int n, int max)
+{
+    // avoid per-pixel switch
+    if (n >= 0 && n < max)
+        return n;
+    return stbir__edge_wrap_slow(edge, n, max);
+}
+
+// What input pixels contribute to this output pixel?
+static void stbir__calculate_sample_range_upsample(int n, float out_filter_radius, float scale_ratio, float out_shift, int* in_first_pixel, int* in_last_pixel, float* in_center_of_out)
+{
+    float out_pixel_center = (float)n + 0.5f;
+    float out_pixel_influence_lowerbound = out_pixel_center - out_filter_radius;
+    float out_pixel_influence_upperbound = out_pixel_center + out_filter_radius;
+
+    float in_pixel_influence_lowerbound = (out_pixel_influence_lowerbound + out_shift) / scale_ratio;
+    float in_pixel_influence_upperbound = (out_pixel_influence_upperbound + out_shift) / scale_ratio;
+
+    *in_center_of_out = (out_pixel_center + out_shift) / scale_ratio;
+    *in_first_pixel = (int)(floor(in_pixel_influence_lowerbound + 0.5));
+    *in_last_pixel = (int)(floor(in_pixel_influence_upperbound - 0.5));
+}
+
+// What output pixels does this input pixel contribute to?
+static void stbir__calculate_sample_range_downsample(int n, float in_pixels_radius, float scale_ratio, float out_shift, int* out_first_pixel, int* out_last_pixel, float* out_center_of_in)
+{
+    float in_pixel_center = (float)n + 0.5f;
+    float in_pixel_influence_lowerbound = in_pixel_center - in_pixels_radius;
+    float in_pixel_influence_upperbound = in_pixel_center + in_pixels_radius;
+
+    float out_pixel_influence_lowerbound = in_pixel_influence_lowerbound * scale_ratio - out_shift;
+    float out_pixel_influence_upperbound = in_pixel_influence_upperbound * scale_ratio - out_shift;
+
+    *out_center_of_in = in_pixel_center * scale_ratio - out_shift;
+    *out_first_pixel = (int)(floor(out_pixel_influence_lowerbound + 0.5));
+    *out_last_pixel = (int)(floor(out_pixel_influence_upperbound - 0.5));
+}
+
+static void stbir__calculate_coefficients_upsample(stbir_filter filter, float scale, int in_first_pixel, int in_last_pixel, float in_center_of_out, stbir__contributors* contributor, float* coefficient_group)
+{
+    int i;
+    float total_filter = 0;
+    float filter_scale;
+
+    STBIR_ASSERT(in_last_pixel - in_first_pixel <= (int)ceil(stbir__filter_info_table[filter].support(1/scale) * 2)); // Taken directly from stbir__get_coefficient_width() which we can't call because we don't know if we're horizontal or vertical.
+
+    contributor->n0 = in_first_pixel;
+    contributor->n1 = in_last_pixel;
+
+    STBIR_ASSERT(contributor->n1 >= contributor->n0);
+
+    for (i = 0; i <= in_last_pixel - in_first_pixel; i++)
+    {
+        float in_pixel_center = (float)(i + in_first_pixel) + 0.5f;
+        coefficient_group[i] = stbir__filter_info_table[filter].kernel(in_center_of_out - in_pixel_center, 1 / scale);
+
+        // If the coefficient is zero, skip it. (Don't do the <0 check here, we want the influence of those outside pixels.)
+        if (i == 0 && !coefficient_group[i])
+        {
+            contributor->n0 = ++in_first_pixel;
+            i--;
+            continue;
+        }
+
+        total_filter += coefficient_group[i];
+    }
+
+    // NOTE(fg): Not actually true in general, nor is there any reason to expect it should be.
+    // It would be true in exact math but is at best approximately true in floating-point math,
+    // and it would not make sense to try and put actual bounds on this here because it depends
+    // on the image aspect ratio which can get pretty extreme.
+    //STBIR_ASSERT(stbir__filter_info_table[filter].kernel((float)(in_last_pixel + 1) + 0.5f - in_center_of_out, 1/scale) == 0);
+
+    STBIR_ASSERT(total_filter > 0.9);
+    STBIR_ASSERT(total_filter < 1.1f); // Make sure it's not way off.
+
+    // Make sure the sum of all coefficients is 1.
+    filter_scale = 1 / total_filter;
+
+    for (i = 0; i <= in_last_pixel - in_first_pixel; i++)
+        coefficient_group[i] *= filter_scale;
+
+    for (i = in_last_pixel - in_first_pixel; i >= 0; i--)
+    {
+        if (coefficient_group[i])
+            break;
+
+        // This line has no weight. We can skip it.
+        contributor->n1 = contributor->n0 + i - 1;
+    }
+}
+
+static void stbir__calculate_coefficients_downsample(stbir_filter filter, float scale_ratio, int out_first_pixel, int out_last_pixel, float out_center_of_in, stbir__contributors* contributor, float* coefficient_group)
+{
+    int i;
+
+    STBIR_ASSERT(out_last_pixel - out_first_pixel <= (int)ceil(stbir__filter_info_table[filter].support(scale_ratio) * 2)); // Taken directly from stbir__get_coefficient_width() which we can't call because we don't know if we're horizontal or vertical.
+
+    contributor->n0 = out_first_pixel;
+    contributor->n1 = out_last_pixel;
+
+    STBIR_ASSERT(contributor->n1 >= contributor->n0);
+
+    for (i = 0; i <= out_last_pixel - out_first_pixel; i++)
+    {
+        float out_pixel_center = (float)(i + out_first_pixel) + 0.5f;
+        float x = out_pixel_center - out_center_of_in;
+        coefficient_group[i] = stbir__filter_info_table[filter].kernel(x, scale_ratio) * scale_ratio;
+    }
+
+    // NOTE(fg): Not actually true in general, nor is there any reason to expect it should be.
+    // It would be true in exact math but is at best approximately true in floating-point math,
+    // and it would not make sense to try and put actual bounds on this here because it depends
+    // on the image aspect ratio which can get pretty extreme.
+    //STBIR_ASSERT(stbir__filter_info_table[filter].kernel((float)(out_last_pixel + 1) + 0.5f - out_center_of_in, scale_ratio) == 0);
+
+    for (i = out_last_pixel - out_first_pixel; i >= 0; i--)
+    {
+        if (coefficient_group[i])
+            break;
+
+        // This line has no weight. We can skip it.
+        contributor->n1 = contributor->n0 + i - 1;
+    }
+}
+
+static void stbir__normalize_downsample_coefficients(stbir__contributors* contributors, float* coefficients, stbir_filter filter, float scale_ratio, int input_size, int output_size)
+{
+    int num_contributors = stbir__get_contributors(scale_ratio, filter, input_size, output_size);
+    int num_coefficients = stbir__get_coefficient_width(filter, scale_ratio);
+    int i, j;
+    int skip;
+
+    for (i = 0; i < output_size; i++)
+    {
+        float scale;
+        float total = 0;
+
+        for (j = 0; j < num_contributors; j++)
+        {
+            if (i >= contributors[j].n0 && i <= contributors[j].n1)
+            {
+                float coefficient = *stbir__get_coefficient(coefficients, filter, scale_ratio, j, i - contributors[j].n0);
+                total += coefficient;
+            }
+            else if (i < contributors[j].n0)
+                break;
+        }
+
+        STBIR_ASSERT(total > 0.9f);
+        STBIR_ASSERT(total < 1.1f);
+
+        scale = 1 / total;
+
+        for (j = 0; j < num_contributors; j++)
+        {
+            if (i >= contributors[j].n0 && i <= contributors[j].n1)
+                *stbir__get_coefficient(coefficients, filter, scale_ratio, j, i - contributors[j].n0) *= scale;
+            else if (i < contributors[j].n0)
+                break;
+        }
+    }
+
+    // Optimize: Skip zero coefficients and contributions outside of image bounds.
+    // Do this after normalizing because normalization depends on the n0/n1 values.
+    for (j = 0; j < num_contributors; j++)
+    {
+        int range, max, width;
+
+        skip = 0;
+        while (*stbir__get_coefficient(coefficients, filter, scale_ratio, j, skip) == 0)
+            skip++;
+
+        contributors[j].n0 += skip;
+
+        while (contributors[j].n0 < 0)
+        {
+            contributors[j].n0++;
+            skip++;
+        }
+
+        range = contributors[j].n1 - contributors[j].n0 + 1;
+        max = stbir__min(num_coefficients, range);
+
+        width = stbir__get_coefficient_width(filter, scale_ratio);
+        for (i = 0; i < max; i++)
+        {
+            if (i + skip >= width)
+                break;
+
+            *stbir__get_coefficient(coefficients, filter, scale_ratio, j, i) = *stbir__get_coefficient(coefficients, filter, scale_ratio, j, i + skip);
+        }
+
+        continue;
+    }
+
+    // Using min to avoid writing into invalid pixels.
+    for (i = 0; i < num_contributors; i++)
+        contributors[i].n1 = stbir__min(contributors[i].n1, output_size - 1);
+}
+
+// Each scan line uses the same kernel values so we should calculate the kernel
+// values once and then we can use them for every scan line.
+static void stbir__calculate_filters(stbir__contributors* contributors, float* coefficients, stbir_filter filter, float scale_ratio, float shift, int input_size, int output_size)
+{
+    int n;
+    int total_contributors = stbir__get_contributors(scale_ratio, filter, input_size, output_size);
+
+    if (stbir__use_upsampling(scale_ratio))
+    {
+        float out_pixels_radius = stbir__filter_info_table[filter].support(1 / scale_ratio) * scale_ratio;
+
+        // Looping through out pixels
+        for (n = 0; n < total_contributors; n++)
+        {
+            float in_center_of_out; // Center of the current out pixel in the in pixel space
+            int in_first_pixel, in_last_pixel;
+
+            stbir__calculate_sample_range_upsample(n, out_pixels_radius, scale_ratio, shift, &in_first_pixel, &in_last_pixel, &in_center_of_out);
+
+            stbir__calculate_coefficients_upsample(filter, scale_ratio, in_first_pixel, in_last_pixel, in_center_of_out, stbir__get_contributor(contributors, n), stbir__get_coefficient(coefficients, filter, scale_ratio, n, 0));
+        }
+    }
+    else
+    {
+        float in_pixels_radius = stbir__filter_info_table[filter].support(scale_ratio) / scale_ratio;
+
+        // Looping through in pixels
+        for (n = 0; n < total_contributors; n++)
+        {
+            float out_center_of_in; // Center of the current out pixel in the in pixel space
+            int out_first_pixel, out_last_pixel;
+            int n_adjusted = n - stbir__get_filter_pixel_margin(filter, scale_ratio);
+
+            stbir__calculate_sample_range_downsample(n_adjusted, in_pixels_radius, scale_ratio, shift, &out_first_pixel, &out_last_pixel, &out_center_of_in);
+
+            stbir__calculate_coefficients_downsample(filter, scale_ratio, out_first_pixel, out_last_pixel, out_center_of_in, stbir__get_contributor(contributors, n), stbir__get_coefficient(coefficients, filter, scale_ratio, n, 0));
+        }
+
+        stbir__normalize_downsample_coefficients(contributors, coefficients, filter, scale_ratio, input_size, output_size);
+    }
+}
+
+static float* stbir__get_decode_buffer(stbir__info* stbir_info)
+{
+    // The 0 index of the decode buffer starts after the margin. This makes
+    // it okay to use negative indexes on the decode buffer.
+    return &stbir_info->decode_buffer[stbir_info->horizontal_filter_pixel_margin * stbir_info->channels];
+}
+
+#define STBIR__DECODE(type, colorspace) ((int)(type) * (STBIR_MAX_COLORSPACES) + (int)(colorspace))
+
+static void stbir__decode_scanline(stbir__info* stbir_info, int n)
+{
+    int c;
+    int channels = stbir_info->channels;
+    int alpha_channel = stbir_info->alpha_channel;
+    int type = stbir_info->type;
+    int colorspace = stbir_info->colorspace;
+    int input_w = stbir_info->input_w;
+    size_t input_stride_bytes = stbir_info->input_stride_bytes;
+    float* decode_buffer = stbir__get_decode_buffer(stbir_info);
+    stbir_edge edge_horizontal = stbir_info->edge_horizontal;
+    stbir_edge edge_vertical = stbir_info->edge_vertical;
+    size_t in_buffer_row_offset = stbir__edge_wrap(edge_vertical, n, stbir_info->input_h) * input_stride_bytes;
+    const void* input_data = (char *) stbir_info->input_data + in_buffer_row_offset;
+    int max_x = input_w + stbir_info->horizontal_filter_pixel_margin;
+    int decode = STBIR__DECODE(type, colorspace);
+
+    int x = -stbir_info->horizontal_filter_pixel_margin;
+
+    // special handling for STBIR_EDGE_ZERO because it needs to return an item that doesn't appear in the input,
+    // and we want to avoid paying overhead on every pixel if not STBIR_EDGE_ZERO
+    if (edge_vertical == STBIR_EDGE_ZERO && (n < 0 || n >= stbir_info->input_h))
+    {
+        for (; x < max_x; x++)
+            for (c = 0; c < channels; c++)
+                decode_buffer[x*channels + c] = 0;
+        return;
+    }
+
+    switch (decode)
+    {
+    case STBIR__DECODE(STBIR_TYPE_UINT8, STBIR_COLORSPACE_LINEAR):
+        for (; x < max_x; x++)
+        {
+            int decode_pixel_index = x * channels;
+            int input_pixel_index = stbir__edge_wrap(edge_horizontal, x, input_w) * channels;
+            for (c = 0; c < channels; c++)
+                decode_buffer[decode_pixel_index + c] = ((float)((const unsigned char*)input_data)[input_pixel_index + c]) / stbir__max_uint8_as_float;
+        }
+        break;
+
+    case STBIR__DECODE(STBIR_TYPE_UINT8, STBIR_COLORSPACE_SRGB):
+        for (; x < max_x; x++)
+        {
+            int decode_pixel_index = x * channels;
+            int input_pixel_index = stbir__edge_wrap(edge_horizontal, x, input_w) * channels;
+            for (c = 0; c < channels; c++)
+                decode_buffer[decode_pixel_index + c] = stbir__srgb_uchar_to_linear_float[((const unsigned char*)input_data)[input_pixel_index + c]];
+
+            if (!(stbir_info->flags&STBIR_FLAG_ALPHA_USES_COLORSPACE))
+                decode_buffer[decode_pixel_index + alpha_channel] = ((float)((const unsigned char*)input_data)[input_pixel_index + alpha_channel]) / stbir__max_uint8_as_float;
+        }
+        break;
+
+    case STBIR__DECODE(STBIR_TYPE_UINT16, STBIR_COLORSPACE_LINEAR):
+        for (; x < max_x; x++)
+        {
+            int decode_pixel_index = x * channels;
+            int input_pixel_index = stbir__edge_wrap(edge_horizontal, x, input_w) * channels;
+            for (c = 0; c < channels; c++)
+                decode_buffer[decode_pixel_index + c] = ((float)((const unsigned short*)input_data)[input_pixel_index + c]) / stbir__max_uint16_as_float;
+        }
+        break;
+
+    case STBIR__DECODE(STBIR_TYPE_UINT16, STBIR_COLORSPACE_SRGB):
+        for (; x < max_x; x++)
+        {
+            int decode_pixel_index = x * channels;
+            int input_pixel_index = stbir__edge_wrap(edge_horizontal, x, input_w) * channels;
+            for (c = 0; c < channels; c++)
+                decode_buffer[decode_pixel_index + c] = stbir__srgb_to_linear(((float)((const unsigned short*)input_data)[input_pixel_index + c]) / stbir__max_uint16_as_float);
+
+            if (!(stbir_info->flags&STBIR_FLAG_ALPHA_USES_COLORSPACE))
+                decode_buffer[decode_pixel_index + alpha_channel] = ((float)((const unsigned short*)input_data)[input_pixel_index + alpha_channel]) / stbir__max_uint16_as_float;
+        }
+        break;
+
+    case STBIR__DECODE(STBIR_TYPE_UINT32, STBIR_COLORSPACE_LINEAR):
+        for (; x < max_x; x++)
+        {
+            int decode_pixel_index = x * channels;
+            int input_pixel_index = stbir__edge_wrap(edge_horizontal, x, input_w) * channels;
+            for (c = 0; c < channels; c++)
+                decode_buffer[decode_pixel_index + c] = (float)(((double)((const unsigned int*)input_data)[input_pixel_index + c]) / stbir__max_uint32_as_float);
+        }
+        break;
+
+    case STBIR__DECODE(STBIR_TYPE_UINT32, STBIR_COLORSPACE_SRGB):
+        for (; x < max_x; x++)
+        {
+            int decode_pixel_index = x * channels;
+            int input_pixel_index = stbir__edge_wrap(edge_horizontal, x, input_w) * channels;
+            for (c = 0; c < channels; c++)
+                decode_buffer[decode_pixel_index + c] = stbir__srgb_to_linear((float)(((double)((const unsigned int*)input_data)[input_pixel_index + c]) / stbir__max_uint32_as_float));
+
+            if (!(stbir_info->flags&STBIR_FLAG_ALPHA_USES_COLORSPACE))
+                decode_buffer[decode_pixel_index + alpha_channel] = (float)(((double)((const unsigned int*)input_data)[input_pixel_index + alpha_channel]) / stbir__max_uint32_as_float);
+        }
+        break;
+
+    case STBIR__DECODE(STBIR_TYPE_FLOAT, STBIR_COLORSPACE_LINEAR):
+        for (; x < max_x; x++)
+        {
+            int decode_pixel_index = x * channels;
+            int input_pixel_index = stbir__edge_wrap(edge_horizontal, x, input_w) * channels;
+            for (c = 0; c < channels; c++)
+                decode_buffer[decode_pixel_index + c] = ((const float*)input_data)[input_pixel_index + c];
+        }
+        break;
+
+    case STBIR__DECODE(STBIR_TYPE_FLOAT, STBIR_COLORSPACE_SRGB):
+        for (; x < max_x; x++)
+        {
+            int decode_pixel_index = x * channels;
+            int input_pixel_index = stbir__edge_wrap(edge_horizontal, x, input_w) * channels;
+            for (c = 0; c < channels; c++)
+                decode_buffer[decode_pixel_index + c] = stbir__srgb_to_linear(((const float*)input_data)[input_pixel_index + c]);
+
+            if (!(stbir_info->flags&STBIR_FLAG_ALPHA_USES_COLORSPACE))
+                decode_buffer[decode_pixel_index + alpha_channel] = ((const float*)input_data)[input_pixel_index + alpha_channel];
+        }
+
+        break;
+
+    default:
+        STBIR_ASSERT(!"Unknown type/colorspace/channels combination.");
+        break;
+    }
+
+    if (!(stbir_info->flags & STBIR_FLAG_ALPHA_PREMULTIPLIED))
+    {
+        for (x = -stbir_info->horizontal_filter_pixel_margin; x < max_x; x++)
+        {
+            int decode_pixel_index = x * channels;
+
+            // If the alpha value is 0 it will clobber the color values. Make sure it's not.
+            float alpha = decode_buffer[decode_pixel_index + alpha_channel];
+#ifndef STBIR_NO_ALPHA_EPSILON
+            if (stbir_info->type != STBIR_TYPE_FLOAT) {
+                alpha += STBIR_ALPHA_EPSILON;
+                decode_buffer[decode_pixel_index + alpha_channel] = alpha;
+            }
+#endif
+            for (c = 0; c < channels; c++)
+            {
+                if (c == alpha_channel)
+                    continue;
+
+                decode_buffer[decode_pixel_index + c] *= alpha;
+            }
+        }
+    }
+
+    if (edge_horizontal == STBIR_EDGE_ZERO)
+    {
+        for (x = -stbir_info->horizontal_filter_pixel_margin; x < 0; x++)
+        {
+            for (c = 0; c < channels; c++)
+                decode_buffer[x*channels + c] = 0;
+        }
+        for (x = input_w; x < max_x; x++)
+        {
+            for (c = 0; c < channels; c++)
+                decode_buffer[x*channels + c] = 0;
+        }
+    }
+}
+
+static float* stbir__get_ring_buffer_entry(float* ring_buffer, int index, int ring_buffer_length)
+{
+    return &ring_buffer[index * ring_buffer_length];
+}
+
+static float* stbir__add_empty_ring_buffer_entry(stbir__info* stbir_info, int n)
+{
+    int ring_buffer_index;
+    float* ring_buffer;
+
+    stbir_info->ring_buffer_last_scanline = n;
+
+    if (stbir_info->ring_buffer_begin_index < 0)
+    {
+        ring_buffer_index = stbir_info->ring_buffer_begin_index = 0;
+        stbir_info->ring_buffer_first_scanline = n;
+    }
+    else
+    {
+        ring_buffer_index = (stbir_info->ring_buffer_begin_index + (stbir_info->ring_buffer_last_scanline - stbir_info->ring_buffer_first_scanline)) % stbir_info->ring_buffer_num_entries;
+        STBIR_ASSERT(ring_buffer_index != stbir_info->ring_buffer_begin_index);
+    }
+
+    ring_buffer = stbir__get_ring_buffer_entry(stbir_info->ring_buffer, ring_buffer_index, stbir_info->ring_buffer_length_bytes / sizeof(float));
+    memset(ring_buffer, 0, stbir_info->ring_buffer_length_bytes);
+
+    return ring_buffer;
+}
+
+
+static void stbir__resample_horizontal_upsample(stbir__info* stbir_info, float* output_buffer)
+{
+    int x, k;
+    int output_w = stbir_info->output_w;
+    int channels = stbir_info->channels;
+    float* decode_buffer = stbir__get_decode_buffer(stbir_info);
+    stbir__contributors* horizontal_contributors = stbir_info->horizontal_contributors;
+    float* horizontal_coefficients = stbir_info->horizontal_coefficients;
+    int coefficient_width = stbir_info->horizontal_coefficient_width;
+
+    for (x = 0; x < output_w; x++)
+    {
+        int n0 = horizontal_contributors[x].n0;
+        int n1 = horizontal_contributors[x].n1;
+
+        int out_pixel_index = x * channels;
+        int coefficient_group = coefficient_width * x;
+        int coefficient_counter = 0;
+
+        STBIR_ASSERT(n1 >= n0);
+        STBIR_ASSERT(n0 >= -stbir_info->horizontal_filter_pixel_margin);
+        STBIR_ASSERT(n1 >= -stbir_info->horizontal_filter_pixel_margin);
+        STBIR_ASSERT(n0 < stbir_info->input_w + stbir_info->horizontal_filter_pixel_margin);
+        STBIR_ASSERT(n1 < stbir_info->input_w + stbir_info->horizontal_filter_pixel_margin);
+
+        switch (channels) {
+            case 1:
+                for (k = n0; k <= n1; k++)
+                {
+                    int in_pixel_index = k * 1;
+                    float coefficient = horizontal_coefficients[coefficient_group + coefficient_counter++];
+                    STBIR_ASSERT(coefficient != 0);
+                    output_buffer[out_pixel_index + 0] += decode_buffer[in_pixel_index + 0] * coefficient;
+                }
+                break;
+            case 2:
+                for (k = n0; k <= n1; k++)
+                {
+                    int in_pixel_index = k * 2;
+                    float coefficient = horizontal_coefficients[coefficient_group + coefficient_counter++];
+                    STBIR_ASSERT(coefficient != 0);
+                    output_buffer[out_pixel_index + 0] += decode_buffer[in_pixel_index + 0] * coefficient;
+                    output_buffer[out_pixel_index + 1] += decode_buffer[in_pixel_index + 1] * coefficient;
+                }
+                break;
+            case 3:
+                for (k = n0; k <= n1; k++)
+                {
+                    int in_pixel_index = k * 3;
+                    float coefficient = horizontal_coefficients[coefficient_group + coefficient_counter++];
+                    STBIR_ASSERT(coefficient != 0);
+                    output_buffer[out_pixel_index + 0] += decode_buffer[in_pixel_index + 0] * coefficient;
+                    output_buffer[out_pixel_index + 1] += decode_buffer[in_pixel_index + 1] * coefficient;
+                    output_buffer[out_pixel_index + 2] += decode_buffer[in_pixel_index + 2] * coefficient;
+                }
+                break;
+            case 4:
+                for (k = n0; k <= n1; k++)
+                {
+                    int in_pixel_index = k * 4;
+                    float coefficient = horizontal_coefficients[coefficient_group + coefficient_counter++];
+                    STBIR_ASSERT(coefficient != 0);
+                    output_buffer[out_pixel_index + 0] += decode_buffer[in_pixel_index + 0] * coefficient;
+                    output_buffer[out_pixel_index + 1] += decode_buffer[in_pixel_index + 1] * coefficient;
+                    output_buffer[out_pixel_index + 2] += decode_buffer[in_pixel_index + 2] * coefficient;
+                    output_buffer[out_pixel_index + 3] += decode_buffer[in_pixel_index + 3] * coefficient;
+                }
+                break;
+            default:
+                for (k = n0; k <= n1; k++)
+                {
+                    int in_pixel_index = k * channels;
+                    float coefficient = horizontal_coefficients[coefficient_group + coefficient_counter++];
+                    int c;
+                    STBIR_ASSERT(coefficient != 0);
+                    for (c = 0; c < channels; c++)
+                        output_buffer[out_pixel_index + c] += decode_buffer[in_pixel_index + c] * coefficient;
+                }
+                break;
+        }
+    }
+}
+
+static void stbir__resample_horizontal_downsample(stbir__info* stbir_info, float* output_buffer)
+{
+    int x, k;
+    int input_w = stbir_info->input_w;
+    int channels = stbir_info->channels;
+    float* decode_buffer = stbir__get_decode_buffer(stbir_info);
+    stbir__contributors* horizontal_contributors = stbir_info->horizontal_contributors;
+    float* horizontal_coefficients = stbir_info->horizontal_coefficients;
+    int coefficient_width = stbir_info->horizontal_coefficient_width;
+    int filter_pixel_margin = stbir_info->horizontal_filter_pixel_margin;
+    int max_x = input_w + filter_pixel_margin * 2;
+
+    STBIR_ASSERT(!stbir__use_width_upsampling(stbir_info));
+
+    switch (channels) {
+        case 1:
+            for (x = 0; x < max_x; x++)
+            {
+                int n0 = horizontal_contributors[x].n0;
+                int n1 = horizontal_contributors[x].n1;
+
+                int in_x = x - filter_pixel_margin;
+                int in_pixel_index = in_x * 1;
+                int max_n = n1;
+                int coefficient_group = coefficient_width * x;
+
+                for (k = n0; k <= max_n; k++)
+                {
+                    int out_pixel_index = k * 1;
+                    float coefficient = horizontal_coefficients[coefficient_group + k - n0];
+                    output_buffer[out_pixel_index + 0] += decode_buffer[in_pixel_index + 0] * coefficient;
+                }
+            }
+            break;
+
+        case 2:
+            for (x = 0; x < max_x; x++)
+            {
+                int n0 = horizontal_contributors[x].n0;
+                int n1 = horizontal_contributors[x].n1;
+
+                int in_x = x - filter_pixel_margin;
+                int in_pixel_index = in_x * 2;
+                int max_n = n1;
+                int coefficient_group = coefficient_width * x;
+
+                for (k = n0; k <= max_n; k++)
+                {
+                    int out_pixel_index = k * 2;
+                    float coefficient = horizontal_coefficients[coefficient_group + k - n0];
+                    output_buffer[out_pixel_index + 0] += decode_buffer[in_pixel_index + 0] * coefficient;
+                    output_buffer[out_pixel_index + 1] += decode_buffer[in_pixel_index + 1] * coefficient;
+                }
+            }
+            break;
+
+        case 3:
+            for (x = 0; x < max_x; x++)
+            {
+                int n0 = horizontal_contributors[x].n0;
+                int n1 = horizontal_contributors[x].n1;
+
+                int in_x = x - filter_pixel_margin;
+                int in_pixel_index = in_x * 3;
+                int max_n = n1;
+                int coefficient_group = coefficient_width * x;
+
+                for (k = n0; k <= max_n; k++)
+                {
+                    int out_pixel_index = k * 3;
+                    float coefficient = horizontal_coefficients[coefficient_group + k - n0];
+                    output_buffer[out_pixel_index + 0] += decode_buffer[in_pixel_index + 0] * coefficient;
+                    output_buffer[out_pixel_index + 1] += decode_buffer[in_pixel_index + 1] * coefficient;
+                    output_buffer[out_pixel_index + 2] += decode_buffer[in_pixel_index + 2] * coefficient;
+                }
+            }
+            break;
+
+        case 4:
+            for (x = 0; x < max_x; x++)
+            {
+                int n0 = horizontal_contributors[x].n0;
+                int n1 = horizontal_contributors[x].n1;
+
+                int in_x = x - filter_pixel_margin;
+                int in_pixel_index = in_x * 4;
+                int max_n = n1;
+                int coefficient_group = coefficient_width * x;
+
+                for (k = n0; k <= max_n; k++)
+                {
+                    int out_pixel_index = k * 4;
+                    float coefficient = horizontal_coefficients[coefficient_group + k - n0];
+                    output_buffer[out_pixel_index + 0] += decode_buffer[in_pixel_index + 0] * coefficient;
+                    output_buffer[out_pixel_index + 1] += decode_buffer[in_pixel_index + 1] * coefficient;
+                    output_buffer[out_pixel_index + 2] += decode_buffer[in_pixel_index + 2] * coefficient;
+                    output_buffer[out_pixel_index + 3] += decode_buffer[in_pixel_index + 3] * coefficient;
+                }
+            }
+            break;
+
+        default:
+            for (x = 0; x < max_x; x++)
+            {
+                int n0 = horizontal_contributors[x].n0;
+                int n1 = horizontal_contributors[x].n1;
+
+                int in_x = x - filter_pixel_margin;
+                int in_pixel_index = in_x * channels;
+                int max_n = n1;
+                int coefficient_group = coefficient_width * x;
+
+                for (k = n0; k <= max_n; k++)
+                {
+                    int c;
+                    int out_pixel_index = k * channels;
+                    float coefficient = horizontal_coefficients[coefficient_group + k - n0];
+                    for (c = 0; c < channels; c++)
+                        output_buffer[out_pixel_index + c] += decode_buffer[in_pixel_index + c] * coefficient;
+                }
+            }
+            break;
+    }
+}
+
+static void stbir__decode_and_resample_upsample(stbir__info* stbir_info, int n)
+{
+    // Decode the nth scanline from the source image into the decode buffer.
+    stbir__decode_scanline(stbir_info, n);
+
+    // Now resample it into the ring buffer.
+    if (stbir__use_width_upsampling(stbir_info))
+        stbir__resample_horizontal_upsample(stbir_info, stbir__add_empty_ring_buffer_entry(stbir_info, n));
+    else
+        stbir__resample_horizontal_downsample(stbir_info, stbir__add_empty_ring_buffer_entry(stbir_info, n));
+
+    // Now it's sitting in the ring buffer ready to be used as source for the vertical sampling.
+}
+
+static void stbir__decode_and_resample_downsample(stbir__info* stbir_info, int n)
+{
+    // Decode the nth scanline from the source image into the decode buffer.
+    stbir__decode_scanline(stbir_info, n);
+
+    memset(stbir_info->horizontal_buffer, 0, stbir_info->output_w * stbir_info->channels * sizeof(float));
+
+    // Now resample it into the horizontal buffer.
+    if (stbir__use_width_upsampling(stbir_info))
+        stbir__resample_horizontal_upsample(stbir_info, stbir_info->horizontal_buffer);
+    else
+        stbir__resample_horizontal_downsample(stbir_info, stbir_info->horizontal_buffer);
+
+    // Now it's sitting in the horizontal buffer ready to be distributed into the ring buffers.
+}
+
+// Get the specified scan line from the ring buffer.
+static float* stbir__get_ring_buffer_scanline(int get_scanline, float* ring_buffer, int begin_index, int first_scanline, int ring_buffer_num_entries, int ring_buffer_length)
+{
+    int ring_buffer_index = (begin_index + (get_scanline - first_scanline)) % ring_buffer_num_entries;
+    return stbir__get_ring_buffer_entry(ring_buffer, ring_buffer_index, ring_buffer_length);
+}
+
+
+static void stbir__encode_scanline(stbir__info* stbir_info, int num_pixels, void *output_buffer, float *encode_buffer, int channels, int alpha_channel, int decode)
+{
+    int x;
+    int n;
+    int num_nonalpha;
+    stbir_uint16 nonalpha[STBIR_MAX_CHANNELS];
+
+    if (!(stbir_info->flags&STBIR_FLAG_ALPHA_PREMULTIPLIED))
+    {
+        for (x=0; x < num_pixels; ++x)
+        {
+            int pixel_index = x*channels;
+
+            float alpha = encode_buffer[pixel_index + alpha_channel];
+            float reciprocal_alpha = alpha ? 1.0f / alpha : 0;
+
+            // unrolling this produced a 1% slowdown upscaling a large RGBA linear-space image on my machine - stb
+            for (n = 0; n < channels; n++)
+                if (n != alpha_channel)
+                    encode_buffer[pixel_index + n] *= reciprocal_alpha;
+
+            // We added in a small epsilon to prevent the color channel from being deleted with zero alpha.
+            // Because we only add it for integer types, it will automatically be discarded on integer
+            // conversion, so we don't need to subtract it back out (which would be problematic for
+            // numeric precision reasons).
+        }
+    }
+
+    // build a table of all channels that need colorspace correction, so
+    // we don't perform colorspace correction on channels that don't need it.
+    for (x = 0, num_nonalpha = 0; x < channels; ++x)
+    {
+        if (x != alpha_channel || (stbir_info->flags & STBIR_FLAG_ALPHA_USES_COLORSPACE))
+        {
+            nonalpha[num_nonalpha++] = (stbir_uint16)x;
+        }
+    }
+
+    #define STBIR__ROUND_INT(f)    ((int)          ((f)+0.5))
+    #define STBIR__ROUND_UINT(f)   ((stbir_uint32) ((f)+0.5))
+
+    #ifdef STBIR__SATURATE_INT
+    #define STBIR__ENCODE_LINEAR8(f)   stbir__saturate8 (STBIR__ROUND_INT((f) * stbir__max_uint8_as_float ))
+    #define STBIR__ENCODE_LINEAR16(f)  stbir__saturate16(STBIR__ROUND_INT((f) * stbir__max_uint16_as_float))
+    #else
+    #define STBIR__ENCODE_LINEAR8(f)   (unsigned char ) STBIR__ROUND_INT(stbir__saturate(f) * stbir__max_uint8_as_float )
+    #define STBIR__ENCODE_LINEAR16(f)  (unsigned short) STBIR__ROUND_INT(stbir__saturate(f) * stbir__max_uint16_as_float)
+    #endif
+
+    switch (decode)
+    {
+        case STBIR__DECODE(STBIR_TYPE_UINT8, STBIR_COLORSPACE_LINEAR):
+            for (x=0; x < num_pixels; ++x)
+            {
+                int pixel_index = x*channels;
+
+                for (n = 0; n < channels; n++)
+                {
+                    int index = pixel_index + n;
+                    ((unsigned char*)output_buffer)[index] = STBIR__ENCODE_LINEAR8(encode_buffer[index]);
+                }
+            }
+            break;
+
+        case STBIR__DECODE(STBIR_TYPE_UINT8, STBIR_COLORSPACE_SRGB):
+            for (x=0; x < num_pixels; ++x)
+            {
+                int pixel_index = x*channels;
+
+                for (n = 0; n < num_nonalpha; n++)
+                {
+                    int index = pixel_index + nonalpha[n];
+                    ((unsigned char*)output_buffer)[index] = stbir__linear_to_srgb_uchar(encode_buffer[index]);
+                }
+
+                if (!(stbir_info->flags & STBIR_FLAG_ALPHA_USES_COLORSPACE))
+                    ((unsigned char *)output_buffer)[pixel_index + alpha_channel] = STBIR__ENCODE_LINEAR8(encode_buffer[pixel_index+alpha_channel]);
+            }
+            break;
+
+        case STBIR__DECODE(STBIR_TYPE_UINT16, STBIR_COLORSPACE_LINEAR):
+            for (x=0; x < num_pixels; ++x)
+            {
+                int pixel_index = x*channels;
+
+                for (n = 0; n < channels; n++)
+                {
+                    int index = pixel_index + n;
+                    ((unsigned short*)output_buffer)[index] = STBIR__ENCODE_LINEAR16(encode_buffer[index]);
+                }
+            }
+            break;
+
+        case STBIR__DECODE(STBIR_TYPE_UINT16, STBIR_COLORSPACE_SRGB):
+            for (x=0; x < num_pixels; ++x)
+            {
+                int pixel_index = x*channels;
+
+                for (n = 0; n < num_nonalpha; n++)
+                {
+                    int index = pixel_index + nonalpha[n];
+                    ((unsigned short*)output_buffer)[index] = (unsigned short)STBIR__ROUND_INT(stbir__linear_to_srgb(stbir__saturate(encode_buffer[index])) * stbir__max_uint16_as_float);
+                }
+
+                if (!(stbir_info->flags&STBIR_FLAG_ALPHA_USES_COLORSPACE))
+                    ((unsigned short*)output_buffer)[pixel_index + alpha_channel] = STBIR__ENCODE_LINEAR16(encode_buffer[pixel_index + alpha_channel]);
+            }
+
+            break;
+
+        case STBIR__DECODE(STBIR_TYPE_UINT32, STBIR_COLORSPACE_LINEAR):
+            for (x=0; x < num_pixels; ++x)
+            {
+                int pixel_index = x*channels;
+
+                for (n = 0; n < channels; n++)
+                {
+                    int index = pixel_index + n;
+                    ((unsigned int*)output_buffer)[index] = (unsigned int)STBIR__ROUND_UINT(((double)stbir__saturate(encode_buffer[index])) * stbir__max_uint32_as_float);
+                }
+            }
+            break;
+
+        case STBIR__DECODE(STBIR_TYPE_UINT32, STBIR_COLORSPACE_SRGB):
+            for (x=0; x < num_pixels; ++x)
+            {
+                int pixel_index = x*channels;
+
+                for (n = 0; n < num_nonalpha; n++)
+                {
+                    int index = pixel_index + nonalpha[n];
+                    ((unsigned int*)output_buffer)[index] = (unsigned int)STBIR__ROUND_UINT(((double)stbir__linear_to_srgb(stbir__saturate(encode_buffer[index]))) * stbir__max_uint32_as_float);
+                }
+
+                if (!(stbir_info->flags&STBIR_FLAG_ALPHA_USES_COLORSPACE))
+                    ((unsigned int*)output_buffer)[pixel_index + alpha_channel] = (unsigned int)STBIR__ROUND_INT(((double)stbir__saturate(encode_buffer[pixel_index + alpha_channel])) * stbir__max_uint32_as_float);
+            }
+            break;
+
+        case STBIR__DECODE(STBIR_TYPE_FLOAT, STBIR_COLORSPACE_LINEAR):
+            for (x=0; x < num_pixels; ++x)
+            {
+                int pixel_index = x*channels;
+
+                for (n = 0; n < channels; n++)
+                {
+                    int index = pixel_index + n;
+                    ((float*)output_buffer)[index] = encode_buffer[index];
+                }
+            }
+            break;
+
+        case STBIR__DECODE(STBIR_TYPE_FLOAT, STBIR_COLORSPACE_SRGB):
+            for (x=0; x < num_pixels; ++x)
+            {
+                int pixel_index = x*channels;
+
+                for (n = 0; n < num_nonalpha; n++)
+                {
+                    int index = pixel_index + nonalpha[n];
+                    ((float*)output_buffer)[index] = stbir__linear_to_srgb(encode_buffer[index]);
+                }
+
+                if (!(stbir_info->flags&STBIR_FLAG_ALPHA_USES_COLORSPACE))
+                    ((float*)output_buffer)[pixel_index + alpha_channel] = encode_buffer[pixel_index + alpha_channel];
+            }
+            break;
+
+        default:
+            STBIR_ASSERT(!"Unknown type/colorspace/channels combination.");
+            break;
+    }
+}
+
+static void stbir__resample_vertical_upsample(stbir__info* stbir_info, int n)
+{
+    int x, k;
+    int output_w = stbir_info->output_w;
+    stbir__contributors* vertical_contributors = stbir_info->vertical_contributors;
+    float* vertical_coefficients = stbir_info->vertical_coefficients;
+    int channels = stbir_info->channels;
+    int alpha_channel = stbir_info->alpha_channel;
+    int type = stbir_info->type;
+    int colorspace = stbir_info->colorspace;
+    int ring_buffer_entries = stbir_info->ring_buffer_num_entries;
+    void* output_data = stbir_info->output_data;
+    float* encode_buffer = stbir_info->encode_buffer;
+    int decode = STBIR__DECODE(type, colorspace);
+    int coefficient_width = stbir_info->vertical_coefficient_width;
+    int coefficient_counter;
+    int contributor = n;
+
+    float* ring_buffer = stbir_info->ring_buffer;
+    int ring_buffer_begin_index = stbir_info->ring_buffer_begin_index;
+    int ring_buffer_first_scanline = stbir_info->ring_buffer_first_scanline;
+    int ring_buffer_length = stbir_info->ring_buffer_length_bytes/sizeof(float);
+
+    int n0,n1, output_row_start;
+    int coefficient_group = coefficient_width * contributor;
+
+    n0 = vertical_contributors[contributor].n0;
+    n1 = vertical_contributors[contributor].n1;
+
+    output_row_start = n * stbir_info->output_stride_bytes;
+
+    STBIR_ASSERT(stbir__use_height_upsampling(stbir_info));
+
+    memset(encode_buffer, 0, output_w * sizeof(float) * channels);
+
+    // I tried reblocking this for better cache usage of encode_buffer
+    // (using x_outer, k, x_inner), but it lost speed. -- stb
+
+    coefficient_counter = 0;
+    switch (channels) {
+        case 1:
+            for (k = n0; k <= n1; k++)
+            {
+                int coefficient_index = coefficient_counter++;
+                float* ring_buffer_entry = stbir__get_ring_buffer_scanline(k, ring_buffer, ring_buffer_begin_index, ring_buffer_first_scanline, ring_buffer_entries, ring_buffer_length);
+                float coefficient = vertical_coefficients[coefficient_group + coefficient_index];
+                for (x = 0; x < output_w; ++x)
+                {
+                    int in_pixel_index = x * 1;
+                    encode_buffer[in_pixel_index + 0] += ring_buffer_entry[in_pixel_index + 0] * coefficient;
+                }
+            }
+            break;
+        case 2:
+            for (k = n0; k <= n1; k++)
+            {
+                int coefficient_index = coefficient_counter++;
+                float* ring_buffer_entry = stbir__get_ring_buffer_scanline(k, ring_buffer, ring_buffer_begin_index, ring_buffer_first_scanline, ring_buffer_entries, ring_buffer_length);
+                float coefficient = vertical_coefficients[coefficient_group + coefficient_index];
+                for (x = 0; x < output_w; ++x)
+                {
+                    int in_pixel_index = x * 2;
+                    encode_buffer[in_pixel_index + 0] += ring_buffer_entry[in_pixel_index + 0] * coefficient;
+                    encode_buffer[in_pixel_index + 1] += ring_buffer_entry[in_pixel_index + 1] * coefficient;
+                }
+            }
+            break;
+        case 3:
+            for (k = n0; k <= n1; k++)
+            {
+                int coefficient_index = coefficient_counter++;
+                float* ring_buffer_entry = stbir__get_ring_buffer_scanline(k, ring_buffer, ring_buffer_begin_index, ring_buffer_first_scanline, ring_buffer_entries, ring_buffer_length);
+                float coefficient = vertical_coefficients[coefficient_group + coefficient_index];
+                for (x = 0; x < output_w; ++x)
+                {
+                    int in_pixel_index = x * 3;
+                    encode_buffer[in_pixel_index + 0] += ring_buffer_entry[in_pixel_index + 0] * coefficient;
+                    encode_buffer[in_pixel_index + 1] += ring_buffer_entry[in_pixel_index + 1] * coefficient;
+                    encode_buffer[in_pixel_index + 2] += ring_buffer_entry[in_pixel_index + 2] * coefficient;
+                }
+            }
+            break;
+        case 4:
+            for (k = n0; k <= n1; k++)
+            {
+                int coefficient_index = coefficient_counter++;
+                float* ring_buffer_entry = stbir__get_ring_buffer_scanline(k, ring_buffer, ring_buffer_begin_index, ring_buffer_first_scanline, ring_buffer_entries, ring_buffer_length);
+                float coefficient = vertical_coefficients[coefficient_group + coefficient_index];
+                for (x = 0; x < output_w; ++x)
+                {
+                    int in_pixel_index = x * 4;
+                    encode_buffer[in_pixel_index + 0] += ring_buffer_entry[in_pixel_index + 0] * coefficient;
+                    encode_buffer[in_pixel_index + 1] += ring_buffer_entry[in_pixel_index + 1] * coefficient;
+                    encode_buffer[in_pixel_index + 2] += ring_buffer_entry[in_pixel_index + 2] * coefficient;
+                    encode_buffer[in_pixel_index + 3] += ring_buffer_entry[in_pixel_index + 3] * coefficient;
+                }
+            }
+            break;
+        default:
+            for (k = n0; k <= n1; k++)
+            {
+                int coefficient_index = coefficient_counter++;
+                float* ring_buffer_entry = stbir__get_ring_buffer_scanline(k, ring_buffer, ring_buffer_begin_index, ring_buffer_first_scanline, ring_buffer_entries, ring_buffer_length);
+                float coefficient = vertical_coefficients[coefficient_group + coefficient_index];
+                for (x = 0; x < output_w; ++x)
+                {
+                    int in_pixel_index = x * channels;
+                    int c;
+                    for (c = 0; c < channels; c++)
+                        encode_buffer[in_pixel_index + c] += ring_buffer_entry[in_pixel_index + c] * coefficient;
+                }
+            }
+            break;
+    }
+    stbir__encode_scanline(stbir_info, output_w, (char *) output_data + output_row_start, encode_buffer, channels, alpha_channel, decode);
+}
+
+static void stbir__resample_vertical_downsample(stbir__info* stbir_info, int n)
+{
+    int x, k;
+    int output_w = stbir_info->output_w;
+    stbir__contributors* vertical_contributors = stbir_info->vertical_contributors;
+    float* vertical_coefficients = stbir_info->vertical_coefficients;
+    int channels = stbir_info->channels;
+    int ring_buffer_entries = stbir_info->ring_buffer_num_entries;
+    float* horizontal_buffer = stbir_info->horizontal_buffer;
+    int coefficient_width = stbir_info->vertical_coefficient_width;
+    int contributor = n + stbir_info->vertical_filter_pixel_margin;
+
+    float* ring_buffer = stbir_info->ring_buffer;
+    int ring_buffer_begin_index = stbir_info->ring_buffer_begin_index;
+    int ring_buffer_first_scanline = stbir_info->ring_buffer_first_scanline;
+    int ring_buffer_length = stbir_info->ring_buffer_length_bytes/sizeof(float);
+    int n0,n1;
+
+    n0 = vertical_contributors[contributor].n0;
+    n1 = vertical_contributors[contributor].n1;
+
+    STBIR_ASSERT(!stbir__use_height_upsampling(stbir_info));
+
+    for (k = n0; k <= n1; k++)
+    {
+        int coefficient_index = k - n0;
+        int coefficient_group = coefficient_width * contributor;
+        float coefficient = vertical_coefficients[coefficient_group + coefficient_index];
+
+        float* ring_buffer_entry = stbir__get_ring_buffer_scanline(k, ring_buffer, ring_buffer_begin_index, ring_buffer_first_scanline, ring_buffer_entries, ring_buffer_length);
+
+        switch (channels) {
+            case 1:
+                for (x = 0; x < output_w; x++)
+                {
+                    int in_pixel_index = x * 1;
+                    ring_buffer_entry[in_pixel_index + 0] += horizontal_buffer[in_pixel_index + 0] * coefficient;
+                }
+                break;
+            case 2:
+                for (x = 0; x < output_w; x++)
+                {
+                    int in_pixel_index = x * 2;
+                    ring_buffer_entry[in_pixel_index + 0] += horizontal_buffer[in_pixel_index + 0] * coefficient;
+                    ring_buffer_entry[in_pixel_index + 1] += horizontal_buffer[in_pixel_index + 1] * coefficient;
+                }
+                break;
+            case 3:
+                for (x = 0; x < output_w; x++)
+                {
+                    int in_pixel_index = x * 3;
+                    ring_buffer_entry[in_pixel_index + 0] += horizontal_buffer[in_pixel_index + 0] * coefficient;
+                    ring_buffer_entry[in_pixel_index + 1] += horizontal_buffer[in_pixel_index + 1] * coefficient;
+                    ring_buffer_entry[in_pixel_index + 2] += horizontal_buffer[in_pixel_index + 2] * coefficient;
+                }
+                break;
+            case 4:
+                for (x = 0; x < output_w; x++)
+                {
+                    int in_pixel_index = x * 4;
+                    ring_buffer_entry[in_pixel_index + 0] += horizontal_buffer[in_pixel_index + 0] * coefficient;
+                    ring_buffer_entry[in_pixel_index + 1] += horizontal_buffer[in_pixel_index + 1] * coefficient;
+                    ring_buffer_entry[in_pixel_index + 2] += horizontal_buffer[in_pixel_index + 2] * coefficient;
+                    ring_buffer_entry[in_pixel_index + 3] += horizontal_buffer[in_pixel_index + 3] * coefficient;
+                }
+                break;
+            default:
+                for (x = 0; x < output_w; x++)
+                {
+                    int in_pixel_index = x * channels;
+
+                    int c;
+                    for (c = 0; c < channels; c++)
+                        ring_buffer_entry[in_pixel_index + c] += horizontal_buffer[in_pixel_index + c] * coefficient;
+                }
+                break;
+        }
+    }
+}
+
+static void stbir__buffer_loop_upsample(stbir__info* stbir_info)
+{
+    int y;
+    float scale_ratio = stbir_info->vertical_scale;
+    float out_scanlines_radius = stbir__filter_info_table[stbir_info->vertical_filter].support(1/scale_ratio) * scale_ratio;
+
+    STBIR_ASSERT(stbir__use_height_upsampling(stbir_info));
+
+    for (y = 0; y < stbir_info->output_h; y++)
+    {
+        float in_center_of_out = 0; // Center of the current out scanline in the in scanline space
+        int in_first_scanline = 0, in_last_scanline = 0;
+
+        stbir__calculate_sample_range_upsample(y, out_scanlines_radius, scale_ratio, stbir_info->vertical_shift, &in_first_scanline, &in_last_scanline, &in_center_of_out);
+
+        STBIR_ASSERT(in_last_scanline - in_first_scanline + 1 <= stbir_info->ring_buffer_num_entries);
+
+        if (stbir_info->ring_buffer_begin_index >= 0)
+        {
+            // Get rid of whatever we don't need anymore.
+            while (in_first_scanline > stbir_info->ring_buffer_first_scanline)
+            {
+                if (stbir_info->ring_buffer_first_scanline == stbir_info->ring_buffer_last_scanline)
+                {
+                    // We just popped the last scanline off the ring buffer.
+                    // Reset it to the empty state.
+                    stbir_info->ring_buffer_begin_index = -1;
+                    stbir_info->ring_buffer_first_scanline = 0;
+                    stbir_info->ring_buffer_last_scanline = 0;
+                    break;
+                }
+                else
+                {
+                    stbir_info->ring_buffer_first_scanline++;
+                    stbir_info->ring_buffer_begin_index = (stbir_info->ring_buffer_begin_index + 1) % stbir_info->ring_buffer_num_entries;
+                }
+            }
+        }
+
+        // Load in new ones.
+        if (stbir_info->ring_buffer_begin_index < 0)
+            stbir__decode_and_resample_upsample(stbir_info, in_first_scanline);
+
+        while (in_last_scanline > stbir_info->ring_buffer_last_scanline)
+            stbir__decode_and_resample_upsample(stbir_info, stbir_info->ring_buffer_last_scanline + 1);
+
+        // Now all buffers should be ready to write a row of vertical sampling.
+        stbir__resample_vertical_upsample(stbir_info, y);
+
+        STBIR_PROGRESS_REPORT((float)y / stbir_info->output_h);
+    }
+}
+
+static void stbir__empty_ring_buffer(stbir__info* stbir_info, int first_necessary_scanline)
+{
+    int output_stride_bytes = stbir_info->output_stride_bytes;
+    int channels = stbir_info->channels;
+    int alpha_channel = stbir_info->alpha_channel;
+    int type = stbir_info->type;
+    int colorspace = stbir_info->colorspace;
+    int output_w = stbir_info->output_w;
+    void* output_data = stbir_info->output_data;
+    int decode = STBIR__DECODE(type, colorspace);
+
+    float* ring_buffer = stbir_info->ring_buffer;
+    int ring_buffer_length = stbir_info->ring_buffer_length_bytes/sizeof(float);
+
+    if (stbir_info->ring_buffer_begin_index >= 0)
+    {
+        // Get rid of whatever we don't need anymore.
+        while (first_necessary_scanline > stbir_info->ring_buffer_first_scanline)
+        {
+            if (stbir_info->ring_buffer_first_scanline >= 0 && stbir_info->ring_buffer_first_scanline < stbir_info->output_h)
+            {
+                int output_row_start = stbir_info->ring_buffer_first_scanline * output_stride_bytes;
+                float* ring_buffer_entry = stbir__get_ring_buffer_entry(ring_buffer, stbir_info->ring_buffer_begin_index, ring_buffer_length);
+                stbir__encode_scanline(stbir_info, output_w, (char *) output_data + output_row_start, ring_buffer_entry, channels, alpha_channel, decode);
+                STBIR_PROGRESS_REPORT((float)stbir_info->ring_buffer_first_scanline / stbir_info->output_h);
+            }
+
+            if (stbir_info->ring_buffer_first_scanline == stbir_info->ring_buffer_last_scanline)
+            {
+                // We just popped the last scanline off the ring buffer.
+                // Reset it to the empty state.
+                stbir_info->ring_buffer_begin_index = -1;
+                stbir_info->ring_buffer_first_scanline = 0;
+                stbir_info->ring_buffer_last_scanline = 0;
+                break;
+            }
+            else
+            {
+                stbir_info->ring_buffer_first_scanline++;
+                stbir_info->ring_buffer_begin_index = (stbir_info->ring_buffer_begin_index + 1) % stbir_info->ring_buffer_num_entries;
+            }
+        }
+    }
+}
+
+static void stbir__buffer_loop_downsample(stbir__info* stbir_info)
+{
+    int y;
+    float scale_ratio = stbir_info->vertical_scale;
+    int output_h = stbir_info->output_h;
+    float in_pixels_radius = stbir__filter_info_table[stbir_info->vertical_filter].support(scale_ratio) / scale_ratio;
+    int pixel_margin = stbir_info->vertical_filter_pixel_margin;
+    int max_y = stbir_info->input_h + pixel_margin;
+
+    STBIR_ASSERT(!stbir__use_height_upsampling(stbir_info));
+
+    for (y = -pixel_margin; y < max_y; y++)
+    {
+        float out_center_of_in; // Center of the current out scanline in the in scanline space
+        int out_first_scanline, out_last_scanline;
+
+        stbir__calculate_sample_range_downsample(y, in_pixels_radius, scale_ratio, stbir_info->vertical_shift, &out_first_scanline, &out_last_scanline, &out_center_of_in);
+
+        STBIR_ASSERT(out_last_scanline - out_first_scanline + 1 <= stbir_info->ring_buffer_num_entries);
+
+        if (out_last_scanline < 0 || out_first_scanline >= output_h)
+            continue;
+
+        stbir__empty_ring_buffer(stbir_info, out_first_scanline);
+
+        stbir__decode_and_resample_downsample(stbir_info, y);
+
+        // Load in new ones.
+        if (stbir_info->ring_buffer_begin_index < 0)
+            stbir__add_empty_ring_buffer_entry(stbir_info, out_first_scanline);
+
+        while (out_last_scanline > stbir_info->ring_buffer_last_scanline)
+            stbir__add_empty_ring_buffer_entry(stbir_info, stbir_info->ring_buffer_last_scanline + 1);
+
+        // Now the horizontal buffer is ready to write to all ring buffer rows.
+        stbir__resample_vertical_downsample(stbir_info, y);
+    }
+
+    stbir__empty_ring_buffer(stbir_info, stbir_info->output_h);
+}
+
+static void stbir__setup(stbir__info *info, int input_w, int input_h, int output_w, int output_h, int channels)
+{
+    info->input_w = input_w;
+    info->input_h = input_h;
+    info->output_w = output_w;
+    info->output_h = output_h;
+    info->channels = channels;
+}
+
+static void stbir__calculate_transform(stbir__info *info, float s0, float t0, float s1, float t1, float *transform)
+{
+    info->s0 = s0;
+    info->t0 = t0;
+    info->s1 = s1;
+    info->t1 = t1;
+
+    if (transform)
+    {
+        info->horizontal_scale = transform[0];
+        info->vertical_scale   = transform[1];
+        info->horizontal_shift = transform[2];
+        info->vertical_shift   = transform[3];
+    }
+    else
+    {
+        info->horizontal_scale = ((float)info->output_w / info->input_w) / (s1 - s0);
+        info->vertical_scale = ((float)info->output_h / info->input_h) / (t1 - t0);
+
+        info->horizontal_shift = s0 * info->output_w / (s1 - s0);
+        info->vertical_shift = t0 * info->output_h / (t1 - t0);
+    }
+}
+
+static void stbir__choose_filter(stbir__info *info, stbir_filter h_filter, stbir_filter v_filter)
+{
+    if (h_filter == 0)
+        h_filter = stbir__use_upsampling(info->horizontal_scale) ? STBIR_DEFAULT_FILTER_UPSAMPLE : STBIR_DEFAULT_FILTER_DOWNSAMPLE;
+    if (v_filter == 0)
+        v_filter = stbir__use_upsampling(info->vertical_scale)   ? STBIR_DEFAULT_FILTER_UPSAMPLE : STBIR_DEFAULT_FILTER_DOWNSAMPLE;
+    info->horizontal_filter = h_filter;
+    info->vertical_filter = v_filter;
+}
+
+static stbir_uint32 stbir__calculate_memory(stbir__info *info)
+{
+    int pixel_margin = stbir__get_filter_pixel_margin(info->horizontal_filter, info->horizontal_scale);
+    int filter_height = stbir__get_filter_pixel_width(info->vertical_filter, info->vertical_scale);
+
+    info->horizontal_num_contributors = stbir__get_contributors(info->horizontal_scale, info->horizontal_filter, info->input_w, info->output_w);
+    info->vertical_num_contributors   = stbir__get_contributors(info->vertical_scale  , info->vertical_filter  , info->input_h, info->output_h);
+
+    // One extra entry because floating point precision problems sometimes cause an extra to be necessary.
+    info->ring_buffer_num_entries = filter_height + 1;
+
+    info->horizontal_contributors_size = info->horizontal_num_contributors * sizeof(stbir__contributors);
+    info->horizontal_coefficients_size = stbir__get_total_horizontal_coefficients(info) * sizeof(float);
+    info->vertical_contributors_size = info->vertical_num_contributors * sizeof(stbir__contributors);
+    info->vertical_coefficients_size = stbir__get_total_vertical_coefficients(info) * sizeof(float);
+    info->decode_buffer_size = (info->input_w + pixel_margin * 2) * info->channels * sizeof(float);
+    info->horizontal_buffer_size = info->output_w * info->channels * sizeof(float);
+    info->ring_buffer_size = info->output_w * info->channels * info->ring_buffer_num_entries * sizeof(float);
+    info->encode_buffer_size = info->output_w * info->channels * sizeof(float);
+
+    STBIR_ASSERT(info->horizontal_filter != 0);
+    STBIR_ASSERT(info->horizontal_filter < STBIR__ARRAY_SIZE(stbir__filter_info_table)); // this now happens too late
+    STBIR_ASSERT(info->vertical_filter != 0);
+    STBIR_ASSERT(info->vertical_filter < STBIR__ARRAY_SIZE(stbir__filter_info_table)); // this now happens too late
+
+    if (stbir__use_height_upsampling(info))
+        // The horizontal buffer is for when we're downsampling the height and we
+        // can't output the result of sampling the decode buffer directly into the
+        // ring buffers.
+        info->horizontal_buffer_size = 0;
+    else
+        // The encode buffer is to retain precision in the height upsampling method
+        // and isn't used when height downsampling.
+        info->encode_buffer_size = 0;
+
+    return info->horizontal_contributors_size + info->horizontal_coefficients_size
+        + info->vertical_contributors_size + info->vertical_coefficients_size
+        + info->decode_buffer_size + info->horizontal_buffer_size
+        + info->ring_buffer_size + info->encode_buffer_size;
+}
+
+static int stbir__resize_allocated(stbir__info *info,
+    const void* input_data, int input_stride_in_bytes,
+    void* output_data, int output_stride_in_bytes,
+    int alpha_channel, stbir_uint32 flags, stbir_datatype type,
+    stbir_edge edge_horizontal, stbir_edge edge_vertical, stbir_colorspace colorspace,
+    void* tempmem, size_t tempmem_size_in_bytes)
+{
+    size_t memory_required = stbir__calculate_memory(info);
+
+    int width_stride_input = input_stride_in_bytes ? input_stride_in_bytes : info->channels * info->input_w * stbir__type_size[type];
+    int width_stride_output = output_stride_in_bytes ? output_stride_in_bytes : info->channels * info->output_w * stbir__type_size[type];
+
+#ifdef STBIR_DEBUG_OVERWRITE_TEST
+#define OVERWRITE_ARRAY_SIZE 8
+    unsigned char overwrite_output_before_pre[OVERWRITE_ARRAY_SIZE];
+    unsigned char overwrite_tempmem_before_pre[OVERWRITE_ARRAY_SIZE];
+    unsigned char overwrite_output_after_pre[OVERWRITE_ARRAY_SIZE];
+    unsigned char overwrite_tempmem_after_pre[OVERWRITE_ARRAY_SIZE];
+
+    size_t begin_forbidden = width_stride_output * (info->output_h - 1) + info->output_w * info->channels * stbir__type_size[type];
+    memcpy(overwrite_output_before_pre, &((unsigned char*)output_data)[-OVERWRITE_ARRAY_SIZE], OVERWRITE_ARRAY_SIZE);
+    memcpy(overwrite_output_after_pre, &((unsigned char*)output_data)[begin_forbidden], OVERWRITE_ARRAY_SIZE);
+    memcpy(overwrite_tempmem_before_pre, &((unsigned char*)tempmem)[-OVERWRITE_ARRAY_SIZE], OVERWRITE_ARRAY_SIZE);
+    memcpy(overwrite_tempmem_after_pre, &((unsigned char*)tempmem)[tempmem_size_in_bytes], OVERWRITE_ARRAY_SIZE);
+#endif
+
+    STBIR_ASSERT(info->channels >= 0);
+    STBIR_ASSERT(info->channels <= STBIR_MAX_CHANNELS);
+
+    if (info->channels < 0 || info->channels > STBIR_MAX_CHANNELS)
+        return 0;
+
+    STBIR_ASSERT(info->horizontal_filter < STBIR__ARRAY_SIZE(stbir__filter_info_table));
+    STBIR_ASSERT(info->vertical_filter < STBIR__ARRAY_SIZE(stbir__filter_info_table));
+
+    if (info->horizontal_filter >= STBIR__ARRAY_SIZE(stbir__filter_info_table))
+        return 0;
+    if (info->vertical_filter >= STBIR__ARRAY_SIZE(stbir__filter_info_table))
+        return 0;
+
+    if (alpha_channel < 0)
+        flags |= STBIR_FLAG_ALPHA_USES_COLORSPACE | STBIR_FLAG_ALPHA_PREMULTIPLIED;
+
+    if (!(flags&STBIR_FLAG_ALPHA_USES_COLORSPACE) || !(flags&STBIR_FLAG_ALPHA_PREMULTIPLIED)) {
+        STBIR_ASSERT(alpha_channel >= 0 && alpha_channel < info->channels);
+    }
+
+    if (alpha_channel >= info->channels)
+        return 0;
+
+    STBIR_ASSERT(tempmem);
+
+    if (!tempmem)
+        return 0;
+
+    STBIR_ASSERT(tempmem_size_in_bytes >= memory_required);
+
+    if (tempmem_size_in_bytes < memory_required)
+        return 0;
+
+    memset(tempmem, 0, tempmem_size_in_bytes);
+
+    info->input_data = input_data;
+    info->input_stride_bytes = width_stride_input;
+
+    info->output_data = output_data;
+    info->output_stride_bytes = width_stride_output;
+
+    info->alpha_channel = alpha_channel;
+    info->flags = flags;
+    info->type = type;
+    info->edge_horizontal = edge_horizontal;
+    info->edge_vertical = edge_vertical;
+    info->colorspace = colorspace;
+
+    info->horizontal_coefficient_width   = stbir__get_coefficient_width  (info->horizontal_filter, info->horizontal_scale);
+    info->vertical_coefficient_width     = stbir__get_coefficient_width  (info->vertical_filter  , info->vertical_scale  );
+    info->horizontal_filter_pixel_width  = stbir__get_filter_pixel_width (info->horizontal_filter, info->horizontal_scale);
+    info->vertical_filter_pixel_width    = stbir__get_filter_pixel_width (info->vertical_filter  , info->vertical_scale  );
+    info->horizontal_filter_pixel_margin = stbir__get_filter_pixel_margin(info->horizontal_filter, info->horizontal_scale);
+    info->vertical_filter_pixel_margin   = stbir__get_filter_pixel_margin(info->vertical_filter  , info->vertical_scale  );
+
+    info->ring_buffer_length_bytes = info->output_w * info->channels * sizeof(float);
+    info->decode_buffer_pixels = info->input_w + info->horizontal_filter_pixel_margin * 2;
+
+#define STBIR__NEXT_MEMPTR(current, newtype) (newtype*)(((unsigned char*)current) + current##_size)
+
+    info->horizontal_contributors = (stbir__contributors *) tempmem;
+    info->horizontal_coefficients = STBIR__NEXT_MEMPTR(info->horizontal_contributors, float);
+    info->vertical_contributors = STBIR__NEXT_MEMPTR(info->horizontal_coefficients, stbir__contributors);
+    info->vertical_coefficients = STBIR__NEXT_MEMPTR(info->vertical_contributors, float);
+    info->decode_buffer = STBIR__NEXT_MEMPTR(info->vertical_coefficients, float);
+
+    if (stbir__use_height_upsampling(info))
+    {
+        info->horizontal_buffer = NULL;
+        info->ring_buffer = STBIR__NEXT_MEMPTR(info->decode_buffer, float);
+        info->encode_buffer = STBIR__NEXT_MEMPTR(info->ring_buffer, float);
+
+        STBIR_ASSERT((size_t)STBIR__NEXT_MEMPTR(info->encode_buffer, unsigned char) == (size_t)tempmem + tempmem_size_in_bytes);
+    }
+    else
+    {
+        info->horizontal_buffer = STBIR__NEXT_MEMPTR(info->decode_buffer, float);
+        info->ring_buffer = STBIR__NEXT_MEMPTR(info->horizontal_buffer, float);
+        info->encode_buffer = NULL;
+
+        STBIR_ASSERT((size_t)STBIR__NEXT_MEMPTR(info->ring_buffer, unsigned char) == (size_t)tempmem + tempmem_size_in_bytes);
+    }
+
+#undef STBIR__NEXT_MEMPTR
+
+    // This signals that the ring buffer is empty
+    info->ring_buffer_begin_index = -1;
+
+    stbir__calculate_filters(info->horizontal_contributors, info->horizontal_coefficients, info->horizontal_filter, info->horizontal_scale, info->horizontal_shift, info->input_w, info->output_w);
+    stbir__calculate_filters(info->vertical_contributors, info->vertical_coefficients, info->vertical_filter, info->vertical_scale, info->vertical_shift, info->input_h, info->output_h);
+
+    STBIR_PROGRESS_REPORT(0);
+
+    if (stbir__use_height_upsampling(info))
+        stbir__buffer_loop_upsample(info);
+    else
+        stbir__buffer_loop_downsample(info);
+
+    STBIR_PROGRESS_REPORT(1);
+
+#ifdef STBIR_DEBUG_OVERWRITE_TEST
+    STBIR_ASSERT(memcmp(overwrite_output_before_pre, &((unsigned char*)output_data)[-OVERWRITE_ARRAY_SIZE], OVERWRITE_ARRAY_SIZE) == 0);
+    STBIR_ASSERT(memcmp(overwrite_output_after_pre, &((unsigned char*)output_data)[begin_forbidden], OVERWRITE_ARRAY_SIZE) == 0);
+    STBIR_ASSERT(memcmp(overwrite_tempmem_before_pre, &((unsigned char*)tempmem)[-OVERWRITE_ARRAY_SIZE], OVERWRITE_ARRAY_SIZE) == 0);
+    STBIR_ASSERT(memcmp(overwrite_tempmem_after_pre, &((unsigned char*)tempmem)[tempmem_size_in_bytes], OVERWRITE_ARRAY_SIZE) == 0);
+#endif
+
+    return 1;
+}
+
+
+static int stbir__resize_arbitrary(
+    void *alloc_context,
+    const void* input_data, int input_w, int input_h, int input_stride_in_bytes,
+    void* output_data, int output_w, int output_h, int output_stride_in_bytes,
+    float s0, float t0, float s1, float t1, float *transform,
+    int channels, int alpha_channel, stbir_uint32 flags, stbir_datatype type,
+    stbir_filter h_filter, stbir_filter v_filter,
+    stbir_edge edge_horizontal, stbir_edge edge_vertical, stbir_colorspace colorspace)
+{
+    stbir__info info;
+    int result;
+    size_t memory_required;
+    void* extra_memory;
+
+    stbir__setup(&info, input_w, input_h, output_w, output_h, channels);
+    stbir__calculate_transform(&info, s0,t0,s1,t1,transform);
+    stbir__choose_filter(&info, h_filter, v_filter);
+    memory_required = stbir__calculate_memory(&info);
+    extra_memory = STBIR_MALLOC(memory_required, alloc_context);
+
+    if (!extra_memory)
+        return 0;
+
+    result = stbir__resize_allocated(&info, input_data, input_stride_in_bytes,
+                                            output_data, output_stride_in_bytes,
+                                            alpha_channel, flags, type,
+                                            edge_horizontal, edge_vertical,
+                                            colorspace, extra_memory, memory_required);
+
+    STBIR_FREE(extra_memory, alloc_context);
+
+    return result;
+}
+
+STBIRDEF int stbir_resize_uint8(     const unsigned char *input_pixels , int input_w , int input_h , int input_stride_in_bytes,
+                                           unsigned char *output_pixels, int output_w, int output_h, int output_stride_in_bytes,
+                                     int num_channels)
+{
+    return stbir__resize_arbitrary(NULL, input_pixels, input_w, input_h, input_stride_in_bytes,
+        output_pixels, output_w, output_h, output_stride_in_bytes,
+        0,0,1,1,NULL,num_channels,-1,0, STBIR_TYPE_UINT8, STBIR_FILTER_DEFAULT, STBIR_FILTER_DEFAULT,
+        STBIR_EDGE_CLAMP, STBIR_EDGE_CLAMP, STBIR_COLORSPACE_LINEAR);
+}
+
+STBIRDEF int stbir_resize_float(     const float *input_pixels , int input_w , int input_h , int input_stride_in_bytes,
+                                           float *output_pixels, int output_w, int output_h, int output_stride_in_bytes,
+                                     int num_channels)
+{
+    return stbir__resize_arbitrary(NULL, input_pixels, input_w, input_h, input_stride_in_bytes,
+        output_pixels, output_w, output_h, output_stride_in_bytes,
+        0,0,1,1,NULL,num_channels,-1,0, STBIR_TYPE_FLOAT, STBIR_FILTER_DEFAULT, STBIR_FILTER_DEFAULT,
+        STBIR_EDGE_CLAMP, STBIR_EDGE_CLAMP, STBIR_COLORSPACE_LINEAR);
+}
+
+STBIRDEF int stbir_resize_uint8_srgb(const unsigned char *input_pixels , int input_w , int input_h , int input_stride_in_bytes,
+                                           unsigned char *output_pixels, int output_w, int output_h, int output_stride_in_bytes,
+                                     int num_channels, int alpha_channel, int flags)
+{
+    return stbir__resize_arbitrary(NULL, input_pixels, input_w, input_h, input_stride_in_bytes,
+        output_pixels, output_w, output_h, output_stride_in_bytes,
+        0,0,1,1,NULL,num_channels,alpha_channel,flags, STBIR_TYPE_UINT8, STBIR_FILTER_DEFAULT, STBIR_FILTER_DEFAULT,
+        STBIR_EDGE_CLAMP, STBIR_EDGE_CLAMP, STBIR_COLORSPACE_SRGB);
+}
+
+STBIRDEF int stbir_resize_uint8_srgb_edgemode(const unsigned char *input_pixels , int input_w , int input_h , int input_stride_in_bytes,
+                                                    unsigned char *output_pixels, int output_w, int output_h, int output_stride_in_bytes,
+                                              int num_channels, int alpha_channel, int flags,
+                                              stbir_edge edge_wrap_mode)
+{
+    return stbir__resize_arbitrary(NULL, input_pixels, input_w, input_h, input_stride_in_bytes,
+        output_pixels, output_w, output_h, output_stride_in_bytes,
+        0,0,1,1,NULL,num_channels,alpha_channel,flags, STBIR_TYPE_UINT8, STBIR_FILTER_DEFAULT, STBIR_FILTER_DEFAULT,
+        edge_wrap_mode, edge_wrap_mode, STBIR_COLORSPACE_SRGB);
+}
+
+STBIRDEF int stbir_resize_uint8_generic( const unsigned char *input_pixels , int input_w , int input_h , int input_stride_in_bytes,
+                                               unsigned char *output_pixels, int output_w, int output_h, int output_stride_in_bytes,
+                                         int num_channels, int alpha_channel, int flags,
+                                         stbir_edge edge_wrap_mode, stbir_filter filter, stbir_colorspace space,
+                                         void *alloc_context)
+{
+    return stbir__resize_arbitrary(alloc_context, input_pixels, input_w, input_h, input_stride_in_bytes,
+        output_pixels, output_w, output_h, output_stride_in_bytes,
+        0,0,1,1,NULL,num_channels,alpha_channel,flags, STBIR_TYPE_UINT8, filter, filter,
+        edge_wrap_mode, edge_wrap_mode, space);
+}
+
+STBIRDEF int stbir_resize_uint16_generic(const stbir_uint16 *input_pixels  , int input_w , int input_h , int input_stride_in_bytes,
+                                               stbir_uint16 *output_pixels , int output_w, int output_h, int output_stride_in_bytes,
+                                         int num_channels, int alpha_channel, int flags,
+                                         stbir_edge edge_wrap_mode, stbir_filter filter, stbir_colorspace space,
+                                         void *alloc_context)
+{
+    return stbir__resize_arbitrary(alloc_context, input_pixels, input_w, input_h, input_stride_in_bytes,
+        output_pixels, output_w, output_h, output_stride_in_bytes,
+        0,0,1,1,NULL,num_channels,alpha_channel,flags, STBIR_TYPE_UINT16, filter, filter,
+        edge_wrap_mode, edge_wrap_mode, space);
+}
+
+
+STBIRDEF int stbir_resize_float_generic( const float *input_pixels         , int input_w , int input_h , int input_stride_in_bytes,
+                                               float *output_pixels        , int output_w, int output_h, int output_stride_in_bytes,
+                                         int num_channels, int alpha_channel, int flags,
+                                         stbir_edge edge_wrap_mode, stbir_filter filter, stbir_colorspace space,
+                                         void *alloc_context)
+{
+    return stbir__resize_arbitrary(alloc_context, input_pixels, input_w, input_h, input_stride_in_bytes,
+        output_pixels, output_w, output_h, output_stride_in_bytes,
+        0,0,1,1,NULL,num_channels,alpha_channel,flags, STBIR_TYPE_FLOAT, filter, filter,
+        edge_wrap_mode, edge_wrap_mode, space);
+}
+
+
+STBIRDEF int stbir_resize(         const void *input_pixels , int input_w , int input_h , int input_stride_in_bytes,
+                                         void *output_pixels, int output_w, int output_h, int output_stride_in_bytes,
+                                   stbir_datatype datatype,
+                                   int num_channels, int alpha_channel, int flags,
+                                   stbir_edge edge_mode_horizontal, stbir_edge edge_mode_vertical,
+                                   stbir_filter filter_horizontal,  stbir_filter filter_vertical,
+                                   stbir_colorspace space, void *alloc_context)
+{
+    return stbir__resize_arbitrary(alloc_context, input_pixels, input_w, input_h, input_stride_in_bytes,
+        output_pixels, output_w, output_h, output_stride_in_bytes,
+        0,0,1,1,NULL,num_channels,alpha_channel,flags, datatype, filter_horizontal, filter_vertical,
+        edge_mode_horizontal, edge_mode_vertical, space);
+}
+
+
+STBIRDEF int stbir_resize_subpixel(const void *input_pixels , int input_w , int input_h , int input_stride_in_bytes,
+                                         void *output_pixels, int output_w, int output_h, int output_stride_in_bytes,
+                                   stbir_datatype datatype,
+                                   int num_channels, int alpha_channel, int flags,
+                                   stbir_edge edge_mode_horizontal, stbir_edge edge_mode_vertical,
+                                   stbir_filter filter_horizontal,  stbir_filter filter_vertical,
+                                   stbir_colorspace space, void *alloc_context,
+                                   float x_scale, float y_scale,
+                                   float x_offset, float y_offset)
+{
+    float transform[4];
+    transform[0] = x_scale;
+    transform[1] = y_scale;
+    transform[2] = x_offset;
+    transform[3] = y_offset;
+    return stbir__resize_arbitrary(alloc_context, input_pixels, input_w, input_h, input_stride_in_bytes,
+        output_pixels, output_w, output_h, output_stride_in_bytes,
+        0,0,1,1,transform,num_channels,alpha_channel,flags, datatype, filter_horizontal, filter_vertical,
+        edge_mode_horizontal, edge_mode_vertical, space);
+}
+
+STBIRDEF int stbir_resize_region(  const void *input_pixels , int input_w , int input_h , int input_stride_in_bytes,
+                                         void *output_pixels, int output_w, int output_h, int output_stride_in_bytes,
+                                   stbir_datatype datatype,
+                                   int num_channels, int alpha_channel, int flags,
+                                   stbir_edge edge_mode_horizontal, stbir_edge edge_mode_vertical,
+                                   stbir_filter filter_horizontal,  stbir_filter filter_vertical,
+                                   stbir_colorspace space, void *alloc_context,
+                                   float s0, float t0, float s1, float t1)
+{
+    return stbir__resize_arbitrary(alloc_context, input_pixels, input_w, input_h, input_stride_in_bytes,
+        output_pixels, output_w, output_h, output_stride_in_bytes,
+        s0,t0,s1,t1,NULL,num_channels,alpha_channel,flags, datatype, filter_horizontal, filter_vertical,
+        edge_mode_horizontal, edge_mode_vertical, space);
+}
+
+#endif // STB_IMAGE_RESIZE_IMPLEMENTATION
+
+/*
+------------------------------------------------------------------------------
+This software is available under 2 licenses -- choose whichever you prefer.
+------------------------------------------------------------------------------
+ALTERNATIVE A - MIT License
+Copyright (c) 2017 Sean Barrett
+Permission is hereby granted, free of charge, to any person obtaining a copy of
+this software and associated documentation files (the "Software"), to deal in
+the Software without restriction, including without limitation the rights to
+use, copy, modify, merge, publish, distribute, sublicense, and/or sell copies
+of the Software, and to permit persons to whom the Software is furnished to do
+so, subject to the following conditions:
+The above copyright notice and this permission notice shall be included in all
+copies or substantial portions of the Software.
+THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+SOFTWARE.
+------------------------------------------------------------------------------
+ALTERNATIVE B - Public Domain (www.unlicense.org)
+This is free and unencumbered software released into the public domain.
+Anyone is free to copy, modify, publish, use, compile, sell, or distribute this
+software, either in source code form or as a compiled binary, for any purpose,
+commercial or non-commercial, and by any means.
+In jurisdictions that recognize copyright laws, the author or authors of this
+software dedicate any and all copyright interest in the software to the public
+domain. We make this dedication for the benefit of the public at large and to
+the detriment of our heirs and successors. We intend this dedication to be an
+overt act of relinquishment in perpetuity of all present and future rights to
+this software under copyright law.
+THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+AUTHORS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
+ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION
+WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+------------------------------------------------------------------------------
+*/
diff --git a/externals/stb/stb_image_write.h b/externals/stb/stb_image_write.h
new file mode 100644
index 000000000..e4b32ed1b
--- /dev/null
+++ b/externals/stb/stb_image_write.h
@@ -0,0 +1,1724 @@
+/* stb_image_write - v1.16 - public domain - http://nothings.org/stb
+   writes out PNG/BMP/TGA/JPEG/HDR images to C stdio - Sean Barrett 2010-2015
+                                     no warranty implied; use at your own risk
+
+   Before #including,
+
+       #define STB_IMAGE_WRITE_IMPLEMENTATION
+
+   in the file that you want to have the implementation.
+
+   Will probably not work correctly with strict-aliasing optimizations.
+
+ABOUT:
+
+   This header file is a library for writing images to C stdio or a callback.
+
+   The PNG output is not optimal; it is 20-50% larger than the file
+   written by a decent optimizing implementation; though providing a custom
+   zlib compress function (see STBIW_ZLIB_COMPRESS) can mitigate that.
+   This library is designed for source code compactness and simplicity,
+   not optimal image file size or run-time performance.
+
+BUILDING:
+
+   You can #define STBIW_ASSERT(x) before the #include to avoid using assert.h.
+   You can #define STBIW_MALLOC(), STBIW_REALLOC(), and STBIW_FREE() to replace
+   malloc,realloc,free.
+   You can #define STBIW_MEMMOVE() to replace memmove()
+   You can #define STBIW_ZLIB_COMPRESS to use a custom zlib-style compress function
+   for PNG compression (instead of the builtin one), it must have the following signature:
+   unsigned char * my_compress(unsigned char *data, int data_len, int *out_len, int quality);
+   The returned data will be freed with STBIW_FREE() (free() by default),
+   so it must be heap allocated with STBIW_MALLOC() (malloc() by default),
+
+UNICODE:
+
+   If compiling for Windows and you wish to use Unicode filenames, compile
+   with
+       #define STBIW_WINDOWS_UTF8
+   and pass utf8-encoded filenames. Call stbiw_convert_wchar_to_utf8 to convert
+   Windows wchar_t filenames to utf8.
+
+USAGE:
+
+   There are five functions, one for each image file format:
+
+     int stbi_write_png(char const *filename, int w, int h, int comp, const void *data, int stride_in_bytes);
+     int stbi_write_bmp(char const *filename, int w, int h, int comp, const void *data);
+     int stbi_write_tga(char const *filename, int w, int h, int comp, const void *data);
+     int stbi_write_jpg(char const *filename, int w, int h, int comp, const void *data, int quality);
+     int stbi_write_hdr(char const *filename, int w, int h, int comp, const float *data);
+
+     void stbi_flip_vertically_on_write(int flag); // flag is non-zero to flip data vertically
+
+   There are also five equivalent functions that use an arbitrary write function. You are
+   expected to open/close your file-equivalent before and after calling these:
+
+     int stbi_write_png_to_func(stbi_write_func *func, void *context, int w, int h, int comp, const void  *data, int stride_in_bytes);
+     int stbi_write_bmp_to_func(stbi_write_func *func, void *context, int w, int h, int comp, const void  *data);
+     int stbi_write_tga_to_func(stbi_write_func *func, void *context, int w, int h, int comp, const void  *data);
+     int stbi_write_hdr_to_func(stbi_write_func *func, void *context, int w, int h, int comp, const float *data);
+     int stbi_write_jpg_to_func(stbi_write_func *func, void *context, int x, int y, int comp, const void *data, int quality);
+
+   where the callback is:
+      void stbi_write_func(void *context, void *data, int size);
+
+   You can configure it with these global variables:
+      int stbi_write_tga_with_rle;             // defaults to true; set to 0 to disable RLE
+      int stbi_write_png_compression_level;    // defaults to 8; set to higher for more compression
+      int stbi_write_force_png_filter;         // defaults to -1; set to 0..5 to force a filter mode
+
+
+   You can define STBI_WRITE_NO_STDIO to disable the file variant of these
+   functions, so the library will not use stdio.h at all. However, this will
+   also disable HDR writing, because it requires stdio for formatted output.
+
+   Each function returns 0 on failure and non-0 on success.
+
+   The functions create an image file defined by the parameters. The image
+   is a rectangle of pixels stored from left-to-right, top-to-bottom.
+   Each pixel contains 'comp' channels of data stored interleaved with 8-bits
+   per channel, in the following order: 1=Y, 2=YA, 3=RGB, 4=RGBA. (Y is
+   monochrome color.) The rectangle is 'w' pixels wide and 'h' pixels tall.
+   The *data pointer points to the first byte of the top-left-most pixel.
+   For PNG, "stride_in_bytes" is the distance in bytes from the first byte of
+   a row of pixels to the first byte of the next row of pixels.
+
+   PNG creates output files with the same number of components as the input.
+   The BMP format expands Y to RGB in the file format and does not
+   output alpha.
+
+   PNG supports writing rectangles of data even when the bytes storing rows of
+   data are not consecutive in memory (e.g. sub-rectangles of a larger image),
+   by supplying the stride between the beginning of adjacent rows. The other
+   formats do not. (Thus you cannot write a native-format BMP through the BMP
+   writer, both because it is in BGR order and because it may have padding
+   at the end of the line.)
+
+   PNG allows you to set the deflate compression level by setting the global
+   variable 'stbi_write_png_compression_level' (it defaults to 8).
+
+   HDR expects linear float data. Since the format is always 32-bit rgb(e)
+   data, alpha (if provided) is discarded, and for monochrome data it is
+   replicated across all three channels.
+
+   TGA supports RLE or non-RLE compressed data. To use non-RLE-compressed
+   data, set the global variable 'stbi_write_tga_with_rle' to 0.
+
+   JPEG does ignore alpha channels in input data; quality is between 1 and 100.
+   Higher quality looks better but results in a bigger image.
+   JPEG baseline (no JPEG progressive).
+
+CREDITS:
+
+
+   Sean Barrett           -    PNG/BMP/TGA
+   Baldur Karlsson        -    HDR
+   Jean-Sebastien Guay    -    TGA monochrome
+   Tim Kelsey             -    misc enhancements
+   Alan Hickman           -    TGA RLE
+   Emmanuel Julien        -    initial file IO callback implementation
+   Jon Olick              -    original jo_jpeg.cpp code
+   Daniel Gibson          -    integrate JPEG, allow external zlib
+   Aarni Koskela          -    allow choosing PNG filter
+
+   bugfixes:
+      github:Chribba
+      Guillaume Chereau
+      github:jry2
+      github:romigrou
+      Sergio Gonzalez
+      Jonas Karlsson
+      Filip Wasil
+      Thatcher Ulrich
+      github:poppolopoppo
+      Patrick Boettcher
+      github:xeekworx
+      Cap Petschulat
+      Simon Rodriguez
+      Ivan Tikhonov
+      github:ignotion
+      Adam Schackart
+      Andrew Kensler
+
+LICENSE
+
+  See end of file for license information.
+
+*/
+
+#ifndef INCLUDE_STB_IMAGE_WRITE_H
+#define INCLUDE_STB_IMAGE_WRITE_H
+
+#include <stdlib.h>
+
+// if STB_IMAGE_WRITE_STATIC causes problems, try defining STBIWDEF to 'inline' or 'static inline'
+#ifndef STBIWDEF
+#ifdef STB_IMAGE_WRITE_STATIC
+#define STBIWDEF  static
+#else
+#ifdef __cplusplus
+#define STBIWDEF  extern "C"
+#else
+#define STBIWDEF  extern
+#endif
+#endif
+#endif
+
+#ifndef STB_IMAGE_WRITE_STATIC  // C++ forbids static forward declarations
+STBIWDEF int stbi_write_tga_with_rle;
+STBIWDEF int stbi_write_png_compression_level;
+STBIWDEF int stbi_write_force_png_filter;
+#endif
+
+#ifndef STBI_WRITE_NO_STDIO
+STBIWDEF int stbi_write_png(char const *filename, int w, int h, int comp, const void  *data, int stride_in_bytes);
+STBIWDEF int stbi_write_bmp(char const *filename, int w, int h, int comp, const void  *data);
+STBIWDEF int stbi_write_tga(char const *filename, int w, int h, int comp, const void  *data);
+STBIWDEF int stbi_write_hdr(char const *filename, int w, int h, int comp, const float *data);
+STBIWDEF int stbi_write_jpg(char const *filename, int x, int y, int comp, const void  *data, int quality);
+
+#ifdef STBIW_WINDOWS_UTF8
+STBIWDEF int stbiw_convert_wchar_to_utf8(char *buffer, size_t bufferlen, const wchar_t* input);
+#endif
+#endif
+
+typedef void stbi_write_func(void *context, void *data, int size);
+
+STBIWDEF int stbi_write_png_to_func(stbi_write_func *func, void *context, int w, int h, int comp, const void  *data, int stride_in_bytes);
+STBIWDEF int stbi_write_bmp_to_func(stbi_write_func *func, void *context, int w, int h, int comp, const void  *data);
+STBIWDEF int stbi_write_tga_to_func(stbi_write_func *func, void *context, int w, int h, int comp, const void  *data);
+STBIWDEF int stbi_write_hdr_to_func(stbi_write_func *func, void *context, int w, int h, int comp, const float *data);
+STBIWDEF int stbi_write_jpg_to_func(stbi_write_func *func, void *context, int x, int y, int comp, const void  *data, int quality);
+
+STBIWDEF void stbi_flip_vertically_on_write(int flip_boolean);
+
+#endif//INCLUDE_STB_IMAGE_WRITE_H
+
+#ifdef STB_IMAGE_WRITE_IMPLEMENTATION
+
+#ifdef _WIN32
+   #ifndef _CRT_SECURE_NO_WARNINGS
+   #define _CRT_SECURE_NO_WARNINGS
+   #endif
+   #ifndef _CRT_NONSTDC_NO_DEPRECATE
+   #define _CRT_NONSTDC_NO_DEPRECATE
+   #endif
+#endif
+
+#ifndef STBI_WRITE_NO_STDIO
+#include <stdio.h>
+#endif // STBI_WRITE_NO_STDIO
+
+#include <stdarg.h>
+#include <stdlib.h>
+#include <string.h>
+#include <math.h>
+
+#if defined(STBIW_MALLOC) && defined(STBIW_FREE) && (defined(STBIW_REALLOC) || defined(STBIW_REALLOC_SIZED))
+// ok
+#elif !defined(STBIW_MALLOC) && !defined(STBIW_FREE) && !defined(STBIW_REALLOC) && !defined(STBIW_REALLOC_SIZED)
+// ok
+#else
+#error "Must define all or none of STBIW_MALLOC, STBIW_FREE, and STBIW_REALLOC (or STBIW_REALLOC_SIZED)."
+#endif
+
+#ifndef STBIW_MALLOC
+#define STBIW_MALLOC(sz)        malloc(sz)
+#define STBIW_REALLOC(p,newsz)  realloc(p,newsz)
+#define STBIW_FREE(p)           free(p)
+#endif
+
+#ifndef STBIW_REALLOC_SIZED
+#define STBIW_REALLOC_SIZED(p,oldsz,newsz) STBIW_REALLOC(p,newsz)
+#endif
+
+
+#ifndef STBIW_MEMMOVE
+#define STBIW_MEMMOVE(a,b,sz) memmove(a,b,sz)
+#endif
+
+
+#ifndef STBIW_ASSERT
+#include <assert.h>
+#define STBIW_ASSERT(x) assert(x)
+#endif
+
+#define STBIW_UCHAR(x) (unsigned char) ((x) & 0xff)
+
+#ifdef STB_IMAGE_WRITE_STATIC
+static int stbi_write_png_compression_level = 8;
+static int stbi_write_tga_with_rle = 1;
+static int stbi_write_force_png_filter = -1;
+#else
+int stbi_write_png_compression_level = 8;
+int stbi_write_tga_with_rle = 1;
+int stbi_write_force_png_filter = -1;
+#endif
+
+static int stbi__flip_vertically_on_write = 0;
+
+STBIWDEF void stbi_flip_vertically_on_write(int flag)
+{
+   stbi__flip_vertically_on_write = flag;
+}
+
+typedef struct
+{
+   stbi_write_func *func;
+   void *context;
+   unsigned char buffer[64];
+   int buf_used;
+} stbi__write_context;
+
+// initialize a callback-based context
+static void stbi__start_write_callbacks(stbi__write_context *s, stbi_write_func *c, void *context)
+{
+   s->func    = c;
+   s->context = context;
+}
+
+#ifndef STBI_WRITE_NO_STDIO
+
+static void stbi__stdio_write(void *context, void *data, int size)
+{
+   fwrite(data,1,size,(FILE*) context);
+}
+
+#if defined(_WIN32) && defined(STBIW_WINDOWS_UTF8)
+#ifdef __cplusplus
+#define STBIW_EXTERN extern "C"
+#else
+#define STBIW_EXTERN extern
+#endif
+STBIW_EXTERN __declspec(dllimport) int __stdcall MultiByteToWideChar(unsigned int cp, unsigned long flags, const char *str, int cbmb, wchar_t *widestr, int cchwide);
+STBIW_EXTERN __declspec(dllimport) int __stdcall WideCharToMultiByte(unsigned int cp, unsigned long flags, const wchar_t *widestr, int cchwide, char *str, int cbmb, const char *defchar, int *used_default);
+
+STBIWDEF int stbiw_convert_wchar_to_utf8(char *buffer, size_t bufferlen, const wchar_t* input)
+{
+   return WideCharToMultiByte(65001 /* UTF8 */, 0, input, -1, buffer, (int) bufferlen, NULL, NULL);
+}
+#endif
+
+static FILE *stbiw__fopen(char const *filename, char const *mode)
+{
+   FILE *f;
+#if defined(_WIN32) && defined(STBIW_WINDOWS_UTF8)
+   wchar_t wMode[64];
+   wchar_t wFilename[1024];
+   if (0 == MultiByteToWideChar(65001 /* UTF8 */, 0, filename, -1, wFilename, sizeof(wFilename)/sizeof(*wFilename)))
+      return 0;
+
+   if (0 == MultiByteToWideChar(65001 /* UTF8 */, 0, mode, -1, wMode, sizeof(wMode)/sizeof(*wMode)))
+      return 0;
+
+#if defined(_MSC_VER) && _MSC_VER >= 1400
+   if (0 != _wfopen_s(&f, wFilename, wMode))
+      f = 0;
+#else
+   f = _wfopen(wFilename, wMode);
+#endif
+
+#elif defined(_MSC_VER) && _MSC_VER >= 1400
+   if (0 != fopen_s(&f, filename, mode))
+      f=0;
+#else
+   f = fopen(filename, mode);
+#endif
+   return f;
+}
+
+static int stbi__start_write_file(stbi__write_context *s, const char *filename)
+{
+   FILE *f = stbiw__fopen(filename, "wb");
+   stbi__start_write_callbacks(s, stbi__stdio_write, (void *) f);
+   return f != NULL;
+}
+
+static void stbi__end_write_file(stbi__write_context *s)
+{
+   fclose((FILE *)s->context);
+}
+
+#endif // !STBI_WRITE_NO_STDIO
+
+typedef unsigned int stbiw_uint32;
+typedef int stb_image_write_test[sizeof(stbiw_uint32)==4 ? 1 : -1];
+
+static void stbiw__writefv(stbi__write_context *s, const char *fmt, va_list v)
+{
+   while (*fmt) {
+      switch (*fmt++) {
+         case ' ': break;
+         case '1': { unsigned char x = STBIW_UCHAR(va_arg(v, int));
+                     s->func(s->context,&x,1);
+                     break; }
+         case '2': { int x = va_arg(v,int);
+                     unsigned char b[2];
+                     b[0] = STBIW_UCHAR(x);
+                     b[1] = STBIW_UCHAR(x>>8);
+                     s->func(s->context,b,2);
+                     break; }
+         case '4': { stbiw_uint32 x = va_arg(v,int);
+                     unsigned char b[4];
+                     b[0]=STBIW_UCHAR(x);
+                     b[1]=STBIW_UCHAR(x>>8);
+                     b[2]=STBIW_UCHAR(x>>16);
+                     b[3]=STBIW_UCHAR(x>>24);
+                     s->func(s->context,b,4);
+                     break; }
+         default:
+            STBIW_ASSERT(0);
+            return;
+      }
+   }
+}
+
+static void stbiw__writef(stbi__write_context *s, const char *fmt, ...)
+{
+   va_list v;
+   va_start(v, fmt);
+   stbiw__writefv(s, fmt, v);
+   va_end(v);
+}
+
+static void stbiw__write_flush(stbi__write_context *s)
+{
+   if (s->buf_used) {
+      s->func(s->context, &s->buffer, s->buf_used);
+      s->buf_used = 0;
+   }
+}
+
+static void stbiw__putc(stbi__write_context *s, unsigned char c)
+{
+   s->func(s->context, &c, 1);
+}
+
+static void stbiw__write1(stbi__write_context *s, unsigned char a)
+{
+   if ((size_t)s->buf_used + 1 > sizeof(s->buffer))
+      stbiw__write_flush(s);
+   s->buffer[s->buf_used++] = a;
+}
+
+static void stbiw__write3(stbi__write_context *s, unsigned char a, unsigned char b, unsigned char c)
+{
+   int n;
+   if ((size_t)s->buf_used + 3 > sizeof(s->buffer))
+      stbiw__write_flush(s);
+   n = s->buf_used;
+   s->buf_used = n+3;
+   s->buffer[n+0] = a;
+   s->buffer[n+1] = b;
+   s->buffer[n+2] = c;
+}
+
+static void stbiw__write_pixel(stbi__write_context *s, int rgb_dir, int comp, int write_alpha, int expand_mono, unsigned char *d)
+{
+   unsigned char bg[3] = { 255, 0, 255}, px[3];
+   int k;
+
+   if (write_alpha < 0)
+      stbiw__write1(s, d[comp - 1]);
+
+   switch (comp) {
+      case 2: // 2 pixels = mono + alpha, alpha is written separately, so same as 1-channel case
+      case 1:
+         if (expand_mono)
+            stbiw__write3(s, d[0], d[0], d[0]); // monochrome bmp
+         else
+            stbiw__write1(s, d[0]);  // monochrome TGA
+         break;
+      case 4:
+         if (!write_alpha) {
+            // composite against pink background
+            for (k = 0; k < 3; ++k)
+               px[k] = bg[k] + ((d[k] - bg[k]) * d[3]) / 255;
+            stbiw__write3(s, px[1 - rgb_dir], px[1], px[1 + rgb_dir]);
+            break;
+         }
+         /* FALLTHROUGH */
+      case 3:
+         stbiw__write3(s, d[1 - rgb_dir], d[1], d[1 + rgb_dir]);
+         break;
+   }
+   if (write_alpha > 0)
+      stbiw__write1(s, d[comp - 1]);
+}
+
+static void stbiw__write_pixels(stbi__write_context *s, int rgb_dir, int vdir, int x, int y, int comp, void *data, int write_alpha, int scanline_pad, int expand_mono)
+{
+   stbiw_uint32 zero = 0;
+   int i,j, j_end;
+
+   if (y <= 0)
+      return;
+
+   if (stbi__flip_vertically_on_write)
+      vdir *= -1;
+
+   if (vdir < 0) {
+      j_end = -1; j = y-1;
+   } else {
+      j_end =  y; j = 0;
+   }
+
+   for (; j != j_end; j += vdir) {
+      for (i=0; i < x; ++i) {
+         unsigned char *d = (unsigned char *) data + (j*x+i)*comp;
+         stbiw__write_pixel(s, rgb_dir, comp, write_alpha, expand_mono, d);
+      }
+      stbiw__write_flush(s);
+      s->func(s->context, &zero, scanline_pad);
+   }
+}
+
+static int stbiw__outfile(stbi__write_context *s, int rgb_dir, int vdir, int x, int y, int comp, int expand_mono, void *data, int alpha, int pad, const char *fmt, ...)
+{
+   if (y < 0 || x < 0) {
+      return 0;
+   } else {
+      va_list v;
+      va_start(v, fmt);
+      stbiw__writefv(s, fmt, v);
+      va_end(v);
+      stbiw__write_pixels(s,rgb_dir,vdir,x,y,comp,data,alpha,pad, expand_mono);
+      return 1;
+   }
+}
+
+static int stbi_write_bmp_core(stbi__write_context *s, int x, int y, int comp, const void *data)
+{
+   if (comp != 4) {
+      // write RGB bitmap
+      int pad = (-x*3) & 3;
+      return stbiw__outfile(s,-1,-1,x,y,comp,1,(void *) data,0,pad,
+              "11 4 22 4" "4 44 22 444444",
+              'B', 'M', 14+40+(x*3+pad)*y, 0,0, 14+40,  // file header
+               40, x,y, 1,24, 0,0,0,0,0,0);             // bitmap header
+   } else {
+      // RGBA bitmaps need a v4 header
+      // use BI_BITFIELDS mode with 32bpp and alpha mask
+      // (straight BI_RGB with alpha mask doesn't work in most readers)
+      return stbiw__outfile(s,-1,-1,x,y,comp,1,(void *)data,1,0,
+         "11 4 22 4" "4 44 22 444444 4444 4 444 444 444 444",
+         'B', 'M', 14+108+x*y*4, 0, 0, 14+108, // file header
+         108, x,y, 1,32, 3,0,0,0,0,0, 0xff0000,0xff00,0xff,0xff000000u, 0, 0,0,0, 0,0,0, 0,0,0, 0,0,0); // bitmap V4 header
+   }
+}
+
+STBIWDEF int stbi_write_bmp_to_func(stbi_write_func *func, void *context, int x, int y, int comp, const void *data)
+{
+   stbi__write_context s = { 0 };
+   stbi__start_write_callbacks(&s, func, context);
+   return stbi_write_bmp_core(&s, x, y, comp, data);
+}
+
+#ifndef STBI_WRITE_NO_STDIO
+STBIWDEF int stbi_write_bmp(char const *filename, int x, int y, int comp, const void *data)
+{
+   stbi__write_context s = { 0 };
+   if (stbi__start_write_file(&s,filename)) {
+      int r = stbi_write_bmp_core(&s, x, y, comp, data);
+      stbi__end_write_file(&s);
+      return r;
+   } else
+      return 0;
+}
+#endif //!STBI_WRITE_NO_STDIO
+
+static int stbi_write_tga_core(stbi__write_context *s, int x, int y, int comp, void *data)
+{
+   int has_alpha = (comp == 2 || comp == 4);
+   int colorbytes = has_alpha ? comp-1 : comp;
+   int format = colorbytes < 2 ? 3 : 2; // 3 color channels (RGB/RGBA) = 2, 1 color channel (Y/YA) = 3
+
+   if (y < 0 || x < 0)
+      return 0;
+
+   if (!stbi_write_tga_with_rle) {
+      return stbiw__outfile(s, -1, -1, x, y, comp, 0, (void *) data, has_alpha, 0,
+         "111 221 2222 11", 0, 0, format, 0, 0, 0, 0, 0, x, y, (colorbytes + has_alpha) * 8, has_alpha * 8);
+   } else {
+      int i,j,k;
+      int jend, jdir;
+
+      stbiw__writef(s, "111 221 2222 11", 0,0,format+8, 0,0,0, 0,0,x,y, (colorbytes + has_alpha) * 8, has_alpha * 8);
+
+      if (stbi__flip_vertically_on_write) {
+         j = 0;
+         jend = y;
+         jdir = 1;
+      } else {
+         j = y-1;
+         jend = -1;
+         jdir = -1;
+      }
+      for (; j != jend; j += jdir) {
+         unsigned char *row = (unsigned char *) data + j * x * comp;
+         int len;
+
+         for (i = 0; i < x; i += len) {
+            unsigned char *begin = row + i * comp;
+            int diff = 1;
+            len = 1;
+
+            if (i < x - 1) {
+               ++len;
+               diff = memcmp(begin, row + (i + 1) * comp, comp);
+               if (diff) {
+                  const unsigned char *prev = begin;
+                  for (k = i + 2; k < x && len < 128; ++k) {
+                     if (memcmp(prev, row + k * comp, comp)) {
+                        prev += comp;
+                        ++len;
+                     } else {
+                        --len;
+                        break;
+                     }
+                  }
+               } else {
+                  for (k = i + 2; k < x && len < 128; ++k) {
+                     if (!memcmp(begin, row + k * comp, comp)) {
+                        ++len;
+                     } else {
+                        break;
+                     }
+                  }
+               }
+            }
+
+            if (diff) {
+               unsigned char header = STBIW_UCHAR(len - 1);
+               stbiw__write1(s, header);
+               for (k = 0; k < len; ++k) {
+                  stbiw__write_pixel(s, -1, comp, has_alpha, 0, begin + k * comp);
+               }
+            } else {
+               unsigned char header = STBIW_UCHAR(len - 129);
+               stbiw__write1(s, header);
+               stbiw__write_pixel(s, -1, comp, has_alpha, 0, begin);
+            }
+         }
+      }
+      stbiw__write_flush(s);
+   }
+   return 1;
+}
+
+STBIWDEF int stbi_write_tga_to_func(stbi_write_func *func, void *context, int x, int y, int comp, const void *data)
+{
+   stbi__write_context s = { 0 };
+   stbi__start_write_callbacks(&s, func, context);
+   return stbi_write_tga_core(&s, x, y, comp, (void *) data);
+}
+
+#ifndef STBI_WRITE_NO_STDIO
+STBIWDEF int stbi_write_tga(char const *filename, int x, int y, int comp, const void *data)
+{
+   stbi__write_context s = { 0 };
+   if (stbi__start_write_file(&s,filename)) {
+      int r = stbi_write_tga_core(&s, x, y, comp, (void *) data);
+      stbi__end_write_file(&s);
+      return r;
+   } else
+      return 0;
+}
+#endif
+
+// *************************************************************************************************
+// Radiance RGBE HDR writer
+// by Baldur Karlsson
+
+#define stbiw__max(a, b)  ((a) > (b) ? (a) : (b))
+
+#ifndef STBI_WRITE_NO_STDIO
+
+static void stbiw__linear_to_rgbe(unsigned char *rgbe, float *linear)
+{
+   int exponent;
+   float maxcomp = stbiw__max(linear[0], stbiw__max(linear[1], linear[2]));
+
+   if (maxcomp < 1e-32f) {
+      rgbe[0] = rgbe[1] = rgbe[2] = rgbe[3] = 0;
+   } else {
+      float normalize = (float) frexp(maxcomp, &exponent) * 256.0f/maxcomp;
+
+      rgbe[0] = (unsigned char)(linear[0] * normalize);
+      rgbe[1] = (unsigned char)(linear[1] * normalize);
+      rgbe[2] = (unsigned char)(linear[2] * normalize);
+      rgbe[3] = (unsigned char)(exponent + 128);
+   }
+}
+
+static void stbiw__write_run_data(stbi__write_context *s, int length, unsigned char databyte)
+{
+   unsigned char lengthbyte = STBIW_UCHAR(length+128);
+   STBIW_ASSERT(length+128 <= 255);
+   s->func(s->context, &lengthbyte, 1);
+   s->func(s->context, &databyte, 1);
+}
+
+static void stbiw__write_dump_data(stbi__write_context *s, int length, unsigned char *data)
+{
+   unsigned char lengthbyte = STBIW_UCHAR(length);
+   STBIW_ASSERT(length <= 128); // inconsistent with spec but consistent with official code
+   s->func(s->context, &lengthbyte, 1);
+   s->func(s->context, data, length);
+}
+
+static void stbiw__write_hdr_scanline(stbi__write_context *s, int width, int ncomp, unsigned char *scratch, float *scanline)
+{
+   unsigned char scanlineheader[4] = { 2, 2, 0, 0 };
+   unsigned char rgbe[4];
+   float linear[3];
+   int x;
+
+   scanlineheader[2] = (width&0xff00)>>8;
+   scanlineheader[3] = (width&0x00ff);
+
+   /* skip RLE for images too small or large */
+   if (width < 8 || width >= 32768) {
+      for (x=0; x < width; x++) {
+         switch (ncomp) {
+            case 4: /* fallthrough */
+            case 3: linear[2] = scanline[x*ncomp + 2];
+                    linear[1] = scanline[x*ncomp + 1];
+                    linear[0] = scanline[x*ncomp + 0];
+                    break;
+            default:
+                    linear[0] = linear[1] = linear[2] = scanline[x*ncomp + 0];
+                    break;
+         }
+         stbiw__linear_to_rgbe(rgbe, linear);
+         s->func(s->context, rgbe, 4);
+      }
+   } else {
+      int c,r;
+      /* encode into scratch buffer */
+      for (x=0; x < width; x++) {
+         switch(ncomp) {
+            case 4: /* fallthrough */
+            case 3: linear[2] = scanline[x*ncomp + 2];
+                    linear[1] = scanline[x*ncomp + 1];
+                    linear[0] = scanline[x*ncomp + 0];
+                    break;
+            default:
+                    linear[0] = linear[1] = linear[2] = scanline[x*ncomp + 0];
+                    break;
+         }
+         stbiw__linear_to_rgbe(rgbe, linear);
+         scratch[x + width*0] = rgbe[0];
+         scratch[x + width*1] = rgbe[1];
+         scratch[x + width*2] = rgbe[2];
+         scratch[x + width*3] = rgbe[3];
+      }
+
+      s->func(s->context, scanlineheader, 4);
+
+      /* RLE each component separately */
+      for (c=0; c < 4; c++) {
+         unsigned char *comp = &scratch[width*c];
+
+         x = 0;
+         while (x < width) {
+            // find first run
+            r = x;
+            while (r+2 < width) {
+               if (comp[r] == comp[r+1] && comp[r] == comp[r+2])
+                  break;
+               ++r;
+            }
+            if (r+2 >= width)
+               r = width;
+            // dump up to first run
+            while (x < r) {
+               int len = r-x;
+               if (len > 128) len = 128;
+               stbiw__write_dump_data(s, len, &comp[x]);
+               x += len;
+            }
+            // if there's a run, output it
+            if (r+2 < width) { // same test as what we break out of in search loop, so only true if we break'd
+               // find next byte after run
+               while (r < width && comp[r] == comp[x])
+                  ++r;
+               // output run up to r
+               while (x < r) {
+                  int len = r-x;
+                  if (len > 127) len = 127;
+                  stbiw__write_run_data(s, len, comp[x]);
+                  x += len;
+               }
+            }
+         }
+      }
+   }
+}
+
+static int stbi_write_hdr_core(stbi__write_context *s, int x, int y, int comp, float *data)
+{
+   if (y <= 0 || x <= 0 || data == NULL)
+      return 0;
+   else {
+      // Each component is stored separately. Allocate scratch space for full output scanline.
+      unsigned char *scratch = (unsigned char *) STBIW_MALLOC(x*4);
+      int i, len;
+      char buffer[128];
+      char header[] = "#?RADIANCE\n# Written by stb_image_write.h\nFORMAT=32-bit_rle_rgbe\n";
+      s->func(s->context, header, sizeof(header)-1);
+
+#ifdef __STDC_LIB_EXT1__
+      len = sprintf_s(buffer, sizeof(buffer), "EXPOSURE=          1.0000000000000\n\n-Y %d +X %d\n", y, x);
+#else
+      len = sprintf(buffer, "EXPOSURE=          1.0000000000000\n\n-Y %d +X %d\n", y, x);
+#endif
+      s->func(s->context, buffer, len);
+
+      for(i=0; i < y; i++)
+         stbiw__write_hdr_scanline(s, x, comp, scratch, data + comp*x*(stbi__flip_vertically_on_write ? y-1-i : i));
+      STBIW_FREE(scratch);
+      return 1;
+   }
+}
+
+STBIWDEF int stbi_write_hdr_to_func(stbi_write_func *func, void *context, int x, int y, int comp, const float *data)
+{
+   stbi__write_context s = { 0 };
+   stbi__start_write_callbacks(&s, func, context);
+   return stbi_write_hdr_core(&s, x, y, comp, (float *) data);
+}
+
+STBIWDEF int stbi_write_hdr(char const *filename, int x, int y, int comp, const float *data)
+{
+   stbi__write_context s = { 0 };
+   if (stbi__start_write_file(&s,filename)) {
+      int r = stbi_write_hdr_core(&s, x, y, comp, (float *) data);
+      stbi__end_write_file(&s);
+      return r;
+   } else
+      return 0;
+}
+#endif // STBI_WRITE_NO_STDIO
+
+
+//////////////////////////////////////////////////////////////////////////////
+//
+// PNG writer
+//
+
+#ifndef STBIW_ZLIB_COMPRESS
+// stretchy buffer; stbiw__sbpush() == vector<>::push_back() -- stbiw__sbcount() == vector<>::size()
+#define stbiw__sbraw(a) ((int *) (void *) (a) - 2)
+#define stbiw__sbm(a)   stbiw__sbraw(a)[0]
+#define stbiw__sbn(a)   stbiw__sbraw(a)[1]
+
+#define stbiw__sbneedgrow(a,n)  ((a)==0 || stbiw__sbn(a)+n >= stbiw__sbm(a))
+#define stbiw__sbmaybegrow(a,n) (stbiw__sbneedgrow(a,(n)) ? stbiw__sbgrow(a,n) : 0)
+#define stbiw__sbgrow(a,n)  stbiw__sbgrowf((void **) &(a), (n), sizeof(*(a)))
+
+#define stbiw__sbpush(a, v)      (stbiw__sbmaybegrow(a,1), (a)[stbiw__sbn(a)++] = (v))
+#define stbiw__sbcount(a)        ((a) ? stbiw__sbn(a) : 0)
+#define stbiw__sbfree(a)         ((a) ? STBIW_FREE(stbiw__sbraw(a)),0 : 0)
+
+static void *stbiw__sbgrowf(void **arr, int increment, int itemsize)
+{
+   int m = *arr ? 2*stbiw__sbm(*arr)+increment : increment+1;
+   void *p = STBIW_REALLOC_SIZED(*arr ? stbiw__sbraw(*arr) : 0, *arr ? (stbiw__sbm(*arr)*itemsize + sizeof(int)*2) : 0, itemsize * m + sizeof(int)*2);
+   STBIW_ASSERT(p);
+   if (p) {
+      if (!*arr) ((int *) p)[1] = 0;
+      *arr = (void *) ((int *) p + 2);
+      stbiw__sbm(*arr) = m;
+   }
+   return *arr;
+}
+
+static unsigned char *stbiw__zlib_flushf(unsigned char *data, unsigned int *bitbuffer, int *bitcount)
+{
+   while (*bitcount >= 8) {
+      stbiw__sbpush(data, STBIW_UCHAR(*bitbuffer));
+      *bitbuffer >>= 8;
+      *bitcount -= 8;
+   }
+   return data;
+}
+
+static int stbiw__zlib_bitrev(int code, int codebits)
+{
+   int res=0;
+   while (codebits--) {
+      res = (res << 1) | (code & 1);
+      code >>= 1;
+   }
+   return res;
+}
+
+static unsigned int stbiw__zlib_countm(unsigned char *a, unsigned char *b, int limit)
+{
+   int i;
+   for (i=0; i < limit && i < 258; ++i)
+      if (a[i] != b[i]) break;
+   return i;
+}
+
+static unsigned int stbiw__zhash(unsigned char *data)
+{
+   stbiw_uint32 hash = data[0] + (data[1] << 8) + (data[2] << 16);
+   hash ^= hash << 3;
+   hash += hash >> 5;
+   hash ^= hash << 4;
+   hash += hash >> 17;
+   hash ^= hash << 25;
+   hash += hash >> 6;
+   return hash;
+}
+
+#define stbiw__zlib_flush() (out = stbiw__zlib_flushf(out, &bitbuf, &bitcount))
+#define stbiw__zlib_add(code,codebits) \
+      (bitbuf |= (code) << bitcount, bitcount += (codebits), stbiw__zlib_flush())
+#define stbiw__zlib_huffa(b,c)  stbiw__zlib_add(stbiw__zlib_bitrev(b,c),c)
+// default huffman tables
+#define stbiw__zlib_huff1(n)  stbiw__zlib_huffa(0x30 + (n), 8)
+#define stbiw__zlib_huff2(n)  stbiw__zlib_huffa(0x190 + (n)-144, 9)
+#define stbiw__zlib_huff3(n)  stbiw__zlib_huffa(0 + (n)-256,7)
+#define stbiw__zlib_huff4(n)  stbiw__zlib_huffa(0xc0 + (n)-280,8)
+#define stbiw__zlib_huff(n)  ((n) <= 143 ? stbiw__zlib_huff1(n) : (n) <= 255 ? stbiw__zlib_huff2(n) : (n) <= 279 ? stbiw__zlib_huff3(n) : stbiw__zlib_huff4(n))
+#define stbiw__zlib_huffb(n) ((n) <= 143 ? stbiw__zlib_huff1(n) : stbiw__zlib_huff2(n))
+
+#define stbiw__ZHASH   16384
+
+#endif // STBIW_ZLIB_COMPRESS
+
+STBIWDEF unsigned char * stbi_zlib_compress(unsigned char *data, int data_len, int *out_len, int quality)
+{
+#ifdef STBIW_ZLIB_COMPRESS
+   // user provided a zlib compress implementation, use that
+   return STBIW_ZLIB_COMPRESS(data, data_len, out_len, quality);
+#else // use builtin
+   static unsigned short lengthc[] = { 3,4,5,6,7,8,9,10,11,13,15,17,19,23,27,31,35,43,51,59,67,83,99,115,131,163,195,227,258, 259 };
+   static unsigned char  lengtheb[]= { 0,0,0,0,0,0,0, 0, 1, 1, 1, 1, 2, 2, 2, 2, 3, 3, 3, 3, 4, 4, 4,  4,  5,  5,  5,  5,  0 };
+   static unsigned short distc[]   = { 1,2,3,4,5,7,9,13,17,25,33,49,65,97,129,193,257,385,513,769,1025,1537,2049,3073,4097,6145,8193,12289,16385,24577, 32768 };
+   static unsigned char  disteb[]  = { 0,0,0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7,8,8,9,9,10,10,11,11,12,12,13,13 };
+   unsigned int bitbuf=0;
+   int i,j, bitcount=0;
+   unsigned char *out = NULL;
+   unsigned char ***hash_table = (unsigned char***) STBIW_MALLOC(stbiw__ZHASH * sizeof(unsigned char**));
+   if (hash_table == NULL)
+      return NULL;
+   if (quality < 5) quality = 5;
+
+   stbiw__sbpush(out, 0x78);   // DEFLATE 32K window
+   stbiw__sbpush(out, 0x5e);   // FLEVEL = 1
+   stbiw__zlib_add(1,1);  // BFINAL = 1
+   stbiw__zlib_add(1,2);  // BTYPE = 1 -- fixed huffman
+
+   for (i=0; i < stbiw__ZHASH; ++i)
+      hash_table[i] = NULL;
+
+   i=0;
+   while (i < data_len-3) {
+      // hash next 3 bytes of data to be compressed
+      int h = stbiw__zhash(data+i)&(stbiw__ZHASH-1), best=3;
+      unsigned char *bestloc = 0;
+      unsigned char **hlist = hash_table[h];
+      int n = stbiw__sbcount(hlist);
+      for (j=0; j < n; ++j) {
+         if (hlist[j]-data > i-32768) { // if entry lies within window
+            int d = stbiw__zlib_countm(hlist[j], data+i, data_len-i);
+            if (d >= best) { best=d; bestloc=hlist[j]; }
+         }
+      }
+      // when hash table entry is too long, delete half the entries
+      if (hash_table[h] && stbiw__sbn(hash_table[h]) == 2*quality) {
+         STBIW_MEMMOVE(hash_table[h], hash_table[h]+quality, sizeof(hash_table[h][0])*quality);
+         stbiw__sbn(hash_table[h]) = quality;
+      }
+      stbiw__sbpush(hash_table[h],data+i);
+
+      if (bestloc) {
+         // "lazy matching" - check match at *next* byte, and if it's better, do cur byte as literal
+         h = stbiw__zhash(data+i+1)&(stbiw__ZHASH-1);
+         hlist = hash_table[h];
+         n = stbiw__sbcount(hlist);
+         for (j=0; j < n; ++j) {
+            if (hlist[j]-data > i-32767) {
+               int e = stbiw__zlib_countm(hlist[j], data+i+1, data_len-i-1);
+               if (e > best) { // if next match is better, bail on current match
+                  bestloc = NULL;
+                  break;
+               }
+            }
+         }
+      }
+
+      if (bestloc) {
+         int d = (int) (data+i - bestloc); // distance back
+         STBIW_ASSERT(d <= 32767 && best <= 258);
+         for (j=0; best > lengthc[j+1]-1; ++j);
+         stbiw__zlib_huff(j+257);
+         if (lengtheb[j]) stbiw__zlib_add(best - lengthc[j], lengtheb[j]);
+         for (j=0; d > distc[j+1]-1; ++j);
+         stbiw__zlib_add(stbiw__zlib_bitrev(j,5),5);
+         if (disteb[j]) stbiw__zlib_add(d - distc[j], disteb[j]);
+         i += best;
+      } else {
+         stbiw__zlib_huffb(data[i]);
+         ++i;
+      }
+   }
+   // write out final bytes
+   for (;i < data_len; ++i)
+      stbiw__zlib_huffb(data[i]);
+   stbiw__zlib_huff(256); // end of block
+   // pad with 0 bits to byte boundary
+   while (bitcount)
+      stbiw__zlib_add(0,1);
+
+   for (i=0; i < stbiw__ZHASH; ++i)
+      (void) stbiw__sbfree(hash_table[i]);
+   STBIW_FREE(hash_table);
+
+   // store uncompressed instead if compression was worse
+   if (stbiw__sbn(out) > data_len + 2 + ((data_len+32766)/32767)*5) {
+      stbiw__sbn(out) = 2;  // truncate to DEFLATE 32K window and FLEVEL = 1
+      for (j = 0; j < data_len;) {
+         int blocklen = data_len - j;
+         if (blocklen > 32767) blocklen = 32767;
+         stbiw__sbpush(out, data_len - j == blocklen); // BFINAL = ?, BTYPE = 0 -- no compression
+         stbiw__sbpush(out, STBIW_UCHAR(blocklen)); // LEN
+         stbiw__sbpush(out, STBIW_UCHAR(blocklen >> 8));
+         stbiw__sbpush(out, STBIW_UCHAR(~blocklen)); // NLEN
+         stbiw__sbpush(out, STBIW_UCHAR(~blocklen >> 8));
+         memcpy(out+stbiw__sbn(out), data+j, blocklen);
+         stbiw__sbn(out) += blocklen;
+         j += blocklen;
+      }
+   }
+
+   {
+      // compute adler32 on input
+      unsigned int s1=1, s2=0;
+      int blocklen = (int) (data_len % 5552);
+      j=0;
+      while (j < data_len) {
+         for (i=0; i < blocklen; ++i) { s1 += data[j+i]; s2 += s1; }
+         s1 %= 65521; s2 %= 65521;
+         j += blocklen;
+         blocklen = 5552;
+      }
+      stbiw__sbpush(out, STBIW_UCHAR(s2 >> 8));
+      stbiw__sbpush(out, STBIW_UCHAR(s2));
+      stbiw__sbpush(out, STBIW_UCHAR(s1 >> 8));
+      stbiw__sbpush(out, STBIW_UCHAR(s1));
+   }
+   *out_len = stbiw__sbn(out);
+   // make returned pointer freeable
+   STBIW_MEMMOVE(stbiw__sbraw(out), out, *out_len);
+   return (unsigned char *) stbiw__sbraw(out);
+#endif // STBIW_ZLIB_COMPRESS
+}
+
+static unsigned int stbiw__crc32(unsigned char *buffer, int len)
+{
+#ifdef STBIW_CRC32
+    return STBIW_CRC32(buffer, len);
+#else
+   static unsigned int crc_table[256] =
+   {
+      0x00000000, 0x77073096, 0xEE0E612C, 0x990951BA, 0x076DC419, 0x706AF48F, 0xE963A535, 0x9E6495A3,
+      0x0eDB8832, 0x79DCB8A4, 0xE0D5E91E, 0x97D2D988, 0x09B64C2B, 0x7EB17CBD, 0xE7B82D07, 0x90BF1D91,
+      0x1DB71064, 0x6AB020F2, 0xF3B97148, 0x84BE41DE, 0x1ADAD47D, 0x6DDDE4EB, 0xF4D4B551, 0x83D385C7,
+      0x136C9856, 0x646BA8C0, 0xFD62F97A, 0x8A65C9EC, 0x14015C4F, 0x63066CD9, 0xFA0F3D63, 0x8D080DF5,
+      0x3B6E20C8, 0x4C69105E, 0xD56041E4, 0xA2677172, 0x3C03E4D1, 0x4B04D447, 0xD20D85FD, 0xA50AB56B,
+      0x35B5A8FA, 0x42B2986C, 0xDBBBC9D6, 0xACBCF940, 0x32D86CE3, 0x45DF5C75, 0xDCD60DCF, 0xABD13D59,
+      0x26D930AC, 0x51DE003A, 0xC8D75180, 0xBFD06116, 0x21B4F4B5, 0x56B3C423, 0xCFBA9599, 0xB8BDA50F,
+      0x2802B89E, 0x5F058808, 0xC60CD9B2, 0xB10BE924, 0x2F6F7C87, 0x58684C11, 0xC1611DAB, 0xB6662D3D,
+      0x76DC4190, 0x01DB7106, 0x98D220BC, 0xEFD5102A, 0x71B18589, 0x06B6B51F, 0x9FBFE4A5, 0xE8B8D433,
+      0x7807C9A2, 0x0F00F934, 0x9609A88E, 0xE10E9818, 0x7F6A0DBB, 0x086D3D2D, 0x91646C97, 0xE6635C01,
+      0x6B6B51F4, 0x1C6C6162, 0x856530D8, 0xF262004E, 0x6C0695ED, 0x1B01A57B, 0x8208F4C1, 0xF50FC457,
+      0x65B0D9C6, 0x12B7E950, 0x8BBEB8EA, 0xFCB9887C, 0x62DD1DDF, 0x15DA2D49, 0x8CD37CF3, 0xFBD44C65,
+      0x4DB26158, 0x3AB551CE, 0xA3BC0074, 0xD4BB30E2, 0x4ADFA541, 0x3DD895D7, 0xA4D1C46D, 0xD3D6F4FB,
+      0x4369E96A, 0x346ED9FC, 0xAD678846, 0xDA60B8D0, 0x44042D73, 0x33031DE5, 0xAA0A4C5F, 0xDD0D7CC9,
+      0x5005713C, 0x270241AA, 0xBE0B1010, 0xC90C2086, 0x5768B525, 0x206F85B3, 0xB966D409, 0xCE61E49F,
+      0x5EDEF90E, 0x29D9C998, 0xB0D09822, 0xC7D7A8B4, 0x59B33D17, 0x2EB40D81, 0xB7BD5C3B, 0xC0BA6CAD,
+      0xEDB88320, 0x9ABFB3B6, 0x03B6E20C, 0x74B1D29A, 0xEAD54739, 0x9DD277AF, 0x04DB2615, 0x73DC1683,
+      0xE3630B12, 0x94643B84, 0x0D6D6A3E, 0x7A6A5AA8, 0xE40ECF0B, 0x9309FF9D, 0x0A00AE27, 0x7D079EB1,
+      0xF00F9344, 0x8708A3D2, 0x1E01F268, 0x6906C2FE, 0xF762575D, 0x806567CB, 0x196C3671, 0x6E6B06E7,
+      0xFED41B76, 0x89D32BE0, 0x10DA7A5A, 0x67DD4ACC, 0xF9B9DF6F, 0x8EBEEFF9, 0x17B7BE43, 0x60B08ED5,
+      0xD6D6A3E8, 0xA1D1937E, 0x38D8C2C4, 0x4FDFF252, 0xD1BB67F1, 0xA6BC5767, 0x3FB506DD, 0x48B2364B,
+      0xD80D2BDA, 0xAF0A1B4C, 0x36034AF6, 0x41047A60, 0xDF60EFC3, 0xA867DF55, 0x316E8EEF, 0x4669BE79,
+      0xCB61B38C, 0xBC66831A, 0x256FD2A0, 0x5268E236, 0xCC0C7795, 0xBB0B4703, 0x220216B9, 0x5505262F,
+      0xC5BA3BBE, 0xB2BD0B28, 0x2BB45A92, 0x5CB36A04, 0xC2D7FFA7, 0xB5D0CF31, 0x2CD99E8B, 0x5BDEAE1D,
+      0x9B64C2B0, 0xEC63F226, 0x756AA39C, 0x026D930A, 0x9C0906A9, 0xEB0E363F, 0x72076785, 0x05005713,
+      0x95BF4A82, 0xE2B87A14, 0x7BB12BAE, 0x0CB61B38, 0x92D28E9B, 0xE5D5BE0D, 0x7CDCEFB7, 0x0BDBDF21,
+      0x86D3D2D4, 0xF1D4E242, 0x68DDB3F8, 0x1FDA836E, 0x81BE16CD, 0xF6B9265B, 0x6FB077E1, 0x18B74777,
+      0x88085AE6, 0xFF0F6A70, 0x66063BCA, 0x11010B5C, 0x8F659EFF, 0xF862AE69, 0x616BFFD3, 0x166CCF45,
+      0xA00AE278, 0xD70DD2EE, 0x4E048354, 0x3903B3C2, 0xA7672661, 0xD06016F7, 0x4969474D, 0x3E6E77DB,
+      0xAED16A4A, 0xD9D65ADC, 0x40DF0B66, 0x37D83BF0, 0xA9BCAE53, 0xDEBB9EC5, 0x47B2CF7F, 0x30B5FFE9,
+      0xBDBDF21C, 0xCABAC28A, 0x53B39330, 0x24B4A3A6, 0xBAD03605, 0xCDD70693, 0x54DE5729, 0x23D967BF,
+      0xB3667A2E, 0xC4614AB8, 0x5D681B02, 0x2A6F2B94, 0xB40BBE37, 0xC30C8EA1, 0x5A05DF1B, 0x2D02EF8D
+   };
+
+   unsigned int crc = ~0u;
+   int i;
+   for (i=0; i < len; ++i)
+      crc = (crc >> 8) ^ crc_table[buffer[i] ^ (crc & 0xff)];
+   return ~crc;
+#endif
+}
+
+#define stbiw__wpng4(o,a,b,c,d) ((o)[0]=STBIW_UCHAR(a),(o)[1]=STBIW_UCHAR(b),(o)[2]=STBIW_UCHAR(c),(o)[3]=STBIW_UCHAR(d),(o)+=4)
+#define stbiw__wp32(data,v) stbiw__wpng4(data, (v)>>24,(v)>>16,(v)>>8,(v));
+#define stbiw__wptag(data,s) stbiw__wpng4(data, s[0],s[1],s[2],s[3])
+
+static void stbiw__wpcrc(unsigned char **data, int len)
+{
+   unsigned int crc = stbiw__crc32(*data - len - 4, len+4);
+   stbiw__wp32(*data, crc);
+}
+
+static unsigned char stbiw__paeth(int a, int b, int c)
+{
+   int p = a + b - c, pa = abs(p-a), pb = abs(p-b), pc = abs(p-c);
+   if (pa <= pb && pa <= pc) return STBIW_UCHAR(a);
+   if (pb <= pc) return STBIW_UCHAR(b);
+   return STBIW_UCHAR(c);
+}
+
+// @OPTIMIZE: provide an option that always forces left-predict or paeth predict
+static void stbiw__encode_png_line(unsigned char *pixels, int stride_bytes, int width, int height, int y, int n, int filter_type, signed char *line_buffer)
+{
+   static int mapping[] = { 0,1,2,3,4 };
+   static int firstmap[] = { 0,1,0,5,6 };
+   int *mymap = (y != 0) ? mapping : firstmap;
+   int i;
+   int type = mymap[filter_type];
+   unsigned char *z = pixels + stride_bytes * (stbi__flip_vertically_on_write ? height-1-y : y);
+   int signed_stride = stbi__flip_vertically_on_write ? -stride_bytes : stride_bytes;
+
+   if (type==0) {
+      memcpy(line_buffer, z, width*n);
+      return;
+   }
+
+   // first loop isn't optimized since it's just one pixel
+   for (i = 0; i < n; ++i) {
+      switch (type) {
+         case 1: line_buffer[i] = z[i]; break;
+         case 2: line_buffer[i] = z[i] - z[i-signed_stride]; break;
+         case 3: line_buffer[i] = z[i] - (z[i-signed_stride]>>1); break;
+         case 4: line_buffer[i] = (signed char) (z[i] - stbiw__paeth(0,z[i-signed_stride],0)); break;
+         case 5: line_buffer[i] = z[i]; break;
+         case 6: line_buffer[i] = z[i]; break;
+      }
+   }
+   switch (type) {
+      case 1: for (i=n; i < width*n; ++i) line_buffer[i] = z[i] - z[i-n]; break;
+      case 2: for (i=n; i < width*n; ++i) line_buffer[i] = z[i] - z[i-signed_stride]; break;
+      case 3: for (i=n; i < width*n; ++i) line_buffer[i] = z[i] - ((z[i-n] + z[i-signed_stride])>>1); break;
+      case 4: for (i=n; i < width*n; ++i) line_buffer[i] = z[i] - stbiw__paeth(z[i-n], z[i-signed_stride], z[i-signed_stride-n]); break;
+      case 5: for (i=n; i < width*n; ++i) line_buffer[i] = z[i] - (z[i-n]>>1); break;
+      case 6: for (i=n; i < width*n; ++i) line_buffer[i] = z[i] - stbiw__paeth(z[i-n], 0,0); break;
+   }
+}
+
+STBIWDEF unsigned char *stbi_write_png_to_mem(const unsigned char *pixels, int stride_bytes, int x, int y, int n, int *out_len)
+{
+   int force_filter = stbi_write_force_png_filter;
+   int ctype[5] = { -1, 0, 4, 2, 6 };
+   unsigned char sig[8] = { 137,80,78,71,13,10,26,10 };
+   unsigned char *out,*o, *filt, *zlib;
+   signed char *line_buffer;
+   int j,zlen;
+
+   if (stride_bytes == 0)
+      stride_bytes = x * n;
+
+   if (force_filter >= 5) {
+      force_filter = -1;
+   }
+
+   filt = (unsigned char *) STBIW_MALLOC((x*n+1) * y); if (!filt) return 0;
+   line_buffer = (signed char *) STBIW_MALLOC(x * n); if (!line_buffer) { STBIW_FREE(filt); return 0; }
+   for (j=0; j < y; ++j) {
+      int filter_type;
+      if (force_filter > -1) {
+         filter_type = force_filter;
+         stbiw__encode_png_line((unsigned char*)(pixels), stride_bytes, x, y, j, n, force_filter, line_buffer);
+      } else { // Estimate the best filter by running through all of them:
+         int best_filter = 0, best_filter_val = 0x7fffffff, est, i;
+         for (filter_type = 0; filter_type < 5; filter_type++) {
+            stbiw__encode_png_line((unsigned char*)(pixels), stride_bytes, x, y, j, n, filter_type, line_buffer);
+
+            // Estimate the entropy of the line using this filter; the less, the better.
+            est = 0;
+            for (i = 0; i < x*n; ++i) {
+               est += abs((signed char) line_buffer[i]);
+            }
+            if (est < best_filter_val) {
+               best_filter_val = est;
+               best_filter = filter_type;
+            }
+         }
+         if (filter_type != best_filter) {  // If the last iteration already got us the best filter, don't redo it
+            stbiw__encode_png_line((unsigned char*)(pixels), stride_bytes, x, y, j, n, best_filter, line_buffer);
+            filter_type = best_filter;
+         }
+      }
+      // when we get here, filter_type contains the filter type, and line_buffer contains the data
+      filt[j*(x*n+1)] = (unsigned char) filter_type;
+      STBIW_MEMMOVE(filt+j*(x*n+1)+1, line_buffer, x*n);
+   }
+   STBIW_FREE(line_buffer);
+   zlib = stbi_zlib_compress(filt, y*( x*n+1), &zlen, stbi_write_png_compression_level);
+   STBIW_FREE(filt);
+   if (!zlib) return 0;
+
+   // each tag requires 12 bytes of overhead
+   out = (unsigned char *) STBIW_MALLOC(8 + 12+13 + 12+zlen + 12);
+   if (!out) return 0;
+   *out_len = 8 + 12+13 + 12+zlen + 12;
+
+   o=out;
+   STBIW_MEMMOVE(o,sig,8); o+= 8;
+   stbiw__wp32(o, 13); // header length
+   stbiw__wptag(o, "IHDR");
+   stbiw__wp32(o, x);
+   stbiw__wp32(o, y);
+   *o++ = 8;
+   *o++ = STBIW_UCHAR(ctype[n]);
+   *o++ = 0;
+   *o++ = 0;
+   *o++ = 0;
+   stbiw__wpcrc(&o,13);
+
+   stbiw__wp32(o, zlen);
+   stbiw__wptag(o, "IDAT");
+   STBIW_MEMMOVE(o, zlib, zlen);
+   o += zlen;
+   STBIW_FREE(zlib);
+   stbiw__wpcrc(&o, zlen);
+
+   stbiw__wp32(o,0);
+   stbiw__wptag(o, "IEND");
+   stbiw__wpcrc(&o,0);
+
+   STBIW_ASSERT(o == out + *out_len);
+
+   return out;
+}
+
+#ifndef STBI_WRITE_NO_STDIO
+STBIWDEF int stbi_write_png(char const *filename, int x, int y, int comp, const void *data, int stride_bytes)
+{
+   FILE *f;
+   int len;
+   unsigned char *png = stbi_write_png_to_mem((const unsigned char *) data, stride_bytes, x, y, comp, &len);
+   if (png == NULL) return 0;
+
+   f = stbiw__fopen(filename, "wb");
+   if (!f) { STBIW_FREE(png); return 0; }
+   fwrite(png, 1, len, f);
+   fclose(f);
+   STBIW_FREE(png);
+   return 1;
+}
+#endif
+
+STBIWDEF int stbi_write_png_to_func(stbi_write_func *func, void *context, int x, int y, int comp, const void *data, int stride_bytes)
+{
+   int len;
+   unsigned char *png = stbi_write_png_to_mem((const unsigned char *) data, stride_bytes, x, y, comp, &len);
+   if (png == NULL) return 0;
+   func(context, png, len);
+   STBIW_FREE(png);
+   return 1;
+}
+
+
+/* ***************************************************************************
+ *
+ * JPEG writer
+ *
+ * This is based on Jon Olick's jo_jpeg.cpp:
+ * public domain Simple, Minimalistic JPEG writer - http://www.jonolick.com/code.html
+ */
+
+static const unsigned char stbiw__jpg_ZigZag[] = { 0,1,5,6,14,15,27,28,2,4,7,13,16,26,29,42,3,8,12,17,25,30,41,43,9,11,18,
+      24,31,40,44,53,10,19,23,32,39,45,52,54,20,22,33,38,46,51,55,60,21,34,37,47,50,56,59,61,35,36,48,49,57,58,62,63 };
+
+static void stbiw__jpg_writeBits(stbi__write_context *s, int *bitBufP, int *bitCntP, const unsigned short *bs) {
+   int bitBuf = *bitBufP, bitCnt = *bitCntP;
+   bitCnt += bs[1];
+   bitBuf |= bs[0] << (24 - bitCnt);
+   while(bitCnt >= 8) {
+      unsigned char c = (bitBuf >> 16) & 255;
+      stbiw__putc(s, c);
+      if(c == 255) {
+         stbiw__putc(s, 0);
+      }
+      bitBuf <<= 8;
+      bitCnt -= 8;
+   }
+   *bitBufP = bitBuf;
+   *bitCntP = bitCnt;
+}
+
+static void stbiw__jpg_DCT(float *d0p, float *d1p, float *d2p, float *d3p, float *d4p, float *d5p, float *d6p, float *d7p) {
+   float d0 = *d0p, d1 = *d1p, d2 = *d2p, d3 = *d3p, d4 = *d4p, d5 = *d5p, d6 = *d6p, d7 = *d7p;
+   float z1, z2, z3, z4, z5, z11, z13;
+
+   float tmp0 = d0 + d7;
+   float tmp7 = d0 - d7;
+   float tmp1 = d1 + d6;
+   float tmp6 = d1 - d6;
+   float tmp2 = d2 + d5;
+   float tmp5 = d2 - d5;
+   float tmp3 = d3 + d4;
+   float tmp4 = d3 - d4;
+
+   // Even part
+   float tmp10 = tmp0 + tmp3;   // phase 2
+   float tmp13 = tmp0 - tmp3;
+   float tmp11 = tmp1 + tmp2;
+   float tmp12 = tmp1 - tmp2;
+
+   d0 = tmp10 + tmp11;       // phase 3
+   d4 = tmp10 - tmp11;
+
+   z1 = (tmp12 + tmp13) * 0.707106781f; // c4
+   d2 = tmp13 + z1;       // phase 5
+   d6 = tmp13 - z1;
+
+   // Odd part
+   tmp10 = tmp4 + tmp5;       // phase 2
+   tmp11 = tmp5 + tmp6;
+   tmp12 = tmp6 + tmp7;
+
+   // The rotator is modified from fig 4-8 to avoid extra negations.
+   z5 = (tmp10 - tmp12) * 0.382683433f; // c6
+   z2 = tmp10 * 0.541196100f + z5; // c2-c6
+   z4 = tmp12 * 1.306562965f + z5; // c2+c6
+   z3 = tmp11 * 0.707106781f; // c4
+
+   z11 = tmp7 + z3;      // phase 5
+   z13 = tmp7 - z3;
+
+   *d5p = z13 + z2;         // phase 6
+   *d3p = z13 - z2;
+   *d1p = z11 + z4;
+   *d7p = z11 - z4;
+
+   *d0p = d0;  *d2p = d2;  *d4p = d4;  *d6p = d6;
+}
+
+static void stbiw__jpg_calcBits(int val, unsigned short bits[2]) {
+   int tmp1 = val < 0 ? -val : val;
+   val = val < 0 ? val-1 : val;
+   bits[1] = 1;
+   while(tmp1 >>= 1) {
+      ++bits[1];
+   }
+   bits[0] = val & ((1<<bits[1])-1);
+}
+
+static int stbiw__jpg_processDU(stbi__write_context *s, int *bitBuf, int *bitCnt, float *CDU, int du_stride, float *fdtbl, int DC, const unsigned short HTDC[256][2], const unsigned short HTAC[256][2]) {
+   const unsigned short EOB[2] = { HTAC[0x00][0], HTAC[0x00][1] };
+   const unsigned short M16zeroes[2] = { HTAC[0xF0][0], HTAC[0xF0][1] };
+   int dataOff, i, j, n, diff, end0pos, x, y;
+   int DU[64];
+
+   // DCT rows
+   for(dataOff=0, n=du_stride*8; dataOff<n; dataOff+=du_stride) {
+      stbiw__jpg_DCT(&CDU[dataOff], &CDU[dataOff+1], &CDU[dataOff+2], &CDU[dataOff+3], &CDU[dataOff+4], &CDU[dataOff+5], &CDU[dataOff+6], &CDU[dataOff+7]);
+   }
+   // DCT columns
+   for(dataOff=0; dataOff<8; ++dataOff) {
+      stbiw__jpg_DCT(&CDU[dataOff], &CDU[dataOff+du_stride], &CDU[dataOff+du_stride*2], &CDU[dataOff+du_stride*3], &CDU[dataOff+du_stride*4],
+                     &CDU[dataOff+du_stride*5], &CDU[dataOff+du_stride*6], &CDU[dataOff+du_stride*7]);
+   }
+   // Quantize/descale/zigzag the coefficients
+   for(y = 0, j=0; y < 8; ++y) {
+      for(x = 0; x < 8; ++x,++j) {
+         float v;
+         i = y*du_stride+x;
+         v = CDU[i]*fdtbl[j];
+         // DU[stbiw__jpg_ZigZag[j]] = (int)(v < 0 ? ceilf(v - 0.5f) : floorf(v + 0.5f));
+         // ceilf() and floorf() are C99, not C89, but I /think/ they're not needed here anyway?
+         DU[stbiw__jpg_ZigZag[j]] = (int)(v < 0 ? v - 0.5f : v + 0.5f);
+      }
+   }
+
+   // Encode DC
+   diff = DU[0] - DC;
+   if (diff == 0) {
+      stbiw__jpg_writeBits(s, bitBuf, bitCnt, HTDC[0]);
+   } else {
+      unsigned short bits[2];
+      stbiw__jpg_calcBits(diff, bits);
+      stbiw__jpg_writeBits(s, bitBuf, bitCnt, HTDC[bits[1]]);
+      stbiw__jpg_writeBits(s, bitBuf, bitCnt, bits);
+   }
+   // Encode ACs
+   end0pos = 63;
+   for(; (end0pos>0)&&(DU[end0pos]==0); --end0pos) {
+   }
+   // end0pos = first element in reverse order !=0
+   if(end0pos == 0) {
+      stbiw__jpg_writeBits(s, bitBuf, bitCnt, EOB);
+      return DU[0];
+   }
+   for(i = 1; i <= end0pos; ++i) {
+      int startpos = i;
+      int nrzeroes;
+      unsigned short bits[2];
+      for (; DU[i]==0 && i<=end0pos; ++i) {
+      }
+      nrzeroes = i-startpos;
+      if ( nrzeroes >= 16 ) {
+         int lng = nrzeroes>>4;
+         int nrmarker;
+         for (nrmarker=1; nrmarker <= lng; ++nrmarker)
+            stbiw__jpg_writeBits(s, bitBuf, bitCnt, M16zeroes);
+         nrzeroes &= 15;
+      }
+      stbiw__jpg_calcBits(DU[i], bits);
+      stbiw__jpg_writeBits(s, bitBuf, bitCnt, HTAC[(nrzeroes<<4)+bits[1]]);
+      stbiw__jpg_writeBits(s, bitBuf, bitCnt, bits);
+   }
+   if(end0pos != 63) {
+      stbiw__jpg_writeBits(s, bitBuf, bitCnt, EOB);
+   }
+   return DU[0];
+}
+
+static int stbi_write_jpg_core(stbi__write_context *s, int width, int height, int comp, const void* data, int quality) {
+   // Constants that don't pollute global namespace
+   static const unsigned char std_dc_luminance_nrcodes[] = {0,0,1,5,1,1,1,1,1,1,0,0,0,0,0,0,0};
+   static const unsigned char std_dc_luminance_values[] = {0,1,2,3,4,5,6,7,8,9,10,11};
+   static const unsigned char std_ac_luminance_nrcodes[] = {0,0,2,1,3,3,2,4,3,5,5,4,4,0,0,1,0x7d};
+   static const unsigned char std_ac_luminance_values[] = {
+      0x01,0x02,0x03,0x00,0x04,0x11,0x05,0x12,0x21,0x31,0x41,0x06,0x13,0x51,0x61,0x07,0x22,0x71,0x14,0x32,0x81,0x91,0xa1,0x08,
+      0x23,0x42,0xb1,0xc1,0x15,0x52,0xd1,0xf0,0x24,0x33,0x62,0x72,0x82,0x09,0x0a,0x16,0x17,0x18,0x19,0x1a,0x25,0x26,0x27,0x28,
+      0x29,0x2a,0x34,0x35,0x36,0x37,0x38,0x39,0x3a,0x43,0x44,0x45,0x46,0x47,0x48,0x49,0x4a,0x53,0x54,0x55,0x56,0x57,0x58,0x59,
+      0x5a,0x63,0x64,0x65,0x66,0x67,0x68,0x69,0x6a,0x73,0x74,0x75,0x76,0x77,0x78,0x79,0x7a,0x83,0x84,0x85,0x86,0x87,0x88,0x89,
+      0x8a,0x92,0x93,0x94,0x95,0x96,0x97,0x98,0x99,0x9a,0xa2,0xa3,0xa4,0xa5,0xa6,0xa7,0xa8,0xa9,0xaa,0xb2,0xb3,0xb4,0xb5,0xb6,
+      0xb7,0xb8,0xb9,0xba,0xc2,0xc3,0xc4,0xc5,0xc6,0xc7,0xc8,0xc9,0xca,0xd2,0xd3,0xd4,0xd5,0xd6,0xd7,0xd8,0xd9,0xda,0xe1,0xe2,
+      0xe3,0xe4,0xe5,0xe6,0xe7,0xe8,0xe9,0xea,0xf1,0xf2,0xf3,0xf4,0xf5,0xf6,0xf7,0xf8,0xf9,0xfa
+   };
+   static const unsigned char std_dc_chrominance_nrcodes[] = {0,0,3,1,1,1,1,1,1,1,1,1,0,0,0,0,0};
+   static const unsigned char std_dc_chrominance_values[] = {0,1,2,3,4,5,6,7,8,9,10,11};
+   static const unsigned char std_ac_chrominance_nrcodes[] = {0,0,2,1,2,4,4,3,4,7,5,4,4,0,1,2,0x77};
+   static const unsigned char std_ac_chrominance_values[] = {
+      0x00,0x01,0x02,0x03,0x11,0x04,0x05,0x21,0x31,0x06,0x12,0x41,0x51,0x07,0x61,0x71,0x13,0x22,0x32,0x81,0x08,0x14,0x42,0x91,
+      0xa1,0xb1,0xc1,0x09,0x23,0x33,0x52,0xf0,0x15,0x62,0x72,0xd1,0x0a,0x16,0x24,0x34,0xe1,0x25,0xf1,0x17,0x18,0x19,0x1a,0x26,
+      0x27,0x28,0x29,0x2a,0x35,0x36,0x37,0x38,0x39,0x3a,0x43,0x44,0x45,0x46,0x47,0x48,0x49,0x4a,0x53,0x54,0x55,0x56,0x57,0x58,
+      0x59,0x5a,0x63,0x64,0x65,0x66,0x67,0x68,0x69,0x6a,0x73,0x74,0x75,0x76,0x77,0x78,0x79,0x7a,0x82,0x83,0x84,0x85,0x86,0x87,
+      0x88,0x89,0x8a,0x92,0x93,0x94,0x95,0x96,0x97,0x98,0x99,0x9a,0xa2,0xa3,0xa4,0xa5,0xa6,0xa7,0xa8,0xa9,0xaa,0xb2,0xb3,0xb4,
+      0xb5,0xb6,0xb7,0xb8,0xb9,0xba,0xc2,0xc3,0xc4,0xc5,0xc6,0xc7,0xc8,0xc9,0xca,0xd2,0xd3,0xd4,0xd5,0xd6,0xd7,0xd8,0xd9,0xda,
+      0xe2,0xe3,0xe4,0xe5,0xe6,0xe7,0xe8,0xe9,0xea,0xf2,0xf3,0xf4,0xf5,0xf6,0xf7,0xf8,0xf9,0xfa
+   };
+   // Huffman tables
+   static const unsigned short YDC_HT[256][2] = { {0,2},{2,3},{3,3},{4,3},{5,3},{6,3},{14,4},{30,5},{62,6},{126,7},{254,8},{510,9}};
+   static const unsigned short UVDC_HT[256][2] = { {0,2},{1,2},{2,2},{6,3},{14,4},{30,5},{62,6},{126,7},{254,8},{510,9},{1022,10},{2046,11}};
+   static const unsigned short YAC_HT[256][2] = {
+      {10,4},{0,2},{1,2},{4,3},{11,4},{26,5},{120,7},{248,8},{1014,10},{65410,16},{65411,16},{0,0},{0,0},{0,0},{0,0},{0,0},{0,0},
+      {12,4},{27,5},{121,7},{502,9},{2038,11},{65412,16},{65413,16},{65414,16},{65415,16},{65416,16},{0,0},{0,0},{0,0},{0,0},{0,0},{0,0},
+      {28,5},{249,8},{1015,10},{4084,12},{65417,16},{65418,16},{65419,16},{65420,16},{65421,16},{65422,16},{0,0},{0,0},{0,0},{0,0},{0,0},{0,0},
+      {58,6},{503,9},{4085,12},{65423,16},{65424,16},{65425,16},{65426,16},{65427,16},{65428,16},{65429,16},{0,0},{0,0},{0,0},{0,0},{0,0},{0,0},
+      {59,6},{1016,10},{65430,16},{65431,16},{65432,16},{65433,16},{65434,16},{65435,16},{65436,16},{65437,16},{0,0},{0,0},{0,0},{0,0},{0,0},{0,0},
+      {122,7},{2039,11},{65438,16},{65439,16},{65440,16},{65441,16},{65442,16},{65443,16},{65444,16},{65445,16},{0,0},{0,0},{0,0},{0,0},{0,0},{0,0},
+      {123,7},{4086,12},{65446,16},{65447,16},{65448,16},{65449,16},{65450,16},{65451,16},{65452,16},{65453,16},{0,0},{0,0},{0,0},{0,0},{0,0},{0,0},
+      {250,8},{4087,12},{65454,16},{65455,16},{65456,16},{65457,16},{65458,16},{65459,16},{65460,16},{65461,16},{0,0},{0,0},{0,0},{0,0},{0,0},{0,0},
+      {504,9},{32704,15},{65462,16},{65463,16},{65464,16},{65465,16},{65466,16},{65467,16},{65468,16},{65469,16},{0,0},{0,0},{0,0},{0,0},{0,0},{0,0},
+      {505,9},{65470,16},{65471,16},{65472,16},{65473,16},{65474,16},{65475,16},{65476,16},{65477,16},{65478,16},{0,0},{0,0},{0,0},{0,0},{0,0},{0,0},
+      {506,9},{65479,16},{65480,16},{65481,16},{65482,16},{65483,16},{65484,16},{65485,16},{65486,16},{65487,16},{0,0},{0,0},{0,0},{0,0},{0,0},{0,0},
+      {1017,10},{65488,16},{65489,16},{65490,16},{65491,16},{65492,16},{65493,16},{65494,16},{65495,16},{65496,16},{0,0},{0,0},{0,0},{0,0},{0,0},{0,0},
+      {1018,10},{65497,16},{65498,16},{65499,16},{65500,16},{65501,16},{65502,16},{65503,16},{65504,16},{65505,16},{0,0},{0,0},{0,0},{0,0},{0,0},{0,0},
+      {2040,11},{65506,16},{65507,16},{65508,16},{65509,16},{65510,16},{65511,16},{65512,16},{65513,16},{65514,16},{0,0},{0,0},{0,0},{0,0},{0,0},{0,0},
+      {65515,16},{65516,16},{65517,16},{65518,16},{65519,16},{65520,16},{65521,16},{65522,16},{65523,16},{65524,16},{0,0},{0,0},{0,0},{0,0},{0,0},
+      {2041,11},{65525,16},{65526,16},{65527,16},{65528,16},{65529,16},{65530,16},{65531,16},{65532,16},{65533,16},{65534,16},{0,0},{0,0},{0,0},{0,0},{0,0}
+   };
+   static const unsigned short UVAC_HT[256][2] = {
+      {0,2},{1,2},{4,3},{10,4},{24,5},{25,5},{56,6},{120,7},{500,9},{1014,10},{4084,12},{0,0},{0,0},{0,0},{0,0},{0,0},{0,0},
+      {11,4},{57,6},{246,8},{501,9},{2038,11},{4085,12},{65416,16},{65417,16},{65418,16},{65419,16},{0,0},{0,0},{0,0},{0,0},{0,0},{0,0},
+      {26,5},{247,8},{1015,10},{4086,12},{32706,15},{65420,16},{65421,16},{65422,16},{65423,16},{65424,16},{0,0},{0,0},{0,0},{0,0},{0,0},{0,0},
+      {27,5},{248,8},{1016,10},{4087,12},{65425,16},{65426,16},{65427,16},{65428,16},{65429,16},{65430,16},{0,0},{0,0},{0,0},{0,0},{0,0},{0,0},
+      {58,6},{502,9},{65431,16},{65432,16},{65433,16},{65434,16},{65435,16},{65436,16},{65437,16},{65438,16},{0,0},{0,0},{0,0},{0,0},{0,0},{0,0},
+      {59,6},{1017,10},{65439,16},{65440,16},{65441,16},{65442,16},{65443,16},{65444,16},{65445,16},{65446,16},{0,0},{0,0},{0,0},{0,0},{0,0},{0,0},
+      {121,7},{2039,11},{65447,16},{65448,16},{65449,16},{65450,16},{65451,16},{65452,16},{65453,16},{65454,16},{0,0},{0,0},{0,0},{0,0},{0,0},{0,0},
+      {122,7},{2040,11},{65455,16},{65456,16},{65457,16},{65458,16},{65459,16},{65460,16},{65461,16},{65462,16},{0,0},{0,0},{0,0},{0,0},{0,0},{0,0},
+      {249,8},{65463,16},{65464,16},{65465,16},{65466,16},{65467,16},{65468,16},{65469,16},{65470,16},{65471,16},{0,0},{0,0},{0,0},{0,0},{0,0},{0,0},
+      {503,9},{65472,16},{65473,16},{65474,16},{65475,16},{65476,16},{65477,16},{65478,16},{65479,16},{65480,16},{0,0},{0,0},{0,0},{0,0},{0,0},{0,0},
+      {504,9},{65481,16},{65482,16},{65483,16},{65484,16},{65485,16},{65486,16},{65487,16},{65488,16},{65489,16},{0,0},{0,0},{0,0},{0,0},{0,0},{0,0},
+      {505,9},{65490,16},{65491,16},{65492,16},{65493,16},{65494,16},{65495,16},{65496,16},{65497,16},{65498,16},{0,0},{0,0},{0,0},{0,0},{0,0},{0,0},
+      {506,9},{65499,16},{65500,16},{65501,16},{65502,16},{65503,16},{65504,16},{65505,16},{65506,16},{65507,16},{0,0},{0,0},{0,0},{0,0},{0,0},{0,0},
+      {2041,11},{65508,16},{65509,16},{65510,16},{65511,16},{65512,16},{65513,16},{65514,16},{65515,16},{65516,16},{0,0},{0,0},{0,0},{0,0},{0,0},{0,0},
+      {16352,14},{65517,16},{65518,16},{65519,16},{65520,16},{65521,16},{65522,16},{65523,16},{65524,16},{65525,16},{0,0},{0,0},{0,0},{0,0},{0,0},
+      {1018,10},{32707,15},{65526,16},{65527,16},{65528,16},{65529,16},{65530,16},{65531,16},{65532,16},{65533,16},{65534,16},{0,0},{0,0},{0,0},{0,0},{0,0}
+   };
+   static const int YQT[] = {16,11,10,16,24,40,51,61,12,12,14,19,26,58,60,55,14,13,16,24,40,57,69,56,14,17,22,29,51,87,80,62,18,22,
+                             37,56,68,109,103,77,24,35,55,64,81,104,113,92,49,64,78,87,103,121,120,101,72,92,95,98,112,100,103,99};
+   static const int UVQT[] = {17,18,24,47,99,99,99,99,18,21,26,66,99,99,99,99,24,26,56,99,99,99,99,99,47,66,99,99,99,99,99,99,
+                              99,99,99,99,99,99,99,99,99,99,99,99,99,99,99,99,99,99,99,99,99,99,99,99,99,99,99,99,99,99,99,99};
+   static const float aasf[] = { 1.0f * 2.828427125f, 1.387039845f * 2.828427125f, 1.306562965f * 2.828427125f, 1.175875602f * 2.828427125f,
+                                 1.0f * 2.828427125f, 0.785694958f * 2.828427125f, 0.541196100f * 2.828427125f, 0.275899379f * 2.828427125f };
+
+   int row, col, i, k, subsample;
+   float fdtbl_Y[64], fdtbl_UV[64];
+   unsigned char YTable[64], UVTable[64];
+
+   if(!data || !width || !height || comp > 4 || comp < 1) {
+      return 0;
+   }
+
+   quality = quality ? quality : 90;
+   subsample = quality <= 90 ? 1 : 0;
+   quality = quality < 1 ? 1 : quality > 100 ? 100 : quality;
+   quality = quality < 50 ? 5000 / quality : 200 - quality * 2;
+
+   for(i = 0; i < 64; ++i) {
+      int uvti, yti = (YQT[i]*quality+50)/100;
+      YTable[stbiw__jpg_ZigZag[i]] = (unsigned char) (yti < 1 ? 1 : yti > 255 ? 255 : yti);
+      uvti = (UVQT[i]*quality+50)/100;
+      UVTable[stbiw__jpg_ZigZag[i]] = (unsigned char) (uvti < 1 ? 1 : uvti > 255 ? 255 : uvti);
+   }
+
+   for(row = 0, k = 0; row < 8; ++row) {
+      for(col = 0; col < 8; ++col, ++k) {
+         fdtbl_Y[k]  = 1 / (YTable [stbiw__jpg_ZigZag[k]] * aasf[row] * aasf[col]);
+         fdtbl_UV[k] = 1 / (UVTable[stbiw__jpg_ZigZag[k]] * aasf[row] * aasf[col]);
+      }
+   }
+
+   // Write Headers
+   {
+      static const unsigned char head0[] = { 0xFF,0xD8,0xFF,0xE0,0,0x10,'J','F','I','F',0,1,1,0,0,1,0,1,0,0,0xFF,0xDB,0,0x84,0 };
+      static const unsigned char head2[] = { 0xFF,0xDA,0,0xC,3,1,0,2,0x11,3,0x11,0,0x3F,0 };
+      const unsigned char head1[] = { 0xFF,0xC0,0,0x11,8,(unsigned char)(height>>8),STBIW_UCHAR(height),(unsigned char)(width>>8),STBIW_UCHAR(width),
+                                      3,1,(unsigned char)(subsample?0x22:0x11),0,2,0x11,1,3,0x11,1,0xFF,0xC4,0x01,0xA2,0 };
+      s->func(s->context, (void*)head0, sizeof(head0));
+      s->func(s->context, (void*)YTable, sizeof(YTable));
+      stbiw__putc(s, 1);
+      s->func(s->context, UVTable, sizeof(UVTable));
+      s->func(s->context, (void*)head1, sizeof(head1));
+      s->func(s->context, (void*)(std_dc_luminance_nrcodes+1), sizeof(std_dc_luminance_nrcodes)-1);
+      s->func(s->context, (void*)std_dc_luminance_values, sizeof(std_dc_luminance_values));
+      stbiw__putc(s, 0x10); // HTYACinfo
+      s->func(s->context, (void*)(std_ac_luminance_nrcodes+1), sizeof(std_ac_luminance_nrcodes)-1);
+      s->func(s->context, (void*)std_ac_luminance_values, sizeof(std_ac_luminance_values));
+      stbiw__putc(s, 1); // HTUDCinfo
+      s->func(s->context, (void*)(std_dc_chrominance_nrcodes+1), sizeof(std_dc_chrominance_nrcodes)-1);
+      s->func(s->context, (void*)std_dc_chrominance_values, sizeof(std_dc_chrominance_values));
+      stbiw__putc(s, 0x11); // HTUACinfo
+      s->func(s->context, (void*)(std_ac_chrominance_nrcodes+1), sizeof(std_ac_chrominance_nrcodes)-1);
+      s->func(s->context, (void*)std_ac_chrominance_values, sizeof(std_ac_chrominance_values));
+      s->func(s->context, (void*)head2, sizeof(head2));
+   }
+
+   // Encode 8x8 macroblocks
+   {
+      static const unsigned short fillBits[] = {0x7F, 7};
+      int DCY=0, DCU=0, DCV=0;
+      int bitBuf=0, bitCnt=0;
+      // comp == 2 is grey+alpha (alpha is ignored)
+      int ofsG = comp > 2 ? 1 : 0, ofsB = comp > 2 ? 2 : 0;
+      const unsigned char *dataR = (const unsigned char *)data;
+      const unsigned char *dataG = dataR + ofsG;
+      const unsigned char *dataB = dataR + ofsB;
+      int x, y, pos;
+      if(subsample) {
+         for(y = 0; y < height; y += 16) {
+            for(x = 0; x < width; x += 16) {
+               float Y[256], U[256], V[256];
+               for(row = y, pos = 0; row < y+16; ++row) {
+                  // row >= height => use last input row
+                  int clamped_row = (row < height) ? row : height - 1;
+                  int base_p = (stbi__flip_vertically_on_write ? (height-1-clamped_row) : clamped_row)*width*comp;
+                  for(col = x; col < x+16; ++col, ++pos) {
+                     // if col >= width => use pixel from last input column
+                     int p = base_p + ((col < width) ? col : (width-1))*comp;
+                     float r = dataR[p], g = dataG[p], b = dataB[p];
+                     Y[pos]= +0.29900f*r + 0.58700f*g + 0.11400f*b - 128;
+                     U[pos]= -0.16874f*r - 0.33126f*g + 0.50000f*b;
+                     V[pos]= +0.50000f*r - 0.41869f*g - 0.08131f*b;
+                  }
+               }
+               DCY = stbiw__jpg_processDU(s, &bitBuf, &bitCnt, Y+0,   16, fdtbl_Y, DCY, YDC_HT, YAC_HT);
+               DCY = stbiw__jpg_processDU(s, &bitBuf, &bitCnt, Y+8,   16, fdtbl_Y, DCY, YDC_HT, YAC_HT);
+               DCY = stbiw__jpg_processDU(s, &bitBuf, &bitCnt, Y+128, 16, fdtbl_Y, DCY, YDC_HT, YAC_HT);
+               DCY = stbiw__jpg_processDU(s, &bitBuf, &bitCnt, Y+136, 16, fdtbl_Y, DCY, YDC_HT, YAC_HT);
+
+               // subsample U,V
+               {
+                  float subU[64], subV[64];
+                  int yy, xx;
+                  for(yy = 0, pos = 0; yy < 8; ++yy) {
+                     for(xx = 0; xx < 8; ++xx, ++pos) {
+                        int j = yy*32+xx*2;
+                        subU[pos] = (U[j+0] + U[j+1] + U[j+16] + U[j+17]) * 0.25f;
+                        subV[pos] = (V[j+0] + V[j+1] + V[j+16] + V[j+17]) * 0.25f;
+                     }
+                  }
+                  DCU = stbiw__jpg_processDU(s, &bitBuf, &bitCnt, subU, 8, fdtbl_UV, DCU, UVDC_HT, UVAC_HT);
+                  DCV = stbiw__jpg_processDU(s, &bitBuf, &bitCnt, subV, 8, fdtbl_UV, DCV, UVDC_HT, UVAC_HT);
+               }
+            }
+         }
+      } else {
+         for(y = 0; y < height; y += 8) {
+            for(x = 0; x < width; x += 8) {
+               float Y[64], U[64], V[64];
+               for(row = y, pos = 0; row < y+8; ++row) {
+                  // row >= height => use last input row
+                  int clamped_row = (row < height) ? row : height - 1;
+                  int base_p = (stbi__flip_vertically_on_write ? (height-1-clamped_row) : clamped_row)*width*comp;
+                  for(col = x; col < x+8; ++col, ++pos) {
+                     // if col >= width => use pixel from last input column
+                     int p = base_p + ((col < width) ? col : (width-1))*comp;
+                     float r = dataR[p], g = dataG[p], b = dataB[p];
+                     Y[pos]= +0.29900f*r + 0.58700f*g + 0.11400f*b - 128;
+                     U[pos]= -0.16874f*r - 0.33126f*g + 0.50000f*b;
+                     V[pos]= +0.50000f*r - 0.41869f*g - 0.08131f*b;
+                  }
+               }
+
+               DCY = stbiw__jpg_processDU(s, &bitBuf, &bitCnt, Y, 8, fdtbl_Y,  DCY, YDC_HT, YAC_HT);
+               DCU = stbiw__jpg_processDU(s, &bitBuf, &bitCnt, U, 8, fdtbl_UV, DCU, UVDC_HT, UVAC_HT);
+               DCV = stbiw__jpg_processDU(s, &bitBuf, &bitCnt, V, 8, fdtbl_UV, DCV, UVDC_HT, UVAC_HT);
+            }
+         }
+      }
+
+      // Do the bit alignment of the EOI marker
+      stbiw__jpg_writeBits(s, &bitBuf, &bitCnt, fillBits);
+   }
+
+   // EOI
+   stbiw__putc(s, 0xFF);
+   stbiw__putc(s, 0xD9);
+
+   return 1;
+}
+
+STBIWDEF int stbi_write_jpg_to_func(stbi_write_func *func, void *context, int x, int y, int comp, const void *data, int quality)
+{
+   stbi__write_context s = { 0 };
+   stbi__start_write_callbacks(&s, func, context);
+   return stbi_write_jpg_core(&s, x, y, comp, (void *) data, quality);
+}
+
+
+#ifndef STBI_WRITE_NO_STDIO
+STBIWDEF int stbi_write_jpg(char const *filename, int x, int y, int comp, const void *data, int quality)
+{
+   stbi__write_context s = { 0 };
+   if (stbi__start_write_file(&s,filename)) {
+      int r = stbi_write_jpg_core(&s, x, y, comp, data, quality);
+      stbi__end_write_file(&s);
+      return r;
+   } else
+      return 0;
+}
+#endif
+
+#endif // STB_IMAGE_WRITE_IMPLEMENTATION
+
+/* Revision history
+      1.16  (2021-07-11)
+             make Deflate code emit uncompressed blocks when it would otherwise expand
+             support writing BMPs with alpha channel
+      1.15  (2020-07-13) unknown
+      1.14  (2020-02-02) updated JPEG writer to downsample chroma channels
+      1.13
+      1.12
+      1.11  (2019-08-11)
+
+      1.10  (2019-02-07)
+             support utf8 filenames in Windows; fix warnings and platform ifdefs
+      1.09  (2018-02-11)
+             fix typo in zlib quality API, improve STB_I_W_STATIC in C++
+      1.08  (2018-01-29)
+             add stbi__flip_vertically_on_write, external zlib, zlib quality, choose PNG filter
+      1.07  (2017-07-24)
+             doc fix
+      1.06 (2017-07-23)
+             writing JPEG (using Jon Olick's code)
+      1.05   ???
+      1.04 (2017-03-03)
+             monochrome BMP expansion
+      1.03   ???
+      1.02 (2016-04-02)
+             avoid allocating large structures on the stack
+      1.01 (2016-01-16)
+             STBIW_REALLOC_SIZED: support allocators with no realloc support
+             avoid race-condition in crc initialization
+             minor compile issues
+      1.00 (2015-09-14)
+             installable file IO function
+      0.99 (2015-09-13)
+             warning fixes; TGA rle support
+      0.98 (2015-04-08)
+             added STBIW_MALLOC, STBIW_ASSERT etc
+      0.97 (2015-01-18)
+             fixed HDR asserts, rewrote HDR rle logic
+      0.96 (2015-01-17)
+             add HDR output
+             fix monochrome BMP
+      0.95 (2014-08-17)
+             add monochrome TGA output
+      0.94 (2014-05-31)
+             rename private functions to avoid conflicts with stb_image.h
+      0.93 (2014-05-27)
+             warning fixes
+      0.92 (2010-08-01)
+             casts to unsigned char to fix warnings
+      0.91 (2010-07-17)
+             first public release
+      0.90   first internal release
+*/
+
+/*
+------------------------------------------------------------------------------
+This software is available under 2 licenses -- choose whichever you prefer.
+------------------------------------------------------------------------------
+ALTERNATIVE A - MIT License
+Copyright (c) 2017 Sean Barrett
+Permission is hereby granted, free of charge, to any person obtaining a copy of
+this software and associated documentation files (the "Software"), to deal in
+the Software without restriction, including without limitation the rights to
+use, copy, modify, merge, publish, distribute, sublicense, and/or sell copies
+of the Software, and to permit persons to whom the Software is furnished to do
+so, subject to the following conditions:
+The above copyright notice and this permission notice shall be included in all
+copies or substantial portions of the Software.
+THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+SOFTWARE.
+------------------------------------------------------------------------------
+ALTERNATIVE B - Public Domain (www.unlicense.org)
+This is free and unencumbered software released into the public domain.
+Anyone is free to copy, modify, publish, use, compile, sell, or distribute this
+software, either in source code form or as a compiled binary, for any purpose,
+commercial or non-commercial, and by any means.
+In jurisdictions that recognize copyright laws, the author or authors of this
+software dedicate any and all copyright interest in the software to the public
+domain. We make this dedication for the benefit of the public at large and to
+the detriment of our heirs and successors. We intend this dedication to be an
+overt act of relinquishment in perpetuity of all present and future rights to
+this software under copyright law.
+THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+AUTHORS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
+ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION
+WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+------------------------------------------------------------------------------
+*/
diff --git a/src/android/app/build.gradle.kts b/src/android/app/build.gradle.kts
index 84a3308b7..ac43d84b7 100644
--- a/src/android/app/build.gradle.kts
+++ b/src/android/app/build.gradle.kts
@@ -27,7 +27,7 @@ android {
     namespace = "org.yuzu.yuzu_emu"
 
     compileSdkVersion = "android-34"
-    ndkVersion = "25.2.9519653"
+    ndkVersion = "26.1.10909125"
 
     buildFeatures {
         viewBinding = true
@@ -203,23 +203,23 @@ ktlint {
 }
 
 dependencies {
-    implementation("androidx.core:core-ktx:1.10.1")
+    implementation("androidx.core:core-ktx:1.12.0")
     implementation("androidx.appcompat:appcompat:1.6.1")
-    implementation("androidx.recyclerview:recyclerview:1.3.0")
+    implementation("androidx.recyclerview:recyclerview:1.3.1")
     implementation("androidx.constraintlayout:constraintlayout:2.1.4")
-    implementation("androidx.fragment:fragment-ktx:1.6.0")
+    implementation("androidx.fragment:fragment-ktx:1.6.1")
     implementation("androidx.documentfile:documentfile:1.0.1")
     implementation("com.google.android.material:material:1.9.0")
-    implementation("androidx.preference:preference:1.2.0")
-    implementation("androidx.lifecycle:lifecycle-viewmodel-ktx:2.6.1")
+    implementation("androidx.preference:preference-ktx:1.2.1")
+    implementation("androidx.lifecycle:lifecycle-viewmodel-ktx:2.6.2")
     implementation("io.coil-kt:coil:2.2.2")
     implementation("androidx.core:core-splashscreen:1.0.1")
     implementation("androidx.window:window:1.2.0-beta03")
     implementation("org.ini4j:ini4j:0.5.4")
     implementation("androidx.constraintlayout:constraintlayout:2.1.4")
     implementation("androidx.swiperefreshlayout:swiperefreshlayout:1.1.0")
-    implementation("androidx.navigation:navigation-fragment-ktx:2.6.0")
-    implementation("androidx.navigation:navigation-ui-ktx:2.6.0")
+    implementation("androidx.navigation:navigation-fragment-ktx:2.7.4")
+    implementation("androidx.navigation:navigation-ui-ktx:2.7.4")
     implementation("info.debatty:java-string-similarity:2.0.0")
     implementation("org.jetbrains.kotlinx:kotlinx-serialization-json:1.5.0")
 }
diff --git a/src/android/app/src/main/AndroidManifest.xml b/src/android/app/src/main/AndroidManifest.xml
index 832c08e15..a67351727 100644
--- a/src/android/app/src/main/AndroidManifest.xml
+++ b/src/android/app/src/main/AndroidManifest.xml
@@ -28,7 +28,6 @@ SPDX-License-Identifier: GPL-3.0-or-later
         android:appCategory="game"
         android:localeConfig="@xml/locales_config"
         android:banner="@drawable/tv_banner"
-        android:extractNativeLibs="true"
         android:fullBackupContent="@xml/data_extraction_rules"
         android:dataExtractionRules="@xml/data_extraction_rules_api_31"
         android:enableOnBackInvokedCallback="true">
diff --git a/src/android/app/src/main/java/org/yuzu/yuzu_emu/NativeLibrary.kt b/src/android/app/src/main/java/org/yuzu/yuzu_emu/NativeLibrary.kt
index 6e39e542b..115f72710 100644
--- a/src/android/app/src/main/java/org/yuzu/yuzu_emu/NativeLibrary.kt
+++ b/src/android/app/src/main/java/org/yuzu/yuzu_emu/NativeLibrary.kt
@@ -15,13 +15,9 @@ import androidx.annotation.Keep
 import androidx.fragment.app.DialogFragment
 import com.google.android.material.dialog.MaterialAlertDialogBuilder
 import java.lang.ref.WeakReference
-import org.yuzu.yuzu_emu.YuzuApplication.Companion.appContext
 import org.yuzu.yuzu_emu.activities.EmulationActivity
 import org.yuzu.yuzu_emu.utils.DocumentsTree.Companion.isNativePath
-import org.yuzu.yuzu_emu.utils.FileUtil.exists
-import org.yuzu.yuzu_emu.utils.FileUtil.getFileSize
-import org.yuzu.yuzu_emu.utils.FileUtil.isDirectory
-import org.yuzu.yuzu_emu.utils.FileUtil.openContentUri
+import org.yuzu.yuzu_emu.utils.FileUtil
 import org.yuzu.yuzu_emu.utils.Log
 import org.yuzu.yuzu_emu.utils.SerializableHelper.serializable
 
@@ -75,7 +71,7 @@ object NativeLibrary {
         return if (isNativePath(path!!)) {
             YuzuApplication.documentsTree!!.openContentUri(path, openmode)
         } else {
-            openContentUri(appContext, path, openmode)
+            FileUtil.openContentUri(path, openmode)
         }
     }
 
@@ -85,7 +81,7 @@ object NativeLibrary {
         return if (isNativePath(path!!)) {
             YuzuApplication.documentsTree!!.getFileSize(path)
         } else {
-            getFileSize(appContext, path)
+            FileUtil.getFileSize(path)
         }
     }
 
@@ -95,7 +91,7 @@ object NativeLibrary {
         return if (isNativePath(path!!)) {
             YuzuApplication.documentsTree!!.exists(path)
         } else {
-            exists(appContext, path)
+            FileUtil.exists(path)
         }
     }
 
@@ -105,7 +101,7 @@ object NativeLibrary {
         return if (isNativePath(path!!)) {
             YuzuApplication.documentsTree!!.isDirectory(path)
         } else {
-            isDirectory(appContext, path)
+            FileUtil.isDirectory(path)
         }
     }
 
diff --git a/src/android/app/src/main/java/org/yuzu/yuzu_emu/YuzuApplication.kt b/src/android/app/src/main/java/org/yuzu/yuzu_emu/YuzuApplication.kt
index 9561748cb..8c053670c 100644
--- a/src/android/app/src/main/java/org/yuzu/yuzu_emu/YuzuApplication.kt
+++ b/src/android/app/src/main/java/org/yuzu/yuzu_emu/YuzuApplication.kt
@@ -47,7 +47,7 @@ class YuzuApplication : Application() {
         application = this
         documentsTree = DocumentsTree()
         DirectoryInitialization.start()
-        GpuDriverHelper.initializeDriverParameters(applicationContext)
+        GpuDriverHelper.initializeDriverParameters()
         NativeLibrary.logDeviceInfo()
 
         createNotificationChannels()
diff --git a/src/android/app/src/main/java/org/yuzu/yuzu_emu/adapters/DriverAdapter.kt b/src/android/app/src/main/java/org/yuzu/yuzu_emu/adapters/DriverAdapter.kt
new file mode 100644
index 000000000..0e818cab9
--- /dev/null
+++ b/src/android/app/src/main/java/org/yuzu/yuzu_emu/adapters/DriverAdapter.kt
@@ -0,0 +1,117 @@
+// SPDX-FileCopyrightText: 2023 yuzu Emulator Project
+// SPDX-License-Identifier: GPL-2.0-or-later
+
+package org.yuzu.yuzu_emu.adapters
+
+import android.text.TextUtils
+import android.view.LayoutInflater
+import android.view.View
+import android.view.ViewGroup
+import androidx.recyclerview.widget.AsyncDifferConfig
+import androidx.recyclerview.widget.DiffUtil
+import androidx.recyclerview.widget.ListAdapter
+import androidx.recyclerview.widget.RecyclerView
+import org.yuzu.yuzu_emu.R
+import org.yuzu.yuzu_emu.databinding.CardDriverOptionBinding
+import org.yuzu.yuzu_emu.model.DriverViewModel
+import org.yuzu.yuzu_emu.utils.GpuDriverHelper
+import org.yuzu.yuzu_emu.utils.GpuDriverMetadata
+
+class DriverAdapter(private val driverViewModel: DriverViewModel) :
+    ListAdapter<Pair<String, GpuDriverMetadata>, DriverAdapter.DriverViewHolder>(
+        AsyncDifferConfig.Builder(DiffCallback()).build()
+    ) {
+    override fun onCreateViewHolder(parent: ViewGroup, viewType: Int): DriverViewHolder {
+        val binding =
+            CardDriverOptionBinding.inflate(LayoutInflater.from(parent.context), parent, false)
+        return DriverViewHolder(binding)
+    }
+
+    override fun getItemCount(): Int = currentList.size
+
+    override fun onBindViewHolder(holder: DriverViewHolder, position: Int) =
+        holder.bind(currentList[position])
+
+    private fun onSelectDriver(position: Int) {
+        driverViewModel.setSelectedDriverIndex(position)
+        notifyItemChanged(driverViewModel.previouslySelectedDriver)
+        notifyItemChanged(driverViewModel.selectedDriver)
+    }
+
+    private fun onDeleteDriver(driverData: Pair<String, GpuDriverMetadata>, position: Int) {
+        if (driverViewModel.selectedDriver > position) {
+            driverViewModel.setSelectedDriverIndex(driverViewModel.selectedDriver - 1)
+        }
+        if (GpuDriverHelper.customDriverData == driverData.second) {
+            driverViewModel.setSelectedDriverIndex(0)
+        }
+        driverViewModel.driversToDelete.add(driverData.first)
+        driverViewModel.removeDriver(driverData)
+        notifyItemRemoved(position)
+        notifyItemChanged(driverViewModel.selectedDriver)
+    }
+
+    inner class DriverViewHolder(val binding: CardDriverOptionBinding) :
+        RecyclerView.ViewHolder(binding.root) {
+        private lateinit var driverData: Pair<String, GpuDriverMetadata>
+
+        fun bind(driverData: Pair<String, GpuDriverMetadata>) {
+            this.driverData = driverData
+            val driver = driverData.second
+
+            binding.apply {
+                radioButton.isChecked = driverViewModel.selectedDriver == bindingAdapterPosition
+                root.setOnClickListener {
+                    onSelectDriver(bindingAdapterPosition)
+                }
+                buttonDelete.setOnClickListener {
+                    onDeleteDriver(driverData, bindingAdapterPosition)
+                }
+
+                // Delay marquee by 3s
+                title.postDelayed(
+                    {
+                        title.isSelected = true
+                        title.ellipsize = TextUtils.TruncateAt.MARQUEE
+                        version.isSelected = true
+                        version.ellipsize = TextUtils.TruncateAt.MARQUEE
+                        description.isSelected = true
+                        description.ellipsize = TextUtils.TruncateAt.MARQUEE
+                    },
+                    3000
+                )
+                if (driver.name == null) {
+                    title.setText(R.string.system_gpu_driver)
+                    description.text = ""
+                    version.text = ""
+                    version.visibility = View.GONE
+                    description.visibility = View.GONE
+                    buttonDelete.visibility = View.GONE
+                } else {
+                    title.text = driver.name
+                    version.text = driver.version
+                    description.text = driver.description
+                    version.visibility = View.VISIBLE
+                    description.visibility = View.VISIBLE
+                    buttonDelete.visibility = View.VISIBLE
+                }
+            }
+        }
+    }
+
+    private class DiffCallback : DiffUtil.ItemCallback<Pair<String, GpuDriverMetadata>>() {
+        override fun areItemsTheSame(
+            oldItem: Pair<String, GpuDriverMetadata>,
+            newItem: Pair<String, GpuDriverMetadata>
+        ): Boolean {
+            return oldItem.first == newItem.first
+        }
+
+        override fun areContentsTheSame(
+            oldItem: Pair<String, GpuDriverMetadata>,
+            newItem: Pair<String, GpuDriverMetadata>
+        ): Boolean {
+            return oldItem.second == newItem.second
+        }
+    }
+}
diff --git a/src/android/app/src/main/java/org/yuzu/yuzu_emu/fragments/DriverManagerFragment.kt b/src/android/app/src/main/java/org/yuzu/yuzu_emu/fragments/DriverManagerFragment.kt
new file mode 100644
index 000000000..df21d74b2
--- /dev/null
+++ b/src/android/app/src/main/java/org/yuzu/yuzu_emu/fragments/DriverManagerFragment.kt
@@ -0,0 +1,186 @@
+// SPDX-FileCopyrightText: 2023 yuzu Emulator Project
+// SPDX-License-Identifier: GPL-2.0-or-later
+
+package org.yuzu.yuzu_emu.fragments
+
+import android.os.Bundle
+import android.view.LayoutInflater
+import android.view.View
+import android.view.ViewGroup
+import androidx.activity.result.contract.ActivityResultContracts
+import androidx.core.view.ViewCompat
+import androidx.core.view.WindowInsetsCompat
+import androidx.core.view.updatePadding
+import androidx.fragment.app.Fragment
+import androidx.fragment.app.activityViewModels
+import androidx.lifecycle.lifecycleScope
+import androidx.navigation.findNavController
+import androidx.recyclerview.widget.GridLayoutManager
+import com.google.android.material.transition.MaterialSharedAxis
+import kotlinx.coroutines.flow.collectLatest
+import kotlinx.coroutines.launch
+import org.yuzu.yuzu_emu.R
+import org.yuzu.yuzu_emu.adapters.DriverAdapter
+import org.yuzu.yuzu_emu.databinding.FragmentDriverManagerBinding
+import org.yuzu.yuzu_emu.model.DriverViewModel
+import org.yuzu.yuzu_emu.model.HomeViewModel
+import org.yuzu.yuzu_emu.utils.FileUtil
+import org.yuzu.yuzu_emu.utils.GpuDriverHelper
+import java.io.File
+import java.io.IOException
+
+class DriverManagerFragment : Fragment() {
+    private var _binding: FragmentDriverManagerBinding? = null
+    private val binding get() = _binding!!
+
+    private val homeViewModel: HomeViewModel by activityViewModels()
+    private val driverViewModel: DriverViewModel by activityViewModels()
+
+    override fun onCreate(savedInstanceState: Bundle?) {
+        super.onCreate(savedInstanceState)
+        enterTransition = MaterialSharedAxis(MaterialSharedAxis.X, true)
+        returnTransition = MaterialSharedAxis(MaterialSharedAxis.X, false)
+        reenterTransition = MaterialSharedAxis(MaterialSharedAxis.X, false)
+    }
+
+    override fun onCreateView(
+        inflater: LayoutInflater,
+        container: ViewGroup?,
+        savedInstanceState: Bundle?
+    ): View {
+        _binding = FragmentDriverManagerBinding.inflate(inflater)
+        return binding.root
+    }
+
+    override fun onViewCreated(view: View, savedInstanceState: Bundle?) {
+        super.onViewCreated(view, savedInstanceState)
+        homeViewModel.setNavigationVisibility(visible = false, animated = true)
+        homeViewModel.setStatusBarShadeVisibility(visible = false)
+
+        if (!driverViewModel.isInteractionAllowed) {
+            DriversLoadingDialogFragment().show(
+                childFragmentManager,
+                DriversLoadingDialogFragment.TAG
+            )
+        }
+
+        binding.toolbarDrivers.setNavigationOnClickListener {
+            binding.root.findNavController().popBackStack()
+        }
+
+        binding.buttonInstall.setOnClickListener {
+            getDriver.launch(arrayOf("application/zip"))
+        }
+
+        binding.listDrivers.apply {
+            layoutManager = GridLayoutManager(
+                requireContext(),
+                resources.getInteger(R.integer.grid_columns)
+            )
+            adapter = DriverAdapter(driverViewModel)
+        }
+
+        viewLifecycleOwner.lifecycleScope.apply {
+            launch {
+                driverViewModel.driverList.collectLatest {
+                    (binding.listDrivers.adapter as DriverAdapter).submitList(it)
+                }
+            }
+            launch {
+                driverViewModel.newDriverInstalled.collect {
+                    if (_binding != null && it) {
+                        (binding.listDrivers.adapter as DriverAdapter).apply {
+                            notifyItemChanged(driverViewModel.previouslySelectedDriver)
+                            notifyItemChanged(driverViewModel.selectedDriver)
+                            driverViewModel.setNewDriverInstalled(false)
+                        }
+                    }
+                }
+            }
+        }
+
+        setInsets()
+    }
+
+    // Start installing requested driver
+    override fun onStop() {
+        super.onStop()
+        driverViewModel.onCloseDriverManager()
+    }
+
+    private fun setInsets() =
+        ViewCompat.setOnApplyWindowInsetsListener(
+            binding.root
+        ) { _: View, windowInsets: WindowInsetsCompat ->
+            val barInsets = windowInsets.getInsets(WindowInsetsCompat.Type.systemBars())
+            val cutoutInsets = windowInsets.getInsets(WindowInsetsCompat.Type.displayCutout())
+
+            val leftInsets = barInsets.left + cutoutInsets.left
+            val rightInsets = barInsets.right + cutoutInsets.right
+
+            val mlpAppBar = binding.toolbarDrivers.layoutParams as ViewGroup.MarginLayoutParams
+            mlpAppBar.leftMargin = leftInsets
+            mlpAppBar.rightMargin = rightInsets
+            binding.toolbarDrivers.layoutParams = mlpAppBar
+
+            val mlplistDrivers = binding.listDrivers.layoutParams as ViewGroup.MarginLayoutParams
+            mlplistDrivers.leftMargin = leftInsets
+            mlplistDrivers.rightMargin = rightInsets
+            binding.listDrivers.layoutParams = mlplistDrivers
+
+            val fabSpacing = resources.getDimensionPixelSize(R.dimen.spacing_fab)
+            val mlpFab =
+                binding.buttonInstall.layoutParams as ViewGroup.MarginLayoutParams
+            mlpFab.leftMargin = leftInsets + fabSpacing
+            mlpFab.rightMargin = rightInsets + fabSpacing
+            mlpFab.bottomMargin = barInsets.bottom + fabSpacing
+            binding.buttonInstall.layoutParams = mlpFab
+
+            binding.listDrivers.updatePadding(
+                bottom = barInsets.bottom +
+                    resources.getDimensionPixelSize(R.dimen.spacing_bottom_list_fab)
+            )
+
+            windowInsets
+        }
+
+    private val getDriver =
+        registerForActivityResult(ActivityResultContracts.OpenDocument()) { result ->
+            if (result == null) {
+                return@registerForActivityResult
+            }
+
+            IndeterminateProgressDialogFragment.newInstance(
+                requireActivity(),
+                R.string.installing_driver,
+                false
+            ) {
+                val driverPath =
+                    "${GpuDriverHelper.driverStoragePath}/${FileUtil.getFilename(result)}"
+                val driverFile = File(driverPath)
+
+                // Ignore file exceptions when a user selects an invalid zip
+                try {
+                    if (!GpuDriverHelper.copyDriverToInternalStorage(result)) {
+                        throw IOException("Driver failed validation!")
+                    }
+                } catch (_: IOException) {
+                    if (driverFile.exists()) {
+                        driverFile.delete()
+                    }
+                    return@newInstance getString(R.string.select_gpu_driver_error)
+                }
+
+                val driverData = GpuDriverHelper.getMetadataFromZip(driverFile)
+                val driverInList =
+                    driverViewModel.driverList.value.firstOrNull { it.second == driverData }
+                if (driverInList != null) {
+                    return@newInstance getString(R.string.driver_already_installed)
+                } else {
+                    driverViewModel.addDriver(Pair(driverPath, driverData))
+                    driverViewModel.setNewDriverInstalled(true)
+                }
+                return@newInstance Any()
+            }.show(childFragmentManager, IndeterminateProgressDialogFragment.TAG)
+        }
+}
diff --git a/src/android/app/src/main/java/org/yuzu/yuzu_emu/fragments/DriversLoadingDialogFragment.kt b/src/android/app/src/main/java/org/yuzu/yuzu_emu/fragments/DriversLoadingDialogFragment.kt
new file mode 100644
index 000000000..f8c34346a
--- /dev/null
+++ b/src/android/app/src/main/java/org/yuzu/yuzu_emu/fragments/DriversLoadingDialogFragment.kt
@@ -0,0 +1,75 @@
+// SPDX-FileCopyrightText: 2023 yuzu Emulator Project
+// SPDX-License-Identifier: GPL-2.0-or-later
+
+package org.yuzu.yuzu_emu.fragments
+
+import android.app.Dialog
+import android.os.Bundle
+import android.view.LayoutInflater
+import android.view.View
+import android.view.ViewGroup
+import androidx.fragment.app.DialogFragment
+import androidx.fragment.app.activityViewModels
+import androidx.lifecycle.Lifecycle
+import androidx.lifecycle.lifecycleScope
+import androidx.lifecycle.repeatOnLifecycle
+import com.google.android.material.dialog.MaterialAlertDialogBuilder
+import kotlinx.coroutines.launch
+import org.yuzu.yuzu_emu.R
+import org.yuzu.yuzu_emu.databinding.DialogProgressBarBinding
+import org.yuzu.yuzu_emu.model.DriverViewModel
+
+class DriversLoadingDialogFragment : DialogFragment() {
+    private val driverViewModel: DriverViewModel by activityViewModels()
+
+    private lateinit var binding: DialogProgressBarBinding
+
+    override fun onCreateDialog(savedInstanceState: Bundle?): Dialog {
+        binding = DialogProgressBarBinding.inflate(layoutInflater)
+        binding.progressBar.isIndeterminate = true
+
+        isCancelable = false
+
+        return MaterialAlertDialogBuilder(requireContext())
+            .setTitle(R.string.loading)
+            .setView(binding.root)
+            .create()
+    }
+
+    override fun onCreateView(
+        inflater: LayoutInflater,
+        container: ViewGroup?,
+        savedInstanceState: Bundle?
+    ): View = binding.root
+
+    override fun onViewCreated(view: View, savedInstanceState: Bundle?) {
+        super.onViewCreated(view, savedInstanceState)
+        viewLifecycleOwner.lifecycleScope.apply {
+            launch {
+                repeatOnLifecycle(Lifecycle.State.RESUMED) {
+                    driverViewModel.areDriversLoading.collect { checkForDismiss() }
+                }
+            }
+            launch {
+                repeatOnLifecycle(Lifecycle.State.RESUMED) {
+                    driverViewModel.isDriverReady.collect { checkForDismiss() }
+                }
+            }
+            launch {
+                repeatOnLifecycle(Lifecycle.State.RESUMED) {
+                    driverViewModel.isDeletingDrivers.collect { checkForDismiss() }
+                }
+            }
+        }
+    }
+
+    private fun checkForDismiss() {
+        if (driverViewModel.isInteractionAllowed) {
+            dismiss()
+        }
+    }
+
+    companion object {
+        const val TAG = "DriversLoadingDialogFragment"
+    }
+}
diff --git a/src/android/app/src/main/java/org/yuzu/yuzu_emu/fragments/EmulationFragment.kt b/src/android/app/src/main/java/org/yuzu/yuzu_emu/fragments/EmulationFragment.kt
index e6ad2aa77..598a9d42b 100644
--- a/src/android/app/src/main/java/org/yuzu/yuzu_emu/fragments/EmulationFragment.kt
+++ b/src/android/app/src/main/java/org/yuzu/yuzu_emu/fragments/EmulationFragment.kt
@@ -39,6 +39,7 @@ import androidx.window.layout.WindowLayoutInfo
 import com.google.android.material.dialog.MaterialAlertDialogBuilder
 import com.google.android.material.slider.Slider
 import kotlinx.coroutines.Dispatchers
+import kotlinx.coroutines.flow.collect
 import kotlinx.coroutines.flow.collectLatest
 import kotlinx.coroutines.launch
 import org.yuzu.yuzu_emu.HomeNavigationDirections
@@ -50,6 +51,7 @@ import org.yuzu.yuzu_emu.databinding.DialogOverlayAdjustBinding
 import org.yuzu.yuzu_emu.databinding.FragmentEmulationBinding
 import org.yuzu.yuzu_emu.features.settings.model.IntSetting
 import org.yuzu.yuzu_emu.features.settings.model.Settings
+import org.yuzu.yuzu_emu.model.DriverViewModel
 import org.yuzu.yuzu_emu.model.Game
 import org.yuzu.yuzu_emu.model.EmulationViewModel
 import org.yuzu.yuzu_emu.overlay.InputOverlay
@@ -70,6 +72,7 @@ class EmulationFragment : Fragment(), SurfaceHolder.Callback {
     private lateinit var game: Game
 
     private val emulationViewModel: EmulationViewModel by activityViewModels()
+    private val driverViewModel: DriverViewModel by activityViewModels()
 
     private var isInFoldableLayout = false
 
@@ -299,6 +302,21 @@ class EmulationFragment : Fragment(), SurfaceHolder.Callback {
                     }
                 }
             }
+            launch {
+                repeatOnLifecycle(Lifecycle.State.RESUMED) {
+                    driverViewModel.isDriverReady.collect {
+                        if (it && !emulationState.isRunning) {
+                            if (!DirectoryInitialization.areDirectoriesReady) {
+                                DirectoryInitialization.start()
+                            }
+
+                            updateScreenLayout()
+
+                            emulationState.run(emulationActivity!!.isActivityRecreated)
+                        }
+                    }
+                }
+            }
         }
     }
 
@@ -332,17 +350,6 @@ class EmulationFragment : Fragment(), SurfaceHolder.Callback {
         }
     }
 
-    override fun onResume() {
-        super.onResume()
-        if (!DirectoryInitialization.areDirectoriesReady) {
-            DirectoryInitialization.start()
-        }
-
-        updateScreenLayout()
-
-        emulationState.run(emulationActivity!!.isActivityRecreated)
-    }
-
     override fun onPause() {
         if (emulationState.isRunning && emulationActivity?.isInPictureInPictureMode != true) {
             emulationState.pause()
diff --git a/src/android/app/src/main/java/org/yuzu/yuzu_emu/fragments/HomeSettingsFragment.kt b/src/android/app/src/main/java/org/yuzu/yuzu_emu/fragments/HomeSettingsFragment.kt
index 8923c0ea2..f273c880a 100644
--- a/src/android/app/src/main/java/org/yuzu/yuzu_emu/fragments/HomeSettingsFragment.kt
+++ b/src/android/app/src/main/java/org/yuzu/yuzu_emu/fragments/HomeSettingsFragment.kt
@@ -5,7 +5,6 @@ package org.yuzu.yuzu_emu.fragments
 
 import android.Manifest
 import android.content.ActivityNotFoundException
-import android.content.DialogInterface
 import android.content.Intent
 import android.content.pm.PackageManager
 import android.os.Bundle
@@ -27,8 +26,7 @@ import androidx.fragment.app.Fragment
 import androidx.fragment.app.activityViewModels
 import androidx.navigation.findNavController
 import androidx.navigation.fragment.findNavController
-import androidx.recyclerview.widget.LinearLayoutManager
-import com.google.android.material.dialog.MaterialAlertDialogBuilder
+import androidx.recyclerview.widget.GridLayoutManager
 import com.google.android.material.transition.MaterialSharedAxis
 import org.yuzu.yuzu_emu.BuildConfig
 import org.yuzu.yuzu_emu.HomeNavigationDirections
@@ -37,6 +35,7 @@ import org.yuzu.yuzu_emu.adapters.HomeSettingAdapter
 import org.yuzu.yuzu_emu.databinding.FragmentHomeSettingsBinding
 import org.yuzu.yuzu_emu.features.DocumentProvider
 import org.yuzu.yuzu_emu.features.settings.model.Settings
+import org.yuzu.yuzu_emu.model.DriverViewModel
 import org.yuzu.yuzu_emu.model.HomeSetting
 import org.yuzu.yuzu_emu.model.HomeViewModel
 import org.yuzu.yuzu_emu.ui.main.MainActivity
@@ -50,6 +49,7 @@ class HomeSettingsFragment : Fragment() {
     private lateinit var mainActivity: MainActivity
 
     private val homeViewModel: HomeViewModel by activityViewModels()
+    private val driverViewModel: DriverViewModel by activityViewModels()
 
     override fun onCreate(savedInstanceState: Bundle?) {
         super.onCreate(savedInstanceState)
@@ -107,13 +107,17 @@ class HomeSettingsFragment : Fragment() {
             )
             add(
                 HomeSetting(
-                    R.string.install_gpu_driver,
+                    R.string.gpu_driver_manager,
                     R.string.install_gpu_driver_description,
-                    R.drawable.ic_exit,
-                    { driverInstaller() },
+                    R.drawable.ic_build,
+                    {
+                        binding.root.findNavController()
+                            .navigate(R.id.action_homeSettingsFragment_to_driverManagerFragment)
+                    },
                     { GpuDriverHelper.supportsCustomDriverLoading() },
                     R.string.custom_driver_not_supported,
-                    R.string.custom_driver_not_supported_description
+                    R.string.custom_driver_not_supported_description,
+                    driverViewModel.selectedDriverMetadata
                 )
             )
             add(
@@ -182,7 +186,8 @@ class HomeSettingsFragment : Fragment() {
         }
 
         binding.homeSettingsList.apply {
-            layoutManager = LinearLayoutManager(requireContext())
+            layoutManager =
+                GridLayoutManager(requireContext(), resources.getInteger(R.integer.grid_columns))
             adapter = HomeSettingAdapter(
                 requireActivity() as AppCompatActivity,
                 viewLifecycleOwner,
@@ -292,31 +297,6 @@ class HomeSettingsFragment : Fragment() {
         }
     }
 
-    private fun driverInstaller() {
-        // Get the driver name for the dialog message.
-        var driverName = GpuDriverHelper.customDriverName
-        if (driverName == null) {
-            driverName = getString(R.string.system_gpu_driver)
-        }
-
-        MaterialAlertDialogBuilder(requireContext())
-            .setTitle(getString(R.string.select_gpu_driver_title))
-            .setMessage(driverName)
-            .setNegativeButton(android.R.string.cancel, null)
-            .setNeutralButton(R.string.select_gpu_driver_default) { _: DialogInterface?, _: Int ->
-                GpuDriverHelper.installDefaultDriver(requireContext())
-                Toast.makeText(
-                    requireContext(),
-                    R.string.select_gpu_driver_use_default,
-                    Toast.LENGTH_SHORT
-                ).show()
-            }
-            .setPositiveButton(R.string.select_gpu_driver_install) { _: DialogInterface?, _: Int ->
-                mainActivity.getDriver.launch(arrayOf("application/zip"))
-            }
-            .show()
-    }
-
     private fun shareLog() {
         val file = DocumentFile.fromSingleUri(
             mainActivity,
diff --git a/src/android/app/src/main/java/org/yuzu/yuzu_emu/fragments/IndeterminateProgressDialogFragment.kt b/src/android/app/src/main/java/org/yuzu/yuzu_emu/fragments/IndeterminateProgressDialogFragment.kt
index f128deda8..7e467814d 100644
--- a/src/android/app/src/main/java/org/yuzu/yuzu_emu/fragments/IndeterminateProgressDialogFragment.kt
+++ b/src/android/app/src/main/java/org/yuzu/yuzu_emu/fragments/IndeterminateProgressDialogFragment.kt
@@ -10,8 +10,8 @@ import android.view.View
 import android.view.ViewGroup
 import android.widget.Toast
 import androidx.appcompat.app.AlertDialog
-import androidx.appcompat.app.AppCompatActivity
 import androidx.fragment.app.DialogFragment
+import androidx.fragment.app.FragmentActivity
 import androidx.fragment.app.activityViewModels
 import androidx.lifecycle.Lifecycle
 import androidx.lifecycle.ViewModelProvider
@@ -78,6 +78,10 @@ class IndeterminateProgressDialogFragment : DialogFragment() {
                                     requireActivity().supportFragmentManager,
                                     MessageDialogFragment.TAG
                                 )
+
+                                else -> {
+                                    // Do nothing
+                                }
                             }
                             taskViewModel.clear()
                         }
@@ -115,7 +119,7 @@ class IndeterminateProgressDialogFragment : DialogFragment() {
         private const val CANCELLABLE = "Cancellable"
 
         fun newInstance(
-            activity: AppCompatActivity,
+            activity: FragmentActivity,
             titleId: Int,
             cancellable: Boolean = false,
             task: () -> Any
diff --git a/src/android/app/src/main/java/org/yuzu/yuzu_emu/model/DriverViewModel.kt b/src/android/app/src/main/java/org/yuzu/yuzu_emu/model/DriverViewModel.kt
new file mode 100644
index 000000000..62945ad65
--- /dev/null
+++ b/src/android/app/src/main/java/org/yuzu/yuzu_emu/model/DriverViewModel.kt
@@ -0,0 +1,158 @@
+// SPDX-FileCopyrightText: 2023 yuzu Emulator Project
+// SPDX-License-Identifier: GPL-2.0-or-later
+
+package org.yuzu.yuzu_emu.model
+
+import androidx.lifecycle.ViewModel
+import androidx.lifecycle.viewModelScope
+import kotlinx.coroutines.Dispatchers
+import kotlinx.coroutines.flow.MutableStateFlow
+import kotlinx.coroutines.flow.StateFlow
+import kotlinx.coroutines.launch
+import kotlinx.coroutines.withContext
+import org.yuzu.yuzu_emu.R
+import org.yuzu.yuzu_emu.YuzuApplication
+import org.yuzu.yuzu_emu.utils.FileUtil
+import org.yuzu.yuzu_emu.utils.GpuDriverHelper
+import org.yuzu.yuzu_emu.utils.GpuDriverMetadata
+import java.io.BufferedOutputStream
+import java.io.File
+
+class DriverViewModel : ViewModel() {
+    private val _areDriversLoading = MutableStateFlow(false)
+    val areDriversLoading: StateFlow<Boolean> get() = _areDriversLoading
+
+    private val _isDriverReady = MutableStateFlow(true)
+    val isDriverReady: StateFlow<Boolean> get() = _isDriverReady
+
+    private val _isDeletingDrivers = MutableStateFlow(false)
+    val isDeletingDrivers: StateFlow<Boolean> get() = _isDeletingDrivers
+
+    private val _driverList = MutableStateFlow(mutableListOf<Pair<String, GpuDriverMetadata>>())
+    val driverList: StateFlow<MutableList<Pair<String, GpuDriverMetadata>>> get() = _driverList
+
+    var previouslySelectedDriver = 0
+    var selectedDriver = -1
+
+    private val _selectedDriverMetadata =
+        MutableStateFlow(
+            GpuDriverHelper.customDriverData.name
+                ?: YuzuApplication.appContext.getString(R.string.system_gpu_driver)
+        )
+    val selectedDriverMetadata: StateFlow<String> get() = _selectedDriverMetadata
+
+    private val _newDriverInstalled = MutableStateFlow(false)
+    val newDriverInstalled: StateFlow<Boolean> get() = _newDriverInstalled
+
+    val driversToDelete = mutableListOf<String>()
+
+    val isInteractionAllowed
+        get() = !areDriversLoading.value && isDriverReady.value && !isDeletingDrivers.value
+
+    init {
+        _areDriversLoading.value = true
+        viewModelScope.launch {
+            withContext(Dispatchers.IO) {
+                val drivers = GpuDriverHelper.getDrivers()
+                val currentDriverMetadata = GpuDriverHelper.customDriverData
+                for (i in drivers.indices) {
+                    if (drivers[i].second == currentDriverMetadata) {
+                        setSelectedDriverIndex(i)
+                        break
+                    }
+                }
+
+                // If a user had installed a driver before the manager was implemented, this zips
+                // the installed driver to UserData/gpu_drivers/CustomDriver.zip so that it can
+                // be indexed and exported as expected.
+                if (selectedDriver == -1) {
+                    val driverToSave =
+                        File(GpuDriverHelper.driverStoragePath, "CustomDriver.zip")
+                    driverToSave.createNewFile()
+                    FileUtil.zipFromInternalStorage(
+                        File(GpuDriverHelper.driverInstallationPath!!),
+                        GpuDriverHelper.driverInstallationPath!!,
+                        BufferedOutputStream(driverToSave.outputStream())
+                    )
+                    drivers.add(Pair(driverToSave.path, currentDriverMetadata))
+                    setSelectedDriverIndex(drivers.size - 1)
+                }
+
+                _driverList.value = drivers
+                _areDriversLoading.value = false
+            }
+        }
+    }
+
+    fun setSelectedDriverIndex(value: Int) {
+        if (selectedDriver != -1) {
+            previouslySelectedDriver = selectedDriver
+        }
+        selectedDriver = value
+    }
+
+    fun setNewDriverInstalled(value: Boolean) {
+        _newDriverInstalled.value = value
+    }
+
+    fun addDriver(driverData: Pair<String, GpuDriverMetadata>) {
+        val driverIndex = _driverList.value.indexOfFirst { it == driverData }
+        if (driverIndex == -1) {
+            setSelectedDriverIndex(_driverList.value.size)
+            _driverList.value.add(driverData)
+            _selectedDriverMetadata.value = driverData.second.name
+                ?: YuzuApplication.appContext.getString(R.string.system_gpu_driver)
+        } else {
+            setSelectedDriverIndex(driverIndex)
+        }
+    }
+
+    fun removeDriver(driverData: Pair<String, GpuDriverMetadata>) {
+        _driverList.value.remove(driverData)
+    }
+
+    fun onCloseDriverManager() {
+        _isDeletingDrivers.value = true
+        viewModelScope.launch {
+            withContext(Dispatchers.IO) {
+                driversToDelete.forEach {
+                    val driver = File(it)
+                    if (driver.exists()) {
+                        driver.delete()
+                    }
+                }
+                driversToDelete.clear()
+                _isDeletingDrivers.value = false
+            }
+        }
+
+        if (GpuDriverHelper.customDriverData == driverList.value[selectedDriver].second) {
+            return
+        }
+
+        _isDriverReady.value = false
+        viewModelScope.launch {
+            withContext(Dispatchers.IO) {
+                if (selectedDriver == 0) {
+                    GpuDriverHelper.installDefaultDriver()
+                    setDriverReady()
+                    return@withContext
+                }
+
+                val driverToInstall = File(driverList.value[selectedDriver].first)
+                if (driverToInstall.exists()) {
+                    GpuDriverHelper.installCustomDriver(driverToInstall)
+                } else {
+                    GpuDriverHelper.installDefaultDriver()
+                }
+                setDriverReady()
+            }
+        }
+    }
+
+    private fun setDriverReady() {
+        _isDriverReady.value = true
+        _selectedDriverMetadata.value = GpuDriverHelper.customDriverData.name
+            ?: YuzuApplication.appContext.getString(R.string.system_gpu_driver)
+    }
+}
diff --git a/src/android/app/src/main/java/org/yuzu/yuzu_emu/ui/main/MainActivity.kt b/src/android/app/src/main/java/org/yuzu/yuzu_emu/ui/main/MainActivity.kt
index 0fa5df5e5..233aa4101 100644
--- a/src/android/app/src/main/java/org/yuzu/yuzu_emu/ui/main/MainActivity.kt
+++ b/src/android/app/src/main/java/org/yuzu/yuzu_emu/ui/main/MainActivity.kt
@@ -29,12 +29,10 @@ import androidx.navigation.fragment.NavHostFragment
 import androidx.navigation.ui.setupWithNavController
 import androidx.preference.PreferenceManager
 import com.google.android.material.color.MaterialColors
-import com.google.android.material.dialog.MaterialAlertDialogBuilder
 import com.google.android.material.navigation.NavigationBarView
 import kotlinx.coroutines.CoroutineScope
 import java.io.File
 import java.io.FilenameFilter
-import java.io.IOException
 import kotlinx.coroutines.Dispatchers
 import kotlinx.coroutines.launch
 import kotlinx.coroutines.withContext
@@ -43,7 +41,6 @@ import org.yuzu.yuzu_emu.NativeLibrary
 import org.yuzu.yuzu_emu.R
 import org.yuzu.yuzu_emu.activities.EmulationActivity
 import org.yuzu.yuzu_emu.databinding.ActivityMainBinding
-import org.yuzu.yuzu_emu.databinding.DialogProgressBarBinding
 import org.yuzu.yuzu_emu.features.DocumentProvider
 import org.yuzu.yuzu_emu.features.settings.model.Settings
 import org.yuzu.yuzu_emu.fragments.IndeterminateProgressDialogFragment
@@ -343,11 +340,10 @@ class MainActivity : AppCompatActivity(), ThemeProvider {
 
         val dstPath = DirectoryInitialization.userDirectory + "/keys/"
         if (FileUtil.copyUriToInternalStorage(
-                applicationContext,
                 result,
                 dstPath,
                 "prod.keys"
-            )
+            ) != null
         ) {
             if (NativeLibrary.reloadKeys()) {
                 Toast.makeText(
@@ -446,11 +442,10 @@ class MainActivity : AppCompatActivity(), ThemeProvider {
 
             val dstPath = DirectoryInitialization.userDirectory + "/keys/"
             if (FileUtil.copyUriToInternalStorage(
-                    applicationContext,
                     result,
                     dstPath,
                     "key_retail.bin"
-                )
+                ) != null
             ) {
                 if (NativeLibrary.reloadKeys()) {
                     Toast.makeText(
@@ -469,59 +464,6 @@ class MainActivity : AppCompatActivity(), ThemeProvider {
             }
         }
 
-    val getDriver =
-        registerForActivityResult(ActivityResultContracts.OpenDocument()) { result ->
-            if (result == null) {
-                return@registerForActivityResult
-            }
-
-            val takeFlags =
-                Intent.FLAG_GRANT_WRITE_URI_PERMISSION or Intent.FLAG_GRANT_READ_URI_PERMISSION
-            contentResolver.takePersistableUriPermission(
-                result,
-                takeFlags
-            )
-
-            val progressBinding = DialogProgressBarBinding.inflate(layoutInflater)
-            progressBinding.progressBar.isIndeterminate = true
-            val installationDialog = MaterialAlertDialogBuilder(this)
-                .setTitle(R.string.installing_driver)
-                .setView(progressBinding.root)
-                .show()
-
-            lifecycleScope.launch {
-                withContext(Dispatchers.IO) {
-                    // Ignore file exceptions when a user selects an invalid zip
-                    try {
-                        GpuDriverHelper.installCustomDriver(applicationContext, result)
-                    } catch (_: IOException) {
-                    }
-
-                    withContext(Dispatchers.Main) {
-                        installationDialog.dismiss()
-
-                        val driverName = GpuDriverHelper.customDriverName
-                        if (driverName != null) {
-                            Toast.makeText(
-                                applicationContext,
-                                getString(
-                                    R.string.select_gpu_driver_install_success,
-                                    driverName
-                                ),
-                                Toast.LENGTH_SHORT
-                            ).show()
-                        } else {
-                            Toast.makeText(
-                                applicationContext,
-                                R.string.select_gpu_driver_error,
-                                Toast.LENGTH_LONG
-                            ).show()
-                        }
-                    }
-                }
-            }
-        }
-
     val installGameUpdate = registerForActivityResult(
         ActivityResultContracts.OpenMultipleDocuments()
     ) { documents: List<Uri> ->
diff --git a/src/android/app/src/main/java/org/yuzu/yuzu_emu/utils/DocumentsTree.kt b/src/android/app/src/main/java/org/yuzu/yuzu_emu/utils/DocumentsTree.kt
index cf226ad94..eafcf9e42 100644
--- a/src/android/app/src/main/java/org/yuzu/yuzu_emu/utils/DocumentsTree.kt
+++ b/src/android/app/src/main/java/org/yuzu/yuzu_emu/utils/DocumentsTree.kt
@@ -7,7 +7,6 @@ import android.net.Uri
 import androidx.documentfile.provider.DocumentFile
 import java.io.File
 import java.util.*
-import org.yuzu.yuzu_emu.YuzuApplication
 import org.yuzu.yuzu_emu.model.MinimalDocumentFile
 
 class DocumentsTree {
@@ -22,7 +21,7 @@ class DocumentsTree {
 
     fun openContentUri(filepath: String, openMode: String?): Int {
         val node = resolvePath(filepath) ?: return -1
-        return FileUtil.openContentUri(YuzuApplication.appContext, node.uri.toString(), openMode)
+        return FileUtil.openContentUri(node.uri.toString(), openMode)
     }
 
     fun getFileSize(filepath: String): Long {
@@ -30,7 +29,7 @@ class DocumentsTree {
         return if (node == null || node.isDirectory) {
             0
         } else {
-            FileUtil.getFileSize(YuzuApplication.appContext, node.uri.toString())
+            FileUtil.getFileSize(node.uri.toString())
         }
     }
 
@@ -67,7 +66,7 @@ class DocumentsTree {
      * @param parent parent node of this level
      */
     private fun structTree(parent: DocumentsNode) {
-        val documents = FileUtil.listFiles(YuzuApplication.appContext, parent.uri!!)
+        val documents = FileUtil.listFiles(parent.uri!!)
         for (document in documents) {
             val node = DocumentsNode(document)
             node.parent = parent
diff --git a/src/android/app/src/main/java/org/yuzu/yuzu_emu/utils/FileUtil.kt b/src/android/app/src/main/java/org/yuzu/yuzu_emu/utils/FileUtil.kt
index c3f53f1c5..5ee74a52c 100644
--- a/src/android/app/src/main/java/org/yuzu/yuzu_emu/utils/FileUtil.kt
+++ b/src/android/app/src/main/java/org/yuzu/yuzu_emu/utils/FileUtil.kt
@@ -3,7 +3,6 @@
 
 package org.yuzu.yuzu_emu.utils
 
-import android.content.Context
 import android.database.Cursor
 import android.net.Uri
 import android.provider.DocumentsContract
@@ -11,7 +10,6 @@ import androidx.documentfile.provider.DocumentFile
 import kotlinx.coroutines.flow.StateFlow
 import java.io.BufferedInputStream
 import java.io.File
-import java.io.FileOutputStream
 import java.io.IOException
 import java.io.InputStream
 import java.net.URLDecoder
@@ -21,6 +19,8 @@ import org.yuzu.yuzu_emu.YuzuApplication
 import org.yuzu.yuzu_emu.model.MinimalDocumentFile
 import org.yuzu.yuzu_emu.model.TaskState
 import java.io.BufferedOutputStream
+import java.lang.NullPointerException
+import java.nio.charset.StandardCharsets
 import java.util.zip.ZipOutputStream
 
 object FileUtil {
@@ -29,6 +29,8 @@ object FileUtil {
     const val APPLICATION_OCTET_STREAM = "application/octet-stream"
     const val TEXT_PLAIN = "text/plain"
 
+    private val context get() = YuzuApplication.appContext
+
     /**
      * Create a file from directory with filename.
      * @param context Application context
@@ -36,11 +38,11 @@ object FileUtil {
      * @param filename file display name.
      * @return boolean
      */
-    fun createFile(context: Context?, directory: String?, filename: String): DocumentFile? {
+    fun createFile(directory: String?, filename: String): DocumentFile? {
         var decodedFilename = filename
         try {
             val directoryUri = Uri.parse(directory)
-            val parent = DocumentFile.fromTreeUri(context!!, directoryUri) ?: return null
+            val parent = DocumentFile.fromTreeUri(context, directoryUri) ?: return null
             decodedFilename = URLDecoder.decode(decodedFilename, DECODE_METHOD)
             var mimeType = APPLICATION_OCTET_STREAM
             if (decodedFilename.endsWith(".txt")) {
@@ -56,16 +58,15 @@ object FileUtil {
 
     /**
      * Create a directory from directory with filename.
-     * @param context Application context
      * @param directory parent path for directory.
      * @param directoryName directory display name.
      * @return boolean
      */
-    fun createDir(context: Context?, directory: String?, directoryName: String?): DocumentFile? {
+    fun createDir(directory: String?, directoryName: String?): DocumentFile? {
         var decodedDirectoryName = directoryName
         try {
             val directoryUri = Uri.parse(directory)
-            val parent = DocumentFile.fromTreeUri(context!!, directoryUri) ?: return null
+            val parent = DocumentFile.fromTreeUri(context, directoryUri) ?: return null
             decodedDirectoryName = URLDecoder.decode(decodedDirectoryName, DECODE_METHOD)
             val isExist = parent.findFile(decodedDirectoryName)
             return isExist ?: parent.createDirectory(decodedDirectoryName)
@@ -77,13 +78,12 @@ object FileUtil {
 
     /**
      * Open content uri and return file descriptor to JNI.
-     * @param context Application context
      * @param path Native content uri path
      * @param openMode will be one of "r", "r", "rw", "wa", "rwa"
      * @return file descriptor
      */
     @JvmStatic
-    fun openContentUri(context: Context, path: String, openMode: String?): Int {
+    fun openContentUri(path: String, openMode: String?): Int {
         try {
             val uri = Uri.parse(path)
             val parcelFileDescriptor = context.contentResolver.openFileDescriptor(uri, openMode!!)
@@ -103,11 +103,10 @@ object FileUtil {
     /**
      * Reference:  https://stackoverflow.com/questions/42186820/documentfile-is-very-slow
      * This function will be faster than DoucmentFile.listFiles
-     * @param context Application context
      * @param uri Directory uri.
      * @return CheapDocument lists.
      */
-    fun listFiles(context: Context, uri: Uri): Array<MinimalDocumentFile> {
+    fun listFiles(uri: Uri): Array<MinimalDocumentFile> {
         val resolver = context.contentResolver
         val columns = arrayOf(
             DocumentsContract.Document.COLUMN_DOCUMENT_ID,
@@ -145,7 +144,7 @@ object FileUtil {
      * @param path Native content uri path
      * @return bool
      */
-    fun exists(context: Context, path: String?): Boolean {
+    fun exists(path: String?): Boolean {
         var c: Cursor? = null
         try {
             val mUri = Uri.parse(path)
@@ -165,7 +164,7 @@ object FileUtil {
      * @param path content uri path
      * @return bool
      */
-    fun isDirectory(context: Context, path: String): Boolean {
+    fun isDirectory(path: String): Boolean {
         val resolver = context.contentResolver
         val columns = arrayOf(
             DocumentsContract.Document.COLUMN_MIME_TYPE
@@ -210,10 +209,10 @@ object FileUtil {
         return filename
     }
 
-    fun getFilesName(context: Context, path: String): Array<String> {
+    fun getFilesName(path: String): Array<String> {
         val uri = Uri.parse(path)
         val files: MutableList<String> = ArrayList()
-        for (file in listFiles(context, uri)) {
+        for (file in listFiles(uri)) {
             files.add(file.filename)
         }
         return files.toTypedArray()
@@ -225,7 +224,7 @@ object FileUtil {
      * @return long file size
      */
     @JvmStatic
-    fun getFileSize(context: Context, path: String): Long {
+    fun getFileSize(path: String): Long {
         val resolver = context.contentResolver
         val columns = arrayOf(
             DocumentsContract.Document.COLUMN_SIZE
@@ -245,44 +244,38 @@ object FileUtil {
         return size
     }
 
+    /**
+     * Creates an input stream with a given [Uri] and copies its data to the given path. This will
+     * overwrite any pre-existing files.
+     *
+     * @param sourceUri The [Uri] to copy data from
+     * @param destinationParentPath Destination directory
+     * @param destinationFilename Optionally renames the file once copied
+     */
     fun copyUriToInternalStorage(
-        context: Context,
-        sourceUri: Uri?,
+        sourceUri: Uri,
         destinationParentPath: String,
-        destinationFilename: String
-    ): Boolean {
-        var input: InputStream? = null
-        var output: FileOutputStream? = null
+        destinationFilename: String = ""
+    ): File? =
         try {
-            input = context.contentResolver.openInputStream(sourceUri!!)
-            output = FileOutputStream("$destinationParentPath/$destinationFilename")
-            val buffer = ByteArray(1024)
-            var len: Int
-            while (input!!.read(buffer).also { len = it } != -1) {
-                output.write(buffer, 0, len)
-            }
-            output.flush()
-            return true
-        } catch (e: Exception) {
-            Log.error("[FileUtil]: Cannot copy file, error: " + e.message)
-        } finally {
-            if (input != null) {
-                try {
-                    input.close()
-                } catch (e: IOException) {
-                    Log.error("[FileUtil]: Cannot close input file, error: " + e.message)
-                }
+            val fileName =
+                if (destinationFilename == "") getFilename(sourceUri) else "/$destinationFilename"
+            val inputStream = context.contentResolver.openInputStream(sourceUri)!!
+
+            val destinationFile = File("$destinationParentPath$fileName")
+            if (destinationFile.exists()) {
+                destinationFile.delete()
             }
-            if (output != null) {
-                try {
-                    output.close()
-                } catch (e: IOException) {
-                    Log.error("[FileUtil]: Cannot close output file, error: " + e.message)
-                }
+
+            destinationFile.outputStream().use { fos ->
+                inputStream.use { it.copyTo(fos) }
             }
+            destinationFile
+        } catch (e: IOException) {
+            null
+        } catch (e: NullPointerException) {
+            null
         }
-        return false
-    }
 
     /**
      * Extracts the given zip file into the given directory.
@@ -368,4 +361,12 @@ object FileUtil {
         return fileName.substring(fileName.lastIndexOf(".") + 1)
             .lowercase()
     }
+
+    @Throws(IOException::class)
+    fun getStringFromFile(file: File): String =
+        String(file.readBytes(), StandardCharsets.UTF_8)
+
+    @Throws(IOException::class)
+    fun getStringFromInputStream(stream: InputStream): String =
+        String(stream.readBytes(), StandardCharsets.UTF_8)
 }
diff --git a/src/android/app/src/main/java/org/yuzu/yuzu_emu/utils/GameHelper.kt b/src/android/app/src/main/java/org/yuzu/yuzu_emu/utils/GameHelper.kt
index e0ee29c9b..9001ca9ab 100644
--- a/src/android/app/src/main/java/org/yuzu/yuzu_emu/utils/GameHelper.kt
+++ b/src/android/app/src/main/java/org/yuzu/yuzu_emu/utils/GameHelper.kt
@@ -30,7 +30,7 @@ object GameHelper {
         // Ensure keys are loaded so that ROM metadata can be decrypted.
         NativeLibrary.reloadKeys()
 
-        addGamesRecursive(games, FileUtil.listFiles(context, gamesUri), 3)
+        addGamesRecursive(games, FileUtil.listFiles(gamesUri), 3)
 
         // Cache list of games found on disk
         val serializedGames = mutableSetOf<String>()
@@ -58,7 +58,7 @@ object GameHelper {
             if (it.isDirectory) {
                 addGamesRecursive(
                     games,
-                    FileUtil.listFiles(YuzuApplication.appContext, it.uri),
+                    FileUtil.listFiles(it.uri),
                     depth - 1
                 )
             } else {
diff --git a/src/android/app/src/main/java/org/yuzu/yuzu_emu/utils/GpuDriverHelper.kt b/src/android/app/src/main/java/org/yuzu/yuzu_emu/utils/GpuDriverHelper.kt
index 1d4695a2a..f6882ce6c 100644
--- a/src/android/app/src/main/java/org/yuzu/yuzu_emu/utils/GpuDriverHelper.kt
+++ b/src/android/app/src/main/java/org/yuzu/yuzu_emu/utils/GpuDriverHelper.kt
@@ -3,64 +3,33 @@
 
 package org.yuzu.yuzu_emu.utils
 
-import android.content.Context
 import android.net.Uri
+import android.os.Build
 import java.io.BufferedInputStream
 import java.io.File
-import java.io.FileInputStream
-import java.io.FileOutputStream
 import java.io.IOException
-import java.util.zip.ZipInputStream
 import org.yuzu.yuzu_emu.NativeLibrary
-import org.yuzu.yuzu_emu.utils.FileUtil.copyUriToInternalStorage
+import org.yuzu.yuzu_emu.YuzuApplication
+import java.util.zip.ZipException
+import java.util.zip.ZipFile
 
 object GpuDriverHelper {
     private const val META_JSON_FILENAME = "meta.json"
-    private const val DRIVER_INTERNAL_FILENAME = "gpu_driver.zip"
     private var fileRedirectionPath: String? = null
-    private var driverInstallationPath: String? = null
+    var driverInstallationPath: String? = null
     private var hookLibPath: String? = null
 
-    @Throws(IOException::class)
-    private fun unzip(zipFilePath: String, destDir: String) {
-        val dir = File(destDir)
-
-        // Create output directory if it doesn't exist
-        if (!dir.exists()) dir.mkdirs()
-
-        // Unpack the files.
-        val inputStream = FileInputStream(zipFilePath)
-        val zis = ZipInputStream(BufferedInputStream(inputStream))
-        val buffer = ByteArray(1024)
-        var ze = zis.nextEntry
-        while (ze != null) {
-            val newFile = File(destDir, ze.name)
-            val canonicalPath = newFile.canonicalPath
-            if (!canonicalPath.startsWith(destDir + ze.name)) {
-                throw SecurityException("Zip file attempted path traversal! " + ze.name)
-            }
-
-            newFile.parentFile!!.mkdirs()
-            val fos = FileOutputStream(newFile)
-            var len: Int
-            while (zis.read(buffer).also { len = it } > 0) {
-                fos.write(buffer, 0, len)
-            }
-            fos.close()
-            zis.closeEntry()
-            ze = zis.nextEntry
-        }
-        zis.closeEntry()
-    }
+    val driverStoragePath get() = DirectoryInitialization.userDirectory!! + "/gpu_drivers/"
 
-    fun initializeDriverParameters(context: Context) {
+    fun initializeDriverParameters() {
         try {
             // Initialize the file redirection directory.
-            fileRedirectionPath =
-                context.getExternalFilesDir(null)!!.canonicalPath + "/gpu/vk_file_redirect/"
+            fileRedirectionPath = YuzuApplication.appContext
+                .getExternalFilesDir(null)!!.canonicalPath + "/gpu/vk_file_redirect/"
 
             // Initialize the driver installation directory.
-            driverInstallationPath = context.filesDir.canonicalPath + "/gpu_driver/"
+            driverInstallationPath = YuzuApplication.appContext
+                .filesDir.canonicalPath + "/gpu_driver/"
         } catch (e: IOException) {
             throw RuntimeException(e)
         }
@@ -69,68 +38,169 @@ object GpuDriverHelper {
         initializeDirectories()
 
         // Initialize hook libraries directory.
-        hookLibPath = context.applicationInfo.nativeLibraryDir + "/"
+        hookLibPath = YuzuApplication.appContext.applicationInfo.nativeLibraryDir + "/"
 
         // Initialize GPU driver.
         NativeLibrary.initializeGpuDriver(
             hookLibPath,
             driverInstallationPath,
-            customDriverLibraryName,
+            customDriverData.libraryName,
             fileRedirectionPath
         )
     }
 
-    fun installDefaultDriver(context: Context) {
+    fun getDrivers(): MutableList<Pair<String, GpuDriverMetadata>> {
+        val driverZips = File(driverStoragePath).listFiles()
+        val drivers: MutableList<Pair<String, GpuDriverMetadata>> =
+            driverZips
+                ?.mapNotNull {
+                    val metadata = getMetadataFromZip(it)
+                    metadata.name?.let { _ -> Pair(it.path, metadata) }
+                }
+                ?.sortedByDescending { it: Pair<String, GpuDriverMetadata> -> it.second.name }
+                ?.distinct()
+                ?.toMutableList() ?: mutableListOf()
+
+        // TODO: Get system driver information
+        drivers.add(0, Pair("", GpuDriverMetadata()))
+        return drivers
+    }
+
+    fun installDefaultDriver() {
         // Removing the installed driver will result in the backend using the default system driver.
-        val driverInstallationDir = File(driverInstallationPath!!)
-        deleteRecursive(driverInstallationDir)
-        initializeDriverParameters(context)
+        File(driverInstallationPath!!).deleteRecursively()
+        initializeDriverParameters()
+    }
+
+    fun copyDriverToInternalStorage(driverUri: Uri): Boolean {
+        // Ensure we have directories.
+        initializeDirectories()
+
+        // Copy the zip file URI to user data
+        val copiedFile =
+            FileUtil.copyUriToInternalStorage(driverUri, driverStoragePath) ?: return false
+
+        // Validate driver
+        val metadata = getMetadataFromZip(copiedFile)
+        if (metadata.name == null) {
+            copiedFile.delete()
+            return false
+        }
+
+        if (metadata.minApi > Build.VERSION.SDK_INT) {
+            copiedFile.delete()
+            return false
+        }
+        return true
     }
 
-    fun installCustomDriver(context: Context, driverPathUri: Uri?) {
+    /**
+     * Copies driver zip into user data directory so that it can be exported along with
+     * other user data and also unzipped into the installation directory
+     */
+    fun installCustomDriver(driverUri: Uri): Boolean {
         // Revert to system default in the event the specified driver is bad.
-        installDefaultDriver(context)
+        installDefaultDriver()
 
         // Ensure we have directories.
         initializeDirectories()
 
-        // Copy the zip file URI into our private storage.
-        copyUriToInternalStorage(
-            context,
-            driverPathUri,
-            driverInstallationPath!!,
-            DRIVER_INTERNAL_FILENAME
-        )
+        // Copy the zip file URI to user data
+        val copiedFile =
+            FileUtil.copyUriToInternalStorage(driverUri, driverStoragePath) ?: return false
+
+        // Validate driver
+        val metadata = getMetadataFromZip(copiedFile)
+        if (metadata.name == null) {
+            copiedFile.delete()
+            return false
+        }
+
+        if (metadata.minApi > Build.VERSION.SDK_INT) {
+            copiedFile.delete()
+            return false
+        }
 
         // Unzip the driver.
         try {
-            unzip(driverInstallationPath + DRIVER_INTERNAL_FILENAME, driverInstallationPath!!)
+            FileUtil.unzipToInternalStorage(
+                BufferedInputStream(copiedFile.inputStream()),
+                File(driverInstallationPath!!)
+            )
         } catch (e: SecurityException) {
-            return
+            return false
         }
 
         // Initialize the driver parameters.
-        initializeDriverParameters(context)
+        initializeDriverParameters()
+
+        return true
     }
 
-    external fun supportsCustomDriverLoading(): Boolean
+    /**
+     * Unzips driver into installation directory
+     */
+    fun installCustomDriver(driver: File): Boolean {
+        // Revert to system default in the event the specified driver is bad.
+        installDefaultDriver()
 
-    // Parse the custom driver metadata to retrieve the name.
-    val customDriverName: String?
-        get() {
-            val metadata = GpuDriverMetadata(driverInstallationPath + META_JSON_FILENAME)
-            return metadata.name
+        // Ensure we have directories.
+        initializeDirectories()
+
+        // Validate driver
+        val metadata = getMetadataFromZip(driver)
+        if (metadata.name == null) {
+            driver.delete()
+            return false
         }
 
-    // Parse the custom driver metadata to retrieve the library name.
-    private val customDriverLibraryName: String?
-        get() {
-            // Parse the custom driver metadata to retrieve the library name.
-            val metadata = GpuDriverMetadata(driverInstallationPath + META_JSON_FILENAME)
-            return metadata.libraryName
+        // Unzip the driver to the private installation directory
+        try {
+            FileUtil.unzipToInternalStorage(
+                BufferedInputStream(driver.inputStream()),
+                File(driverInstallationPath!!)
+            )
+        } catch (e: SecurityException) {
+            return false
         }
 
-    private fun initializeDirectories() {
+        // Initialize the driver parameters.
+        initializeDriverParameters()
+
+        return true
+    }
+
+    /**
+     * Takes in a zip file and reads the meta.json file for presentation to the UI
+     *
+     * @param driver Zip containing driver and meta.json file
+     * @return A non-null [GpuDriverMetadata] instance that may have null members
+     */
+    fun getMetadataFromZip(driver: File): GpuDriverMetadata {
+        try {
+            ZipFile(driver).use { zf ->
+                val entries = zf.entries()
+                while (entries.hasMoreElements()) {
+                    val entry = entries.nextElement()
+                    if (!entry.isDirectory && entry.name.lowercase().contains(".json")) {
+                        zf.getInputStream(entry).use {
+                            return GpuDriverMetadata(it, entry.size)
+                        }
+                    }
+                }
+            }
+        } catch (_: ZipException) {
+        }
+        return GpuDriverMetadata()
+    }
+
+    external fun supportsCustomDriverLoading(): Boolean
+
+    // Parse the custom driver metadata to retrieve the name.
+    val customDriverData: GpuDriverMetadata
+        get() = GpuDriverMetadata(File(driverInstallationPath + META_JSON_FILENAME))
+
+    fun initializeDirectories() {
         // Ensure the file redirection directory exists.
         val fileRedirectionDir = File(fileRedirectionPath!!)
         if (!fileRedirectionDir.exists()) {
@@ -141,14 +211,10 @@ object GpuDriverHelper {
         if (!driverInstallationDir.exists()) {
             driverInstallationDir.mkdirs()
         }
-    }
-
-    private fun deleteRecursive(fileOrDirectory: File) {
-        if (fileOrDirectory.isDirectory) {
-            for (child in fileOrDirectory.listFiles()!!) {
-                deleteRecursive(child)
-            }
+        // Ensure the driver storage directory exists
+        val driverStorageDirectory = File(driverStoragePath)
+        if (!driverStorageDirectory.exists()) {
+            driverStorageDirectory.mkdirs()
         }
-        fileOrDirectory.delete()
     }
 }
diff --git a/src/android/app/src/main/java/org/yuzu/yuzu_emu/utils/GpuDriverMetadata.kt b/src/android/app/src/main/java/org/yuzu/yuzu_emu/utils/GpuDriverMetadata.kt
index a4e64070a..511a4171a 100644
--- a/src/android/app/src/main/java/org/yuzu/yuzu_emu/utils/GpuDriverMetadata.kt
+++ b/src/android/app/src/main/java/org/yuzu/yuzu_emu/utils/GpuDriverMetadata.kt
@@ -4,29 +4,29 @@
 package org.yuzu.yuzu_emu.utils
 
 import java.io.IOException
-import java.nio.charset.StandardCharsets
-import java.nio.file.Files
-import java.nio.file.Paths
 import org.json.JSONException
 import org.json.JSONObject
+import java.io.File
+import java.io.InputStream
 
-class GpuDriverMetadata(metadataFilePath: String) {
-    var name: String? = null
-    var description: String? = null
-    var author: String? = null
-    var vendor: String? = null
-    var driverVersion: String? = null
-    var minApi = 0
-    var libraryName: String? = null
+class GpuDriverMetadata {
+    /**
+     * Tries to get driver metadata information from a meta.json [File]
+     *
+     * @param metadataFile meta.json file provided with a GPU driver
+     */
+    constructor(metadataFile: File) {
+        if (metadataFile.length() > MAX_META_SIZE_BYTES) {
+            return
+        }
 
-    init {
         try {
-            val json = JSONObject(getStringFromFile(metadataFilePath))
+            val json = JSONObject(FileUtil.getStringFromFile(metadataFile))
             name = json.getString("name")
             description = json.getString("description")
             author = json.getString("author")
             vendor = json.getString("vendor")
-            driverVersion = json.getString("driverVersion")
+            version = json.getString("driverVersion")
             minApi = json.getInt("minApi")
             libraryName = json.getString("libraryName")
         } catch (e: JSONException) {
@@ -36,12 +36,84 @@ class GpuDriverMetadata(metadataFilePath: String) {
         }
     }
 
-    companion object {
-        @Throws(IOException::class)
-        private fun getStringFromFile(filePath: String): String {
-            val path = Paths.get(filePath)
-            val bytes = Files.readAllBytes(path)
-            return String(bytes, StandardCharsets.UTF_8)
+    /**
+     * Tries to get driver metadata information from an input stream that's intended to be
+     * from a zip file
+     *
+     * @param metadataStream ZipEntry input stream
+     * @param size Size of the file in bytes
+     */
+    constructor(metadataStream: InputStream, size: Long) {
+        if (size > MAX_META_SIZE_BYTES) {
+            return
         }
+
+        try {
+            val json = JSONObject(FileUtil.getStringFromInputStream(metadataStream))
+            name = json.getString("name")
+            description = json.getString("description")
+            author = json.getString("author")
+            vendor = json.getString("vendor")
+            version = json.getString("driverVersion")
+            minApi = json.getInt("minApi")
+            libraryName = json.getString("libraryName")
+        } catch (e: JSONException) {
+            // JSON is malformed, ignore and treat as unsupported metadata.
+        } catch (e: IOException) {
+            // File is inaccessible, ignore and treat as unsupported metadata.
+        }
+    }
+
+    /**
+     * Creates an empty metadata instance
+     */
+    constructor()
+
+    override fun equals(other: Any?): Boolean {
+        if (other !is GpuDriverMetadata) {
+            return false
+        }
+
+        return other.name == name &&
+            other.description == description &&
+            other.author == author &&
+            other.vendor == vendor &&
+            other.version == version &&
+            other.minApi == minApi &&
+            other.libraryName == libraryName
+    }
+
+    override fun hashCode(): Int {
+        var result = name?.hashCode() ?: 0
+        result = 31 * result + (description?.hashCode() ?: 0)
+        result = 31 * result + (author?.hashCode() ?: 0)
+        result = 31 * result + (vendor?.hashCode() ?: 0)
+        result = 31 * result + (version?.hashCode() ?: 0)
+        result = 31 * result + minApi
+        result = 31 * result + (libraryName?.hashCode() ?: 0)
+        return result
+    }
+
+    override fun toString(): String =
+        """
+            Name - $name
+            Description - $description
+            Author - $author
+            Vendor - $vendor
+            Version - $version
+            Min API - $minApi
+            Library Name - $libraryName
+        """.trimMargin().trimIndent()
+
+    var name: String? = null
+    var description: String? = null
+    var author: String? = null
+    var vendor: String? = null
+    var version: String? = null
+    var minApi = 0
+    var libraryName: String? = null
+
+    companion object {
+        private const val MAX_META_SIZE_BYTES = 500000
     }
 }
diff --git a/src/android/app/src/main/jni/native.cpp b/src/android/app/src/main/jni/native.cpp
index 9cf71680c..598f4e8bf 100644
--- a/src/android/app/src/main/jni/native.cpp
+++ b/src/android/app/src/main/jni/native.cpp
@@ -218,7 +218,6 @@ public:
             return;
         }
         m_window->OnSurfaceChanged(m_native_window);
-        m_system.Renderer().NotifySurfaceChanged();
     }
 
     void ConfigureFilesystemProvider(const std::string& filepath) {
diff --git a/src/android/app/src/main/res/drawable/ic_build.xml b/src/android/app/src/main/res/drawable/ic_build.xml
new file mode 100644
index 000000000..91d52f1b8
--- /dev/null
+++ b/src/android/app/src/main/res/drawable/ic_build.xml
@@ -0,0 +1,9 @@
+<vector xmlns:android="http://schemas.android.com/apk/res/android"
+    android:width="24dp"
+    android:height="24dp"
+    android:viewportWidth="24"
+    android:viewportHeight="24">
+    <path
+        android:fillColor="?attr/colorControlNormal"
+        android:pathData="M22.7,19l-9.1,-9.1c0.9,-2.3 0.4,-5 -1.5,-6.9 -2,-2 -5,-2.4 -7.4,-1.3L9,6 6,9 1.6,4.7C0.4,7.1 0.9,10.1 2.9,12.1c1.9,1.9 4.6,2.4 6.9,1.5l9.1,9.1c0.4,0.4 1,0.4 1.4,0l2.3,-2.3c0.5,-0.4 0.5,-1.1 0.1,-1.4z" />
+</vector>
diff --git a/src/android/app/src/main/res/drawable/ic_delete.xml b/src/android/app/src/main/res/drawable/ic_delete.xml
new file mode 100644
index 000000000..d26a79711
--- /dev/null
+++ b/src/android/app/src/main/res/drawable/ic_delete.xml
@@ -0,0 +1,9 @@
+<vector xmlns:android="http://schemas.android.com/apk/res/android"
+    android:width="24dp"
+    android:height="24dp"
+    android:viewportWidth="24"
+    android:viewportHeight="24">
+    <path
+        android:fillColor="?attr/colorControlNormal"
+        android:pathData="M6,19c0,1.1 0.9,2 2,2h8c1.1,0 2,-0.9 2,-2V7H6v12zM19,4h-3.5l-1,-1h-5l-1,1H5v2h14V4z" />
+</vector>
diff --git a/src/android/app/src/main/res/layout/card_driver_option.xml b/src/android/app/src/main/res/layout/card_driver_option.xml
new file mode 100644
index 000000000..1dd9a6d7d
--- /dev/null
+++ b/src/android/app/src/main/res/layout/card_driver_option.xml
@@ -0,0 +1,89 @@
+<?xml version="1.0" encoding="utf-8"?>
+<com.google.android.material.card.MaterialCardView xmlns:android="http://schemas.android.com/apk/res/android"
+    xmlns:app="http://schemas.android.com/apk/res-auto"
+    xmlns:tools="http://schemas.android.com/tools"
+    style="?attr/materialCardViewOutlinedStyle"
+    android:layout_width="match_parent"
+    android:layout_height="wrap_content"
+    android:layout_marginHorizontal="16dp"
+    android:layout_marginVertical="12dp"
+    android:background="?attr/selectableItemBackground"
+    android:clickable="true"
+    android:focusable="true">
+
+    <LinearLayout
+        android:layout_width="match_parent"
+        android:layout_height="wrap_content"
+        android:orientation="horizontal"
+        android:layout_gravity="center"
+        android:padding="16dp">
+
+        <RadioButton
+            android:id="@+id/radio_button"
+            android:layout_width="wrap_content"
+            android:layout_height="wrap_content"
+            android:layout_gravity="center_vertical"
+            android:clickable="false"
+            android:checked="false" />
+
+        <LinearLayout
+            android:layout_width="0dp"
+            android:layout_height="wrap_content"
+            android:layout_weight="1"
+            android:orientation="vertical"
+            android:layout_gravity="center_vertical">
+
+            <com.google.android.material.textview.MaterialTextView
+                android:id="@+id/title"
+                style="@style/TextAppearance.Material3.TitleMedium"
+                android:layout_width="match_parent"
+                android:layout_height="wrap_content"
+                android:ellipsize="none"
+                android:marqueeRepeatLimit="marquee_forever"
+                android:requiresFadingEdge="horizontal"
+                android:singleLine="true"
+                android:textAlignment="viewStart"
+                tools:text="@string/select_gpu_driver_default" />
+
+            <com.google.android.material.textview.MaterialTextView
+                android:id="@+id/version"
+                style="@style/TextAppearance.Material3.BodyMedium"
+                android:layout_width="match_parent"
+                android:layout_height="wrap_content"
+                android:layout_marginTop="6dp"
+                android:ellipsize="none"
+                android:marqueeRepeatLimit="marquee_forever"
+                android:requiresFadingEdge="horizontal"
+                android:singleLine="true"
+                android:textAlignment="viewStart"
+                tools:text="@string/install_gpu_driver_description" />
+
+            <com.google.android.material.textview.MaterialTextView
+                android:id="@+id/description"
+                style="@style/TextAppearance.Material3.BodyMedium"
+                android:layout_width="match_parent"
+                android:layout_height="wrap_content"
+                android:layout_marginTop="6dp"
+                android:ellipsize="none"
+                android:marqueeRepeatLimit="marquee_forever"
+                android:requiresFadingEdge="horizontal"
+                android:singleLine="true"
+                android:textAlignment="viewStart"
+                tools:text="@string/install_gpu_driver_description" />
+
+        </LinearLayout>
+
+        <Button
+            android:id="@+id/button_delete"
+            style="@style/Widget.Material3.Button.IconButton"
+            android:layout_width="wrap_content"
+            android:layout_height="wrap_content"
+            android:layout_gravity="center_vertical"
+            android:contentDescription="@string/delete"
+            android:tooltipText="@string/delete"
+            app:icon="@drawable/ic_delete"
+            app:iconTint="?attr/colorControlNormal" />
+
+    </LinearLayout>
+
+</com.google.android.material.card.MaterialCardView>
diff --git a/src/android/app/src/main/res/layout/card_home_option.xml b/src/android/app/src/main/res/layout/card_home_option.xml
index f9f1d89fb..6e8a232f9 100644
--- a/src/android/app/src/main/res/layout/card_home_option.xml
+++ b/src/android/app/src/main/res/layout/card_home_option.xml
@@ -16,7 +16,8 @@
     <LinearLayout
         android:id="@+id/option_layout"
         android:layout_width="match_parent"
-        android:layout_height="wrap_content">
+        android:layout_height="wrap_content"
+        android:layout_gravity="center_vertical">
 
         <ImageView
             android:id="@+id/option_icon"
diff --git a/src/android/app/src/main/res/layout/fragment_driver_manager.xml b/src/android/app/src/main/res/layout/fragment_driver_manager.xml
new file mode 100644
index 000000000..6cea2d164
--- /dev/null
+++ b/src/android/app/src/main/res/layout/fragment_driver_manager.xml
@@ -0,0 +1,48 @@
+<?xml version="1.0" encoding="utf-8"?>
+<androidx.constraintlayout.widget.ConstraintLayout xmlns:android="http://schemas.android.com/apk/res/android"
+    xmlns:app="http://schemas.android.com/apk/res-auto"
+    android:id="@+id/coordinator_licenses"
+    android:layout_width="match_parent"
+    android:layout_height="match_parent"
+    android:background="?attr/colorSurface">
+
+    <androidx.coordinatorlayout.widget.CoordinatorLayout
+        android:layout_width="match_parent"
+        android:layout_height="match_parent">
+
+        <com.google.android.material.appbar.AppBarLayout
+            android:id="@+id/appbar_drivers"
+            android:layout_width="match_parent"
+            android:layout_height="wrap_content"
+            android:fitsSystemWindows="true"
+            app:liftOnScrollTargetViewId="@id/list_drivers">
+
+            <com.google.android.material.appbar.MaterialToolbar
+                android:id="@+id/toolbar_drivers"
+                android:layout_width="match_parent"
+                android:layout_height="?attr/actionBarSize"
+                app:navigationIcon="@drawable/ic_back"
+                app:title="@string/gpu_driver_manager" />
+
+        </com.google.android.material.appbar.AppBarLayout>
+
+        <androidx.recyclerview.widget.RecyclerView
+            android:id="@+id/list_drivers"
+            android:layout_width="match_parent"
+            android:layout_height="match_parent"
+            android:clipToPadding="false"
+            app:layout_behavior="@string/appbar_scrolling_view_behavior" />
+
+    </androidx.coordinatorlayout.widget.CoordinatorLayout>
+
+    <com.google.android.material.floatingactionbutton.ExtendedFloatingActionButton
+        android:id="@+id/button_install"
+        android:layout_width="wrap_content"
+        android:layout_height="wrap_content"
+        android:layout_gravity="bottom|end"
+        android:text="@string/install"
+        app:icon="@drawable/ic_add"
+        app:layout_constraintBottom_toBottomOf="parent"
+        app:layout_constraintEnd_toEndOf="parent" />
+
+</androidx.constraintlayout.widget.ConstraintLayout>
diff --git a/src/android/app/src/main/res/navigation/home_navigation.xml b/src/android/app/src/main/res/navigation/home_navigation.xml
index 2356b802b..82749359d 100644
--- a/src/android/app/src/main/res/navigation/home_navigation.xml
+++ b/src/android/app/src/main/res/navigation/home_navigation.xml
@@ -22,6 +22,9 @@
         <action
             android:id="@+id/action_homeSettingsFragment_to_installableFragment"
             app:destination="@id/installableFragment" />
+        <action
+            android:id="@+id/action_homeSettingsFragment_to_driverManagerFragment"
+            app:destination="@id/driverManagerFragment" />
     </fragment>
 
     <fragment
@@ -95,5 +98,9 @@
         android:id="@+id/installableFragment"
         android:name="org.yuzu.yuzu_emu.fragments.InstallableFragment"
         android:label="InstallableFragment" />
+    <fragment
+        android:id="@+id/driverManagerFragment"
+        android:name="org.yuzu.yuzu_emu.fragments.DriverManagerFragment"
+        android:label="DriverManagerFragment" />
 
 </navigation>
diff --git a/src/android/app/src/main/res/values-de/strings.xml b/src/android/app/src/main/res/values-de/strings.xml
index dd0f36392..72a47fbdb 100644
--- a/src/android/app/src/main/res/values-de/strings.xml
+++ b/src/android/app/src/main/res/values-de/strings.xml
@@ -168,9 +168,7 @@
     <string name="select_gpu_driver_title">Möchtest du deinen aktuellen GPU-Treiber ersetzen?</string>
     <string name="select_gpu_driver_install">Installieren</string>
     <string name="select_gpu_driver_default">Standard</string>
-    <string name="select_gpu_driver_install_success">%s wurde installiert</string>
     <string name="select_gpu_driver_use_default">Standard GPU-Treiber wird verwendet</string>
-    <string name="select_gpu_driver_error">Ungültiger Treiber ausgewählt, Standard-Treiber wird verwendet!</string>
     <string name="system_gpu_driver">System GPU-Treiber</string>
     <string name="installing_driver">Treiber wird installiert...</string>
 
diff --git a/src/android/app/src/main/res/values-es/strings.xml b/src/android/app/src/main/res/values-es/strings.xml
index d398f862f..e5bdd5889 100644
--- a/src/android/app/src/main/res/values-es/strings.xml
+++ b/src/android/app/src/main/res/values-es/strings.xml
@@ -171,9 +171,7 @@
     <string name="select_gpu_driver_title">¿Quiere reemplazar el driver de GPU actual?</string>
     <string name="select_gpu_driver_install">Instalar</string>
     <string name="select_gpu_driver_default">Predeterminado</string>
-    <string name="select_gpu_driver_install_success">Instalado %s</string>
     <string name="select_gpu_driver_use_default">Usando el driver de GPU por defecto </string>
-    <string name="select_gpu_driver_error">¡Driver no válido, utilizando el predeterminado del sistema!</string>
     <string name="system_gpu_driver">Driver GPU del sistema</string>
     <string name="installing_driver">Instalando driver...</string>
 
diff --git a/src/android/app/src/main/res/values-fr/strings.xml b/src/android/app/src/main/res/values-fr/strings.xml
index a7abd9077..1e02828aa 100644
--- a/src/android/app/src/main/res/values-fr/strings.xml
+++ b/src/android/app/src/main/res/values-fr/strings.xml
@@ -171,9 +171,7 @@
     <string name="select_gpu_driver_title">Souhaitez vous remplacer votre pilote actuel ?</string>
     <string name="select_gpu_driver_install">Installer</string>
     <string name="select_gpu_driver_default">Défaut</string>
-    <string name="select_gpu_driver_install_success">%s Installé</string>
     <string name="select_gpu_driver_use_default">Utilisation du pilote de GPU par défaut</string>
-    <string name="select_gpu_driver_error">Pilote non valide sélectionné, utilisation du paramètre par défaut du système !</string>
     <string name="system_gpu_driver">Pilote du GPU du système</string>
     <string name="installing_driver">Installation du pilote...</string>
 
diff --git a/src/android/app/src/main/res/values-it/strings.xml b/src/android/app/src/main/res/values-it/strings.xml
index b18161801..09c9345b0 100644
--- a/src/android/app/src/main/res/values-it/strings.xml
+++ b/src/android/app/src/main/res/values-it/strings.xml
@@ -171,9 +171,7 @@
     <string name="select_gpu_driver_title">Vuoi sostituire il driver della tua GPU attuale?</string>
     <string name="select_gpu_driver_install">Installa</string>
     <string name="select_gpu_driver_default">Predefinito</string>
-    <string name="select_gpu_driver_install_success">Installato%s</string>
     <string name="select_gpu_driver_use_default">Utilizza il driver predefinito della GPU.</string>
-    <string name="select_gpu_driver_error">Il driver selezionato è invalido, è in utilizzo quello predefinito di sistema!</string>
     <string name="system_gpu_driver">Driver GPU del sistema</string>
     <string name="installing_driver">Installando i driver...</string>
 
diff --git a/src/android/app/src/main/res/values-ja/strings.xml b/src/android/app/src/main/res/values-ja/strings.xml
index 88fa5a0bb..a0ea78bef 100644
--- a/src/android/app/src/main/res/values-ja/strings.xml
+++ b/src/android/app/src/main/res/values-ja/strings.xml
@@ -170,9 +170,7 @@
     <string name="select_gpu_driver_title">現在のGPUドライバーを置き換えますか？</string>
     <string name="select_gpu_driver_install">インストール</string>
     <string name="select_gpu_driver_default">デフォルト</string>
-    <string name="select_gpu_driver_install_success">%s をインストールしました</string>
     <string name="select_gpu_driver_use_default">デフォルトのGPUドライバーを使用します</string>
-    <string name="select_gpu_driver_error">選択されたドライバが無効なため、システムのデフォルトを使用します！</string>
     <string name="system_gpu_driver">システムのGPUドライバ</string>
     <string name="installing_driver">インストール中…</string>
 
diff --git a/src/android/app/src/main/res/values-ko/strings.xml b/src/android/app/src/main/res/values-ko/strings.xml
index 4b658255c..214f95706 100644
--- a/src/android/app/src/main/res/values-ko/strings.xml
+++ b/src/android/app/src/main/res/values-ko/strings.xml
@@ -171,9 +171,7 @@
     <string name="select_gpu_driver_title">현재 사용 중인 GPU 드라이버를 교체하겠습니까?</string>
     <string name="select_gpu_driver_install">설치</string>
     <string name="select_gpu_driver_default">기본값</string>
-    <string name="select_gpu_driver_install_success">설치된 %s</string>
     <string name="select_gpu_driver_use_default">기본 GPU 드라이버 사용</string>
-    <string name="select_gpu_driver_error">시스템 기본값을 사용하여 잘못된 드라이버를 선택했습니다!</string>
     <string name="system_gpu_driver">시스템 GPU 드라이버</string>
     <string name="installing_driver">드라이버 설치 중...</string>
 
diff --git a/src/android/app/src/main/res/values-nb/strings.xml b/src/android/app/src/main/res/values-nb/strings.xml
index dd602a389..5443cef42 100644
--- a/src/android/app/src/main/res/values-nb/strings.xml
+++ b/src/android/app/src/main/res/values-nb/strings.xml
@@ -171,9 +171,7 @@
     <string name="select_gpu_driver_title">Ønsker du å bytte ut din nåværende GPU-driver?</string>
     <string name="select_gpu_driver_install">Installer</string>
     <string name="select_gpu_driver_default">Standard</string>
-    <string name="select_gpu_driver_install_success">Installert %s</string>
     <string name="select_gpu_driver_use_default">Bruk av standard GPU-driver</string>
-    <string name="select_gpu_driver_error">Ugyldig driver valgt, bruker systemstandard!</string>
     <string name="system_gpu_driver">Systemets GPU-driver</string>
     <string name="installing_driver">Installerer driver...</string>
 
diff --git a/src/android/app/src/main/res/values-pl/strings.xml b/src/android/app/src/main/res/values-pl/strings.xml
index 2fdd1f952..899e233d0 100644
--- a/src/android/app/src/main/res/values-pl/strings.xml
+++ b/src/android/app/src/main/res/values-pl/strings.xml
@@ -171,9 +171,7 @@
     <string name="select_gpu_driver_title">Chcesz zastąpić obecny sterownik układu graficznego?</string>
     <string name="select_gpu_driver_install">Zainstaluj</string>
     <string name="select_gpu_driver_default">Domyślne</string>
-    <string name="select_gpu_driver_install_success">Zainstalowano %s</string>
     <string name="select_gpu_driver_use_default">Aktywny domyślny sterownik GPU</string>
-    <string name="select_gpu_driver_error">Wybrano błędny sterownik, powrót do domyślnego. </string>
     <string name="system_gpu_driver">Systemowy sterownik GPU</string>
     <string name="installing_driver">Instalowanie sterownika...</string>
 
diff --git a/src/android/app/src/main/res/values-pt-rBR/strings.xml b/src/android/app/src/main/res/values-pt-rBR/strings.xml
index 2f26367fe..caa095364 100644
--- a/src/android/app/src/main/res/values-pt-rBR/strings.xml
+++ b/src/android/app/src/main/res/values-pt-rBR/strings.xml
@@ -171,9 +171,7 @@
     <string name="select_gpu_driver_title">Queres substituir o driver do GPU atual? </string>
     <string name="select_gpu_driver_install">Instalar</string>
     <string name="select_gpu_driver_default">Padrão</string>
-    <string name="select_gpu_driver_install_success">Instalado%s</string>
     <string name="select_gpu_driver_use_default">Usar o driver padrão do GPU</string>
-    <string name="select_gpu_driver_error">Driver selecionado inválido, a usar o padrão do sistema!</string>
     <string name="system_gpu_driver">Driver do GPU padrão</string>
     <string name="installing_driver">A instalar o Driver...</string>
 
diff --git a/src/android/app/src/main/res/values-pt-rPT/strings.xml b/src/android/app/src/main/res/values-pt-rPT/strings.xml
index 4e1eb4cd7..0a1a47fbb 100644
--- a/src/android/app/src/main/res/values-pt-rPT/strings.xml
+++ b/src/android/app/src/main/res/values-pt-rPT/strings.xml
@@ -171,9 +171,7 @@
     <string name="select_gpu_driver_title">Queres substituir o driver do GPU atual? </string>
     <string name="select_gpu_driver_install">Instalar</string>
     <string name="select_gpu_driver_default">Padrão</string>
-    <string name="select_gpu_driver_install_success">Instalado%s</string>
     <string name="select_gpu_driver_use_default">Usar o driver padrão do GPU</string>
-    <string name="select_gpu_driver_error">Driver selecionado inválido, a usar o padrão do sistema!</string>
     <string name="system_gpu_driver">Driver do GPU padrão</string>
     <string name="installing_driver">A instalar o Driver...</string>
 
diff --git a/src/android/app/src/main/res/values-ru/strings.xml b/src/android/app/src/main/res/values-ru/strings.xml
index f5695dc93..0bef035d6 100644
--- a/src/android/app/src/main/res/values-ru/strings.xml
+++ b/src/android/app/src/main/res/values-ru/strings.xml
@@ -171,9 +171,7 @@
     <string name="select_gpu_driver_title">Хотите заменить текущий драйвер ГП?</string>
     <string name="select_gpu_driver_install">Установить</string>
     <string name="select_gpu_driver_default">По умолчанию</string>
-    <string name="select_gpu_driver_install_success">Установлено %s</string>
     <string name="select_gpu_driver_use_default">Используется стандартный драйвер ГП </string>
-    <string name="select_gpu_driver_error">Выбран неверный драйвер, используется стандартный системный!</string>
     <string name="system_gpu_driver">Системный драйвер ГП</string>
     <string name="installing_driver">Установка драйвера...</string>
 
diff --git a/src/android/app/src/main/res/values-uk/strings.xml b/src/android/app/src/main/res/values-uk/strings.xml
index 061bc6f04..5b789ee98 100644
--- a/src/android/app/src/main/res/values-uk/strings.xml
+++ b/src/android/app/src/main/res/values-uk/strings.xml
@@ -171,9 +171,7 @@
     <string name="select_gpu_driver_title">Хочете замінити поточний драйвер ГП?</string>
     <string name="select_gpu_driver_install">Встановити</string>
     <string name="select_gpu_driver_default">За замовчуванням</string>
-    <string name="select_gpu_driver_install_success">Встановлено %s</string>
     <string name="select_gpu_driver_use_default">Використовується стандартний драйвер ГП</string>
-    <string name="select_gpu_driver_error">Обрано неправильний драйвер, використовується стандартний системний!</string>
     <string name="system_gpu_driver">Системний драйвер ГП</string>
     <string name="installing_driver">Встановлення драйвера...</string>
 
diff --git a/src/android/app/src/main/res/values-zh-rCN/strings.xml b/src/android/app/src/main/res/values-zh-rCN/strings.xml
index fe6dd5eaa..c0e885751 100644
--- a/src/android/app/src/main/res/values-zh-rCN/strings.xml
+++ b/src/android/app/src/main/res/values-zh-rCN/strings.xml
@@ -171,9 +171,7 @@
     <string name="select_gpu_driver_title">要取代您当前的 GPU 驱动程序吗？</string>
     <string name="select_gpu_driver_install">安装</string>
     <string name="select_gpu_driver_default">系统默认</string>
-    <string name="select_gpu_driver_install_success">已安装 %s</string>
     <string name="select_gpu_driver_use_default">使用默认 GPU 驱动程序</string>
-    <string name="select_gpu_driver_error">选择的驱动程序无效，将使用系统默认的驱动程序！</string>
     <string name="system_gpu_driver">系统 GPU 驱动程序</string>
     <string name="installing_driver">正在安装驱动程序…</string>
 
diff --git a/src/android/app/src/main/res/values-zh-rTW/strings.xml b/src/android/app/src/main/res/values-zh-rTW/strings.xml
index 9b3e54224..4a21bf893 100644
--- a/src/android/app/src/main/res/values-zh-rTW/strings.xml
+++ b/src/android/app/src/main/res/values-zh-rTW/strings.xml
@@ -171,9 +171,7 @@
     <string name="select_gpu_driver_title">要取代您目前的 GPU 驅動程式嗎？</string>
     <string name="select_gpu_driver_install">安裝</string>
     <string name="select_gpu_driver_default">預設</string>
-    <string name="select_gpu_driver_install_success">已安裝 %s</string>
     <string name="select_gpu_driver_use_default">使用預設 GPU 驅動程式</string>
-    <string name="select_gpu_driver_error">選取的驅動程式無效，將使用系統預設驅動程式！</string>
     <string name="system_gpu_driver">系統 GPU 驅動程式</string>
     <string name="installing_driver">正在安裝驅動程式…</string>
 
diff --git a/src/android/app/src/main/res/values/dimens.xml b/src/android/app/src/main/res/values/dimens.xml
index 7b2296d95..ef855ea6f 100644
--- a/src/android/app/src/main/res/values/dimens.xml
+++ b/src/android/app/src/main/res/values/dimens.xml
@@ -13,6 +13,8 @@
     <dimen name="menu_width">256dp</dimen>
     <dimen name="card_width">165dp</dimen>
     <dimen name="icon_inset">24dp</dimen>
+    <dimen name="spacing_bottom_list_fab">72dp</dimen>
+    <dimen name="spacing_fab">24dp</dimen>
 
     <dimen name="dialog_margin">20dp</dimen>
     <dimen name="elevated_app_bar">3dp</dimen>
diff --git a/src/android/app/src/main/res/values/strings.xml b/src/android/app/src/main/res/values/strings.xml
index e51edf872..9e4854221 100644
--- a/src/android/app/src/main/res/values/strings.xml
+++ b/src/android/app/src/main/res/values/strings.xml
@@ -72,6 +72,7 @@
     <string name="invalid_keys_error">Invalid encryption keys</string>
     <string name="dumping_keys_quickstart_link">https://yuzu-emu.org/help/quickstart/#dumping-decryption-keys</string>
     <string name="install_keys_failure_description">The selected file is incorrect or corrupt. Please redump your keys.</string>
+    <string name="gpu_driver_manager">GPU Driver Manager</string>
     <string name="install_gpu_driver">Install GPU driver</string>
     <string name="install_gpu_driver_description">Install alternative drivers for potentially better performance or accuracy</string>
     <string name="advanced_settings">Advanced settings</string>
@@ -234,15 +235,17 @@
     <string name="export_failed">Export failed</string>
     <string name="import_failed">Import failed</string>
     <string name="cancelling">Cancelling</string>
+    <string name="install">Install</string>
+    <string name="delete">Delete</string>
 
     <!-- GPU driver installation -->
     <string name="select_gpu_driver">Select GPU driver</string>
     <string name="select_gpu_driver_title">Would you like to replace your current GPU driver?</string>
     <string name="select_gpu_driver_install">Install</string>
     <string name="select_gpu_driver_default">Default</string>
-    <string name="select_gpu_driver_install_success">Installed %s</string>
     <string name="select_gpu_driver_use_default">Using default GPU driver</string>
-    <string name="select_gpu_driver_error">Invalid driver selected, using system default!</string>
+    <string name="select_gpu_driver_error">Invalid driver selected</string>
+    <string name="driver_already_installed">Driver already installed</string>
     <string name="system_gpu_driver">System GPU driver</string>
     <string name="installing_driver">Installing driver…</string>
 
diff --git a/src/android/build.gradle.kts b/src/android/build.gradle.kts
index 80f370c16..51e559321 100644
--- a/src/android/build.gradle.kts
+++ b/src/android/build.gradle.kts
@@ -3,8 +3,8 @@
 
 // Top-level build file where you can add configuration options common to all sub-projects/modules.
 plugins {
-    id("com.android.application") version "8.0.2" apply false
-    id("com.android.library") version "8.0.2" apply false
+    id("com.android.application") version "8.1.2" apply false
+    id("com.android.library") version "8.1.2" apply false
     id("org.jetbrains.kotlin.android") version "1.8.21" apply false
 }
 
diff --git a/src/audio_core/adsp/apps/audio_renderer/audio_renderer.cpp b/src/audio_core/adsp/apps/audio_renderer/audio_renderer.cpp
index 972d5e45b..ef301d8b4 100644
--- a/src/audio_core/adsp/apps/audio_renderer/audio_renderer.cpp
+++ b/src/audio_core/adsp/apps/audio_renderer/audio_renderer.cpp
@@ -77,6 +77,7 @@ void AudioRenderer::Wait() {
                   "{}, got {}",
                   Message::RenderResponse, msg);
     }
+    PostDSPClearCommandBuffer();
 }
 
 void AudioRenderer::Send(Direction dir, u32 message) {
@@ -96,6 +97,14 @@ void AudioRenderer::SetCommandBuffer(s32 session_id, CpuAddr buffer, u64 size, u
     command_buffers[session_id].reset_buffer = reset;
 }
 
+void AudioRenderer::PostDSPClearCommandBuffer() noexcept {
+    for (auto& buffer : command_buffers) {
+        buffer.buffer = 0;
+        buffer.size = 0;
+        buffer.reset_buffer = false;
+    }
+}
+
 u32 AudioRenderer::GetRemainCommandCount(s32 session_id) const noexcept {
     return command_buffers[session_id].remaining_command_count;
 }
diff --git a/src/audio_core/adsp/apps/audio_renderer/audio_renderer.h b/src/audio_core/adsp/apps/audio_renderer/audio_renderer.h
index 85874d88a..57b89d9fe 100644
--- a/src/audio_core/adsp/apps/audio_renderer/audio_renderer.h
+++ b/src/audio_core/adsp/apps/audio_renderer/audio_renderer.h
@@ -85,6 +85,8 @@ private:
      */
     void CreateSinkStreams();
 
+    void PostDSPClearCommandBuffer() noexcept;
+
     /// Core system
     Core::System& system;
     /// The output sink the AudioRenderer will send samples to
diff --git a/src/audio_core/sink/sink_stream.cpp b/src/audio_core/sink/sink_stream.cpp
index 6081352a2..d66d04fae 100644
--- a/src/audio_core/sink/sink_stream.cpp
+++ b/src/audio_core/sink/sink_stream.cpp
@@ -204,6 +204,10 @@ void SinkStream::ProcessAudioOutAndRender(std::span<s16> output_buffer, std::siz
     // paused and we'll desync, so just play silence.
     if (system.IsPaused() || system.IsShuttingDown()) {
         if (system.IsShuttingDown()) {
+            {
+                std::scoped_lock lk{release_mutex};
+                queued_buffers.store(0);
+            }
             release_cv.notify_one();
         }
 
diff --git a/src/common/CMakeLists.txt b/src/common/CMakeLists.txt
index 416203c59..e216eb3de 100644
--- a/src/common/CMakeLists.txt
+++ b/src/common/CMakeLists.txt
@@ -120,6 +120,8 @@ add_library(common STATIC
     socket_types.h
     spin_lock.cpp
     spin_lock.h
+    stb.cpp
+    stb.h
     steady_clock.cpp
     steady_clock.h
     stream.cpp
@@ -189,6 +191,14 @@ if(ARCHITECTURE_x86_64)
     target_link_libraries(common PRIVATE xbyak::xbyak)
 endif()
 
+if (ARCHITECTURE_arm64 AND (ANDROID OR LINUX))
+    target_sources(common
+        PRIVATE
+            arm64/native_clock.cpp
+            arm64/native_clock.h
+    )
+endif()
+
 if (MSVC)
   target_compile_definitions(common PRIVATE
     # The standard library doesn't provide any replacement for codecvt yet
@@ -200,6 +210,8 @@ if (MSVC)
     /we4254 # 'operator': conversion from 'type1:field_bits' to 'type2:field_bits', possible loss of data
     /we4800 # Implicit conversion from 'type' to bool. Possible information loss
   )
+else()
+  set_source_files_properties(stb.cpp PROPERTIES COMPILE_OPTIONS "-Wno-implicit-fallthrough;-Wno-missing-declarations;-Wno-missing-field-initializers")
 endif()
 
 if (CMAKE_CXX_COMPILER_ID STREQUAL "Clang")
@@ -215,7 +227,7 @@ endif()
 
 create_target_directory_groups(common)
 
-target_link_libraries(common PUBLIC Boost::context Boost::headers fmt::fmt microprofile Threads::Threads)
+target_link_libraries(common PUBLIC Boost::context Boost::headers fmt::fmt microprofile stb::headers Threads::Threads)
 target_link_libraries(common PRIVATE lz4::lz4 zstd::zstd LLVM::Demangle)
 
 if (ANDROID)
diff --git a/src/common/arm64/native_clock.cpp b/src/common/arm64/native_clock.cpp
new file mode 100644
index 000000000..88fdba527
--- /dev/null
+++ b/src/common/arm64/native_clock.cpp
@@ -0,0 +1,72 @@
+// SPDX-FileCopyrightText: Copyright 2023 yuzu Emulator Project
+// SPDX-License-Identifier: GPL-2.0-or-later
+
+#include "common/arm64/native_clock.h"
+
+namespace Common::Arm64 {
+
+namespace {
+
+NativeClock::FactorType GetFixedPointFactor(u64 num, u64 den) {
+    return (static_cast<NativeClock::FactorType>(num) << 64) / den;
+}
+
+u64 MultiplyHigh(u64 m, NativeClock::FactorType factor) {
+    return static_cast<u64>((m * factor) >> 64);
+}
+
+} // namespace
+
+NativeClock::NativeClock() {
+    const u64 host_cntfrq = GetHostCNTFRQ();
+    ns_cntfrq_factor = GetFixedPointFactor(NsRatio::den, host_cntfrq);
+    us_cntfrq_factor = GetFixedPointFactor(UsRatio::den, host_cntfrq);
+    ms_cntfrq_factor = GetFixedPointFactor(MsRatio::den, host_cntfrq);
+    guest_cntfrq_factor = GetFixedPointFactor(CNTFRQ, host_cntfrq);
+    gputick_cntfrq_factor = GetFixedPointFactor(GPUTickFreq, host_cntfrq);
+}
+
+std::chrono::nanoseconds NativeClock::GetTimeNS() const {
+    return std::chrono::nanoseconds{MultiplyHigh(GetHostTicksElapsed(), ns_cntfrq_factor)};
+}
+
+std::chrono::microseconds NativeClock::GetTimeUS() const {
+    return std::chrono::microseconds{MultiplyHigh(GetHostTicksElapsed(), us_cntfrq_factor)};
+}
+
+std::chrono::milliseconds NativeClock::GetTimeMS() const {
+    return std::chrono::milliseconds{MultiplyHigh(GetHostTicksElapsed(), ms_cntfrq_factor)};
+}
+
+u64 NativeClock::GetCNTPCT() const {
+    return MultiplyHigh(GetHostTicksElapsed(), guest_cntfrq_factor);
+}
+
+u64 NativeClock::GetGPUTick() const {
+    return MultiplyHigh(GetHostTicksElapsed(), gputick_cntfrq_factor);
+}
+
+u64 NativeClock::GetHostTicksNow() const {
+    u64 cntvct_el0 = 0;
+    asm volatile("dsb ish\n\t"
+                 "mrs %[cntvct_el0], cntvct_el0\n\t"
+                 "dsb ish\n\t"
+                 : [cntvct_el0] "=r"(cntvct_el0));
+    return cntvct_el0;
+}
+
+u64 NativeClock::GetHostTicksElapsed() const {
+    return GetHostTicksNow();
+}
+
+bool NativeClock::IsNative() const {
+    return true;
+}
+
+u64 NativeClock::GetHostCNTFRQ() {
+    u64 cntfrq_el0 = 0;
+    asm("mrs %[cntfrq_el0], cntfrq_el0" : [cntfrq_el0] "=r"(cntfrq_el0));
+    return cntfrq_el0;
+}
+
+} // namespace Common::Arm64
diff --git a/src/common/arm64/native_clock.h b/src/common/arm64/native_clock.h
new file mode 100644
index 000000000..a28b419f2
--- /dev/null
+++ b/src/common/arm64/native_clock.h
@@ -0,0 +1,47 @@
+// SPDX-FileCopyrightText: Copyright 2023 yuzu Emulator Project
+// SPDX-License-Identifier: GPL-2.0-or-later
+
+#pragma once
+
+#include "common/wall_clock.h"
+
+namespace Common::Arm64 {
+
+class NativeClock final : public WallClock {
+public:
+    explicit NativeClock();
+
+    std::chrono::nanoseconds GetTimeNS() const override;
+
+    std::chrono::microseconds GetTimeUS() const override;
+
+    std::chrono::milliseconds GetTimeMS() const override;
+
+    u64 GetCNTPCT() const override;
+
+    u64 GetGPUTick() const override;
+
+    u64 GetHostTicksNow() const override;
+
+    u64 GetHostTicksElapsed() const override;
+
+    bool IsNative() const override;
+
+    static u64 GetHostCNTFRQ();
+
+public:
+    using FactorType = unsigned __int128;
+
+    FactorType GetGuestCNTFRQFactor() const {
+        return guest_cntfrq_factor;
+    }
+
+private:
+    FactorType ns_cntfrq_factor;
+    FactorType us_cntfrq_factor;
+    FactorType ms_cntfrq_factor;
+    FactorType guest_cntfrq_factor;
+    FactorType gputick_cntfrq_factor;
+};
+
+} // namespace Common::Arm64
diff --git a/src/common/common_funcs.h b/src/common/common_funcs.h
index 0dad9338a..47d028d48 100644
--- a/src/common/common_funcs.h
+++ b/src/common/common_funcs.h
@@ -39,8 +39,12 @@
 #define Crash() exit(1)
 #endif
 
+#define LTO_NOINLINE __attribute__((noinline))
+
 #else // _MSC_VER
 
+#define LTO_NOINLINE
+
 // Locale Cross-Compatibility
 #define locale_t _locale_t
 
diff --git a/src/common/elf.h b/src/common/elf.h
index 14a5e9597..0b728dc54 100644
--- a/src/common/elf.h
+++ b/src/common/elf.h
@@ -211,6 +211,11 @@ struct Elf64_Rela {
     Elf64_Sxword r_addend; /* Addend */
 };
 
+/* RELR relocation table entry */
+
+using Elf32_Relr = Elf32_Word;
+using Elf64_Relr = Elf64_Xword;
+
 /* How to extract and insert information held in the r_info field.  */
 
 static inline u32 Elf32RelSymIndex(Elf32_Word r_info) {
@@ -328,6 +333,9 @@ constexpr u32 ElfDtFiniArray = 26;   /* Array with addresses of fini fct */
 constexpr u32 ElfDtInitArraySz = 27; /* Size in bytes of DT_INIT_ARRAY */
 constexpr u32 ElfDtFiniArraySz = 28; /* Size in bytes of DT_FINI_ARRAY */
 constexpr u32 ElfDtSymtabShndx = 34; /* Address of SYMTAB_SHNDX section */
+constexpr u32 ElfDtRelrsz = 35;      /* Size of RELR relative relocations */
+constexpr u32 ElfDtRelr = 36;        /* Address of RELR relative relocations */
+constexpr u32 ElfDtRelrent = 37;     /* Size of one RELR relative relocation */
 
 } // namespace ELF
 } // namespace Common
diff --git a/src/common/fs/fs_paths.h b/src/common/fs/fs_paths.h
index 1a3f6ab45..bcf447089 100644
--- a/src/common/fs/fs_paths.h
+++ b/src/common/fs/fs_paths.h
@@ -19,10 +19,12 @@
 #define LOAD_DIR "load"
 #define LOG_DIR "log"
 #define NAND_DIR "nand"
+#define PLAY_TIME_DIR "play_time"
 #define SCREENSHOTS_DIR "screenshots"
 #define SDMC_DIR "sdmc"
 #define SHADER_DIR "shader"
 #define TAS_DIR "tas"
+#define ICONS_DIR "icons"
 
 // yuzu-specific files
 
diff --git a/src/common/fs/path_util.cpp b/src/common/fs/path_util.cpp
index fbac4d80c..0c4c88cde 100644
--- a/src/common/fs/path_util.cpp
+++ b/src/common/fs/path_util.cpp
@@ -125,10 +125,12 @@ public:
         GenerateYuzuPath(YuzuPath::LoadDir, yuzu_path / LOAD_DIR);
         GenerateYuzuPath(YuzuPath::LogDir, yuzu_path / LOG_DIR);
         GenerateYuzuPath(YuzuPath::NANDDir, yuzu_path / NAND_DIR);
+        GenerateYuzuPath(YuzuPath::PlayTimeDir, yuzu_path / PLAY_TIME_DIR);
         GenerateYuzuPath(YuzuPath::ScreenshotsDir, yuzu_path / SCREENSHOTS_DIR);
         GenerateYuzuPath(YuzuPath::SDMCDir, yuzu_path / SDMC_DIR);
         GenerateYuzuPath(YuzuPath::ShaderDir, yuzu_path / SHADER_DIR);
         GenerateYuzuPath(YuzuPath::TASDir, yuzu_path / TAS_DIR);
+        GenerateYuzuPath(YuzuPath::IconsDir, yuzu_path / ICONS_DIR);
     }
 
 private:
diff --git a/src/common/fs/path_util.h b/src/common/fs/path_util.h
index 036e475aa..2874ea738 100644
--- a/src/common/fs/path_util.h
+++ b/src/common/fs/path_util.h
@@ -21,10 +21,12 @@ enum class YuzuPath {
     LoadDir,        // Where cheat/mod files are stored.
     LogDir,         // Where log files are stored.
     NANDDir,        // Where the emulated NAND is stored.
+    PlayTimeDir,    // Where play time data is stored.
     ScreenshotsDir, // Where yuzu screenshots are stored.
     SDMCDir,        // Where the emulated SDMC is stored.
     ShaderDir,      // Where shaders are stored.
     TASDir,         // Where TAS scripts are stored.
+    IconsDir,       // Where Icons for Windows shortcuts are stored.
 };
 
 /**
diff --git a/src/common/nvidia_flags.cpp b/src/common/nvidia_flags.cpp
index 7ed7690ee..fa3747782 100644
--- a/src/common/nvidia_flags.cpp
+++ b/src/common/nvidia_flags.cpp
@@ -25,6 +25,7 @@ void ConfigureNvidiaEnvironmentFlags() {
 
     void(_putenv(fmt::format("__GL_SHADER_DISK_CACHE_PATH={}", windows_path_string).c_str()));
     void(_putenv("__GL_SHADER_DISK_CACHE_SKIP_CLEANUP=1"));
+    void(_putenv("__GL_THREADED_OPTIMIZATIONS=1"));
 #endif
 }
 
diff --git a/src/common/polyfill_thread.h b/src/common/polyfill_thread.h
index 41cbb9ed5..12e59a893 100644
--- a/src/common/polyfill_thread.h
+++ b/src/common/polyfill_thread.h
@@ -15,12 +15,13 @@
 #include <condition_variable>
 #include <stop_token>
 #include <thread>
+#include <utility>
 
 namespace Common {
 
 template <typename Condvar, typename Lock, typename Pred>
 void CondvarWait(Condvar& cv, std::unique_lock<Lock>& lk, std::stop_token token, Pred&& pred) {
-    cv.wait(lk, token, std::move(pred));
+    cv.wait(lk, token, std::forward<Pred>(pred));
 }
 
 template <typename Rep, typename Period>
@@ -109,7 +110,7 @@ public:
 
         // Insert the callback.
         stop_state_callback ret = ++m_next_callback;
-        m_callbacks.emplace(ret, move(f));
+        m_callbacks.emplace(ret, std::move(f));
         return ret;
     }
 
@@ -162,7 +163,7 @@ private:
     friend class stop_source;
     template <typename Callback>
     friend class stop_callback;
-    stop_token(shared_ptr<polyfill::stop_state> stop_state) : m_stop_state(move(stop_state)) {}
+    stop_token(shared_ptr<polyfill::stop_state> stop_state) : m_stop_state(std::move(stop_state)) {}
 
 private:
     shared_ptr<polyfill::stop_state> m_stop_state;
@@ -198,7 +199,7 @@ public:
 private:
     friend class jthread;
     explicit stop_source(shared_ptr<polyfill::stop_state> stop_state)
-        : m_stop_state(move(stop_state)) {}
+        : m_stop_state(std::move(stop_state)) {}
 
 private:
     shared_ptr<polyfill::stop_state> m_stop_state;
@@ -218,16 +219,16 @@ public:
                            C&& cb) noexcept(is_nothrow_constructible_v<Callback, C>)
         : m_stop_state(st.m_stop_state) {
         if (m_stop_state) {
-            m_callback = m_stop_state->insert_callback(move(cb));
+            m_callback = m_stop_state->insert_callback(std::move(cb));
         }
     }
     template <typename C>
         requires constructible_from<Callback, C>
     explicit stop_callback(stop_token&& st,
                            C&& cb) noexcept(is_nothrow_constructible_v<Callback, C>)
-        : m_stop_state(move(st.m_stop_state)) {
+        : m_stop_state(std::move(st.m_stop_state)) {
         if (m_stop_state) {
-            m_callback = m_stop_state->insert_callback(move(cb));
+            m_callback = m_stop_state->insert_callback(std::move(cb));
         }
     }
     ~stop_callback() {
@@ -260,7 +261,7 @@ public:
               typename = enable_if_t<!is_same_v<remove_cvref_t<F>, jthread>>>
     explicit jthread(F&& f, Args&&... args)
         : m_stop_state(make_shared<polyfill::stop_state>()),
-          m_thread(make_thread(move(f), move(args)...)) {}
+          m_thread(make_thread(std::forward<F>(f), std::forward<Args>(args)...)) {}
 
     ~jthread() {
         if (joinable()) {
@@ -317,9 +318,9 @@ private:
     template <typename F, typename... Args>
     thread make_thread(F&& f, Args&&... args) {
         if constexpr (is_invocable_v<decay_t<F>, stop_token, decay_t<Args>...>) {
-            return thread(move(f), get_stop_token(), move(args)...);
+            return thread(std::forward<F>(f), get_stop_token(), std::forward<Args>(args)...);
         } else {
-            return thread(move(f), move(args)...);
+            return thread(std::forward<F>(f), std::forward<Args>(args)...);
         }
     }
 
diff --git a/src/common/settings.cpp b/src/common/settings.cpp
index 3fde3cae6..98b43e49c 100644
--- a/src/common/settings.cpp
+++ b/src/common/settings.cpp
@@ -45,6 +45,7 @@ SWITCHABLE(CpuAccuracy, true);
 SWITCHABLE(FullscreenMode, true);
 SWITCHABLE(GpuAccuracy, true);
 SWITCHABLE(Language, true);
+SWITCHABLE(MemoryLayout, true);
 SWITCHABLE(NvdecEmulation, false);
 SWITCHABLE(Region, true);
 SWITCHABLE(RendererBackend, true);
@@ -61,6 +62,10 @@ SWITCHABLE(u32, false);
 SWITCHABLE(u8, false);
 SWITCHABLE(u8, true);
 
+// Used in UISettings
+// TODO see if we can move this to uisettings.cpp
+SWITCHABLE(ConfirmStop, true);
+
 #undef SETTING
 #undef SWITCHABLE
 #endif
diff --git a/src/common/settings.h b/src/common/settings.h
index 6a3fe47c9..9317075f7 100644
--- a/src/common/settings.h
+++ b/src/common/settings.h
@@ -67,6 +67,7 @@ SWITCHABLE(CpuAccuracy, true);
 SWITCHABLE(FullscreenMode, true);
 SWITCHABLE(GpuAccuracy, true);
 SWITCHABLE(Language, true);
+SWITCHABLE(MemoryLayout, true);
 SWITCHABLE(NvdecEmulation, false);
 SWITCHABLE(Region, true);
 SWITCHABLE(RendererBackend, true);
@@ -83,6 +84,10 @@ SWITCHABLE(u32, false);
 SWITCHABLE(u8, false);
 SWITCHABLE(u8, true);
 
+// Used in UISettings
+// TODO see if we can move this to uisettings.h
+SWITCHABLE(ConfirmStop, true);
+
 #undef SETTING
 #undef SWITCHABLE
 #endif
diff --git a/src/common/settings_enums.h b/src/common/settings_enums.h
index 815cafe15..11429d7a8 100644
--- a/src/common/settings_enums.h
+++ b/src/common/settings_enums.h
@@ -133,6 +133,8 @@ ENUM(CpuAccuracy, Auto, Accurate, Unsafe, Paranoid);
 
 ENUM(MemoryLayout, Memory_4Gb, Memory_6Gb, Memory_8Gb);
 
+ENUM(ConfirmStop, Ask_Always, Ask_Based_On_Game, Ask_Never);
+
 ENUM(FullscreenMode, Borderless, Exclusive);
 
 ENUM(NvdecEmulation, Off, Cpu, Gpu);
diff --git a/src/common/stb.cpp b/src/common/stb.cpp
new file mode 100644
index 000000000..d3b16665d
--- /dev/null
+++ b/src/common/stb.cpp
@@ -0,0 +1,8 @@
+// SPDX-FileCopyrightText: Copyright 2023 yuzu Emulator Project
+// SPDX-License-Identifier: GPL-2.0-or-later
+
+#define STB_IMAGE_IMPLEMENTATION
+#define STB_IMAGE_RESIZE_IMPLEMENTATION
+#define STB_IMAGE_WRITE_IMPLEMENTATION
+
+#include "common/stb.h"
diff --git a/src/common/stb.h b/src/common/stb.h
new file mode 100644
index 000000000..e5c197c11
--- /dev/null
+++ b/src/common/stb.h
@@ -0,0 +1,8 @@
+// SPDX-FileCopyrightText: Copyright 2023 yuzu Emulator Project
+// SPDX-License-Identifier: GPL-2.0-or-later
+
+#pragma once
+
+#include <stb_image.h>
+#include <stb_image_resize.h>
+#include <stb_image_write.h>
diff --git a/src/common/string_util.cpp b/src/common/string_util.cpp
index feab1653d..4c7aba3f5 100644
--- a/src/common/string_util.cpp
+++ b/src/common/string_util.cpp
@@ -135,6 +135,11 @@ std::u16string UTF8ToUTF16(std::string_view input) {
     return convert.from_bytes(input.data(), input.data() + input.size());
 }
 
+std::u32string UTF8ToUTF32(std::string_view input) {
+    std::wstring_convert<std::codecvt_utf8<char32_t>, char32_t> convert;
+    return convert.from_bytes(input.data(), input.data() + input.size());
+}
+
 #ifdef _WIN32
 static std::wstring CPToUTF16(u32 code_page, std::string_view input) {
     const auto size =
diff --git a/src/common/string_util.h b/src/common/string_util.h
index c351f1a0c..9da1ca4e9 100644
--- a/src/common/string_util.h
+++ b/src/common/string_util.h
@@ -38,6 +38,7 @@ bool SplitPath(const std::string& full_path, std::string* _pPath, std::string* _
 
 [[nodiscard]] std::string UTF16ToUTF8(std::u16string_view input);
 [[nodiscard]] std::u16string UTF8ToUTF16(std::string_view input);
+[[nodiscard]] std::u32string UTF8ToUTF32(std::string_view input);
 
 #ifdef _WIN32
 [[nodiscard]] std::string UTF16ToUTF8(std::wstring_view input);
diff --git a/src/common/thread.cpp b/src/common/thread.cpp
index 919e33af9..34cc1527b 100644
--- a/src/common/thread.cpp
+++ b/src/common/thread.cpp
@@ -11,6 +11,7 @@
 #include <mach/mach.h>
 #elif defined(_WIN32)
 #include <windows.h>
+#include "common/string_util.h"
 #else
 #if defined(__Bitrig__) || defined(__DragonFly__) || defined(__FreeBSD__) || defined(__OpenBSD__)
 #include <pthread_np.h>
@@ -82,29 +83,8 @@ void SetCurrentThreadPriority(ThreadPriority new_priority) {
 #ifdef _MSC_VER
 
 // Sets the debugger-visible name of the current thread.
-// Uses trick documented in:
-// https://docs.microsoft.com/en-us/visualstudio/debugger/how-to-set-a-thread-name-in-native-code
 void SetCurrentThreadName(const char* name) {
-    static const DWORD MS_VC_EXCEPTION = 0x406D1388;
-
-#pragma pack(push, 8)
-    struct THREADNAME_INFO {
-        DWORD dwType;     // must be 0x1000
-        LPCSTR szName;    // pointer to name (in user addr space)
-        DWORD dwThreadID; // thread ID (-1=caller thread)
-        DWORD dwFlags;    // reserved for future use, must be zero
-    } info;
-#pragma pack(pop)
-
-    info.dwType = 0x1000;
-    info.szName = name;
-    info.dwThreadID = std::numeric_limits<DWORD>::max();
-    info.dwFlags = 0;
-
-    __try {
-        RaiseException(MS_VC_EXCEPTION, 0, sizeof(info) / sizeof(ULONG_PTR), (ULONG_PTR*)&info);
-    } __except (EXCEPTION_CONTINUE_EXECUTION) {
-    }
+    SetThreadDescription(GetCurrentThread(), UTF8ToUTF16W(name).data());
 }
 
 #else // !MSVC_VER, so must be POSIX threads
diff --git a/src/common/wall_clock.cpp b/src/common/wall_clock.cpp
index 71e15ab4c..caca9a123 100644
--- a/src/common/wall_clock.cpp
+++ b/src/common/wall_clock.cpp
@@ -10,6 +10,10 @@
 #include "common/x64/rdtsc.h"
 #endif
 
+#if defined(ARCHITECTURE_arm64) && defined(__linux__)
+#include "common/arm64/native_clock.h"
+#endif
+
 namespace Common {
 
 class StandardWallClock final : public WallClock {
@@ -53,7 +57,7 @@ private:
 };
 
 std::unique_ptr<WallClock> CreateOptimalClock() {
-#ifdef ARCHITECTURE_x86_64
+#if defined(ARCHITECTURE_x86_64)
     const auto& caps = GetCPUCaps();
 
     if (caps.invariant_tsc && caps.tsc_frequency >= std::nano::den) {
@@ -64,6 +68,8 @@ std::unique_ptr<WallClock> CreateOptimalClock() {
         // - Is not more precise than 1 GHz (1ns resolution)
         return std::make_unique<StandardWallClock>();
     }
+#elif defined(ARCHITECTURE_arm64) && defined(__linux__)
+    return std::make_unique<Arm64::NativeClock>();
 #else
     return std::make_unique<StandardWallClock>();
 #endif
diff --git a/src/core/CMakeLists.txt b/src/core/CMakeLists.txt
index e02ededfc..e4f499135 100644
--- a/src/core/CMakeLists.txt
+++ b/src/core/CMakeLists.txt
@@ -466,14 +466,18 @@ add_library(core STATIC
     hle/service/caps/caps_a.h
     hle/service/caps/caps_c.cpp
     hle/service/caps/caps_c.h
-    hle/service/caps/caps_u.cpp
-    hle/service/caps/caps_u.h
+    hle/service/caps/caps_manager.cpp
+    hle/service/caps/caps_manager.h
+    hle/service/caps/caps_result.h
     hle/service/caps/caps_sc.cpp
     hle/service/caps/caps_sc.h
     hle/service/caps/caps_ss.cpp
     hle/service/caps/caps_ss.h
     hle/service/caps/caps_su.cpp
     hle/service/caps/caps_su.h
+    hle/service/caps/caps_types.h
+    hle/service/caps/caps_u.cpp
+    hle/service/caps/caps_u.h
     hle/service/erpt/erpt.cpp
     hle/service/erpt/erpt.h
     hle/service/es/es.cpp
diff --git a/src/core/arm/arm_interface.cpp b/src/core/arm/arm_interface.cpp
index 0c012f094..5e27dde58 100644
--- a/src/core/arm/arm_interface.cpp
+++ b/src/core/arm/arm_interface.cpp
@@ -86,9 +86,9 @@ void ARM_Interface::SymbolicateBacktrace(Core::System& system, std::vector<Backt
 
     std::map<std::string, Symbols::Symbols> symbols;
     for (const auto& module : modules) {
-        symbols.insert_or_assign(
-            module.second, Symbols::GetSymbols(module.first, system.ApplicationMemory(),
-                                               system.ApplicationProcess()->Is64BitProcess()));
+        symbols.insert_or_assign(module.second,
+                                 Symbols::GetSymbols(module.first, system.ApplicationMemory(),
+                                                     system.ApplicationProcess()->Is64Bit()));
     }
 
     for (auto& entry : out) {
diff --git a/src/core/core.cpp b/src/core/core.cpp
index 08cbb8978..14d6c8c27 100644
--- a/src/core/core.cpp
+++ b/src/core/core.cpp
@@ -116,11 +116,8 @@ FileSys::VirtualFile GetGameFileFromPath(const FileSys::VirtualFilesystem& vfs,
             }
         }
 
-        if (concat.empty()) {
-            return nullptr;
-        }
-
-        return FileSys::ConcatenatedVfsFile::MakeConcatenatedFile(concat, dir->GetName());
+        return FileSys::ConcatenatedVfsFile::MakeConcatenatedFile(dir->GetName(),
+                                                                  std::move(concat));
     }
 
     if (Common::FS::IsDir(path)) {
@@ -312,17 +309,10 @@ struct System::Impl {
 
         telemetry_session->AddInitialInfo(*app_loader, fs_controller, *content_provider);
 
-        // Create a resource limit for the process.
-        const auto physical_memory_size =
-            kernel.MemoryManager().GetSize(Kernel::KMemoryManager::Pool::Application);
-        auto* resource_limit = Kernel::CreateResourceLimitForProcess(system, physical_memory_size);
-
         // Create the process.
         auto main_process = Kernel::KProcess::Create(system.Kernel());
-        ASSERT(Kernel::KProcess::Initialize(main_process, system, "main",
-                                            Kernel::KProcess::ProcessType::Userland, resource_limit)
-                   .IsSuccess());
         Kernel::KProcess::Register(system.Kernel(), main_process);
+        kernel.AppendNewProcess(main_process);
         kernel.MakeApplicationProcess(main_process);
         const auto [load_result, load_parameters] = app_loader->Load(*main_process, system);
         if (load_result != Loader::ResultStatus::Success) {
@@ -421,6 +411,7 @@ struct System::Impl {
             services->KillNVNFlinger();
         }
         kernel.CloseServices();
+        kernel.ShutdownCores();
         services.reset();
         service_manager.reset();
         cheat_engine.reset();
@@ -432,7 +423,6 @@ struct System::Impl {
         gpu_core.reset();
         host1x_core.reset();
         perf_stats.reset();
-        kernel.ShutdownCores();
         cpu_manager.Shutdown();
         debugger.reset();
         kernel.Shutdown();
@@ -1078,6 +1068,10 @@ void System::ApplySettings() {
     impl->RefreshTime();
 
     if (IsPoweredOn()) {
+        if (Settings::values.custom_rtc_enabled) {
+            const s64 posix_time{Settings::values.custom_rtc.GetValue()};
+            GetTimeManager().UpdateLocalSystemClockTime(posix_time);
+        }
         Renderer().RefreshBaseSettings();
     }
 }
diff --git a/src/core/debugger/debugger.cpp b/src/core/debugger/debugger.cpp
index a1589fecb..0e270eb50 100644
--- a/src/core/debugger/debugger.cpp
+++ b/src/core/debugger/debugger.cpp
@@ -258,20 +258,20 @@ private:
         Kernel::KScopedSchedulerLock sl{system.Kernel()};
 
         // Put all threads to sleep on next scheduler round.
-        for (auto* thread : ThreadList()) {
-            thread->RequestSuspend(Kernel::SuspendType::Debug);
+        for (auto& thread : ThreadList()) {
+            thread.RequestSuspend(Kernel::SuspendType::Debug);
         }
     }
 
     void ResumeEmulation(Kernel::KThread* except = nullptr) {
         // Wake up all threads.
-        for (auto* thread : ThreadList()) {
-            if (thread == except) {
+        for (auto& thread : ThreadList()) {
+            if (std::addressof(thread) == except) {
                 continue;
             }
 
-            thread->SetStepState(Kernel::StepState::NotStepping);
-            thread->Resume(Kernel::SuspendType::Debug);
+            thread.SetStepState(Kernel::StepState::NotStepping);
+            thread.Resume(Kernel::SuspendType::Debug);
         }
     }
 
@@ -283,13 +283,17 @@ private:
     }
 
     void UpdateActiveThread() {
-        const auto& threads{ThreadList()};
-        if (std::find(threads.begin(), threads.end(), state->active_thread) == threads.end()) {
-            state->active_thread = threads.front();
+        auto& threads{ThreadList()};
+        for (auto& thread : threads) {
+            if (std::addressof(thread) == state->active_thread) {
+                // Thread is still alive, no need to update.
+                return;
+            }
         }
+        state->active_thread = std::addressof(threads.front());
     }
 
-    const std::list<Kernel::KThread*>& ThreadList() {
+    Kernel::KProcess::ThreadList& ThreadList() {
         return system.ApplicationProcess()->GetThreadList();
     }
 
diff --git a/src/core/debugger/gdbstub.cpp b/src/core/debugger/gdbstub.cpp
index e55831f27..6f5f5156b 100644
--- a/src/core/debugger/gdbstub.cpp
+++ b/src/core/debugger/gdbstub.cpp
@@ -2,6 +2,8 @@
 // SPDX-License-Identifier: GPL-2.0-or-later
 
 #include <atomic>
+#include <codecvt>
+#include <locale>
 #include <numeric>
 #include <optional>
 #include <thread>
@@ -12,6 +14,7 @@
 #include "common/logging/log.h"
 #include "common/scope_exit.h"
 #include "common/settings.h"
+#include "common/string_util.h"
 #include "core/arm/arm_interface.h"
 #include "core/core.h"
 #include "core/debugger/gdbstub.h"
@@ -68,10 +71,16 @@ static std::string EscapeGDB(std::string_view data) {
 }
 
 static std::string EscapeXML(std::string_view data) {
+    std::u32string converted = U"[Encoding error]";
+    try {
+        converted = Common::UTF8ToUTF32(data);
+    } catch (std::range_error&) {
+    }
+
     std::string escaped;
     escaped.reserve(data.size());
 
-    for (char c : data) {
+    for (char32_t c : converted) {
         switch (c) {
         case '&':
             escaped += "&amp;";
@@ -86,7 +95,11 @@ static std::string EscapeXML(std::string_view data) {
             escaped += "&gt;";
             break;
         default:
-            escaped += c;
+            if (c > 0x7f) {
+                escaped += fmt::format("&#{};", static_cast<u32>(c));
+            } else {
+                escaped += static_cast<char>(c);
+            }
             break;
         }
     }
@@ -96,7 +109,7 @@ static std::string EscapeXML(std::string_view data) {
 
 GDBStub::GDBStub(DebuggerBackend& backend_, Core::System& system_)
     : DebuggerFrontend(backend_), system{system_} {
-    if (system.ApplicationProcess()->Is64BitProcess()) {
+    if (system.ApplicationProcess()->Is64Bit()) {
         arch = std::make_unique<GDBStubA64>();
     } else {
         arch = std::make_unique<GDBStubA32>();
@@ -433,10 +446,10 @@ void GDBStub::HandleBreakpointRemove(std::string_view command) {
 // See osdbg_thread_local_region.os.horizon.hpp and osdbg_thread_type.os.horizon.hpp
 
 static std::optional<std::string> GetNameFromThreadType32(Core::Memory::Memory& memory,
-                                                          const Kernel::KThread* thread) {
+                                                          const Kernel::KThread& thread) {
     // Read thread type from TLS
-    const VAddr tls_thread_type{memory.Read32(thread->GetTlsAddress() + 0x1fc)};
-    const VAddr argument_thread_type{thread->GetArgument()};
+    const VAddr tls_thread_type{memory.Read32(thread.GetTlsAddress() + 0x1fc)};
+    const VAddr argument_thread_type{thread.GetArgument()};
 
     if (argument_thread_type && tls_thread_type != argument_thread_type) {
         // Probably not created by nnsdk, no name available.
@@ -464,10 +477,10 @@ static std::optional<std::string> GetNameFromThreadType32(Core::Memory::Memory&
 }
 
 static std::optional<std::string> GetNameFromThreadType64(Core::Memory::Memory& memory,
-                                                          const Kernel::KThread* thread) {
+                                                          const Kernel::KThread& thread) {
     // Read thread type from TLS
-    const VAddr tls_thread_type{memory.Read64(thread->GetTlsAddress() + 0x1f8)};
-    const VAddr argument_thread_type{thread->GetArgument()};
+    const VAddr tls_thread_type{memory.Read64(thread.GetTlsAddress() + 0x1f8)};
+    const VAddr argument_thread_type{thread.GetArgument()};
 
     if (argument_thread_type && tls_thread_type != argument_thread_type) {
         // Probably not created by nnsdk, no name available.
@@ -495,16 +508,16 @@ static std::optional<std::string> GetNameFromThreadType64(Core::Memory::Memory&
 }
 
 static std::optional<std::string> GetThreadName(Core::System& system,
-                                                const Kernel::KThread* thread) {
-    if (system.ApplicationProcess()->Is64BitProcess()) {
+                                                const Kernel::KThread& thread) {
+    if (system.ApplicationProcess()->Is64Bit()) {
         return GetNameFromThreadType64(system.ApplicationMemory(), thread);
     } else {
         return GetNameFromThreadType32(system.ApplicationMemory(), thread);
     }
 }
 
-static std::string_view GetThreadWaitReason(const Kernel::KThread* thread) {
-    switch (thread->GetWaitReasonForDebugging()) {
+static std::string_view GetThreadWaitReason(const Kernel::KThread& thread) {
+    switch (thread.GetWaitReasonForDebugging()) {
     case Kernel::ThreadWaitReasonForDebugging::Sleep:
         return "Sleep";
     case Kernel::ThreadWaitReasonForDebugging::IPC:
@@ -522,8 +535,8 @@ static std::string_view GetThreadWaitReason(const Kernel::KThread* thread) {
     }
 }
 
-static std::string GetThreadState(const Kernel::KThread* thread) {
-    switch (thread->GetState()) {
+static std::string GetThreadState(const Kernel::KThread& thread) {
+    switch (thread.GetState()) {
     case Kernel::ThreadState::Initialized:
         return "Initialized";
     case Kernel::ThreadState::Waiting:
@@ -591,7 +604,7 @@ void GDBStub::HandleQuery(std::string_view command) {
         const auto& threads = system.ApplicationProcess()->GetThreadList();
         std::vector<std::string> thread_ids;
         for (const auto& thread : threads) {
-            thread_ids.push_back(fmt::format("{:x}", thread->GetThreadId()));
+            thread_ids.push_back(fmt::format("{:x}", thread.GetThreadId()));
         }
         SendReply(fmt::format("m{}", fmt::join(thread_ids, ",")));
     } else if (command.starts_with("sThreadInfo")) {
@@ -603,14 +616,14 @@ void GDBStub::HandleQuery(std::string_view command) {
         buffer += "<threads>";
 
         const auto& threads = system.ApplicationProcess()->GetThreadList();
-        for (const auto* thread : threads) {
+        for (const auto& thread : threads) {
             auto thread_name{GetThreadName(system, thread)};
             if (!thread_name) {
-                thread_name = fmt::format("Thread {:d}", thread->GetThreadId());
+                thread_name = fmt::format("Thread {:d}", thread.GetThreadId());
             }
 
             buffer += fmt::format(R"(<thread id="{:x}" core="{:d}" name="{}">{}</thread>)",
-                                  thread->GetThreadId(), thread->GetActiveCore(),
+                                  thread.GetThreadId(), thread.GetActiveCore(),
                                   EscapeXML(*thread_name), GetThreadState(thread));
         }
 
@@ -809,11 +822,13 @@ void GDBStub::HandleRcmd(const std::vector<u8>& command) {
                 const char i = True(mem_info.attribute & MemoryAttribute::IpcLocked) ? 'I' : '-';
                 const char d = True(mem_info.attribute & MemoryAttribute::DeviceShared) ? 'D' : '-';
                 const char u = True(mem_info.attribute & MemoryAttribute::Uncached) ? 'U' : '-';
+                const char p =
+                    True(mem_info.attribute & MemoryAttribute::PermissionLocked) ? 'P' : '-';
 
-                reply +=
-                    fmt::format("  {:#012x} - {:#012x} {} {} {}{}{}{} [{}, {}]\n",
-                                mem_info.base_address, mem_info.base_address + mem_info.size - 1,
-                                perm, state, l, i, d, u, mem_info.ipc_count, mem_info.device_count);
+                reply += fmt::format("  {:#012x} - {:#012x} {} {} {}{}{}{}{} [{}, {}]\n",
+                                     mem_info.base_address,
+                                     mem_info.base_address + mem_info.size - 1, perm, state, l, i,
+                                     d, u, p, mem_info.ipc_count, mem_info.device_count);
             }
 
             const uintptr_t next_address = mem_info.base_address + mem_info.size;
@@ -835,10 +850,10 @@ void GDBStub::HandleRcmd(const std::vector<u8>& command) {
 }
 
 Kernel::KThread* GDBStub::GetThreadByID(u64 thread_id) {
-    const auto& threads{system.ApplicationProcess()->GetThreadList()};
-    for (auto* thread : threads) {
-        if (thread->GetThreadId() == thread_id) {
-            return thread;
+    auto& threads{system.ApplicationProcess()->GetThreadList()};
+    for (auto& thread : threads) {
+        if (thread.GetThreadId() == thread_id) {
+            return std::addressof(thread);
         }
     }
 
diff --git a/src/core/file_sys/fsmitm_romfsbuild.cpp b/src/core/file_sys/fsmitm_romfsbuild.cpp
index e39c7b62b..f1d3e4129 100644
--- a/src/core/file_sys/fsmitm_romfsbuild.cpp
+++ b/src/core/file_sys/fsmitm_romfsbuild.cpp
@@ -107,62 +107,56 @@ static u64 romfs_get_hash_table_count(u64 num_entries) {
 
 void RomFSBuildContext::VisitDirectory(VirtualDir romfs_dir, VirtualDir ext_dir,
                                        std::shared_ptr<RomFSBuildDirectoryContext> parent) {
-    std::vector<std::shared_ptr<RomFSBuildDirectoryContext>> child_dirs;
+    for (auto& child_romfs_file : romfs_dir->GetFiles()) {
+        const auto name = child_romfs_file->GetName();
+        const auto child = std::make_shared<RomFSBuildFileContext>();
+        // Set child's path.
+        child->cur_path_ofs = parent->path_len + 1;
+        child->path_len = child->cur_path_ofs + static_cast<u32>(name.size());
+        child->path = parent->path + "/" + name;
+
+        if (ext_dir != nullptr && ext_dir->GetFile(name + ".stub") != nullptr) {
+            continue;
+        }
 
-    const auto entries = romfs_dir->GetEntries();
+        // Sanity check on path_len
+        ASSERT(child->path_len < FS_MAX_PATH);
 
-    for (const auto& kv : entries) {
-        if (kv.second == VfsEntryType::Directory) {
-            const auto child = std::make_shared<RomFSBuildDirectoryContext>();
-            // Set child's path.
-            child->cur_path_ofs = parent->path_len + 1;
-            child->path_len = child->cur_path_ofs + static_cast<u32>(kv.first.size());
-            child->path = parent->path + "/" + kv.first;
+        child->source = std::move(child_romfs_file);
 
-            if (ext_dir != nullptr && ext_dir->GetFile(kv.first + ".stub") != nullptr) {
-                continue;
+        if (ext_dir != nullptr) {
+            if (const auto ips = ext_dir->GetFile(name + ".ips")) {
+                if (auto patched = PatchIPS(child->source, ips)) {
+                    child->source = std::move(patched);
+                }
             }
+        }
 
-            // Sanity check on path_len
-            ASSERT(child->path_len < FS_MAX_PATH);
-
-            if (AddDirectory(parent, child)) {
-                child_dirs.push_back(child);
-            }
-        } else {
-            const auto child = std::make_shared<RomFSBuildFileContext>();
-            // Set child's path.
-            child->cur_path_ofs = parent->path_len + 1;
-            child->path_len = child->cur_path_ofs + static_cast<u32>(kv.first.size());
-            child->path = parent->path + "/" + kv.first;
-
-            if (ext_dir != nullptr && ext_dir->GetFile(kv.first + ".stub") != nullptr) {
-                continue;
-            }
+        child->size = child->source->GetSize();
 
-            // Sanity check on path_len
-            ASSERT(child->path_len < FS_MAX_PATH);
+        AddFile(parent, child);
+    }
 
-            child->source = romfs_dir->GetFile(kv.first);
+    for (auto& child_romfs_dir : romfs_dir->GetSubdirectories()) {
+        const auto name = child_romfs_dir->GetName();
+        const auto child = std::make_shared<RomFSBuildDirectoryContext>();
+        // Set child's path.
+        child->cur_path_ofs = parent->path_len + 1;
+        child->path_len = child->cur_path_ofs + static_cast<u32>(name.size());
+        child->path = parent->path + "/" + name;
 
-            if (ext_dir != nullptr) {
-                if (const auto ips = ext_dir->GetFile(kv.first + ".ips")) {
-                    if (auto patched = PatchIPS(child->source, ips)) {
-                        child->source = std::move(patched);
-                    }
-                }
-            }
+        if (ext_dir != nullptr && ext_dir->GetFile(name + ".stub") != nullptr) {
+            continue;
+        }
 
-            child->size = child->source->GetSize();
+        // Sanity check on path_len
+        ASSERT(child->path_len < FS_MAX_PATH);
 
-            AddFile(parent, child);
+        if (!AddDirectory(parent, child)) {
+            continue;
         }
-    }
 
-    for (auto& child : child_dirs) {
-        auto subdir_name = std::string_view(child->path).substr(child->cur_path_ofs);
-        auto child_romfs_dir = romfs_dir->GetSubdirectory(subdir_name);
-        auto child_ext_dir = ext_dir != nullptr ? ext_dir->GetSubdirectory(subdir_name) : nullptr;
+        auto child_ext_dir = ext_dir != nullptr ? ext_dir->GetSubdirectory(name) : nullptr;
         this->VisitDirectory(child_romfs_dir, child_ext_dir, child);
     }
 }
@@ -293,7 +287,7 @@ std::multimap<u64, VirtualFile> RomFSBuildContext::Build() {
 
         cur_entry.name_size = name_size;
 
-        out.emplace(cur_file->offset + ROMFS_FILEPARTITION_OFS, cur_file->source);
+        out.emplace(cur_file->offset + ROMFS_FILEPARTITION_OFS, std::move(cur_file->source));
         std::memcpy(file_table.data() + cur_file->entry_offset, &cur_entry, sizeof(RomFSFileEntry));
         std::memset(file_table.data() + cur_file->entry_offset + sizeof(RomFSFileEntry), 0,
                     Common::AlignUp(cur_entry.name_size, 4));
diff --git a/src/core/file_sys/patch_manager.cpp b/src/core/file_sys/patch_manager.cpp
index 8e475f25a..0bca05587 100644
--- a/src/core/file_sys/patch_manager.cpp
+++ b/src/core/file_sys/patch_manager.cpp
@@ -377,16 +377,16 @@ static void ApplyLayeredFS(VirtualFile& romfs, u64 title_id, ContentRecordType t
 
         auto romfs_dir = FindSubdirectoryCaseless(subdir, "romfs");
         if (romfs_dir != nullptr)
-            layers.push_back(std::make_shared<CachedVfsDirectory>(romfs_dir));
+            layers.emplace_back(std::make_shared<CachedVfsDirectory>(std::move(romfs_dir)));
 
         auto ext_dir = FindSubdirectoryCaseless(subdir, "romfs_ext");
         if (ext_dir != nullptr)
-            layers_ext.push_back(std::make_shared<CachedVfsDirectory>(ext_dir));
+            layers_ext.emplace_back(std::make_shared<CachedVfsDirectory>(std::move(ext_dir)));
 
         if (type == ContentRecordType::HtmlDocument) {
             auto manual_dir = FindSubdirectoryCaseless(subdir, "manual_html");
             if (manual_dir != nullptr)
-                layers.push_back(std::make_shared<CachedVfsDirectory>(manual_dir));
+                layers.emplace_back(std::make_shared<CachedVfsDirectory>(std::move(manual_dir)));
         }
     }
 
@@ -400,7 +400,7 @@ static void ApplyLayeredFS(VirtualFile& romfs, u64 title_id, ContentRecordType t
         return;
     }
 
-    layers.push_back(std::move(extracted));
+    layers.emplace_back(std::move(extracted));
 
     auto layered = LayeredVfsDirectory::MakeLayeredDirectory(std::move(layers));
     if (layered == nullptr) {
diff --git a/src/core/file_sys/program_metadata.cpp b/src/core/file_sys/program_metadata.cpp
index f00479bd3..763a44fee 100644
--- a/src/core/file_sys/program_metadata.cpp
+++ b/src/core/file_sys/program_metadata.cpp
@@ -5,6 +5,7 @@
 #include <vector>
 
 #include "common/logging/log.h"
+#include "common/scope_exit.h"
 #include "core/file_sys/program_metadata.h"
 #include "core/file_sys/vfs.h"
 #include "core/loader/loader.h"
@@ -95,17 +96,24 @@ Loader::ResultStatus ProgramMetadata::Load(VirtualFile file) {
     return Loader::ResultStatus::Success;
 }
 
+Loader::ResultStatus ProgramMetadata::Reload(VirtualFile file) {
+    const u64 original_program_id = aci_header.title_id;
+    SCOPE_EXIT({ aci_header.title_id = original_program_id; });
+
+    return this->Load(file);
+}
+
 /*static*/ ProgramMetadata ProgramMetadata::GetDefault() {
-    // Allow use of cores 0~3 and thread priorities 1~63.
-    constexpr u32 default_thread_info_capability = 0x30007F7;
+    // Allow use of cores 0~3 and thread priorities 16~63.
+    constexpr u32 default_thread_info_capability = 0x30043F7;
 
     ProgramMetadata result;
 
     result.LoadManual(
         true /*is_64_bit*/, FileSys::ProgramAddressSpaceType::Is39Bit /*address_space*/,
-        0x2c /*main_thread_prio*/, 0 /*main_thread_core*/, 0x00100000 /*main_thread_stack_size*/,
-        0 /*title_id*/, 0xFFFFFFFFFFFFFFFF /*filesystem_permissions*/,
-        0x1FE00000 /*system_resource_size*/, {default_thread_info_capability} /*capabilities*/);
+        0x2c /*main_thread_prio*/, 0 /*main_thread_core*/, 0x100000 /*main_thread_stack_size*/,
+        0 /*title_id*/, 0xFFFFFFFFFFFFFFFF /*filesystem_permissions*/, 0 /*system_resource_size*/,
+        {default_thread_info_capability} /*capabilities*/);
 
     return result;
 }
diff --git a/src/core/file_sys/program_metadata.h b/src/core/file_sys/program_metadata.h
index 2e8960b07..76ee97d78 100644
--- a/src/core/file_sys/program_metadata.h
+++ b/src/core/file_sys/program_metadata.h
@@ -56,6 +56,7 @@ public:
     static ProgramMetadata GetDefault();
 
     Loader::ResultStatus Load(VirtualFile file);
+    Loader::ResultStatus Reload(VirtualFile file);
 
     /// Load from parameters instead of NPDM file, used for KIP
     void LoadManual(bool is_64_bit, ProgramAddressSpaceType address_space, s32 main_thread_prio,
@@ -72,6 +73,9 @@ public:
     u64 GetFilesystemPermissions() const;
     u32 GetSystemResourceSize() const;
     const KernelCapabilityDescriptors& GetKernelCapabilities() const;
+    const std::array<u8, 0x10>& GetName() const {
+        return npdm_header.application_name;
+    }
 
     void Print() const;
 
@@ -163,14 +167,14 @@ private:
         u32_le unk_size_2;
     };
 
-    Header npdm_header;
-    AciHeader aci_header;
-    AcidHeader acid_header;
+    Header npdm_header{};
+    AciHeader aci_header{};
+    AcidHeader acid_header{};
 
-    FileAccessControl acid_file_access;
-    FileAccessHeader aci_file_access;
+    FileAccessControl acid_file_access{};
+    FileAccessHeader aci_file_access{};
 
-    KernelCapabilityDescriptors aci_kernel_capabilities;
+    KernelCapabilityDescriptors aci_kernel_capabilities{};
 };
 
 } // namespace FileSys
diff --git a/src/core/file_sys/registered_cache.cpp b/src/core/file_sys/registered_cache.cpp
index 04da93d5c..1cc77ad14 100644
--- a/src/core/file_sys/registered_cache.cpp
+++ b/src/core/file_sys/registered_cache.cpp
@@ -322,7 +322,8 @@ VirtualFile RegisteredCache::OpenFileOrDirectoryConcat(const VirtualDir& open_di
         return nullptr;
     }
 
-    return ConcatenatedVfsFile::MakeConcatenatedFile(concat, concat.front()->GetName());
+    auto name = concat.front()->GetName();
+    return ConcatenatedVfsFile::MakeConcatenatedFile(std::move(name), std::move(concat));
 }
 
 VirtualFile RegisteredCache::GetFileAtID(NcaID id) const {
diff --git a/src/core/file_sys/romfs.cpp b/src/core/file_sys/romfs.cpp
index 614da2130..1c580de57 100644
--- a/src/core/file_sys/romfs.cpp
+++ b/src/core/file_sys/romfs.cpp
@@ -133,7 +133,7 @@ VirtualDir ExtractRomFS(VirtualFile file, RomFSExtractionType type) {
         out = out->GetSubdirectories().front();
     }
 
-    return std::make_shared<CachedVfsDirectory>(out);
+    return std::make_shared<CachedVfsDirectory>(std::move(out));
 }
 
 VirtualFile CreateRomFS(VirtualDir dir, VirtualDir ext) {
@@ -141,8 +141,7 @@ VirtualFile CreateRomFS(VirtualDir dir, VirtualDir ext) {
         return nullptr;
 
     RomFSBuildContext ctx{dir, ext};
-    auto file_map = ctx.Build();
-    return ConcatenatedVfsFile::MakeConcatenatedFile(0, file_map, dir->GetName());
+    return ConcatenatedVfsFile::MakeConcatenatedFile(0, dir->GetName(), ctx.Build());
 }
 
 } // namespace FileSys
diff --git a/src/core/file_sys/system_archive/system_version.cpp b/src/core/file_sys/system_archive/system_version.cpp
index bd493ecca..e4751c2b4 100644
--- a/src/core/file_sys/system_archive/system_version.cpp
+++ b/src/core/file_sys/system_archive/system_version.cpp
@@ -1,6 +1,7 @@
 // SPDX-FileCopyrightText: Copyright 2019 yuzu Emulator Project
 // SPDX-License-Identifier: GPL-2.0-or-later
 
+#include "common/logging/log.h"
 #include "core/file_sys/system_archive/system_version.h"
 #include "core/file_sys/vfs_vector.h"
 #include "core/hle/api_version.h"
@@ -12,6 +13,9 @@ std::string GetLongDisplayVersion() {
 }
 
 VirtualDir SystemVersion() {
+    LOG_WARNING(Common_Filesystem, "called - Using hardcoded firmware version '{}'",
+                GetLongDisplayVersion());
+
     VirtualFile file = std::make_shared<VectorVfsFile>(std::vector<u8>(0x100), "file");
     file->WriteObject(HLE::ApiVersion::HOS_VERSION_MAJOR, 0);
     file->WriteObject(HLE::ApiVersion::HOS_VERSION_MINOR, 1);
diff --git a/src/core/file_sys/vfs_cached.cpp b/src/core/file_sys/vfs_cached.cpp
index c3154ee81..7ee5300e5 100644
--- a/src/core/file_sys/vfs_cached.cpp
+++ b/src/core/file_sys/vfs_cached.cpp
@@ -6,13 +6,13 @@
 
 namespace FileSys {
 
-CachedVfsDirectory::CachedVfsDirectory(VirtualDir& source_dir)
+CachedVfsDirectory::CachedVfsDirectory(VirtualDir&& source_dir)
     : name(source_dir->GetName()), parent(source_dir->GetParentDirectory()) {
     for (auto& dir : source_dir->GetSubdirectories()) {
-        dirs.emplace(dir->GetName(), std::make_shared<CachedVfsDirectory>(dir));
+        dirs.emplace(dir->GetName(), std::make_shared<CachedVfsDirectory>(std::move(dir)));
     }
     for (auto& file : source_dir->GetFiles()) {
-        files.emplace(file->GetName(), file);
+        files.emplace(file->GetName(), std::move(file));
     }
 }
 
diff --git a/src/core/file_sys/vfs_cached.h b/src/core/file_sys/vfs_cached.h
index 113acac12..1e5300784 100644
--- a/src/core/file_sys/vfs_cached.h
+++ b/src/core/file_sys/vfs_cached.h
@@ -11,7 +11,7 @@ namespace FileSys {
 
 class CachedVfsDirectory : public ReadOnlyVfsDirectory {
 public:
-    CachedVfsDirectory(VirtualDir& source_directory);
+    CachedVfsDirectory(VirtualDir&& source_directory);
 
     ~CachedVfsDirectory() override;
     VirtualFile GetFile(std::string_view file_name) const override;
diff --git a/src/core/file_sys/vfs_concat.cpp b/src/core/file_sys/vfs_concat.cpp
index 311a59e5f..168b9cbec 100644
--- a/src/core/file_sys/vfs_concat.cpp
+++ b/src/core/file_sys/vfs_concat.cpp
@@ -10,7 +10,7 @@
 
 namespace FileSys {
 
-ConcatenatedVfsFile::ConcatenatedVfsFile(ConcatenationMap&& concatenation_map_, std::string&& name_)
+ConcatenatedVfsFile::ConcatenatedVfsFile(std::string&& name_, ConcatenationMap&& concatenation_map_)
     : concatenation_map(std::move(concatenation_map_)), name(std::move(name_)) {
     DEBUG_ASSERT(this->VerifyContinuity());
 }
@@ -30,8 +30,8 @@ bool ConcatenatedVfsFile::VerifyContinuity() const {
 
 ConcatenatedVfsFile::~ConcatenatedVfsFile() = default;
 
-VirtualFile ConcatenatedVfsFile::MakeConcatenatedFile(const std::vector<VirtualFile>& files,
-                                                      std::string&& name) {
+VirtualFile ConcatenatedVfsFile::MakeConcatenatedFile(std::string&& name,
+                                                      std::vector<VirtualFile>&& files) {
     // Fold trivial cases.
     if (files.empty()) {
         return nullptr;
@@ -46,20 +46,21 @@ VirtualFile ConcatenatedVfsFile::MakeConcatenatedFile(const std::vector<VirtualF
     u64 last_offset = 0;
 
     for (auto& file : files) {
+        const auto size = file->GetSize();
+
         concatenation_map.emplace_back(ConcatenationEntry{
             .offset = last_offset,
-            .file = file,
+            .file = std::move(file),
         });
 
-        last_offset += file->GetSize();
+        last_offset += size;
     }
 
-    return VirtualFile(new ConcatenatedVfsFile(std::move(concatenation_map), std::move(name)));
+    return VirtualFile(new ConcatenatedVfsFile(std::move(name), std::move(concatenation_map)));
 }
 
-VirtualFile ConcatenatedVfsFile::MakeConcatenatedFile(u8 filler_byte,
-                                                      const std::multimap<u64, VirtualFile>& files,
-                                                      std::string&& name) {
+VirtualFile ConcatenatedVfsFile::MakeConcatenatedFile(u8 filler_byte, std::string&& name,
+                                                      std::multimap<u64, VirtualFile>&& files) {
     // Fold trivial cases.
     if (files.empty()) {
         return nullptr;
@@ -76,6 +77,8 @@ VirtualFile ConcatenatedVfsFile::MakeConcatenatedFile(u8 filler_byte,
 
     // Iteration of a multimap is ordered, so offset will be strictly non-decreasing.
     for (auto& [offset, file] : files) {
+        const auto size = file->GetSize();
+
         if (offset > last_offset) {
             concatenation_map.emplace_back(ConcatenationEntry{
                 .offset = last_offset,
@@ -85,13 +88,13 @@ VirtualFile ConcatenatedVfsFile::MakeConcatenatedFile(u8 filler_byte,
 
         concatenation_map.emplace_back(ConcatenationEntry{
             .offset = offset,
-            .file = file,
+            .file = std::move(file),
         });
 
-        last_offset = offset + file->GetSize();
+        last_offset = offset + size;
     }
 
-    return VirtualFile(new ConcatenatedVfsFile(std::move(concatenation_map), std::move(name)));
+    return VirtualFile(new ConcatenatedVfsFile(std::move(name), std::move(concatenation_map)));
 }
 
 std::string ConcatenatedVfsFile::GetName() const {
diff --git a/src/core/file_sys/vfs_concat.h b/src/core/file_sys/vfs_concat.h
index 6b329d545..cbddd12bd 100644
--- a/src/core/file_sys/vfs_concat.h
+++ b/src/core/file_sys/vfs_concat.h
@@ -24,22 +24,20 @@ private:
     };
     using ConcatenationMap = std::vector<ConcatenationEntry>;
 
-    explicit ConcatenatedVfsFile(std::vector<ConcatenationEntry>&& concatenation_map,
-                                 std::string&& name);
+    explicit ConcatenatedVfsFile(std::string&& name,
+                                 std::vector<ConcatenationEntry>&& concatenation_map);
     bool VerifyContinuity() const;
 
 public:
     ~ConcatenatedVfsFile() override;
 
     /// Wrapper function to allow for more efficient handling of files.size() == 0, 1 cases.
-    static VirtualFile MakeConcatenatedFile(const std::vector<VirtualFile>& files,
-                                            std::string&& name);
+    static VirtualFile MakeConcatenatedFile(std::string&& name, std::vector<VirtualFile>&& files);
 
     /// Convenience function that turns a map of offsets to files into a concatenated file, filling
     /// gaps with a given filler byte.
-    static VirtualFile MakeConcatenatedFile(u8 filler_byte,
-                                            const std::multimap<u64, VirtualFile>& files,
-                                            std::string&& name);
+    static VirtualFile MakeConcatenatedFile(u8 filler_byte, std::string&& name,
+                                            std::multimap<u64, VirtualFile>&& files);
 
     std::string GetName() const override;
     std::size_t GetSize() const override;
diff --git a/src/core/file_sys/vfs_layered.cpp b/src/core/file_sys/vfs_layered.cpp
index 3e6426afc..08daca397 100644
--- a/src/core/file_sys/vfs_layered.cpp
+++ b/src/core/file_sys/vfs_layered.cpp
@@ -38,7 +38,7 @@ VirtualDir LayeredVfsDirectory::GetDirectoryRelative(std::string_view path) cons
     for (const auto& layer : dirs) {
         auto dir = layer->GetDirectoryRelative(path);
         if (dir != nullptr) {
-            out.push_back(std::move(dir));
+            out.emplace_back(std::move(dir));
         }
     }
 
@@ -62,11 +62,11 @@ std::vector<VirtualFile> LayeredVfsDirectory::GetFiles() const {
     std::set<std::string, std::less<>> out_names;
 
     for (const auto& layer : dirs) {
-        for (const auto& file : layer->GetFiles()) {
+        for (auto& file : layer->GetFiles()) {
             auto file_name = file->GetName();
             if (!out_names.contains(file_name)) {
                 out_names.emplace(std::move(file_name));
-                out.push_back(file);
+                out.emplace_back(std::move(file));
             }
         }
     }
@@ -86,7 +86,7 @@ std::vector<VirtualDir> LayeredVfsDirectory::GetSubdirectories() const {
     std::vector<VirtualDir> out;
     out.reserve(names.size());
     for (const auto& subdir : names)
-        out.push_back(GetSubdirectory(subdir));
+        out.emplace_back(GetSubdirectory(subdir));
 
     return out;
 }
diff --git a/src/core/hle/kernel/board/nintendo/nx/k_system_control.cpp b/src/core/hle/kernel/board/nintendo/nx/k_system_control.cpp
index 4cfdf4558..59364efa1 100644
--- a/src/core/hle/kernel/board/nintendo/nx/k_system_control.cpp
+++ b/src/core/hle/kernel/board/nintendo/nx/k_system_control.cpp
@@ -8,7 +8,11 @@
 
 #include "core/hle/kernel/board/nintendo/nx/k_system_control.h"
 #include "core/hle/kernel/board/nintendo/nx/secure_monitor.h"
+#include "core/hle/kernel/k_memory_manager.h"
+#include "core/hle/kernel/k_page_table.h"
 #include "core/hle/kernel/k_trace.h"
+#include "core/hle/kernel/kernel.h"
+#include "core/hle/kernel/svc_results.h"
 
 namespace Kernel::Board::Nintendo::Nx {
 
@@ -30,6 +34,8 @@ constexpr const std::size_t RequiredNonSecureSystemMemorySize =
 constexpr const std::size_t RequiredNonSecureSystemMemorySizeWithFatal =
     RequiredNonSecureSystemMemorySize + impl::RequiredNonSecureSystemMemorySizeViFatal;
 
+constexpr const std::size_t SecureAlignment = 128_KiB;
+
 namespace {
 
 using namespace Common::Literals;
@@ -183,4 +189,57 @@ u64 KSystemControl::GenerateRandomRange(u64 min, u64 max) {
     return GenerateUniformRange(min, max, GenerateRandomU64);
 }
 
+size_t KSystemControl::CalculateRequiredSecureMemorySize(size_t size, u32 pool) {
+    if (pool == static_cast<u32>(KMemoryManager::Pool::Applet)) {
+        return 0;
+    } else {
+        // return KSystemControlBase::CalculateRequiredSecureMemorySize(size, pool);
+        return size;
+    }
+}
+
+Result KSystemControl::AllocateSecureMemory(KernelCore& kernel, KVirtualAddress* out, size_t size,
+                                            u32 pool) {
+    // Applet secure memory is handled separately.
+    UNIMPLEMENTED_IF(pool == static_cast<u32>(KMemoryManager::Pool::Applet));
+
+    // Ensure the size is aligned.
+    const size_t alignment =
+        (pool == static_cast<u32>(KMemoryManager::Pool::System) ? PageSize : SecureAlignment);
+    R_UNLESS(Common::IsAligned(size, alignment), ResultInvalidSize);
+
+    // Allocate the memory.
+    const size_t num_pages = size / PageSize;
+    const KPhysicalAddress paddr = kernel.MemoryManager().AllocateAndOpenContinuous(
+        num_pages, alignment / PageSize,
+        KMemoryManager::EncodeOption(static_cast<KMemoryManager::Pool>(pool),
+                                     KMemoryManager::Direction::FromFront));
+    R_UNLESS(paddr != 0, ResultOutOfMemory);
+
+    // Ensure we don't leak references to the memory on error.
+    ON_RESULT_FAILURE {
+        kernel.MemoryManager().Close(paddr, num_pages);
+    };
+
+    // We succeeded.
+    *out = KPageTable::GetHeapVirtualAddress(kernel.MemoryLayout(), paddr);
+    R_SUCCEED();
+}
+
+void KSystemControl::FreeSecureMemory(KernelCore& kernel, KVirtualAddress address, size_t size,
+                                      u32 pool) {
+    // Applet secure memory is handled separately.
+    UNIMPLEMENTED_IF(pool == static_cast<u32>(KMemoryManager::Pool::Applet));
+
+    // Ensure the size is aligned.
+    const size_t alignment =
+        (pool == static_cast<u32>(KMemoryManager::Pool::System) ? PageSize : SecureAlignment);
+    ASSERT(Common::IsAligned(GetInteger(address), alignment));
+    ASSERT(Common::IsAligned(size, alignment));
+
+    // Close the secure region's pages.
+    kernel.MemoryManager().Close(KPageTable::GetHeapPhysicalAddress(kernel.MemoryLayout(), address),
+                                 size / PageSize);
+}
+
 } // namespace Kernel::Board::Nintendo::Nx
diff --git a/src/core/hle/kernel/board/nintendo/nx/k_system_control.h b/src/core/hle/kernel/board/nintendo/nx/k_system_control.h
index b477e8193..ff1feec70 100644
--- a/src/core/hle/kernel/board/nintendo/nx/k_system_control.h
+++ b/src/core/hle/kernel/board/nintendo/nx/k_system_control.h
@@ -4,6 +4,11 @@
 #pragma once
 
 #include "core/hle/kernel/k_typed_address.h"
+#include "core/hle/result.h"
+
+namespace Kernel {
+class KernelCore;
+}
 
 namespace Kernel::Board::Nintendo::Nx {
 
@@ -25,8 +30,16 @@ public:
         static std::size_t GetMinimumNonSecureSystemPoolSize();
     };
 
+    // Randomness.
     static u64 GenerateRandomRange(u64 min, u64 max);
     static u64 GenerateRandomU64();
+
+    // Secure Memory.
+    static size_t CalculateRequiredSecureMemorySize(size_t size, u32 pool);
+    static Result AllocateSecureMemory(KernelCore& kernel, KVirtualAddress* out, size_t size,
+                                       u32 pool);
+    static void FreeSecureMemory(KernelCore& kernel, KVirtualAddress address, size_t size,
+                                 u32 pool);
 };
 
 } // namespace Kernel::Board::Nintendo::Nx
diff --git a/src/core/hle/kernel/init/init_slab_setup.cpp b/src/core/hle/kernel/init/init_slab_setup.cpp
index 1f2db673c..a0e20bbbb 100644
--- a/src/core/hle/kernel/init/init_slab_setup.cpp
+++ b/src/core/hle/kernel/init/init_slab_setup.cpp
@@ -106,7 +106,7 @@ static_assert(KernelPageBufferAdditionalSize ==
 /// memory.
 static KPhysicalAddress TranslateSlabAddrToPhysical(KMemoryLayout& memory_layout,
                                                     KVirtualAddress slab_addr) {
-    slab_addr -= GetInteger(memory_layout.GetSlabRegionAddress());
+    slab_addr -= memory_layout.GetSlabRegion().GetAddress();
     return GetInteger(slab_addr) + Core::DramMemoryMap::SlabHeapBase;
 }
 
@@ -196,7 +196,12 @@ void InitializeSlabHeaps(Core::System& system, KMemoryLayout& memory_layout) {
     auto& kernel = system.Kernel();
 
     // Get the start of the slab region, since that's where we'll be working.
-    KVirtualAddress address = memory_layout.GetSlabRegionAddress();
+    const KMemoryRegion& slab_region = memory_layout.GetSlabRegion();
+    KVirtualAddress address = slab_region.GetAddress();
+
+    // Clear the slab region.
+    // TODO: implement access to kernel VAs.
+    // std::memset(device_ptr, 0, slab_region.GetSize());
 
     // Initialize slab type array to be in sorted order.
     std::array<KSlabType, KSlabType_Count> slab_types;
diff --git a/src/core/hle/kernel/initial_process.h b/src/core/hle/kernel/initial_process.h
index 82195f4f7..2c95269fc 100644
--- a/src/core/hle/kernel/initial_process.h
+++ b/src/core/hle/kernel/initial_process.h
@@ -19,4 +19,8 @@ static inline KPhysicalAddress GetInitialProcessBinaryPhysicalAddress() {
         MainMemoryAddress);
 }
 
+static inline size_t GetInitialProcessBinarySize() {
+    return InitialProcessBinarySizeMax;
+}
+
 } // namespace Kernel
diff --git a/src/core/hle/kernel/k_capabilities.h b/src/core/hle/kernel/k_capabilities.h
index de766c811..ebd4eedb1 100644
--- a/src/core/hle/kernel/k_capabilities.h
+++ b/src/core/hle/kernel/k_capabilities.h
@@ -200,8 +200,8 @@ private:
 
         RawCapabilityValue raw;
         BitField<0, 15, CapabilityType> id;
-        BitField<15, 4, u32> major_version;
-        BitField<19, 13, u32> minor_version;
+        BitField<15, 4, u32> minor_version;
+        BitField<19, 13, u32> major_version;
     };
 
     union HandleTable {
diff --git a/src/core/hle/kernel/k_condition_variable.cpp b/src/core/hle/kernel/k_condition_variable.cpp
index efbac0e6a..7633a51fb 100644
--- a/src/core/hle/kernel/k_condition_variable.cpp
+++ b/src/core/hle/kernel/k_condition_variable.cpp
@@ -107,12 +107,12 @@ KConditionVariable::KConditionVariable(Core::System& system)
 
 KConditionVariable::~KConditionVariable() = default;
 
-Result KConditionVariable::SignalToAddress(KProcessAddress addr) {
-    KThread* owner_thread = GetCurrentThreadPointer(m_kernel);
+Result KConditionVariable::SignalToAddress(KernelCore& kernel, KProcessAddress addr) {
+    KThread* owner_thread = GetCurrentThreadPointer(kernel);
 
     // Signal the address.
     {
-        KScopedSchedulerLock sl(m_kernel);
+        KScopedSchedulerLock sl(kernel);
 
         // Remove waiter thread.
         bool has_waiters{};
@@ -133,7 +133,7 @@ Result KConditionVariable::SignalToAddress(KProcessAddress addr) {
 
         // Write the value to userspace.
         Result result{ResultSuccess};
-        if (WriteToUser(m_kernel, addr, std::addressof(next_value))) [[likely]] {
+        if (WriteToUser(kernel, addr, std::addressof(next_value))) [[likely]] {
             result = ResultSuccess;
         } else {
             result = ResultInvalidCurrentMemory;
@@ -148,28 +148,28 @@ Result KConditionVariable::SignalToAddress(KProcessAddress addr) {
     }
 }
 
-Result KConditionVariable::WaitForAddress(Handle handle, KProcessAddress addr, u32 value) {
-    KThread* cur_thread = GetCurrentThreadPointer(m_kernel);
-    ThreadQueueImplForKConditionVariableWaitForAddress wait_queue(m_kernel);
+Result KConditionVariable::WaitForAddress(KernelCore& kernel, Handle handle, KProcessAddress addr,
+                                          u32 value) {
+    KThread* cur_thread = GetCurrentThreadPointer(kernel);
+    ThreadQueueImplForKConditionVariableWaitForAddress wait_queue(kernel);
 
     // Wait for the address.
     KThread* owner_thread{};
     {
-        KScopedSchedulerLock sl(m_kernel);
+        KScopedSchedulerLock sl(kernel);
 
         // Check if the thread should terminate.
         R_UNLESS(!cur_thread->IsTerminationRequested(), ResultTerminationRequested);
 
         // Read the tag from userspace.
         u32 test_tag{};
-        R_UNLESS(ReadFromUser(m_kernel, std::addressof(test_tag), addr),
-                 ResultInvalidCurrentMemory);
+        R_UNLESS(ReadFromUser(kernel, std::addressof(test_tag), addr), ResultInvalidCurrentMemory);
 
         // If the tag isn't the handle (with wait mask), we're done.
         R_SUCCEED_IF(test_tag != (handle | Svc::HandleWaitMask));
 
         // Get the lock owner thread.
-        owner_thread = GetCurrentProcess(m_kernel)
+        owner_thread = GetCurrentProcess(kernel)
                            .GetHandleTable()
                            .GetObjectWithoutPseudoHandle<KThread>(handle)
                            .ReleasePointerUnsafe();
diff --git a/src/core/hle/kernel/k_condition_variable.h b/src/core/hle/kernel/k_condition_variable.h
index 8c2f3ae51..2620c8e39 100644
--- a/src/core/hle/kernel/k_condition_variable.h
+++ b/src/core/hle/kernel/k_condition_variable.h
@@ -24,11 +24,12 @@ public:
     explicit KConditionVariable(Core::System& system);
     ~KConditionVariable();
 
-    // Arbitration
-    Result SignalToAddress(KProcessAddress addr);
-    Result WaitForAddress(Handle handle, KProcessAddress addr, u32 value);
+    // Arbitration.
+    static Result SignalToAddress(KernelCore& kernel, KProcessAddress addr);
+    static Result WaitForAddress(KernelCore& kernel, Handle handle, KProcessAddress addr,
+                                 u32 value);
 
-    // Condition variable
+    // Condition variable.
     void Signal(u64 cv_key, s32 count);
     Result Wait(KProcessAddress addr, u64 key, u32 value, s64 timeout);
 
diff --git a/src/core/hle/kernel/k_interrupt_manager.cpp b/src/core/hle/kernel/k_interrupt_manager.cpp
index fe6a20168..22d79569a 100644
--- a/src/core/hle/kernel/k_interrupt_manager.cpp
+++ b/src/core/hle/kernel/k_interrupt_manager.cpp
@@ -22,7 +22,7 @@ void HandleInterrupt(KernelCore& kernel, s32 core_id) {
             KScopedSchedulerLock sl{kernel};
 
             // Pin the current thread.
-            process->PinCurrentThread(core_id);
+            process->PinCurrentThread();
 
             // Set the interrupt flag for the thread.
             GetCurrentThread(kernel).SetInterruptFlag();
diff --git a/src/core/hle/kernel/k_memory_block.h b/src/core/hle/kernel/k_memory_block.h
index 41a29da24..ef3f61321 100644
--- a/src/core/hle/kernel/k_memory_block.h
+++ b/src/core/hle/kernel/k_memory_block.h
@@ -36,6 +36,7 @@ enum class KMemoryState : u32 {
     FlagCanChangeAttribute = (1 << 24),
     FlagCanCodeMemory = (1 << 25),
     FlagLinearMapped = (1 << 26),
+    FlagCanPermissionLock = (1 << 27),
 
     FlagsData = FlagCanReprotect | FlagCanUseIpc | FlagCanUseNonDeviceIpc | FlagCanUseNonSecureIpc |
                 FlagMapped | FlagCanAlias | FlagCanTransfer | FlagCanQueryPhysical |
@@ -50,12 +51,16 @@ enum class KMemoryState : u32 {
                 FlagLinearMapped,
 
     Free = static_cast<u32>(Svc::MemoryState::Free),
-    Io = static_cast<u32>(Svc::MemoryState::Io) | FlagMapped | FlagCanDeviceMap |
-         FlagCanAlignedDeviceMap,
+
+    IoMemory = static_cast<u32>(Svc::MemoryState::Io) | FlagMapped | FlagCanDeviceMap |
+               FlagCanAlignedDeviceMap,
+    IoRegister =
+        static_cast<u32>(Svc::MemoryState::Io) | FlagCanDeviceMap | FlagCanAlignedDeviceMap,
+
     Static = static_cast<u32>(Svc::MemoryState::Static) | FlagMapped | FlagCanQueryPhysical,
     Code = static_cast<u32>(Svc::MemoryState::Code) | FlagsCode | FlagCanMapProcess,
     CodeData = static_cast<u32>(Svc::MemoryState::CodeData) | FlagsData | FlagCanMapProcess |
-               FlagCanCodeMemory,
+               FlagCanCodeMemory | FlagCanPermissionLock,
     Normal = static_cast<u32>(Svc::MemoryState::Normal) | FlagsData | FlagCanCodeMemory,
     Shared = static_cast<u32>(Svc::MemoryState::Shared) | FlagMapped | FlagReferenceCounted |
              FlagLinearMapped,
@@ -65,7 +70,8 @@ enum class KMemoryState : u32 {
     AliasCode = static_cast<u32>(Svc::MemoryState::AliasCode) | FlagsCode | FlagCanMapProcess |
                 FlagCanCodeAlias,
     AliasCodeData = static_cast<u32>(Svc::MemoryState::AliasCodeData) | FlagsData |
-                    FlagCanMapProcess | FlagCanCodeAlias | FlagCanCodeMemory,
+                    FlagCanMapProcess | FlagCanCodeAlias | FlagCanCodeMemory |
+                    FlagCanPermissionLock,
 
     Ipc = static_cast<u32>(Svc::MemoryState::Ipc) | FlagsMisc | FlagCanAlignedDeviceMap |
           FlagCanUseIpc | FlagCanUseNonSecureIpc | FlagCanUseNonDeviceIpc,
@@ -73,7 +79,7 @@ enum class KMemoryState : u32 {
     Stack = static_cast<u32>(Svc::MemoryState::Stack) | FlagsMisc | FlagCanAlignedDeviceMap |
             FlagCanUseIpc | FlagCanUseNonSecureIpc | FlagCanUseNonDeviceIpc,
 
-    ThreadLocal = static_cast<u32>(Svc::MemoryState::ThreadLocal) | FlagMapped | FlagLinearMapped,
+    ThreadLocal = static_cast<u32>(Svc::MemoryState::ThreadLocal) | FlagLinearMapped,
 
     Transfered = static_cast<u32>(Svc::MemoryState::Transfered) | FlagsMisc |
                  FlagCanAlignedDeviceMap | FlagCanChangeAttribute | FlagCanUseIpc |
@@ -94,7 +100,7 @@ enum class KMemoryState : u32 {
     NonDeviceIpc =
         static_cast<u32>(Svc::MemoryState::NonDeviceIpc) | FlagsMisc | FlagCanUseNonDeviceIpc,
 
-    Kernel = static_cast<u32>(Svc::MemoryState::Kernel) | FlagMapped,
+    Kernel = static_cast<u32>(Svc::MemoryState::Kernel),
 
     GeneratedCode = static_cast<u32>(Svc::MemoryState::GeneratedCode) | FlagMapped |
                     FlagReferenceCounted | FlagCanDebug | FlagLinearMapped,
@@ -105,34 +111,36 @@ enum class KMemoryState : u32 {
 
     Insecure = static_cast<u32>(Svc::MemoryState::Insecure) | FlagMapped | FlagReferenceCounted |
                FlagLinearMapped | FlagCanChangeAttribute | FlagCanDeviceMap |
-               FlagCanAlignedDeviceMap | FlagCanUseNonSecureIpc | FlagCanUseNonDeviceIpc,
+               FlagCanAlignedDeviceMap | FlagCanQueryPhysical | FlagCanUseNonSecureIpc |
+               FlagCanUseNonDeviceIpc,
 };
 DECLARE_ENUM_FLAG_OPERATORS(KMemoryState);
 
 static_assert(static_cast<u32>(KMemoryState::Free) == 0x00000000);
-static_assert(static_cast<u32>(KMemoryState::Io) == 0x00182001);
+static_assert(static_cast<u32>(KMemoryState::IoMemory) == 0x00182001);
+static_assert(static_cast<u32>(KMemoryState::IoRegister) == 0x00180001);
 static_assert(static_cast<u32>(KMemoryState::Static) == 0x00042002);
 static_assert(static_cast<u32>(KMemoryState::Code) == 0x04DC7E03);
-static_assert(static_cast<u32>(KMemoryState::CodeData) == 0x07FEBD04);
+static_assert(static_cast<u32>(KMemoryState::CodeData) == 0x0FFEBD04);
 static_assert(static_cast<u32>(KMemoryState::Normal) == 0x077EBD05);
 static_assert(static_cast<u32>(KMemoryState::Shared) == 0x04402006);
 
 static_assert(static_cast<u32>(KMemoryState::AliasCode) == 0x04DD7E08);
-static_assert(static_cast<u32>(KMemoryState::AliasCodeData) == 0x07FFBD09);
+static_assert(static_cast<u32>(KMemoryState::AliasCodeData) == 0x0FFFBD09);
 static_assert(static_cast<u32>(KMemoryState::Ipc) == 0x045C3C0A);
 static_assert(static_cast<u32>(KMemoryState::Stack) == 0x045C3C0B);
-static_assert(static_cast<u32>(KMemoryState::ThreadLocal) == 0x0400200C);
+static_assert(static_cast<u32>(KMemoryState::ThreadLocal) == 0x0400000C);
 static_assert(static_cast<u32>(KMemoryState::Transfered) == 0x055C3C0D);
 static_assert(static_cast<u32>(KMemoryState::SharedTransfered) == 0x045C380E);
 static_assert(static_cast<u32>(KMemoryState::SharedCode) == 0x0440380F);
 static_assert(static_cast<u32>(KMemoryState::Inaccessible) == 0x00000010);
 static_assert(static_cast<u32>(KMemoryState::NonSecureIpc) == 0x045C3811);
 static_assert(static_cast<u32>(KMemoryState::NonDeviceIpc) == 0x044C2812);
-static_assert(static_cast<u32>(KMemoryState::Kernel) == 0x00002013);
+static_assert(static_cast<u32>(KMemoryState::Kernel) == 0x00000013);
 static_assert(static_cast<u32>(KMemoryState::GeneratedCode) == 0x04402214);
 static_assert(static_cast<u32>(KMemoryState::CodeOut) == 0x04402015);
 static_assert(static_cast<u32>(KMemoryState::Coverage) == 0x00002016);
-static_assert(static_cast<u32>(KMemoryState::Insecure) == 0x05583817);
+static_assert(static_cast<u32>(KMemoryState::Insecure) == 0x055C3817);
 
 enum class KMemoryPermission : u8 {
     None = 0,
@@ -182,8 +190,9 @@ enum class KMemoryAttribute : u8 {
     IpcLocked = static_cast<u8>(Svc::MemoryAttribute::IpcLocked),
     DeviceShared = static_cast<u8>(Svc::MemoryAttribute::DeviceShared),
     Uncached = static_cast<u8>(Svc::MemoryAttribute::Uncached),
+    PermissionLocked = static_cast<u8>(Svc::MemoryAttribute::PermissionLocked),
 
-    SetMask = Uncached,
+    SetMask = Uncached | PermissionLocked,
 };
 DECLARE_ENUM_FLAG_OPERATORS(KMemoryAttribute);
 
@@ -261,6 +270,10 @@ struct KMemoryInfo {
         return m_state;
     }
 
+    constexpr Svc::MemoryState GetSvcState() const {
+        return static_cast<Svc::MemoryState>(m_state & KMemoryState::Mask);
+    }
+
     constexpr KMemoryPermission GetPermission() const {
         return m_permission;
     }
@@ -326,6 +339,10 @@ public:
         return this->GetEndAddress() - 1;
     }
 
+    constexpr KMemoryState GetState() const {
+        return m_memory_state;
+    }
+
     constexpr u16 GetIpcLockCount() const {
         return m_ipc_lock_count;
     }
@@ -443,6 +460,13 @@ public:
         }
     }
 
+    constexpr void UpdateAttribute(KMemoryAttribute mask, KMemoryAttribute attr) {
+        ASSERT(False(mask & KMemoryAttribute::IpcLocked));
+        ASSERT(False(mask & KMemoryAttribute::DeviceShared));
+
+        m_attribute = (m_attribute & ~mask) | attr;
+    }
+
     constexpr void Split(KMemoryBlock* block, KProcessAddress addr) {
         ASSERT(this->GetAddress() < addr);
         ASSERT(this->Contains(addr));
diff --git a/src/core/hle/kernel/k_memory_block_manager.cpp b/src/core/hle/kernel/k_memory_block_manager.cpp
index ab75f550e..58a1e7216 100644
--- a/src/core/hle/kernel/k_memory_block_manager.cpp
+++ b/src/core/hle/kernel/k_memory_block_manager.cpp
@@ -160,8 +160,8 @@ void KMemoryBlockManager::Update(KMemoryBlockManagerUpdateAllocator* allocator,
             }
 
             // Update block state.
-            it->Update(state, perm, attr, cur_address == address, static_cast<u8>(set_disable_attr),
-                       static_cast<u8>(clear_disable_attr));
+            it->Update(state, perm, attr, it->GetAddress() == address,
+                       static_cast<u8>(set_disable_attr), static_cast<u8>(clear_disable_attr));
             cur_address += cur_info.GetSize();
             remaining_pages -= cur_info.GetNumPages();
         }
@@ -175,7 +175,9 @@ void KMemoryBlockManager::UpdateIfMatch(KMemoryBlockManagerUpdateAllocator* allo
                                         KProcessAddress address, size_t num_pages,
                                         KMemoryState test_state, KMemoryPermission test_perm,
                                         KMemoryAttribute test_attr, KMemoryState state,
-                                        KMemoryPermission perm, KMemoryAttribute attr) {
+                                        KMemoryPermission perm, KMemoryAttribute attr,
+                                        KMemoryBlockDisableMergeAttribute set_disable_attr,
+                                        KMemoryBlockDisableMergeAttribute clear_disable_attr) {
     // Ensure for auditing that we never end up with an invalid tree.
     KScopedMemoryBlockManagerAuditor auditor(this);
     ASSERT(Common::IsAligned(GetInteger(address), PageSize));
@@ -214,7 +216,8 @@ void KMemoryBlockManager::UpdateIfMatch(KMemoryBlockManagerUpdateAllocator* allo
             }
 
             // Update block state.
-            it->Update(state, perm, attr, false, 0, 0);
+            it->Update(state, perm, attr, false, static_cast<u8>(set_disable_attr),
+                       static_cast<u8>(clear_disable_attr));
             cur_address += cur_info.GetSize();
             remaining_pages -= cur_info.GetNumPages();
         } else {
@@ -284,6 +287,65 @@ void KMemoryBlockManager::UpdateLock(KMemoryBlockManagerUpdateAllocator* allocat
     this->CoalesceForUpdate(allocator, address, num_pages);
 }
 
+void KMemoryBlockManager::UpdateAttribute(KMemoryBlockManagerUpdateAllocator* allocator,
+                                          KProcessAddress address, size_t num_pages,
+                                          KMemoryAttribute mask, KMemoryAttribute attr) {
+    // Ensure for auditing that we never end up with an invalid tree.
+    KScopedMemoryBlockManagerAuditor auditor(this);
+    ASSERT(Common::IsAligned(GetInteger(address), PageSize));
+
+    KProcessAddress cur_address = address;
+    size_t remaining_pages = num_pages;
+    iterator it = this->FindIterator(address);
+
+    while (remaining_pages > 0) {
+        const size_t remaining_size = remaining_pages * PageSize;
+        KMemoryInfo cur_info = it->GetMemoryInfo();
+
+        if ((it->GetAttribute() & mask) != attr) {
+            // If we need to, create a new block before and insert it.
+            if (cur_info.GetAddress() != GetInteger(cur_address)) {
+                KMemoryBlock* new_block = allocator->Allocate();
+
+                it->Split(new_block, cur_address);
+                it = m_memory_block_tree.insert(*new_block);
+                it++;
+
+                cur_info = it->GetMemoryInfo();
+                cur_address = cur_info.GetAddress();
+            }
+
+            // If we need to, create a new block after and insert it.
+            if (cur_info.GetSize() > remaining_size) {
+                KMemoryBlock* new_block = allocator->Allocate();
+
+                it->Split(new_block, cur_address + remaining_size);
+                it = m_memory_block_tree.insert(*new_block);
+
+                cur_info = it->GetMemoryInfo();
+            }
+
+            // Update block state.
+            it->UpdateAttribute(mask, attr);
+            cur_address += cur_info.GetSize();
+            remaining_pages -= cur_info.GetNumPages();
+        } else {
+            // If we already have the right attributes, just advance.
+            if (cur_address + remaining_size < cur_info.GetEndAddress()) {
+                remaining_pages = 0;
+                cur_address += remaining_size;
+            } else {
+                remaining_pages =
+                    (cur_address + remaining_size - cur_info.GetEndAddress()) / PageSize;
+                cur_address = cur_info.GetEndAddress();
+            }
+        }
+        it++;
+    }
+
+    this->CoalesceForUpdate(allocator, address, num_pages);
+}
+
 // Debug.
 bool KMemoryBlockManager::CheckState() const {
     // Loop over every block, ensuring that we are sorted and coalesced.
diff --git a/src/core/hle/kernel/k_memory_block_manager.h b/src/core/hle/kernel/k_memory_block_manager.h
index 96496e990..cb7b6f430 100644
--- a/src/core/hle/kernel/k_memory_block_manager.h
+++ b/src/core/hle/kernel/k_memory_block_manager.h
@@ -115,7 +115,11 @@ public:
     void UpdateIfMatch(KMemoryBlockManagerUpdateAllocator* allocator, KProcessAddress address,
                        size_t num_pages, KMemoryState test_state, KMemoryPermission test_perm,
                        KMemoryAttribute test_attr, KMemoryState state, KMemoryPermission perm,
-                       KMemoryAttribute attr);
+                       KMemoryAttribute attr, KMemoryBlockDisableMergeAttribute set_disable_attr,
+                       KMemoryBlockDisableMergeAttribute clear_disable_attr);
+
+    void UpdateAttribute(KMemoryBlockManagerUpdateAllocator* allocator, KProcessAddress address,
+                         size_t num_pages, KMemoryAttribute mask, KMemoryAttribute attr);
 
     iterator FindIterator(KProcessAddress address) const {
         return m_memory_block_tree.find(KMemoryBlock(
diff --git a/src/core/hle/kernel/k_memory_layout.h b/src/core/hle/kernel/k_memory_layout.h
index 54a71df56..c8122644f 100644
--- a/src/core/hle/kernel/k_memory_layout.h
+++ b/src/core/hle/kernel/k_memory_layout.h
@@ -137,11 +137,9 @@ public:
         return GetStackTopAddress(core_id, KMemoryRegionType_KernelMiscExceptionStack);
     }
 
-    KVirtualAddress GetSlabRegionAddress() const {
-        return Dereference(GetVirtualMemoryRegionTree().FindByType(KMemoryRegionType_KernelSlab))
-            .GetAddress();
+    const KMemoryRegion& GetSlabRegion() const {
+        return Dereference(GetVirtualMemoryRegionTree().FindByType(KMemoryRegionType_KernelSlab));
     }
-
     const KMemoryRegion& GetDeviceRegion(KMemoryRegionType type) const {
         return Dereference(GetPhysicalMemoryRegionTree().FindFirstDerived(type));
     }
diff --git a/src/core/hle/kernel/k_memory_manager.cpp b/src/core/hle/kernel/k_memory_manager.cpp
index 74d8169e0..cdc5572d8 100644
--- a/src/core/hle/kernel/k_memory_manager.cpp
+++ b/src/core/hle/kernel/k_memory_manager.cpp
@@ -11,6 +11,7 @@
 #include "core/hle/kernel/initial_process.h"
 #include "core/hle/kernel/k_memory_manager.h"
 #include "core/hle/kernel/k_page_group.h"
+#include "core/hle/kernel/k_page_table.h"
 #include "core/hle/kernel/kernel.h"
 #include "core/hle/kernel/svc_results.h"
 
@@ -119,7 +120,8 @@ void KMemoryManager::Initialize(KVirtualAddress management_region, size_t manage
     // Free each region to its corresponding heap.
     size_t reserved_sizes[MaxManagerCount] = {};
     const KPhysicalAddress ini_start = GetInitialProcessBinaryPhysicalAddress();
-    const KPhysicalAddress ini_end = ini_start + InitialProcessBinarySizeMax;
+    const size_t ini_size = GetInitialProcessBinarySize();
+    const KPhysicalAddress ini_end = ini_start + ini_size;
     const KPhysicalAddress ini_last = ini_end - 1;
     for (const auto& it : m_system.Kernel().MemoryLayout().GetPhysicalMemoryRegionTree()) {
         if (it.IsDerivedFrom(KMemoryRegionType_DramUserPool)) {
@@ -137,13 +139,13 @@ void KMemoryManager::Initialize(KVirtualAddress management_region, size_t manage
                 }
 
                 // Open/reserve the ini memory.
-                manager.OpenFirst(ini_start, InitialProcessBinarySizeMax / PageSize);
-                reserved_sizes[it.GetAttributes()] += InitialProcessBinarySizeMax;
+                manager.OpenFirst(ini_start, ini_size / PageSize);
+                reserved_sizes[it.GetAttributes()] += ini_size;
 
                 // Free memory after the ini to the heap.
                 if (ini_last != cur_last) {
                     ASSERT(cur_end != 0);
-                    manager.Free(ini_end, cur_end - ini_end);
+                    manager.Free(ini_end, (cur_end - ini_end) / PageSize);
                 }
             } else {
                 // Ensure there's no partial overlap with the ini image.
@@ -167,11 +169,37 @@ void KMemoryManager::Initialize(KVirtualAddress management_region, size_t manage
 }
 
 Result KMemoryManager::InitializeOptimizedMemory(u64 process_id, Pool pool) {
-    UNREACHABLE();
+    const u32 pool_index = static_cast<u32>(pool);
+
+    // Lock the pool.
+    KScopedLightLock lk(m_pool_locks[pool_index]);
+
+    // Check that we don't already have an optimized process.
+    R_UNLESS(!m_has_optimized_process[pool_index], ResultBusy);
+
+    // Set the optimized process id.
+    m_optimized_process_ids[pool_index] = process_id;
+    m_has_optimized_process[pool_index] = true;
+
+    // Clear the management area for the optimized process.
+    for (auto* manager = this->GetFirstManager(pool, Direction::FromFront); manager != nullptr;
+         manager = this->GetNextManager(manager, Direction::FromFront)) {
+        manager->InitializeOptimizedMemory(m_system.Kernel());
+    }
+
+    R_SUCCEED();
 }
 
 void KMemoryManager::FinalizeOptimizedMemory(u64 process_id, Pool pool) {
-    UNREACHABLE();
+    const u32 pool_index = static_cast<u32>(pool);
+
+    // Lock the pool.
+    KScopedLightLock lk(m_pool_locks[pool_index]);
+
+    // If the process was optimized, clear it.
+    if (m_has_optimized_process[pool_index] && m_optimized_process_ids[pool_index] == process_id) {
+        m_has_optimized_process[pool_index] = false;
+    }
 }
 
 KPhysicalAddress KMemoryManager::AllocateAndOpenContinuous(size_t num_pages, size_t align_pages,
@@ -206,7 +234,7 @@ KPhysicalAddress KMemoryManager::AllocateAndOpenContinuous(size_t num_pages, siz
 
     // Maintain the optimized memory bitmap, if we should.
     if (m_has_optimized_process[static_cast<size_t>(pool)]) {
-        UNIMPLEMENTED();
+        chosen_manager->TrackUnoptimizedAllocation(m_system.Kernel(), allocated_block, num_pages);
     }
 
     // Open the first reference to the pages.
@@ -254,7 +282,8 @@ Result KMemoryManager::AllocatePageGroupImpl(KPageGroup* out, size_t num_pages,
 
                 // Maintain the optimized memory bitmap, if we should.
                 if (unoptimized) {
-                    UNIMPLEMENTED();
+                    cur_manager->TrackUnoptimizedAllocation(m_system.Kernel(), allocated_block,
+                                                            pages_per_alloc);
                 }
 
                 num_pages -= pages_per_alloc;
@@ -357,8 +386,8 @@ Result KMemoryManager::AllocateForProcess(KPageGroup* out, size_t num_pages, u32
                     // Process part or all of the block.
                     const size_t cur_pages =
                         std::min(remaining_pages, manager.GetPageOffsetToEnd(cur_address));
-                    any_new =
-                        manager.ProcessOptimizedAllocation(cur_address, cur_pages, fill_pattern);
+                    any_new = manager.ProcessOptimizedAllocation(m_system.Kernel(), cur_address,
+                                                                 cur_pages, fill_pattern);
 
                     // Advance.
                     cur_address += cur_pages * PageSize;
@@ -381,7 +410,7 @@ Result KMemoryManager::AllocateForProcess(KPageGroup* out, size_t num_pages, u32
                     // Track some or all of the current pages.
                     const size_t cur_pages =
                         std::min(remaining_pages, manager.GetPageOffsetToEnd(cur_address));
-                    manager.TrackOptimizedAllocation(cur_address, cur_pages);
+                    manager.TrackOptimizedAllocation(m_system.Kernel(), cur_address, cur_pages);
 
                     // Advance.
                     cur_address += cur_pages * PageSize;
@@ -426,17 +455,86 @@ size_t KMemoryManager::Impl::Initialize(KPhysicalAddress address, size_t size,
     return total_management_size;
 }
 
-void KMemoryManager::Impl::TrackUnoptimizedAllocation(KPhysicalAddress block, size_t num_pages) {
-    UNREACHABLE();
+void KMemoryManager::Impl::InitializeOptimizedMemory(KernelCore& kernel) {
+    auto optimize_pa =
+        KPageTable::GetHeapPhysicalAddress(kernel.MemoryLayout(), m_management_region);
+    auto* optimize_map = kernel.System().DeviceMemory().GetPointer<u64>(optimize_pa);
+
+    std::memset(optimize_map, 0, CalculateOptimizedProcessOverheadSize(m_heap.GetSize()));
 }
 
-void KMemoryManager::Impl::TrackOptimizedAllocation(KPhysicalAddress block, size_t num_pages) {
-    UNREACHABLE();
+void KMemoryManager::Impl::TrackUnoptimizedAllocation(KernelCore& kernel, KPhysicalAddress block,
+                                                      size_t num_pages) {
+    auto optimize_pa =
+        KPageTable::GetHeapPhysicalAddress(kernel.MemoryLayout(), m_management_region);
+    auto* optimize_map = kernel.System().DeviceMemory().GetPointer<u64>(optimize_pa);
+
+    // Get the range we're tracking.
+    size_t offset = this->GetPageOffset(block);
+    const size_t last = offset + num_pages - 1;
+
+    // Track.
+    while (offset <= last) {
+        // Mark the page as not being optimized-allocated.
+        optimize_map[offset / Common::BitSize<u64>()] &=
+            ~(u64(1) << (offset % Common::BitSize<u64>()));
+
+        offset++;
+    }
+}
+
+void KMemoryManager::Impl::TrackOptimizedAllocation(KernelCore& kernel, KPhysicalAddress block,
+                                                    size_t num_pages) {
+    auto optimize_pa =
+        KPageTable::GetHeapPhysicalAddress(kernel.MemoryLayout(), m_management_region);
+    auto* optimize_map = kernel.System().DeviceMemory().GetPointer<u64>(optimize_pa);
+
+    // Get the range we're tracking.
+    size_t offset = this->GetPageOffset(block);
+    const size_t last = offset + num_pages - 1;
+
+    // Track.
+    while (offset <= last) {
+        // Mark the page as being optimized-allocated.
+        optimize_map[offset / Common::BitSize<u64>()] |=
+            (u64(1) << (offset % Common::BitSize<u64>()));
+
+        offset++;
+    }
 }
 
-bool KMemoryManager::Impl::ProcessOptimizedAllocation(KPhysicalAddress block, size_t num_pages,
-                                                      u8 fill_pattern) {
-    UNREACHABLE();
+bool KMemoryManager::Impl::ProcessOptimizedAllocation(KernelCore& kernel, KPhysicalAddress block,
+                                                      size_t num_pages, u8 fill_pattern) {
+    auto& device_memory = kernel.System().DeviceMemory();
+    auto optimize_pa =
+        KPageTable::GetHeapPhysicalAddress(kernel.MemoryLayout(), m_management_region);
+    auto* optimize_map = device_memory.GetPointer<u64>(optimize_pa);
+
+    // We want to return whether any pages were newly allocated.
+    bool any_new = false;
+
+    // Get the range we're processing.
+    size_t offset = this->GetPageOffset(block);
+    const size_t last = offset + num_pages - 1;
+
+    // Process.
+    while (offset <= last) {
+        // Check if the page has been optimized-allocated before.
+        if ((optimize_map[offset / Common::BitSize<u64>()] &
+             (u64(1) << (offset % Common::BitSize<u64>()))) == 0) {
+            // If not, it's new.
+            any_new = true;
+
+            // Fill the page.
+            auto* ptr = device_memory.GetPointer<u8>(m_heap.GetAddress());
+            std::memset(ptr + offset * PageSize, fill_pattern, PageSize);
+        }
+
+        offset++;
+    }
+
+    // Return the number of pages we processed.
+    return any_new;
 }
 
 size_t KMemoryManager::Impl::CalculateManagementOverheadSize(size_t region_size) {
diff --git a/src/core/hle/kernel/k_memory_manager.h b/src/core/hle/kernel/k_memory_manager.h
index 7e4b41319..c5a487af9 100644
--- a/src/core/hle/kernel/k_memory_manager.h
+++ b/src/core/hle/kernel/k_memory_manager.h
@@ -216,14 +216,14 @@ private:
             m_heap.SetInitialUsedSize(reserved_size);
         }
 
-        void InitializeOptimizedMemory() {
-            UNIMPLEMENTED();
-        }
+        void InitializeOptimizedMemory(KernelCore& kernel);
 
-        void TrackUnoptimizedAllocation(KPhysicalAddress block, size_t num_pages);
-        void TrackOptimizedAllocation(KPhysicalAddress block, size_t num_pages);
+        void TrackUnoptimizedAllocation(KernelCore& kernel, KPhysicalAddress block,
+                                        size_t num_pages);
+        void TrackOptimizedAllocation(KernelCore& kernel, KPhysicalAddress block, size_t num_pages);
 
-        bool ProcessOptimizedAllocation(KPhysicalAddress block, size_t num_pages, u8 fill_pattern);
+        bool ProcessOptimizedAllocation(KernelCore& kernel, KPhysicalAddress block,
+                                        size_t num_pages, u8 fill_pattern);
 
         constexpr Pool GetPool() const {
             return m_pool;
diff --git a/src/core/hle/kernel/k_memory_region_type.h b/src/core/hle/kernel/k_memory_region_type.h
index e5630c1ac..bcbf450f0 100644
--- a/src/core/hle/kernel/k_memory_region_type.h
+++ b/src/core/hle/kernel/k_memory_region_type.h
@@ -190,9 +190,15 @@ static_assert(KMemoryRegionType_DramKernelInitPt.GetValue() ==
 constexpr inline auto KMemoryRegionType_DramKernelSecureAppletMemory =
     KMemoryRegionType_DramKernelBase.DeriveSparse(1, 3, 0).SetAttribute(
         KMemoryRegionAttr_LinearMapped);
+constexpr inline const auto KMemoryRegionType_DramKernelSecureUnknown =
+    KMemoryRegionType_DramKernelBase.DeriveSparse(1, 3, 1).SetAttribute(
+        KMemoryRegionAttr_LinearMapped);
 static_assert(KMemoryRegionType_DramKernelSecureAppletMemory.GetValue() ==
               (0x18E | KMemoryRegionAttr_CarveoutProtected | KMemoryRegionAttr_NoUserMap |
                KMemoryRegionAttr_LinearMapped));
+static_assert(KMemoryRegionType_DramKernelSecureUnknown.GetValue() ==
+              (0x28E | KMemoryRegionAttr_CarveoutProtected | KMemoryRegionAttr_NoUserMap |
+               KMemoryRegionAttr_LinearMapped));
 
 constexpr inline auto KMemoryRegionType_DramReservedEarly =
     KMemoryRegionType_DramReservedBase.DeriveAttribute(KMemoryRegionAttr_NoUserMap);
@@ -217,16 +223,18 @@ constexpr inline auto KMemoryRegionType_DramPoolPartition =
 static_assert(KMemoryRegionType_DramPoolPartition.GetValue() ==
               (0x26 | KMemoryRegionAttr_LinearMapped | KMemoryRegionAttr_NoUserMap));
 
-constexpr inline auto KMemoryRegionType_DramPoolManagement =
-    KMemoryRegionType_DramPoolPartition.DeriveTransition(0, 2).DeriveTransition().SetAttribute(
+// UNUSED: .Derive(4, 1);
+// UNUSED: .Derive(4, 2);
+constexpr inline const auto KMemoryRegionType_DramPoolManagement =
+    KMemoryRegionType_DramPoolPartition.Derive(4, 0).SetAttribute(
         KMemoryRegionAttr_CarveoutProtected);
-constexpr inline auto KMemoryRegionType_DramUserPool =
-    KMemoryRegionType_DramPoolPartition.DeriveTransition(1, 2).DeriveTransition();
+constexpr inline const auto KMemoryRegionType_DramUserPool =
+    KMemoryRegionType_DramPoolPartition.Derive(4, 3);
 static_assert(KMemoryRegionType_DramPoolManagement.GetValue() ==
-              (0x166 | KMemoryRegionAttr_LinearMapped | KMemoryRegionAttr_NoUserMap |
+              (0xE6 | KMemoryRegionAttr_LinearMapped | KMemoryRegionAttr_NoUserMap |
                KMemoryRegionAttr_CarveoutProtected));
 static_assert(KMemoryRegionType_DramUserPool.GetValue() ==
-              (0x1A6 | KMemoryRegionAttr_LinearMapped | KMemoryRegionAttr_NoUserMap));
+              (0x266 | KMemoryRegionAttr_LinearMapped | KMemoryRegionAttr_NoUserMap));
 
 constexpr inline auto KMemoryRegionType_DramApplicationPool =
     KMemoryRegionType_DramUserPool.Derive(4, 0);
@@ -237,60 +245,63 @@ constexpr inline auto KMemoryRegionType_DramSystemNonSecurePool =
 constexpr inline auto KMemoryRegionType_DramSystemPool =
     KMemoryRegionType_DramUserPool.Derive(4, 3).SetAttribute(KMemoryRegionAttr_CarveoutProtected);
 static_assert(KMemoryRegionType_DramApplicationPool.GetValue() ==
-              (0x7A6 | KMemoryRegionAttr_LinearMapped | KMemoryRegionAttr_NoUserMap));
+              (0xE66 | KMemoryRegionAttr_LinearMapped | KMemoryRegionAttr_NoUserMap));
 static_assert(KMemoryRegionType_DramAppletPool.GetValue() ==
-              (0xBA6 | KMemoryRegionAttr_LinearMapped | KMemoryRegionAttr_NoUserMap));
+              (0x1666 | KMemoryRegionAttr_LinearMapped | KMemoryRegionAttr_NoUserMap));
 static_assert(KMemoryRegionType_DramSystemNonSecurePool.GetValue() ==
-              (0xDA6 | KMemoryRegionAttr_LinearMapped | KMemoryRegionAttr_NoUserMap));
+              (0x1A66 | KMemoryRegionAttr_LinearMapped | KMemoryRegionAttr_NoUserMap));
 static_assert(KMemoryRegionType_DramSystemPool.GetValue() ==
-              (0x13A6 | KMemoryRegionAttr_LinearMapped | KMemoryRegionAttr_NoUserMap |
+              (0x2666 | KMemoryRegionAttr_LinearMapped | KMemoryRegionAttr_NoUserMap |
                KMemoryRegionAttr_CarveoutProtected));
 
 constexpr inline auto KMemoryRegionType_VirtualDramHeapBase =
-    KMemoryRegionType_Dram.DeriveSparse(1, 3, 0);
+    KMemoryRegionType_Dram.DeriveSparse(1, 4, 0);
 constexpr inline auto KMemoryRegionType_VirtualDramKernelPtHeap =
-    KMemoryRegionType_Dram.DeriveSparse(1, 3, 1);
+    KMemoryRegionType_Dram.DeriveSparse(1, 4, 1);
 constexpr inline auto KMemoryRegionType_VirtualDramKernelTraceBuffer =
-    KMemoryRegionType_Dram.DeriveSparse(1, 3, 2);
+    KMemoryRegionType_Dram.DeriveSparse(1, 4, 2);
 static_assert(KMemoryRegionType_VirtualDramHeapBase.GetValue() == 0x1A);
 static_assert(KMemoryRegionType_VirtualDramKernelPtHeap.GetValue() == 0x2A);
 static_assert(KMemoryRegionType_VirtualDramKernelTraceBuffer.GetValue() == 0x4A);
 
-// UNUSED: .DeriveSparse(2, 2, 0);
-constexpr inline auto KMemoryRegionType_VirtualDramUnknownDebug =
-    KMemoryRegionType_Dram.DeriveSparse(2, 2, 1);
-static_assert(KMemoryRegionType_VirtualDramUnknownDebug.GetValue() == (0x52));
-
-constexpr inline auto KMemoryRegionType_VirtualDramKernelSecureAppletMemory =
-    KMemoryRegionType_Dram.DeriveSparse(3, 1, 0);
-static_assert(KMemoryRegionType_VirtualDramKernelSecureAppletMemory.GetValue() == (0x62));
-
-constexpr inline auto KMemoryRegionType_VirtualDramKernelInitPt =
-    KMemoryRegionType_VirtualDramHeapBase.Derive(3, 0);
-constexpr inline auto KMemoryRegionType_VirtualDramPoolManagement =
-    KMemoryRegionType_VirtualDramHeapBase.Derive(3, 1);
-constexpr inline auto KMemoryRegionType_VirtualDramUserPool =
-    KMemoryRegionType_VirtualDramHeapBase.Derive(3, 2);
-static_assert(KMemoryRegionType_VirtualDramKernelInitPt.GetValue() == 0x19A);
-static_assert(KMemoryRegionType_VirtualDramPoolManagement.GetValue() == 0x29A);
-static_assert(KMemoryRegionType_VirtualDramUserPool.GetValue() == 0x31A);
+// UNUSED: .Derive(4, 2);
+constexpr inline const auto KMemoryRegionType_VirtualDramUnknownDebug =
+    KMemoryRegionType_Dram.Advance(2).Derive(4, 0);
+constexpr inline const auto KMemoryRegionType_VirtualDramKernelSecureAppletMemory =
+    KMemoryRegionType_Dram.Advance(2).Derive(4, 1);
+constexpr inline const auto KMemoryRegionType_VirtualDramKernelSecureUnknown =
+    KMemoryRegionType_Dram.Advance(2).Derive(4, 3);
+static_assert(KMemoryRegionType_VirtualDramUnknownDebug.GetValue() == (0x32));
+static_assert(KMemoryRegionType_VirtualDramKernelSecureAppletMemory.GetValue() == (0x52));
+static_assert(KMemoryRegionType_VirtualDramKernelSecureUnknown.GetValue() == (0x92));
+
+// UNUSED: .Derive(4, 3);
+constexpr inline const auto KMemoryRegionType_VirtualDramKernelInitPt =
+    KMemoryRegionType_VirtualDramHeapBase.Derive(4, 0);
+constexpr inline const auto KMemoryRegionType_VirtualDramPoolManagement =
+    KMemoryRegionType_VirtualDramHeapBase.Derive(4, 1);
+constexpr inline const auto KMemoryRegionType_VirtualDramUserPool =
+    KMemoryRegionType_VirtualDramHeapBase.Derive(4, 2);
+static_assert(KMemoryRegionType_VirtualDramKernelInitPt.GetValue() == 0x31A);
+static_assert(KMemoryRegionType_VirtualDramPoolManagement.GetValue() == 0x51A);
+static_assert(KMemoryRegionType_VirtualDramUserPool.GetValue() == 0x61A);
 
 // NOTE: For unknown reason, the pools are derived out-of-order here.
 // It's worth eventually trying to understand why Nintendo made this choice.
 // UNUSED: .Derive(6, 0);
 // UNUSED: .Derive(6, 1);
-constexpr inline auto KMemoryRegionType_VirtualDramAppletPool =
-    KMemoryRegionType_VirtualDramUserPool.Derive(6, 2);
-constexpr inline auto KMemoryRegionType_VirtualDramApplicationPool =
-    KMemoryRegionType_VirtualDramUserPool.Derive(6, 3);
-constexpr inline auto KMemoryRegionType_VirtualDramSystemNonSecurePool =
-    KMemoryRegionType_VirtualDramUserPool.Derive(6, 4);
-constexpr inline auto KMemoryRegionType_VirtualDramSystemPool =
-    KMemoryRegionType_VirtualDramUserPool.Derive(6, 5);
-static_assert(KMemoryRegionType_VirtualDramAppletPool.GetValue() == 0x1B1A);
-static_assert(KMemoryRegionType_VirtualDramApplicationPool.GetValue() == 0x271A);
-static_assert(KMemoryRegionType_VirtualDramSystemNonSecurePool.GetValue() == 0x2B1A);
-static_assert(KMemoryRegionType_VirtualDramSystemPool.GetValue() == 0x331A);
+constexpr inline const auto KMemoryRegionType_VirtualDramApplicationPool =
+    KMemoryRegionType_VirtualDramUserPool.Derive(4, 0);
+constexpr inline const auto KMemoryRegionType_VirtualDramAppletPool =
+    KMemoryRegionType_VirtualDramUserPool.Derive(4, 1);
+constexpr inline const auto KMemoryRegionType_VirtualDramSystemNonSecurePool =
+    KMemoryRegionType_VirtualDramUserPool.Derive(4, 2);
+constexpr inline const auto KMemoryRegionType_VirtualDramSystemPool =
+    KMemoryRegionType_VirtualDramUserPool.Derive(4, 3);
+static_assert(KMemoryRegionType_VirtualDramApplicationPool.GetValue() == 0x361A);
+static_assert(KMemoryRegionType_VirtualDramAppletPool.GetValue() == 0x561A);
+static_assert(KMemoryRegionType_VirtualDramSystemNonSecurePool.GetValue() == 0x661A);
+static_assert(KMemoryRegionType_VirtualDramSystemPool.GetValue() == 0x961A);
 
 constexpr inline auto KMemoryRegionType_ArchDeviceBase =
     KMemoryRegionType_Kernel.DeriveTransition(0, 1).SetSparseOnly();
@@ -354,12 +365,14 @@ constexpr inline auto KMemoryRegionType_KernelTemp =
 static_assert(KMemoryRegionType_KernelTemp.GetValue() == 0x31);
 
 constexpr KMemoryRegionType GetTypeForVirtualLinearMapping(u32 type_id) {
-    if (KMemoryRegionType_KernelTraceBuffer.IsAncestorOf(type_id)) {
-        return KMemoryRegionType_VirtualDramKernelTraceBuffer;
-    } else if (KMemoryRegionType_DramKernelPtHeap.IsAncestorOf(type_id)) {
+    if (KMemoryRegionType_DramKernelPtHeap.IsAncestorOf(type_id)) {
         return KMemoryRegionType_VirtualDramKernelPtHeap;
     } else if (KMemoryRegionType_DramKernelSecureAppletMemory.IsAncestorOf(type_id)) {
         return KMemoryRegionType_VirtualDramKernelSecureAppletMemory;
+    } else if (KMemoryRegionType_DramKernelSecureUnknown.IsAncestorOf(type_id)) {
+        return KMemoryRegionType_VirtualDramKernelSecureUnknown;
+    } else if (KMemoryRegionType_KernelTraceBuffer.IsAncestorOf(type_id)) {
+        return KMemoryRegionType_VirtualDramKernelTraceBuffer;
     } else if ((type_id | KMemoryRegionAttr_ShouldKernelMap) == type_id) {
         return KMemoryRegionType_VirtualDramUnknownDebug;
     } else {
diff --git a/src/core/hle/kernel/k_page_group.h b/src/core/hle/kernel/k_page_group.h
index b32909f05..de9d63a8d 100644
--- a/src/core/hle/kernel/k_page_group.h
+++ b/src/core/hle/kernel/k_page_group.h
@@ -183,12 +183,17 @@ private:
 
 class KScopedPageGroup {
 public:
-    explicit KScopedPageGroup(const KPageGroup* gp) : m_pg(gp) {
+    explicit KScopedPageGroup(const KPageGroup* gp, bool not_first = true) : m_pg(gp) {
         if (m_pg) {
-            m_pg->Open();
+            if (not_first) {
+                m_pg->Open();
+            } else {
+                m_pg->OpenFirst();
+            }
         }
     }
-    explicit KScopedPageGroup(const KPageGroup& gp) : KScopedPageGroup(std::addressof(gp)) {}
+    explicit KScopedPageGroup(const KPageGroup& gp, bool not_first = true)
+        : KScopedPageGroup(std::addressof(gp), not_first) {}
     ~KScopedPageGroup() {
         if (m_pg) {
             m_pg->Close();
diff --git a/src/core/hle/kernel/k_page_table.cpp b/src/core/hle/kernel/k_page_table.cpp
index 5b51edf30..1d47bdf6b 100644
--- a/src/core/hle/kernel/k_page_table.cpp
+++ b/src/core/hle/kernel/k_page_table.cpp
@@ -82,14 +82,14 @@ public:
 
 using namespace Common::Literals;
 
-constexpr size_t GetAddressSpaceWidthFromType(FileSys::ProgramAddressSpaceType as_type) {
+constexpr size_t GetAddressSpaceWidthFromType(Svc::CreateProcessFlag as_type) {
     switch (as_type) {
-    case FileSys::ProgramAddressSpaceType::Is32Bit:
-    case FileSys::ProgramAddressSpaceType::Is32BitNoMap:
+    case Svc::CreateProcessFlag::AddressSpace32Bit:
+    case Svc::CreateProcessFlag::AddressSpace32BitWithoutAlias:
         return 32;
-    case FileSys::ProgramAddressSpaceType::Is36Bit:
+    case Svc::CreateProcessFlag::AddressSpace64BitDeprecated:
         return 36;
-    case FileSys::ProgramAddressSpaceType::Is39Bit:
+    case Svc::CreateProcessFlag::AddressSpace64Bit:
         return 39;
     default:
         ASSERT(false);
@@ -105,7 +105,7 @@ KPageTable::KPageTable(Core::System& system_)
 
 KPageTable::~KPageTable() = default;
 
-Result KPageTable::InitializeForProcess(FileSys::ProgramAddressSpaceType as_type, bool enable_aslr,
+Result KPageTable::InitializeForProcess(Svc::CreateProcessFlag as_type, bool enable_aslr,
                                         bool enable_das_merge, bool from_back,
                                         KMemoryManager::Pool pool, KProcessAddress code_addr,
                                         size_t code_size, KSystemResource* system_resource,
@@ -133,7 +133,7 @@ Result KPageTable::InitializeForProcess(FileSys::ProgramAddressSpaceType as_type
     ASSERT(code_addr + code_size - 1 <= end - 1);
 
     // Adjust heap/alias size if we don't have an alias region
-    if (as_type == FileSys::ProgramAddressSpaceType::Is32BitNoMap) {
+    if (as_type == Svc::CreateProcessFlag::AddressSpace32BitWithoutAlias) {
         heap_region_size += alias_region_size;
         alias_region_size = 0;
     }
@@ -505,7 +505,7 @@ Result KPageTable::UnmapCodeMemory(KProcessAddress dst_address, KProcessAddress
     R_TRY(this->CheckMemoryStateContiguous(
         std::addressof(num_dst_allocator_blocks), dst_address, size, KMemoryState::FlagCanCodeAlias,
         KMemoryState::FlagCanCodeAlias, KMemoryPermission::None, KMemoryPermission::None,
-        KMemoryAttribute::All, KMemoryAttribute::None));
+        KMemoryAttribute::All & ~KMemoryAttribute::PermissionLocked, KMemoryAttribute::None));
 
     // Determine whether any pages being unmapped are code.
     bool any_code_pages = false;
@@ -1724,29 +1724,43 @@ Result KPageTable::MapPhysicalMemory(KProcessAddress address, size_t size) {
                             PageSize;
 
                         // While we have pages to map, map them.
-                        while (map_pages > 0) {
-                            // Check if we're at the end of the physical block.
-                            if (pg_pages == 0) {
-                                // Ensure there are more pages to map.
-                                ASSERT(pg_it != pg.end());
-
-                                // Advance our physical block.
-                                ++pg_it;
-                                pg_phys_addr = pg_it->GetAddress();
-                                pg_pages = pg_it->GetNumPages();
+                        {
+                            // Create a page group for the current mapping range.
+                            KPageGroup cur_pg(m_kernel, m_block_info_manager);
+                            {
+                                ON_RESULT_FAILURE_2 {
+                                    cur_pg.OpenFirst();
+                                    cur_pg.Close();
+                                };
+
+                                size_t remain_pages = map_pages;
+                                while (remain_pages > 0) {
+                                    // Check if we're at the end of the physical block.
+                                    if (pg_pages == 0) {
+                                        // Ensure there are more pages to map.
+                                        ASSERT(pg_it != pg.end());
+
+                                        // Advance our physical block.
+                                        ++pg_it;
+                                        pg_phys_addr = pg_it->GetAddress();
+                                        pg_pages = pg_it->GetNumPages();
+                                    }
+
+                                    // Add whatever we can to the current block.
+                                    const size_t cur_pages = std::min(pg_pages, remain_pages);
+                                    R_TRY(cur_pg.AddBlock(pg_phys_addr +
+                                                              ((pg_pages - cur_pages) * PageSize),
+                                                          cur_pages));
+
+                                    // Advance.
+                                    remain_pages -= cur_pages;
+                                    pg_pages -= cur_pages;
+                                }
                             }
 
-                            // Map whatever we can.
-                            const size_t cur_pages = std::min(pg_pages, map_pages);
-                            R_TRY(Operate(cur_address, cur_pages, KMemoryPermission::UserReadWrite,
-                                          OperationType::MapFirst, pg_phys_addr));
-
-                            // Advance.
-                            cur_address += cur_pages * PageSize;
-                            map_pages -= cur_pages;
-
-                            pg_phys_addr += cur_pages * PageSize;
-                            pg_pages -= cur_pages;
+                            // Map the pages.
+                            R_TRY(this->Operate(cur_address, map_pages, cur_pg,
+                                                OperationType::MapFirstGroup));
                         }
                     }
 
@@ -1770,7 +1784,11 @@ Result KPageTable::MapPhysicalMemory(KProcessAddress address, size_t size) {
                 m_memory_block_manager.UpdateIfMatch(
                     std::addressof(allocator), address, size / PageSize, KMemoryState::Free,
                     KMemoryPermission::None, KMemoryAttribute::None, KMemoryState::Normal,
-                    KMemoryPermission::UserReadWrite, KMemoryAttribute::None);
+                    KMemoryPermission::UserReadWrite, KMemoryAttribute::None,
+                    address == this->GetAliasRegionStart()
+                        ? KMemoryBlockDisableMergeAttribute::Normal
+                        : KMemoryBlockDisableMergeAttribute::None,
+                    KMemoryBlockDisableMergeAttribute::None);
 
                 R_SUCCEED();
             }
@@ -1868,6 +1886,13 @@ Result KPageTable::UnmapPhysicalMemory(KProcessAddress address, size_t size) {
 
     // Iterate over the memory, unmapping as we go.
     auto it = m_memory_block_manager.FindIterator(cur_address);
+
+    const auto clear_merge_attr =
+        (it->GetState() == KMemoryState::Normal &&
+         it->GetAddress() == this->GetAliasRegionStart() && it->GetAddress() == address)
+            ? KMemoryBlockDisableMergeAttribute::Normal
+            : KMemoryBlockDisableMergeAttribute::None;
+
     while (true) {
         // Check that the iterator is valid.
         ASSERT(it != m_memory_block_manager.end());
@@ -1905,7 +1930,7 @@ Result KPageTable::UnmapPhysicalMemory(KProcessAddress address, size_t size) {
     m_memory_block_manager.Update(std::addressof(allocator), address, size / PageSize,
                                   KMemoryState::Free, KMemoryPermission::None,
                                   KMemoryAttribute::None, KMemoryBlockDisableMergeAttribute::None,
-                                  KMemoryBlockDisableMergeAttribute::None);
+                                  clear_merge_attr);
 
     // We succeeded.
     R_SUCCEED();
@@ -2379,8 +2404,7 @@ Result KPageTable::MapPageGroup(KProcessAddress* out_addr, const KPageGroup& pg,
     KScopedPageTableUpdater updater(this);
 
     // Perform mapping operation.
-    const KPageProperties properties = {perm, state == KMemoryState::Io, false,
-                                        DisableMergeAttribute::DisableHead};
+    const KPageProperties properties = {perm, false, false, DisableMergeAttribute::DisableHead};
     R_TRY(this->MapPageGroupImpl(updater.GetPageList(), addr, pg, properties, false));
 
     // Update the blocks.
@@ -2422,8 +2446,7 @@ Result KPageTable::MapPageGroup(KProcessAddress addr, const KPageGroup& pg, KMem
     KScopedPageTableUpdater updater(this);
 
     // Perform mapping operation.
-    const KPageProperties properties = {perm, state == KMemoryState::Io, false,
-                                        DisableMergeAttribute::DisableHead};
+    const KPageProperties properties = {perm, false, false, DisableMergeAttribute::DisableHead};
     R_TRY(this->MapPageGroupImpl(updater.GetPageList(), addr, pg, properties, false));
 
     // Update the blocks.
@@ -2652,11 +2675,18 @@ Result KPageTable::SetMemoryAttribute(KProcessAddress addr, size_t size, u32 mas
     size_t num_allocator_blocks;
     constexpr auto AttributeTestMask =
         ~(KMemoryAttribute::SetMask | KMemoryAttribute::DeviceShared);
-    R_TRY(this->CheckMemoryState(
-        std::addressof(old_state), std::addressof(old_perm), std::addressof(old_attr),
-        std::addressof(num_allocator_blocks), addr, size, KMemoryState::FlagCanChangeAttribute,
-        KMemoryState::FlagCanChangeAttribute, KMemoryPermission::None, KMemoryPermission::None,
-        AttributeTestMask, KMemoryAttribute::None, ~AttributeTestMask));
+    const KMemoryState state_test_mask =
+        static_cast<KMemoryState>(((mask & static_cast<u32>(KMemoryAttribute::Uncached))
+                                       ? static_cast<u32>(KMemoryState::FlagCanChangeAttribute)
+                                       : 0) |
+                                  ((mask & static_cast<u32>(KMemoryAttribute::PermissionLocked))
+                                       ? static_cast<u32>(KMemoryState::FlagCanPermissionLock)
+                                       : 0));
+    R_TRY(this->CheckMemoryState(std::addressof(old_state), std::addressof(old_perm),
+                                 std::addressof(old_attr), std::addressof(num_allocator_blocks),
+                                 addr, size, state_test_mask, state_test_mask,
+                                 KMemoryPermission::None, KMemoryPermission::None,
+                                 AttributeTestMask, KMemoryAttribute::None, ~AttributeTestMask));
 
     // Create an update allocator.
     Result allocator_result{ResultSuccess};
@@ -2664,18 +2694,17 @@ Result KPageTable::SetMemoryAttribute(KProcessAddress addr, size_t size, u32 mas
                                                  m_memory_block_slab_manager, num_allocator_blocks);
     R_TRY(allocator_result);
 
-    // Determine the new attribute.
-    const KMemoryAttribute new_attr =
-        static_cast<KMemoryAttribute>(((old_attr & static_cast<KMemoryAttribute>(~mask)) |
-                                       static_cast<KMemoryAttribute>(attr & mask)));
-
-    // Perform operation.
-    this->Operate(addr, num_pages, old_perm, OperationType::ChangePermissionsAndRefresh);
+    // If we need to, perform a change attribute operation.
+    if (True(KMemoryAttribute::Uncached & static_cast<KMemoryAttribute>(mask))) {
+        // Perform operation.
+        R_TRY(this->Operate(addr, num_pages, old_perm,
+                            OperationType::ChangePermissionsAndRefreshAndFlush, 0));
+    }
 
     // Update the blocks.
-    m_memory_block_manager.Update(std::addressof(allocator), addr, num_pages, old_state, old_perm,
-                                  new_attr, KMemoryBlockDisableMergeAttribute::None,
-                                  KMemoryBlockDisableMergeAttribute::None);
+    m_memory_block_manager.UpdateAttribute(std::addressof(allocator), addr, num_pages,
+                                           static_cast<KMemoryAttribute>(mask),
+                                           static_cast<KMemoryAttribute>(attr));
 
     R_SUCCEED();
 }
@@ -2863,7 +2892,8 @@ Result KPageTable::LockForMapDeviceAddressSpace(bool* out_is_io, KProcessAddress
                                       &KMemoryBlock::ShareToDevice, KMemoryPermission::None);
 
     // Set whether the locked memory was io.
-    *out_is_io = old_state == KMemoryState::Io;
+    *out_is_io =
+        static_cast<Svc::MemoryState>(old_state & KMemoryState::Mask) == Svc::MemoryState::Io;
 
     R_SUCCEED();
 }
@@ -2949,6 +2979,23 @@ Result KPageTable::UnlockForIpcUserBuffer(KProcessAddress address, size_t size)
                                 KMemoryAttribute::Locked, nullptr));
 }
 
+Result KPageTable::LockForTransferMemory(KPageGroup* out, KProcessAddress address, size_t size,
+                                         KMemoryPermission perm) {
+    R_RETURN(this->LockMemoryAndOpen(out, nullptr, address, size, KMemoryState::FlagCanTransfer,
+                                     KMemoryState::FlagCanTransfer, KMemoryPermission::All,
+                                     KMemoryPermission::UserReadWrite, KMemoryAttribute::All,
+                                     KMemoryAttribute::None, perm, KMemoryAttribute::Locked));
+}
+
+Result KPageTable::UnlockForTransferMemory(KProcessAddress address, size_t size,
+                                           const KPageGroup& pg) {
+    R_RETURN(this->UnlockMemory(address, size, KMemoryState::FlagCanTransfer,
+                                KMemoryState::FlagCanTransfer, KMemoryPermission::None,
+                                KMemoryPermission::None, KMemoryAttribute::All,
+                                KMemoryAttribute::Locked, KMemoryPermission::UserReadWrite,
+                                KMemoryAttribute::Locked, std::addressof(pg)));
+}
+
 Result KPageTable::LockForCodeMemory(KPageGroup* out, KProcessAddress addr, size_t size) {
     R_RETURN(this->LockMemoryAndOpen(
         out, nullptr, addr, size, KMemoryState::FlagCanCodeMemory, KMemoryState::FlagCanCodeMemory,
@@ -3004,9 +3051,10 @@ Result KPageTable::Operate(KProcessAddress addr, size_t num_pages, const KPageGr
     ASSERT(num_pages == page_group.GetNumPages());
 
     switch (operation) {
-    case OperationType::MapGroup: {
+    case OperationType::MapGroup:
+    case OperationType::MapFirstGroup: {
         // We want to maintain a new reference to every page in the group.
-        KScopedPageGroup spg(page_group);
+        KScopedPageGroup spg(page_group, operation != OperationType::MapFirstGroup);
 
         for (const auto& node : page_group) {
             const size_t size{node.GetNumPages() * PageSize};
@@ -3048,7 +3096,6 @@ Result KPageTable::Operate(KProcessAddress addr, size_t num_pages, KMemoryPermis
         m_memory->UnmapRegion(*m_page_table_impl, addr, num_pages * PageSize);
         break;
     }
-    case OperationType::MapFirst:
     case OperationType::Map: {
         ASSERT(map_addr);
         ASSERT(Common::IsAligned(GetInteger(map_addr), PageSize));
@@ -3056,11 +3103,7 @@ Result KPageTable::Operate(KProcessAddress addr, size_t num_pages, KMemoryPermis
 
         // Open references to pages, if we should.
         if (IsHeapPhysicalAddress(m_kernel.MemoryLayout(), map_addr)) {
-            if (operation == OperationType::MapFirst) {
-                m_kernel.MemoryManager().OpenFirst(map_addr, num_pages);
-            } else {
-                m_kernel.MemoryManager().Open(map_addr, num_pages);
-            }
+            m_kernel.MemoryManager().Open(map_addr, num_pages);
         }
         break;
     }
@@ -3070,6 +3113,7 @@ Result KPageTable::Operate(KProcessAddress addr, size_t num_pages, KMemoryPermis
     }
     case OperationType::ChangePermissions:
     case OperationType::ChangePermissionsAndRefresh:
+    case OperationType::ChangePermissionsAndRefreshAndFlush:
         break;
     default:
         ASSERT(false);
@@ -3089,79 +3133,79 @@ void KPageTable::FinalizeUpdate(PageLinkedList* page_list) {
     }
 }
 
-KProcessAddress KPageTable::GetRegionAddress(KMemoryState state) const {
+KProcessAddress KPageTable::GetRegionAddress(Svc::MemoryState state) const {
     switch (state) {
-    case KMemoryState::Free:
-    case KMemoryState::Kernel:
+    case Svc::MemoryState::Free:
+    case Svc::MemoryState::Kernel:
         return m_address_space_start;
-    case KMemoryState::Normal:
+    case Svc::MemoryState::Normal:
         return m_heap_region_start;
-    case KMemoryState::Ipc:
-    case KMemoryState::NonSecureIpc:
-    case KMemoryState::NonDeviceIpc:
+    case Svc::MemoryState::Ipc:
+    case Svc::MemoryState::NonSecureIpc:
+    case Svc::MemoryState::NonDeviceIpc:
         return m_alias_region_start;
-    case KMemoryState::Stack:
+    case Svc::MemoryState::Stack:
         return m_stack_region_start;
-    case KMemoryState::Static:
-    case KMemoryState::ThreadLocal:
+    case Svc::MemoryState::Static:
+    case Svc::MemoryState::ThreadLocal:
         return m_kernel_map_region_start;
-    case KMemoryState::Io:
-    case KMemoryState::Shared:
-    case KMemoryState::AliasCode:
-    case KMemoryState::AliasCodeData:
-    case KMemoryState::Transfered:
-    case KMemoryState::SharedTransfered:
-    case KMemoryState::SharedCode:
-    case KMemoryState::GeneratedCode:
-    case KMemoryState::CodeOut:
-    case KMemoryState::Coverage:
-    case KMemoryState::Insecure:
+    case Svc::MemoryState::Io:
+    case Svc::MemoryState::Shared:
+    case Svc::MemoryState::AliasCode:
+    case Svc::MemoryState::AliasCodeData:
+    case Svc::MemoryState::Transfered:
+    case Svc::MemoryState::SharedTransfered:
+    case Svc::MemoryState::SharedCode:
+    case Svc::MemoryState::GeneratedCode:
+    case Svc::MemoryState::CodeOut:
+    case Svc::MemoryState::Coverage:
+    case Svc::MemoryState::Insecure:
         return m_alias_code_region_start;
-    case KMemoryState::Code:
-    case KMemoryState::CodeData:
+    case Svc::MemoryState::Code:
+    case Svc::MemoryState::CodeData:
         return m_code_region_start;
     default:
         UNREACHABLE();
     }
 }
 
-size_t KPageTable::GetRegionSize(KMemoryState state) const {
+size_t KPageTable::GetRegionSize(Svc::MemoryState state) const {
     switch (state) {
-    case KMemoryState::Free:
-    case KMemoryState::Kernel:
+    case Svc::MemoryState::Free:
+    case Svc::MemoryState::Kernel:
         return m_address_space_end - m_address_space_start;
-    case KMemoryState::Normal:
+    case Svc::MemoryState::Normal:
         return m_heap_region_end - m_heap_region_start;
-    case KMemoryState::Ipc:
-    case KMemoryState::NonSecureIpc:
-    case KMemoryState::NonDeviceIpc:
+    case Svc::MemoryState::Ipc:
+    case Svc::MemoryState::NonSecureIpc:
+    case Svc::MemoryState::NonDeviceIpc:
         return m_alias_region_end - m_alias_region_start;
-    case KMemoryState::Stack:
+    case Svc::MemoryState::Stack:
         return m_stack_region_end - m_stack_region_start;
-    case KMemoryState::Static:
-    case KMemoryState::ThreadLocal:
+    case Svc::MemoryState::Static:
+    case Svc::MemoryState::ThreadLocal:
         return m_kernel_map_region_end - m_kernel_map_region_start;
-    case KMemoryState::Io:
-    case KMemoryState::Shared:
-    case KMemoryState::AliasCode:
-    case KMemoryState::AliasCodeData:
-    case KMemoryState::Transfered:
-    case KMemoryState::SharedTransfered:
-    case KMemoryState::SharedCode:
-    case KMemoryState::GeneratedCode:
-    case KMemoryState::CodeOut:
-    case KMemoryState::Coverage:
-    case KMemoryState::Insecure:
+    case Svc::MemoryState::Io:
+    case Svc::MemoryState::Shared:
+    case Svc::MemoryState::AliasCode:
+    case Svc::MemoryState::AliasCodeData:
+    case Svc::MemoryState::Transfered:
+    case Svc::MemoryState::SharedTransfered:
+    case Svc::MemoryState::SharedCode:
+    case Svc::MemoryState::GeneratedCode:
+    case Svc::MemoryState::CodeOut:
+    case Svc::MemoryState::Coverage:
+    case Svc::MemoryState::Insecure:
         return m_alias_code_region_end - m_alias_code_region_start;
-    case KMemoryState::Code:
-    case KMemoryState::CodeData:
+    case Svc::MemoryState::Code:
+    case Svc::MemoryState::CodeData:
         return m_code_region_end - m_code_region_start;
     default:
         UNREACHABLE();
     }
 }
 
-bool KPageTable::CanContain(KProcessAddress addr, size_t size, KMemoryState state) const {
+bool KPageTable::CanContain(KProcessAddress addr, size_t size, Svc::MemoryState state) const {
     const KProcessAddress end = addr + size;
     const KProcessAddress last = end - 1;
 
@@ -3175,32 +3219,32 @@ bool KPageTable::CanContain(KProcessAddress addr, size_t size, KMemoryState stat
     const bool is_in_alias = !(end <= m_alias_region_start || m_alias_region_end <= addr ||
                                m_alias_region_start == m_alias_region_end);
     switch (state) {
-    case KMemoryState::Free:
-    case KMemoryState::Kernel:
+    case Svc::MemoryState::Free:
+    case Svc::MemoryState::Kernel:
         return is_in_region;
-    case KMemoryState::Io:
-    case KMemoryState::Static:
-    case KMemoryState::Code:
-    case KMemoryState::CodeData:
-    case KMemoryState::Shared:
-    case KMemoryState::AliasCode:
-    case KMemoryState::AliasCodeData:
-    case KMemoryState::Stack:
-    case KMemoryState::ThreadLocal:
-    case KMemoryState::Transfered:
-    case KMemoryState::SharedTransfered:
-    case KMemoryState::SharedCode:
-    case KMemoryState::GeneratedCode:
-    case KMemoryState::CodeOut:
-    case KMemoryState::Coverage:
-    case KMemoryState::Insecure:
+    case Svc::MemoryState::Io:
+    case Svc::MemoryState::Static:
+    case Svc::MemoryState::Code:
+    case Svc::MemoryState::CodeData:
+    case Svc::MemoryState::Shared:
+    case Svc::MemoryState::AliasCode:
+    case Svc::MemoryState::AliasCodeData:
+    case Svc::MemoryState::Stack:
+    case Svc::MemoryState::ThreadLocal:
+    case Svc::MemoryState::Transfered:
+    case Svc::MemoryState::SharedTransfered:
+    case Svc::MemoryState::SharedCode:
+    case Svc::MemoryState::GeneratedCode:
+    case Svc::MemoryState::CodeOut:
+    case Svc::MemoryState::Coverage:
+    case Svc::MemoryState::Insecure:
         return is_in_region && !is_in_heap && !is_in_alias;
-    case KMemoryState::Normal:
+    case Svc::MemoryState::Normal:
         ASSERT(is_in_heap);
         return is_in_region && !is_in_alias;
-    case KMemoryState::Ipc:
-    case KMemoryState::NonSecureIpc:
-    case KMemoryState::NonDeviceIpc:
+    case Svc::MemoryState::Ipc:
+    case Svc::MemoryState::NonSecureIpc:
+    case Svc::MemoryState::NonDeviceIpc:
         ASSERT(is_in_alias);
         return is_in_region && !is_in_heap;
     default:
@@ -3264,21 +3308,16 @@ Result KPageTable::CheckMemoryStateContiguous(size_t* out_blocks_needed, KProces
 
 Result KPageTable::CheckMemoryState(KMemoryState* out_state, KMemoryPermission* out_perm,
                                     KMemoryAttribute* out_attr, size_t* out_blocks_needed,
-                                    KProcessAddress addr, size_t size, KMemoryState state_mask,
+                                    KMemoryBlockManager::const_iterator it,
+                                    KProcessAddress last_addr, KMemoryState state_mask,
                                     KMemoryState state, KMemoryPermission perm_mask,
                                     KMemoryPermission perm, KMemoryAttribute attr_mask,
                                     KMemoryAttribute attr, KMemoryAttribute ignore_attr) const {
     ASSERT(this->IsLockedByCurrentThread());
 
     // Get information about the first block.
-    const KProcessAddress last_addr = addr + size - 1;
-    KMemoryBlockManager::const_iterator it = m_memory_block_manager.FindIterator(addr);
     KMemoryInfo info = it->GetMemoryInfo();
 
-    // If the start address isn't aligned, we need a block.
-    const size_t blocks_for_start_align =
-        (Common::AlignDown(GetInteger(addr), PageSize) != info.GetAddress()) ? 1 : 0;
-
     // Validate all blocks in the range have correct state.
     const KMemoryState first_state = info.m_state;
     const KMemoryPermission first_perm = info.m_permission;
@@ -3304,10 +3343,6 @@ Result KPageTable::CheckMemoryState(KMemoryState* out_state, KMemoryPermission*
         info = it->GetMemoryInfo();
     }
 
-    // If the end address isn't aligned, we need a block.
-    const size_t blocks_for_end_align =
-        (Common::AlignUp(GetInteger(addr) + size, PageSize) != info.GetEndAddress()) ? 1 : 0;
-
     // Write output state.
     if (out_state != nullptr) {
         *out_state = first_state;
@@ -3318,9 +3353,39 @@ Result KPageTable::CheckMemoryState(KMemoryState* out_state, KMemoryPermission*
     if (out_attr != nullptr) {
         *out_attr = static_cast<KMemoryAttribute>(first_attr & ~ignore_attr);
     }
+
+    // If the end address isn't aligned, we need a block.
     if (out_blocks_needed != nullptr) {
-        *out_blocks_needed = blocks_for_start_align + blocks_for_end_align;
+        const size_t blocks_for_end_align =
+            (Common::AlignDown(GetInteger(last_addr), PageSize) + PageSize != info.GetEndAddress())
+                ? 1
+                : 0;
+        *out_blocks_needed = blocks_for_end_align;
     }
+
+    R_SUCCEED();
+}
+
+Result KPageTable::CheckMemoryState(KMemoryState* out_state, KMemoryPermission* out_perm,
+                                    KMemoryAttribute* out_attr, size_t* out_blocks_needed,
+                                    KProcessAddress addr, size_t size, KMemoryState state_mask,
+                                    KMemoryState state, KMemoryPermission perm_mask,
+                                    KMemoryPermission perm, KMemoryAttribute attr_mask,
+                                    KMemoryAttribute attr, KMemoryAttribute ignore_attr) const {
+    ASSERT(this->IsLockedByCurrentThread());
+
+    // Check memory state.
+    const KProcessAddress last_addr = addr + size - 1;
+    KMemoryBlockManager::const_iterator it = m_memory_block_manager.FindIterator(addr);
+    R_TRY(this->CheckMemoryState(out_state, out_perm, out_attr, out_blocks_needed, it, last_addr,
+                                 state_mask, state, perm_mask, perm, attr_mask, attr, ignore_attr));
+
+    // If the start address isn't aligned, we need a block.
+    if (out_blocks_needed != nullptr &&
+        Common::AlignDown(GetInteger(addr), PageSize) != it->GetAddress()) {
+        ++(*out_blocks_needed);
+    }
+
     R_SUCCEED();
 }
 
@@ -3388,6 +3453,11 @@ Result KPageTable::LockMemoryAndOpen(KPageGroup* out_pg, KPhysicalAddress* out_K
                                   new_attr, KMemoryBlockDisableMergeAttribute::Locked,
                                   KMemoryBlockDisableMergeAttribute::None);
 
+    // If we have an output page group, open.
+    if (out_pg) {
+        out_pg->Open();
+    }
+
     R_SUCCEED();
 }
 
diff --git a/src/core/hle/kernel/k_page_table.h b/src/core/hle/kernel/k_page_table.h
index b9e8c6042..66f16faaf 100644
--- a/src/core/hle/kernel/k_page_table.h
+++ b/src/core/hle/kernel/k_page_table.h
@@ -63,7 +63,7 @@ public:
     explicit KPageTable(Core::System& system_);
     ~KPageTable();
 
-    Result InitializeForProcess(FileSys::ProgramAddressSpaceType as_type, bool enable_aslr,
+    Result InitializeForProcess(Svc::CreateProcessFlag as_type, bool enable_aslr,
                                 bool enable_das_merge, bool from_back, KMemoryManager::Pool pool,
                                 KProcessAddress code_addr, size_t code_size,
                                 KSystemResource* system_resource, KResourceLimit* resource_limit,
@@ -104,6 +104,9 @@ public:
     Result CleanupForIpcServer(KProcessAddress address, size_t size, KMemoryState dst_state);
     Result CleanupForIpcClient(KProcessAddress address, size_t size, KMemoryState dst_state);
 
+    Result LockForTransferMemory(KPageGroup* out, KProcessAddress address, size_t size,
+                                 KMemoryPermission perm);
+    Result UnlockForTransferMemory(KProcessAddress address, size_t size, const KPageGroup& pg);
     Result LockForCodeMemory(KPageGroup* out, KProcessAddress addr, size_t size);
     Result UnlockForCodeMemory(KProcessAddress addr, size_t size, const KPageGroup& pg);
     Result MakeAndOpenPageGroup(KPageGroup* out, KProcessAddress address, size_t num_pages,
@@ -123,8 +126,6 @@ public:
         return m_block_info_manager;
     }
 
-    bool CanContain(KProcessAddress addr, size_t size, KMemoryState state) const;
-
     Result MapPages(KProcessAddress* out_addr, size_t num_pages, size_t alignment,
                     KPhysicalAddress phys_addr, KProcessAddress region_start,
                     size_t region_num_pages, KMemoryState state, KMemoryPermission perm) {
@@ -159,6 +160,21 @@ public:
     void RemapPageGroup(PageLinkedList* page_list, KProcessAddress address, size_t size,
                         const KPageGroup& pg);
 
+    KProcessAddress GetRegionAddress(Svc::MemoryState state) const;
+    size_t GetRegionSize(Svc::MemoryState state) const;
+    bool CanContain(KProcessAddress addr, size_t size, Svc::MemoryState state) const;
+
+    KProcessAddress GetRegionAddress(KMemoryState state) const {
+        return this->GetRegionAddress(static_cast<Svc::MemoryState>(state & KMemoryState::Mask));
+    }
+    size_t GetRegionSize(KMemoryState state) const {
+        return this->GetRegionSize(static_cast<Svc::MemoryState>(state & KMemoryState::Mask));
+    }
+    bool CanContain(KProcessAddress addr, size_t size, KMemoryState state) const {
+        return this->CanContain(addr, size,
+                                static_cast<Svc::MemoryState>(state & KMemoryState::Mask));
+    }
+
 protected:
     struct PageLinkedList {
     private:
@@ -201,12 +217,13 @@ protected:
 private:
     enum class OperationType : u32 {
         Map = 0,
-        MapFirst = 1,
-        MapGroup = 2,
+        MapGroup = 1,
+        MapFirstGroup = 2,
         Unmap = 3,
         ChangePermissions = 4,
         ChangePermissionsAndRefresh = 5,
-        Separate = 6,
+        ChangePermissionsAndRefreshAndFlush = 6,
+        Separate = 7,
     };
 
     static constexpr KMemoryAttribute DefaultMemoryIgnoreAttr =
@@ -225,8 +242,6 @@ private:
     Result Operate(KProcessAddress addr, size_t num_pages, KMemoryPermission perm,
                    OperationType operation, KPhysicalAddress map_addr = 0);
     void FinalizeUpdate(PageLinkedList* page_list);
-    KProcessAddress GetRegionAddress(KMemoryState state) const;
-    size_t GetRegionSize(KMemoryState state) const;
 
     KProcessAddress FindFreeArea(KProcessAddress region_start, size_t region_num_pages,
                                  size_t num_pages, size_t alignment, size_t offset,
@@ -249,6 +264,13 @@ private:
                             KMemoryAttribute attr_mask, KMemoryAttribute attr) const;
     Result CheckMemoryState(KMemoryState* out_state, KMemoryPermission* out_perm,
                             KMemoryAttribute* out_attr, size_t* out_blocks_needed,
+                            KMemoryBlockManager::const_iterator it, KProcessAddress last_addr,
+                            KMemoryState state_mask, KMemoryState state,
+                            KMemoryPermission perm_mask, KMemoryPermission perm,
+                            KMemoryAttribute attr_mask, KMemoryAttribute attr,
+                            KMemoryAttribute ignore_attr = DefaultMemoryIgnoreAttr) const;
+    Result CheckMemoryState(KMemoryState* out_state, KMemoryPermission* out_perm,
+                            KMemoryAttribute* out_attr, size_t* out_blocks_needed,
                             KProcessAddress addr, size_t size, KMemoryState state_mask,
                             KMemoryState state, KMemoryPermission perm_mask, KMemoryPermission perm,
                             KMemoryAttribute attr_mask, KMemoryAttribute attr,
@@ -378,7 +400,7 @@ public:
     constexpr size_t GetAliasCodeRegionSize() const {
         return m_alias_code_region_end - m_alias_code_region_start;
     }
-    size_t GetNormalMemorySize() {
+    size_t GetNormalMemorySize() const {
         KScopedLightLock lk(m_general_lock);
         return GetHeapSize() + m_mapped_physical_memory_size;
     }
diff --git a/src/core/hle/kernel/k_process.cpp b/src/core/hle/kernel/k_process.cpp
index 4a099286b..1f4b0755d 100644
--- a/src/core/hle/kernel/k_process.cpp
+++ b/src/core/hle/kernel/k_process.cpp
@@ -1,515 +1,598 @@
-// SPDX-FileCopyrightText: 2015 Citra Emulator Project
+// SPDX-FileCopyrightText: Copyright 2023 yuzu Emulator Project
 // SPDX-License-Identifier: GPL-2.0-or-later
 
-#include <algorithm>
-#include <bitset>
-#include <ctime>
-#include <memory>
 #include <random>
-#include "common/alignment.h"
-#include "common/assert.h"
-#include "common/logging/log.h"
 #include "common/scope_exit.h"
 #include "common/settings.h"
 #include "core/core.h"
-#include "core/file_sys/program_metadata.h"
-#include "core/hle/kernel/code_set.h"
-#include "core/hle/kernel/k_memory_block_manager.h"
-#include "core/hle/kernel/k_page_table.h"
 #include "core/hle/kernel/k_process.h"
-#include "core/hle/kernel/k_resource_limit.h"
-#include "core/hle/kernel/k_scheduler.h"
 #include "core/hle/kernel/k_scoped_resource_reservation.h"
 #include "core/hle/kernel/k_shared_memory.h"
 #include "core/hle/kernel/k_shared_memory_info.h"
-#include "core/hle/kernel/k_thread.h"
-#include "core/hle/kernel/kernel.h"
-#include "core/hle/kernel/svc_results.h"
-#include "core/memory.h"
+#include "core/hle/kernel/k_thread_local_page.h"
+#include "core/hle/kernel/k_thread_queue.h"
+#include "core/hle/kernel/k_worker_task_manager.h"
 
 namespace Kernel {
-namespace {
-/**
- * Sets up the primary application thread
- *
- * @param system The system instance to create the main thread under.
- * @param owner_process The parent process for the main thread
- * @param priority The priority to give the main thread
- */
-void SetupMainThread(Core::System& system, KProcess& owner_process, u32 priority,
-                     KProcessAddress stack_top) {
-    const KProcessAddress entry_point = owner_process.GetEntryPoint();
-    ASSERT(owner_process.GetResourceLimit()->Reserve(LimitableResource::ThreadCountMax, 1));
-
-    KThread* thread = KThread::Create(system.Kernel());
-    SCOPE_EXIT({ thread->Close(); });
-
-    ASSERT(KThread::InitializeUserThread(system, thread, entry_point, 0, stack_top, priority,
-                                         owner_process.GetIdealCoreId(),
-                                         std::addressof(owner_process))
-               .IsSuccess());
-
-    // Register 1 must be a handle to the main thread
-    Handle thread_handle{};
-    owner_process.GetHandleTable().Add(std::addressof(thread_handle), thread);
-
-    thread->GetContext32().cpu_registers[0] = 0;
-    thread->GetContext64().cpu_registers[0] = 0;
-    thread->GetContext32().cpu_registers[1] = thread_handle;
-    thread->GetContext64().cpu_registers[1] = thread_handle;
-
-    if (system.DebuggerEnabled()) {
-        thread->RequestSuspend(SuspendType::Debug);
-    }
 
-    // Run our thread.
-    void(thread->Run());
-}
-} // Anonymous namespace
+namespace {
 
-Result KProcess::Initialize(KProcess* process, Core::System& system, std::string process_name,
-                            ProcessType type, KResourceLimit* res_limit) {
-    auto& kernel = system.Kernel();
+Result TerminateChildren(KernelCore& kernel, KProcess* process,
+                         const KThread* thread_to_not_terminate) {
+    // Request that all children threads terminate.
+    {
+        KScopedLightLock proc_lk(process->GetListLock());
+        KScopedSchedulerLock sl(kernel);
+
+        if (thread_to_not_terminate != nullptr &&
+            process->GetPinnedThread(GetCurrentCoreId(kernel)) == thread_to_not_terminate) {
+            // NOTE: Here Nintendo unpins the current thread instead of the thread_to_not_terminate.
+            // This is valid because the only caller which uses non-nullptr as argument uses
+            // GetCurrentThreadPointer(), but it's still notable because it seems incorrect at
+            // first glance.
+            process->UnpinCurrentThread();
+        }
 
-    process->name = std::move(process_name);
-    process->m_resource_limit = res_limit;
-    process->m_system_resource_address = 0;
-    process->m_state = State::Created;
-    process->m_program_id = 0;
-    process->m_process_id = type == ProcessType::KernelInternal ? kernel.CreateNewKernelProcessID()
-                                                                : kernel.CreateNewUserProcessID();
-    process->m_capabilities.InitializeForMetadatalessProcess();
-    process->m_is_initialized = true;
+        auto& thread_list = process->GetThreadList();
+        for (auto it = thread_list.begin(); it != thread_list.end(); ++it) {
+            if (KThread* thread = std::addressof(*it); thread != thread_to_not_terminate) {
+                if (thread->GetState() != ThreadState::Terminated) {
+                    thread->RequestTerminate();
+                }
+            }
+        }
+    }
 
-    std::mt19937 rng(Settings::values.rng_seed_enabled ? Settings::values.rng_seed.GetValue()
-                                                       : static_cast<u32>(std::time(nullptr)));
-    std::uniform_int_distribution<u64> distribution;
-    std::generate(process->m_random_entropy.begin(), process->m_random_entropy.end(),
-                  [&] { return distribution(rng); });
+    // Wait for all children threads to terminate.
+    while (true) {
+        // Get the next child.
+        KThread* cur_child = nullptr;
+        {
+            KScopedLightLock proc_lk(process->GetListLock());
+
+            auto& thread_list = process->GetThreadList();
+            for (auto it = thread_list.begin(); it != thread_list.end(); ++it) {
+                if (KThread* thread = std::addressof(*it); thread != thread_to_not_terminate) {
+                    if (thread->GetState() != ThreadState::Terminated) {
+                        if (thread->Open()) {
+                            cur_child = thread;
+                            break;
+                        }
+                    }
+                }
+            }
+        }
 
-    kernel.AppendNewProcess(process);
+        // If we didn't find any non-terminated children, we're done.
+        if (cur_child == nullptr) {
+            break;
+        }
 
-    // Clear remaining fields.
-    process->m_num_running_threads = 0;
-    process->m_is_signaled = false;
-    process->m_exception_thread = nullptr;
-    process->m_is_suspended = false;
-    process->m_schedule_count = 0;
-    process->m_is_handle_table_initialized = false;
-    process->m_is_hbl = false;
+        // Terminate and close the thread.
+        SCOPE_EXIT({ cur_child->Close(); });
 
-    // Open a reference to the resource limit.
-    process->m_resource_limit->Open();
+        if (const Result terminate_result = cur_child->Terminate();
+            ResultTerminationRequested == terminate_result) {
+            R_THROW(terminate_result);
+        }
+    }
 
     R_SUCCEED();
 }
 
-void KProcess::DoWorkerTaskImpl() {
-    UNIMPLEMENTED();
-}
-
-KResourceLimit* KProcess::GetResourceLimit() const {
-    return m_resource_limit;
-}
+class ThreadQueueImplForKProcessEnterUserException final : public KThreadQueue {
+private:
+    KThread** m_exception_thread;
 
-void KProcess::IncrementRunningThreadCount() {
-    ASSERT(m_num_running_threads.load() >= 0);
-    ++m_num_running_threads;
-}
+public:
+    explicit ThreadQueueImplForKProcessEnterUserException(KernelCore& kernel, KThread** t)
+        : KThreadQueue(kernel), m_exception_thread(t) {}
 
-void KProcess::DecrementRunningThreadCount() {
-    ASSERT(m_num_running_threads.load() > 0);
+    virtual void EndWait(KThread* waiting_thread, Result wait_result) override {
+        // Set the exception thread.
+        *m_exception_thread = waiting_thread;
 
-    if (const auto prev = m_num_running_threads--; prev == 1) {
-        // TODO(bunnei): Process termination to be implemented when multiprocess is supported.
+        // Invoke the base end wait handler.
+        KThreadQueue::EndWait(waiting_thread, wait_result);
     }
-}
 
-u64 KProcess::GetTotalPhysicalMemoryAvailable() {
-    const u64 capacity{m_resource_limit->GetFreeValue(LimitableResource::PhysicalMemoryMax) +
-                       m_page_table.GetNormalMemorySize() + GetSystemResourceSize() + m_image_size +
-                       m_main_thread_stack_size};
-    if (const auto pool_size = m_kernel.MemoryManager().GetSize(KMemoryManager::Pool::Application);
-        capacity != pool_size) {
-        LOG_WARNING(Kernel, "capacity {} != application pool size {}", capacity, pool_size);
-    }
-    if (capacity < m_memory_usage_capacity) {
-        return capacity;
+    virtual void CancelWait(KThread* waiting_thread, Result wait_result,
+                            bool cancel_timer_task) override {
+        // Remove the thread as a waiter on its mutex owner.
+        waiting_thread->GetLockOwner()->RemoveWaiter(waiting_thread);
+
+        // Invoke the base cancel wait handler.
+        KThreadQueue::CancelWait(waiting_thread, wait_result, cancel_timer_task);
     }
-    return m_memory_usage_capacity;
-}
+};
 
-u64 KProcess::GetTotalPhysicalMemoryAvailableWithoutSystemResource() {
-    return this->GetTotalPhysicalMemoryAvailable() - this->GetSystemResourceSize();
+void GenerateRandom(std::span<u64> out_random) {
+    std::mt19937 rng(Settings::values.rng_seed_enabled ? Settings::values.rng_seed.GetValue()
+                                                       : static_cast<u32>(std::time(nullptr)));
+    std::uniform_int_distribution<u64> distribution;
+    std::generate(out_random.begin(), out_random.end(), [&] { return distribution(rng); });
 }
 
-u64 KProcess::GetTotalPhysicalMemoryUsed() {
-    return m_image_size + m_main_thread_stack_size + m_page_table.GetNormalMemorySize() +
-           this->GetSystemResourceSize();
-}
+} // namespace
 
-u64 KProcess::GetTotalPhysicalMemoryUsedWithoutSystemResource() {
-    return this->GetTotalPhysicalMemoryUsed() - this->GetSystemResourceUsage();
-}
+void KProcess::Finalize() {
+    // Delete the process local region.
+    this->DeleteThreadLocalRegion(m_plr_address);
 
-bool KProcess::ReleaseUserException(KThread* thread) {
-    KScopedSchedulerLock sl{m_kernel};
+    // Get the used memory size.
+    const size_t used_memory_size = this->GetUsedNonSystemUserPhysicalMemorySize();
 
-    if (m_exception_thread == thread) {
-        m_exception_thread = nullptr;
+    // Finalize the page table.
+    m_page_table.Finalize();
 
-        // Remove waiter thread.
-        bool has_waiters{};
-        if (KThread* next = thread->RemoveKernelWaiterByKey(
-                std::addressof(has_waiters),
-                reinterpret_cast<uintptr_t>(std::addressof(m_exception_thread)));
-            next != nullptr) {
-            next->EndWait(ResultSuccess);
+    // Finish using our system resource.
+    if (m_system_resource) {
+        if (m_system_resource->IsSecureResource()) {
+            // Finalize optimized memory. If memory wasn't optimized, this is a no-op.
+            m_kernel.MemoryManager().FinalizeOptimizedMemory(this->GetId(), m_memory_pool);
         }
 
-        KScheduler::SetSchedulerUpdateNeeded(m_kernel);
-
-        return true;
-    } else {
-        return false;
+        m_system_resource->Close();
+        m_system_resource = nullptr;
     }
-}
-
-void KProcess::PinCurrentThread(s32 core_id) {
-    ASSERT(KScheduler::IsSchedulerLockedByCurrentThread(m_kernel));
 
-    // Get the current thread.
-    KThread* cur_thread =
-        m_kernel.Scheduler(static_cast<std::size_t>(core_id)).GetSchedulerCurrentThread();
+    // Free all shared memory infos.
+    {
+        auto it = m_shared_memory_list.begin();
+        while (it != m_shared_memory_list.end()) {
+            KSharedMemoryInfo* info = std::addressof(*it);
+            KSharedMemory* shmem = info->GetSharedMemory();
 
-    // If the thread isn't terminated, pin it.
-    if (!cur_thread->IsTerminationRequested()) {
-        // Pin it.
-        this->PinThread(core_id, cur_thread);
-        cur_thread->Pin(core_id);
+            while (!info->Close()) {
+                shmem->Close();
+            }
+            shmem->Close();
 
-        // An update is needed.
-        KScheduler::SetSchedulerUpdateNeeded(m_kernel);
+            it = m_shared_memory_list.erase(it);
+            KSharedMemoryInfo::Free(m_kernel, info);
+        }
     }
-}
 
-void KProcess::UnpinCurrentThread(s32 core_id) {
-    ASSERT(KScheduler::IsSchedulerLockedByCurrentThread(m_kernel));
-
-    // Get the current thread.
-    KThread* cur_thread =
-        m_kernel.Scheduler(static_cast<std::size_t>(core_id)).GetSchedulerCurrentThread();
+    // Our thread local page list must be empty at this point.
+    ASSERT(m_partially_used_tlp_tree.empty());
+    ASSERT(m_fully_used_tlp_tree.empty());
 
-    // Unpin it.
-    cur_thread->Unpin();
-    this->UnpinThread(core_id, cur_thread);
+    // Release memory to the resource limit.
+    if (m_resource_limit != nullptr) {
+        ASSERT(used_memory_size >= m_memory_release_hint);
+        m_resource_limit->Release(Svc::LimitableResource::PhysicalMemoryMax, used_memory_size,
+                                  used_memory_size - m_memory_release_hint);
+        m_resource_limit->Close();
+    }
 
-    // An update is needed.
-    KScheduler::SetSchedulerUpdateNeeded(m_kernel);
+    // Perform inherited finalization.
+    KSynchronizationObject::Finalize();
 }
 
-void KProcess::UnpinThread(KThread* thread) {
-    ASSERT(KScheduler::IsSchedulerLockedByCurrentThread(m_kernel));
-
-    // Get the thread's core id.
-    const auto core_id = thread->GetActiveCore();
+Result KProcess::Initialize(const Svc::CreateProcessParameter& params, KResourceLimit* res_limit,
+                            bool is_real) {
+    // TODO: remove this special case
+    if (is_real) {
+        // Create and clear the process local region.
+        R_TRY(this->CreateThreadLocalRegion(std::addressof(m_plr_address)));
+        this->GetMemory().ZeroBlock(m_plr_address, Svc::ThreadLocalRegionSize);
+    }
 
-    // Unpin it.
-    this->UnpinThread(core_id, thread);
-    thread->Unpin();
+    // Copy in the name from parameters.
+    static_assert(sizeof(params.name) < sizeof(m_name));
+    std::memcpy(m_name.data(), params.name.data(), sizeof(params.name));
+    m_name[sizeof(params.name)] = 0;
+
+    // Set misc fields.
+    m_state = State::Created;
+    m_main_thread_stack_size = 0;
+    m_used_kernel_memory_size = 0;
+    m_ideal_core_id = 0;
+    m_flags = params.flags;
+    m_version = params.version;
+    m_program_id = params.program_id;
+    m_code_address = params.code_address;
+    m_code_size = params.code_num_pages * PageSize;
+    m_is_application = True(params.flags & Svc::CreateProcessFlag::IsApplication);
+
+    // Set thread fields.
+    for (size_t i = 0; i < Core::Hardware::NUM_CPU_CORES; i++) {
+        m_running_threads[i] = nullptr;
+        m_pinned_threads[i] = nullptr;
+        m_running_thread_idle_counts[i] = 0;
+        m_running_thread_switch_counts[i] = 0;
+    }
 
-    // An update is needed.
-    KScheduler::SetSchedulerUpdateNeeded(m_kernel);
-}
+    // Set max memory based on address space type.
+    switch ((params.flags & Svc::CreateProcessFlag::AddressSpaceMask)) {
+    case Svc::CreateProcessFlag::AddressSpace32Bit:
+    case Svc::CreateProcessFlag::AddressSpace64BitDeprecated:
+    case Svc::CreateProcessFlag::AddressSpace64Bit:
+        m_max_process_memory = m_page_table.GetHeapRegionSize();
+        break;
+    case Svc::CreateProcessFlag::AddressSpace32BitWithoutAlias:
+        m_max_process_memory = m_page_table.GetHeapRegionSize() + m_page_table.GetAliasRegionSize();
+        break;
+    default:
+        UNREACHABLE();
+    }
 
-Result KProcess::AddSharedMemory(KSharedMemory* shmem, [[maybe_unused]] KProcessAddress address,
-                                 [[maybe_unused]] size_t size) {
-    // Lock ourselves, to prevent concurrent access.
-    KScopedLightLock lk(m_state_lock);
+    // Generate random entropy.
+    GenerateRandom(m_entropy);
 
-    // Try to find an existing info for the memory.
-    KSharedMemoryInfo* shemen_info = nullptr;
-    const auto iter = std::find_if(
-        m_shared_memory_list.begin(), m_shared_memory_list.end(),
-        [shmem](const KSharedMemoryInfo* info) { return info->GetSharedMemory() == shmem; });
-    if (iter != m_shared_memory_list.end()) {
-        shemen_info = *iter;
-    }
+    // Clear remaining fields.
+    m_num_running_threads = 0;
+    m_num_process_switches = 0;
+    m_num_thread_switches = 0;
+    m_num_fpu_switches = 0;
+    m_num_supervisor_calls = 0;
+    m_num_ipc_messages = 0;
 
-    if (shemen_info == nullptr) {
-        shemen_info = KSharedMemoryInfo::Allocate(m_kernel);
-        R_UNLESS(shemen_info != nullptr, ResultOutOfMemory);
+    m_is_signaled = false;
+    m_exception_thread = nullptr;
+    m_is_suspended = false;
+    m_memory_release_hint = 0;
+    m_schedule_count = 0;
+    m_is_handle_table_initialized = false;
 
-        shemen_info->Initialize(shmem);
-        m_shared_memory_list.push_back(shemen_info);
-    }
+    // Open a reference to our resource limit.
+    m_resource_limit = res_limit;
+    m_resource_limit->Open();
 
-    // Open a reference to the shared memory and its info.
-    shmem->Open();
-    shemen_info->Open();
+    // We're initialized!
+    m_is_initialized = true;
 
     R_SUCCEED();
 }
 
-void KProcess::RemoveSharedMemory(KSharedMemory* shmem, [[maybe_unused]] KProcessAddress address,
-                                  [[maybe_unused]] size_t size) {
-    // Lock ourselves, to prevent concurrent access.
-    KScopedLightLock lk(m_state_lock);
+Result KProcess::Initialize(const Svc::CreateProcessParameter& params, const KPageGroup& pg,
+                            std::span<const u32> caps, KResourceLimit* res_limit,
+                            KMemoryManager::Pool pool, bool immortal) {
+    ASSERT(res_limit != nullptr);
+    ASSERT((params.code_num_pages * PageSize) / PageSize ==
+           static_cast<size_t>(params.code_num_pages));
+
+    // Set members.
+    m_memory_pool = pool;
+    m_is_default_application_system_resource = false;
+    m_is_immortal = immortal;
+
+    // Setup our system resource.
+    if (const size_t system_resource_num_pages = params.system_resource_num_pages;
+        system_resource_num_pages != 0) {
+        // Create a secure system resource.
+        KSecureSystemResource* secure_resource = KSecureSystemResource::Create(m_kernel);
+        R_UNLESS(secure_resource != nullptr, ResultOutOfResource);
+
+        ON_RESULT_FAILURE {
+            secure_resource->Close();
+        };
+
+        // Initialize the secure resource.
+        R_TRY(secure_resource->Initialize(system_resource_num_pages * PageSize, res_limit,
+                                          m_memory_pool));
+
+        // Set our system resource.
+        m_system_resource = secure_resource;
+    } else {
+        // Use the system-wide system resource.
+        const bool is_app = True(params.flags & Svc::CreateProcessFlag::IsApplication);
+        m_system_resource = std::addressof(is_app ? m_kernel.GetAppSystemResource()
+                                                  : m_kernel.GetSystemSystemResource());
 
-    KSharedMemoryInfo* shemen_info = nullptr;
-    const auto iter = std::find_if(
-        m_shared_memory_list.begin(), m_shared_memory_list.end(),
-        [shmem](const KSharedMemoryInfo* info) { return info->GetSharedMemory() == shmem; });
-    if (iter != m_shared_memory_list.end()) {
-        shemen_info = *iter;
+        m_is_default_application_system_resource = is_app;
+
+        // Open reference to the system resource.
+        m_system_resource->Open();
     }
 
-    ASSERT(shemen_info != nullptr);
+    // Ensure we clean up our secure resource, if we fail.
+    ON_RESULT_FAILURE {
+        m_system_resource->Close();
+        m_system_resource = nullptr;
+    };
 
-    if (shemen_info->Close()) {
-        m_shared_memory_list.erase(iter);
-        KSharedMemoryInfo::Free(m_kernel, shemen_info);
+    // Setup page table.
+    {
+        const auto as_type = params.flags & Svc::CreateProcessFlag::AddressSpaceMask;
+        const bool enable_aslr = True(params.flags & Svc::CreateProcessFlag::EnableAslr);
+        const bool enable_das_merge =
+            False(params.flags & Svc::CreateProcessFlag::DisableDeviceAddressSpaceMerge);
+        R_TRY(m_page_table.InitializeForProcess(
+            as_type, enable_aslr, enable_das_merge, !enable_aslr, pool, params.code_address,
+            params.code_num_pages * PageSize, m_system_resource, res_limit, this->GetMemory()));
     }
+    ON_RESULT_FAILURE_2 {
+        m_page_table.Finalize();
+    };
 
-    // Close a reference to the shared memory.
-    shmem->Close();
-}
+    // Ensure we can insert the code region.
+    R_UNLESS(m_page_table.CanContain(params.code_address, params.code_num_pages * PageSize,
+                                     KMemoryState::Code),
+             ResultInvalidMemoryRegion);
 
-void KProcess::RegisterThread(KThread* thread) {
-    KScopedLightLock lk{m_list_lock};
+    // Map the code region.
+    R_TRY(m_page_table.MapPageGroup(params.code_address, pg, KMemoryState::Code,
+                                    KMemoryPermission::KernelRead));
 
-    m_thread_list.push_back(thread);
-}
+    // Initialize capabilities.
+    R_TRY(m_capabilities.InitializeForKip(caps, std::addressof(m_page_table)));
 
-void KProcess::UnregisterThread(KThread* thread) {
-    KScopedLightLock lk{m_list_lock};
+    // Initialize the process id.
+    m_process_id = m_kernel.CreateNewUserProcessID();
+    ASSERT(InitialProcessIdMin <= m_process_id);
+    ASSERT(m_process_id <= InitialProcessIdMax);
 
-    m_thread_list.remove(thread);
-}
+    // Initialize the rest of the process.
+    R_TRY(this->Initialize(params, res_limit, true));
 
-u64 KProcess::GetFreeThreadCount() const {
-    if (m_resource_limit != nullptr) {
-        const auto current_value =
-            m_resource_limit->GetCurrentValue(LimitableResource::ThreadCountMax);
-        const auto limit_value = m_resource_limit->GetLimitValue(LimitableResource::ThreadCountMax);
-        return limit_value - current_value;
-    } else {
-        return 0;
-    }
+    // We succeeded!
+    R_SUCCEED();
 }
 
-Result KProcess::Reset() {
-    // Lock the process and the scheduler.
-    KScopedLightLock lk(m_state_lock);
-    KScopedSchedulerLock sl{m_kernel};
+Result KProcess::Initialize(const Svc::CreateProcessParameter& params,
+                            std::span<const u32> user_caps, KResourceLimit* res_limit,
+                            KMemoryManager::Pool pool) {
+    ASSERT(res_limit != nullptr);
 
-    // Validate that we're in a state that we can reset.
-    R_UNLESS(m_state != State::Terminated, ResultInvalidState);
-    R_UNLESS(m_is_signaled, ResultInvalidState);
+    // Set members.
+    m_memory_pool = pool;
+    m_is_default_application_system_resource = false;
+    m_is_immortal = false;
 
-    // Clear signaled.
-    m_is_signaled = false;
-    R_SUCCEED();
-}
+    // Get the memory sizes.
+    const size_t code_num_pages = params.code_num_pages;
+    const size_t system_resource_num_pages = params.system_resource_num_pages;
+    const size_t code_size = code_num_pages * PageSize;
+    const size_t system_resource_size = system_resource_num_pages * PageSize;
 
-Result KProcess::SetActivity(ProcessActivity activity) {
-    // Lock ourselves and the scheduler.
-    KScopedLightLock lk{m_state_lock};
-    KScopedLightLock list_lk{m_list_lock};
-    KScopedSchedulerLock sl{m_kernel};
+    // Reserve memory for our code resource.
+    KScopedResourceReservation memory_reservation(
+        res_limit, Svc::LimitableResource::PhysicalMemoryMax, code_size);
+    R_UNLESS(memory_reservation.Succeeded(), ResultLimitReached);
 
-    // Validate our state.
-    R_UNLESS(m_state != State::Terminating, ResultInvalidState);
-    R_UNLESS(m_state != State::Terminated, ResultInvalidState);
+    // Setup our system resource.
+    if (system_resource_num_pages != 0) {
+        // Create a secure system resource.
+        KSecureSystemResource* secure_resource = KSecureSystemResource::Create(m_kernel);
+        R_UNLESS(secure_resource != nullptr, ResultOutOfResource);
 
-    // Either pause or resume.
-    if (activity == ProcessActivity::Paused) {
-        // Verify that we're not suspended.
-        R_UNLESS(!m_is_suspended, ResultInvalidState);
+        ON_RESULT_FAILURE {
+            secure_resource->Close();
+        };
 
-        // Suspend all threads.
-        for (auto* thread : this->GetThreadList()) {
-            thread->RequestSuspend(SuspendType::Process);
-        }
+        // Initialize the secure resource.
+        R_TRY(secure_resource->Initialize(system_resource_size, res_limit, m_memory_pool));
+
+        // Set our system resource.
+        m_system_resource = secure_resource;
 
-        // Set ourselves as suspended.
-        this->SetSuspended(true);
     } else {
-        ASSERT(activity == ProcessActivity::Runnable);
+        // Use the system-wide system resource.
+        const bool is_app = True(params.flags & Svc::CreateProcessFlag::IsApplication);
+        m_system_resource = std::addressof(is_app ? m_kernel.GetAppSystemResource()
+                                                  : m_kernel.GetSystemSystemResource());
 
-        // Verify that we're suspended.
-        R_UNLESS(m_is_suspended, ResultInvalidState);
+        m_is_default_application_system_resource = is_app;
 
-        // Resume all threads.
-        for (auto* thread : this->GetThreadList()) {
-            thread->Resume(SuspendType::Process);
-        }
+        // Open reference to the system resource.
+        m_system_resource->Open();
+    }
 
-        // Set ourselves as resumed.
-        this->SetSuspended(false);
+    // Ensure we clean up our secure resource, if we fail.
+    ON_RESULT_FAILURE {
+        m_system_resource->Close();
+        m_system_resource = nullptr;
+    };
+
+    // Setup page table.
+    {
+        const auto as_type = params.flags & Svc::CreateProcessFlag::AddressSpaceMask;
+        const bool enable_aslr = True(params.flags & Svc::CreateProcessFlag::EnableAslr);
+        const bool enable_das_merge =
+            False(params.flags & Svc::CreateProcessFlag::DisableDeviceAddressSpaceMerge);
+        R_TRY(m_page_table.InitializeForProcess(as_type, enable_aslr, enable_das_merge,
+                                                !enable_aslr, pool, params.code_address, code_size,
+                                                m_system_resource, res_limit, this->GetMemory()));
+    }
+    ON_RESULT_FAILURE_2 {
+        m_page_table.Finalize();
+    };
+
+    // Ensure we can insert the code region.
+    R_UNLESS(m_page_table.CanContain(params.code_address, code_size, KMemoryState::Code),
+             ResultInvalidMemoryRegion);
+
+    // Map the code region.
+    R_TRY(m_page_table.MapPages(params.code_address, code_num_pages, KMemoryState::Code,
+                                KMemoryPermission::KernelRead | KMemoryPermission::NotMapped));
+
+    // Initialize capabilities.
+    R_TRY(m_capabilities.InitializeForUser(user_caps, std::addressof(m_page_table)));
+
+    // Initialize the process id.
+    m_process_id = m_kernel.CreateNewUserProcessID();
+    ASSERT(ProcessIdMin <= m_process_id);
+    ASSERT(m_process_id <= ProcessIdMax);
+
+    // If we should optimize memory allocations, do so.
+    if (m_system_resource->IsSecureResource() &&
+        True(params.flags & Svc::CreateProcessFlag::OptimizeMemoryAllocation)) {
+        R_TRY(m_kernel.MemoryManager().InitializeOptimizedMemory(m_process_id, pool));
     }
 
+    // Initialize the rest of the process.
+    R_TRY(this->Initialize(params, res_limit, true));
+
+    // We succeeded, so commit our memory reservation.
+    memory_reservation.Commit();
     R_SUCCEED();
 }
 
-Result KProcess::LoadFromMetadata(const FileSys::ProgramMetadata& metadata, std::size_t code_size,
-                                  bool is_hbl) {
-    m_program_id = metadata.GetTitleID();
-    m_ideal_core = metadata.GetMainThreadCore();
-    m_is_64bit_process = metadata.Is64BitProgram();
-    m_system_resource_size = metadata.GetSystemResourceSize();
-    m_image_size = code_size;
-    m_is_hbl = is_hbl;
+void KProcess::DoWorkerTaskImpl() {
+    // Terminate child threads.
+    TerminateChildren(m_kernel, this, nullptr);
 
-    if (metadata.GetAddressSpaceType() == FileSys::ProgramAddressSpaceType::Is39Bit) {
-        // For 39-bit processes, the ASLR region starts at 0x800'0000 and is ~512GiB large.
-        // However, some (buggy) programs/libraries like skyline incorrectly depend on the
-        // existence of ASLR pages before the entry point, so we will adjust the load address
-        // to point to about 2GiB into the ASLR region.
-        m_code_address = 0x8000'0000;
-    } else {
-        // All other processes can be mapped at the beginning of the code region.
-        if (metadata.GetAddressSpaceType() == FileSys::ProgramAddressSpaceType::Is36Bit) {
-            m_code_address = 0x800'0000;
-        } else {
-            m_code_address = 0x20'0000;
-        }
+    // Finalize the handle table, if we're not immortal.
+    if (!m_is_immortal && m_is_handle_table_initialized) {
+        this->FinalizeHandleTable();
     }
 
-    KScopedResourceReservation memory_reservation(
-        m_resource_limit, LimitableResource::PhysicalMemoryMax, code_size + m_system_resource_size);
-    if (!memory_reservation.Succeeded()) {
-        LOG_ERROR(Kernel, "Could not reserve process memory requirements of size {:X} bytes",
-                  code_size + m_system_resource_size);
-        R_RETURN(ResultLimitReached);
-    }
-    // Initialize process address space
-    if (const Result result{m_page_table.InitializeForProcess(
-            metadata.GetAddressSpaceType(), false, false, false, KMemoryManager::Pool::Application,
-            this->GetEntryPoint(), code_size, std::addressof(m_kernel.GetAppSystemResource()),
-            m_resource_limit, m_kernel.System().ApplicationMemory())};
-        result.IsError()) {
-        R_RETURN(result);
-    }
-
-    // Map process code region
-    if (const Result result{m_page_table.MapProcessCode(this->GetEntryPoint(), code_size / PageSize,
-                                                        KMemoryState::Code,
-                                                        KMemoryPermission::None)};
-        result.IsError()) {
-        R_RETURN(result);
-    }
-
-    // Initialize process capabilities
-    const auto& caps{metadata.GetKernelCapabilities()};
-    if (const Result result{
-            m_capabilities.InitializeForUserProcess(caps.data(), caps.size(), m_page_table)};
-        result.IsError()) {
-        R_RETURN(result);
-    }
-
-    // Set memory usage capacity
-    switch (metadata.GetAddressSpaceType()) {
-    case FileSys::ProgramAddressSpaceType::Is32Bit:
-    case FileSys::ProgramAddressSpaceType::Is36Bit:
-    case FileSys::ProgramAddressSpaceType::Is39Bit:
-        m_memory_usage_capacity =
-            m_page_table.GetHeapRegionEnd() - m_page_table.GetHeapRegionStart();
-        break;
+    // Finish termination.
+    this->FinishTermination();
+}
 
-    case FileSys::ProgramAddressSpaceType::Is32BitNoMap:
-        m_memory_usage_capacity =
-            (m_page_table.GetHeapRegionEnd() - m_page_table.GetHeapRegionStart()) +
-            (m_page_table.GetAliasRegionEnd() - m_page_table.GetAliasRegionStart());
-        break;
+Result KProcess::StartTermination() {
+    // Finalize the handle table when we're done, if the process isn't immortal.
+    SCOPE_EXIT({
+        if (!m_is_immortal) {
+            this->FinalizeHandleTable();
+        }
+    });
 
-    default:
-        ASSERT(false);
-        break;
-    }
+    // Terminate child threads other than the current one.
+    R_RETURN(TerminateChildren(m_kernel, this, GetCurrentThreadPointer(m_kernel)));
+}
 
-    // Create TLS region
-    R_TRY(this->CreateThreadLocalRegion(std::addressof(m_plr_address)));
-    memory_reservation.Commit();
+void KProcess::FinishTermination() {
+    // Only allow termination to occur if the process isn't immortal.
+    if (!m_is_immortal) {
+        // Release resource limit hint.
+        if (m_resource_limit != nullptr) {
+            m_memory_release_hint = this->GetUsedNonSystemUserPhysicalMemorySize();
+            m_resource_limit->Release(Svc::LimitableResource::PhysicalMemoryMax, 0,
+                                      m_memory_release_hint);
+        }
+
+        // Change state.
+        {
+            KScopedSchedulerLock sl(m_kernel);
+            this->ChangeState(State::Terminated);
+        }
 
-    R_RETURN(m_handle_table.Initialize(m_capabilities.GetHandleTableSize()));
+        // Close.
+        this->Close();
+    }
 }
 
-void KProcess::Run(s32 main_thread_priority, u64 stack_size) {
-    ASSERT(this->AllocateMainThreadStack(stack_size) == ResultSuccess);
-    m_resource_limit->Reserve(LimitableResource::ThreadCountMax, 1);
+void KProcess::Exit() {
+    // Determine whether we need to start terminating
+    bool needs_terminate = false;
+    {
+        KScopedLightLock lk(m_state_lock);
+        KScopedSchedulerLock sl(m_kernel);
+
+        ASSERT(m_state != State::Created);
+        ASSERT(m_state != State::CreatedAttached);
+        ASSERT(m_state != State::Crashed);
+        ASSERT(m_state != State::Terminated);
+        if (m_state == State::Running || m_state == State::RunningAttached ||
+            m_state == State::DebugBreak) {
+            this->ChangeState(State::Terminating);
+            needs_terminate = true;
+        }
+    }
 
-    const std::size_t heap_capacity{m_memory_usage_capacity -
-                                    (m_main_thread_stack_size + m_image_size)};
-    ASSERT(!m_page_table.SetMaxHeapSize(heap_capacity).IsError());
+    // If we need to start termination, do so.
+    if (needs_terminate) {
+        this->StartTermination();
 
-    this->ChangeState(State::Running);
+        // Register the process as a work task.
+        m_kernel.WorkerTaskManager().AddTask(m_kernel, KWorkerTaskManager::WorkerType::Exit, this);
+    }
 
-    SetupMainThread(m_kernel.System(), *this, main_thread_priority, m_main_thread_stack_top);
+    // Exit the current thread.
+    GetCurrentThread(m_kernel).Exit();
 }
 
-void KProcess::PrepareForTermination() {
-    this->ChangeState(State::Terminating);
+Result KProcess::Terminate() {
+    // Determine whether we need to start terminating.
+    bool needs_terminate = false;
+    {
+        KScopedLightLock lk(m_state_lock);
 
-    const auto stop_threads = [this](const std::vector<KThread*>& in_thread_list) {
-        for (auto* thread : in_thread_list) {
-            if (thread->GetOwnerProcess() != this)
-                continue;
+        // Check whether we're allowed to terminate.
+        R_UNLESS(m_state != State::Created, ResultInvalidState);
+        R_UNLESS(m_state != State::CreatedAttached, ResultInvalidState);
 
-            if (thread == GetCurrentThreadPointer(m_kernel))
-                continue;
+        KScopedSchedulerLock sl(m_kernel);
 
-            // TODO(Subv): When are the other running/ready threads terminated?
-            ASSERT_MSG(thread->GetState() == ThreadState::Waiting,
-                       "Exiting processes with non-waiting threads is currently unimplemented");
+        if (m_state == State::Running || m_state == State::RunningAttached ||
+            m_state == State::Crashed || m_state == State::DebugBreak) {
+            this->ChangeState(State::Terminating);
+            needs_terminate = true;
+        }
+    }
 
-            thread->Exit();
+    // If we need to terminate, do so.
+    if (needs_terminate) {
+        // Start termination.
+        if (R_SUCCEEDED(this->StartTermination())) {
+            // Finish termination.
+            this->FinishTermination();
+        } else {
+            // Register the process as a work task.
+            m_kernel.WorkerTaskManager().AddTask(m_kernel, KWorkerTaskManager::WorkerType::Exit,
+                                                 this);
         }
-    };
+    }
 
-    stop_threads(m_kernel.System().GlobalSchedulerContext().GetThreadList());
+    R_SUCCEED();
+}
 
-    this->DeleteThreadLocalRegion(m_plr_address);
-    m_plr_address = 0;
+Result KProcess::AddSharedMemory(KSharedMemory* shmem, KProcessAddress address, size_t size) {
+    // Lock ourselves, to prevent concurrent access.
+    KScopedLightLock lk(m_state_lock);
 
-    if (m_resource_limit) {
-        m_resource_limit->Release(LimitableResource::PhysicalMemoryMax,
-                                  m_main_thread_stack_size + m_image_size);
+    // Try to find an existing info for the memory.
+    KSharedMemoryInfo* info = nullptr;
+    for (auto it = m_shared_memory_list.begin(); it != m_shared_memory_list.end(); ++it) {
+        if (it->GetSharedMemory() == shmem) {
+            info = std::addressof(*it);
+            break;
+        }
     }
 
-    this->ChangeState(State::Terminated);
-}
+    // If we didn't find an info, create one.
+    if (info == nullptr) {
+        // Allocate a new info.
+        info = KSharedMemoryInfo::Allocate(m_kernel);
+        R_UNLESS(info != nullptr, ResultOutOfResource);
 
-void KProcess::Finalize() {
-    // Free all shared memory infos.
-    {
-        auto it = m_shared_memory_list.begin();
-        while (it != m_shared_memory_list.end()) {
-            KSharedMemoryInfo* info = *it;
-            KSharedMemory* shmem = info->GetSharedMemory();
+        // Initialize the info and add it to our list.
+        info->Initialize(shmem);
+        m_shared_memory_list.push_back(*info);
+    }
 
-            while (!info->Close()) {
-                shmem->Close();
-            }
+    // Open a reference to the shared memory and its info.
+    shmem->Open();
+    info->Open();
 
-            shmem->Close();
+    R_SUCCEED();
+}
 
-            it = m_shared_memory_list.erase(it);
-            KSharedMemoryInfo::Free(m_kernel, info);
+void KProcess::RemoveSharedMemory(KSharedMemory* shmem, KProcessAddress address, size_t size) {
+    // Lock ourselves, to prevent concurrent access.
+    KScopedLightLock lk(m_state_lock);
+
+    // Find an existing info for the memory.
+    KSharedMemoryInfo* info = nullptr;
+    auto it = m_shared_memory_list.begin();
+    for (; it != m_shared_memory_list.end(); ++it) {
+        if (it->GetSharedMemory() == shmem) {
+            info = std::addressof(*it);
+            break;
         }
     }
+    ASSERT(info != nullptr);
 
-    // Release memory to the resource limit.
-    if (m_resource_limit != nullptr) {
-        m_resource_limit->Close();
-        m_resource_limit = nullptr;
+    // Close a reference to the info and its memory.
+    if (info->Close()) {
+        m_shared_memory_list.erase(it);
+        KSharedMemoryInfo::Free(m_kernel, info);
     }
 
-    // Finalize the page table.
-    m_page_table.Finalize();
-
-    // Perform inherited finalization.
-    KSynchronizationObject::Finalize();
+    shmem->Close();
 }
 
 Result KProcess::CreateThreadLocalRegion(KProcessAddress* out) {
@@ -518,7 +601,7 @@ Result KProcess::CreateThreadLocalRegion(KProcessAddress* out) {
 
     // See if we can get a region from a partially used TLP.
     {
-        KScopedSchedulerLock sl{m_kernel};
+        KScopedSchedulerLock sl(m_kernel);
 
         if (auto it = m_partially_used_tlp_tree.begin(); it != m_partially_used_tlp_tree.end()) {
             tlr = it->Reserve();
@@ -538,7 +621,9 @@ Result KProcess::CreateThreadLocalRegion(KProcessAddress* out) {
     // Allocate a new page.
     tlp = KThreadLocalPage::Allocate(m_kernel);
     R_UNLESS(tlp != nullptr, ResultOutOfMemory);
-    auto tlp_guard = SCOPE_GUARD({ KThreadLocalPage::Free(m_kernel, tlp); });
+    ON_RESULT_FAILURE {
+        KThreadLocalPage::Free(m_kernel, tlp);
+    };
 
     // Initialize the new page.
     R_TRY(tlp->Initialize(m_kernel, this));
@@ -549,7 +634,7 @@ Result KProcess::CreateThreadLocalRegion(KProcessAddress* out) {
 
     // Insert into our tree.
     {
-        KScopedSchedulerLock sl{m_kernel};
+        KScopedSchedulerLock sl(m_kernel);
         if (tlp->IsAllUsed()) {
             m_fully_used_tlp_tree.insert(*tlp);
         } else {
@@ -558,7 +643,6 @@ Result KProcess::CreateThreadLocalRegion(KProcessAddress* out) {
     }
 
     // We succeeded!
-    tlp_guard.Cancel();
     *out = tlr;
     R_SUCCEED();
 }
@@ -568,7 +652,7 @@ Result KProcess::DeleteThreadLocalRegion(KProcessAddress addr) {
 
     // Release the region.
     {
-        KScopedSchedulerLock sl{m_kernel};
+        KScopedSchedulerLock sl(m_kernel);
 
         // Try to find the page in the partially used list.
         auto it = m_partially_used_tlp_tree.find_key(Common::AlignDown(GetInteger(addr), PageSize));
@@ -611,95 +695,213 @@ Result KProcess::DeleteThreadLocalRegion(KProcessAddress addr) {
     R_SUCCEED();
 }
 
-bool KProcess::InsertWatchpoint(KProcessAddress addr, u64 size, DebugWatchpointType type) {
-    const auto watch{std::find_if(m_watchpoints.begin(), m_watchpoints.end(), [&](const auto& wp) {
-        return wp.type == DebugWatchpointType::None;
-    })};
+bool KProcess::ReserveResource(Svc::LimitableResource which, s64 value) {
+    if (KResourceLimit* rl = this->GetResourceLimit(); rl != nullptr) {
+        return rl->Reserve(which, value);
+    } else {
+        return true;
+    }
+}
 
-    if (watch == m_watchpoints.end()) {
-        return false;
+bool KProcess::ReserveResource(Svc::LimitableResource which, s64 value, s64 timeout) {
+    if (KResourceLimit* rl = this->GetResourceLimit(); rl != nullptr) {
+        return rl->Reserve(which, value, timeout);
+    } else {
+        return true;
     }
+}
 
-    watch->start_address = addr;
-    watch->end_address = addr + size;
-    watch->type = type;
+void KProcess::ReleaseResource(Svc::LimitableResource which, s64 value) {
+    if (KResourceLimit* rl = this->GetResourceLimit(); rl != nullptr) {
+        rl->Release(which, value);
+    }
+}
 
-    for (KProcessAddress page = Common::AlignDown(GetInteger(addr), PageSize); page < addr + size;
-         page += PageSize) {
-        m_debug_page_refcounts[page]++;
-        this->GetMemory().MarkRegionDebug(page, PageSize, true);
+void KProcess::ReleaseResource(Svc::LimitableResource which, s64 value, s64 hint) {
+    if (KResourceLimit* rl = this->GetResourceLimit(); rl != nullptr) {
+        rl->Release(which, value, hint);
     }
+}
 
-    return true;
+void KProcess::IncrementRunningThreadCount() {
+    ASSERT(m_num_running_threads.load() >= 0);
+
+    ++m_num_running_threads;
 }
 
-bool KProcess::RemoveWatchpoint(KProcessAddress addr, u64 size, DebugWatchpointType type) {
-    const auto watch{std::find_if(m_watchpoints.begin(), m_watchpoints.end(), [&](const auto& wp) {
-        return wp.start_address == addr && wp.end_address == addr + size && wp.type == type;
-    })};
+void KProcess::DecrementRunningThreadCount() {
+    ASSERT(m_num_running_threads.load() > 0);
 
-    if (watch == m_watchpoints.end()) {
+    if (const auto prev = m_num_running_threads--; prev == 1) {
+        this->Terminate();
+    }
+}
+
+bool KProcess::EnterUserException() {
+    // Get the current thread.
+    KThread* cur_thread = GetCurrentThreadPointer(m_kernel);
+    ASSERT(this == cur_thread->GetOwnerProcess());
+
+    // Check that we haven't already claimed the exception thread.
+    if (m_exception_thread == cur_thread) {
         return false;
     }
 
-    watch->start_address = 0;
-    watch->end_address = 0;
-    watch->type = DebugWatchpointType::None;
+    // Create the wait queue we'll be using.
+    ThreadQueueImplForKProcessEnterUserException wait_queue(m_kernel,
+                                                            std::addressof(m_exception_thread));
 
-    for (KProcessAddress page = Common::AlignDown(GetInteger(addr), PageSize); page < addr + size;
-         page += PageSize) {
-        m_debug_page_refcounts[page]--;
-        if (!m_debug_page_refcounts[page]) {
-            this->GetMemory().MarkRegionDebug(page, PageSize, false);
+    // Claim the exception thread.
+    {
+        // Lock the scheduler.
+        KScopedSchedulerLock sl(m_kernel);
+
+        // Check that we're not terminating.
+        if (cur_thread->IsTerminationRequested()) {
+            return false;
+        }
+
+        // If we don't have an exception thread, we can just claim it directly.
+        if (m_exception_thread == nullptr) {
+            m_exception_thread = cur_thread;
+            KScheduler::SetSchedulerUpdateNeeded(m_kernel);
+            return true;
         }
+
+        // Otherwise, we need to wait until we don't have an exception thread.
+
+        // Add the current thread as a waiter on the current exception thread.
+        cur_thread->SetKernelAddressKey(
+            reinterpret_cast<uintptr_t>(std::addressof(m_exception_thread)) | 1);
+        m_exception_thread->AddWaiter(cur_thread);
+
+        // Wait to claim the exception thread.
+        cur_thread->BeginWait(std::addressof(wait_queue));
     }
 
-    return true;
+    // If our wait didn't end due to thread termination, we succeeded.
+    return ResultTerminationRequested != cur_thread->GetWaitResult();
 }
 
-void KProcess::LoadModule(CodeSet code_set, KProcessAddress base_addr) {
-    const auto ReprotectSegment = [&](const CodeSet::Segment& segment,
-                                      Svc::MemoryPermission permission) {
-        m_page_table.SetProcessMemoryPermission(segment.addr + base_addr, segment.size, permission);
-    };
+bool KProcess::LeaveUserException() {
+    return this->ReleaseUserException(GetCurrentThreadPointer(m_kernel));
+}
 
-    this->GetMemory().WriteBlock(base_addr, code_set.memory.data(), code_set.memory.size());
+bool KProcess::ReleaseUserException(KThread* thread) {
+    KScopedSchedulerLock sl(m_kernel);
 
-    ReprotectSegment(code_set.CodeSegment(), Svc::MemoryPermission::ReadExecute);
-    ReprotectSegment(code_set.RODataSegment(), Svc::MemoryPermission::Read);
-    ReprotectSegment(code_set.DataSegment(), Svc::MemoryPermission::ReadWrite);
+    if (m_exception_thread == thread) {
+        m_exception_thread = nullptr;
+
+        // Remove waiter thread.
+        bool has_waiters;
+        if (KThread* next = thread->RemoveKernelWaiterByKey(
+                std::addressof(has_waiters),
+                reinterpret_cast<uintptr_t>(std::addressof(m_exception_thread)) | 1);
+            next != nullptr) {
+            next->EndWait(ResultSuccess);
+        }
+
+        KScheduler::SetSchedulerUpdateNeeded(m_kernel);
+
+        return true;
+    } else {
+        return false;
+    }
 }
 
-bool KProcess::IsSignaled() const {
-    ASSERT(KScheduler::IsSchedulerLockedByCurrentThread(m_kernel));
-    return m_is_signaled;
+void KProcess::RegisterThread(KThread* thread) {
+    KScopedLightLock lk(m_list_lock);
+
+    m_thread_list.push_back(*thread);
 }
 
-KProcess::KProcess(KernelCore& kernel)
-    : KAutoObjectWithSlabHeapAndContainer{kernel}, m_page_table{m_kernel.System()},
-      m_handle_table{m_kernel}, m_address_arbiter{m_kernel.System()},
-      m_condition_var{m_kernel.System()}, m_state_lock{m_kernel}, m_list_lock{m_kernel} {}
+void KProcess::UnregisterThread(KThread* thread) {
+    KScopedLightLock lk(m_list_lock);
 
-KProcess::~KProcess() = default;
+    m_thread_list.erase(m_thread_list.iterator_to(*thread));
+}
+
+size_t KProcess::GetUsedUserPhysicalMemorySize() const {
+    const size_t norm_size = m_page_table.GetNormalMemorySize();
+    const size_t other_size = m_code_size + m_main_thread_stack_size;
+    const size_t sec_size = this->GetRequiredSecureMemorySizeNonDefault();
 
-void KProcess::ChangeState(State new_state) {
-    if (m_state == new_state) {
-        return;
+    return norm_size + other_size + sec_size;
+}
+
+size_t KProcess::GetTotalUserPhysicalMemorySize() const {
+    // Get the amount of free and used size.
+    const size_t free_size =
+        m_resource_limit->GetFreeValue(Svc::LimitableResource::PhysicalMemoryMax);
+    const size_t max_size = m_max_process_memory;
+
+    // Determine used size.
+    // NOTE: This does *not* check this->IsDefaultApplicationSystemResource(), unlike
+    // GetUsedUserPhysicalMemorySize().
+    const size_t norm_size = m_page_table.GetNormalMemorySize();
+    const size_t other_size = m_code_size + m_main_thread_stack_size;
+    const size_t sec_size = this->GetRequiredSecureMemorySize();
+    const size_t used_size = norm_size + other_size + sec_size;
+
+    // NOTE: These function calls will recalculate, introducing a race...it is unclear why Nintendo
+    // does it this way.
+    if (used_size + free_size > max_size) {
+        return max_size;
+    } else {
+        return free_size + this->GetUsedUserPhysicalMemorySize();
     }
+}
 
-    m_state = new_state;
-    m_is_signaled = true;
-    this->NotifyAvailable();
+size_t KProcess::GetUsedNonSystemUserPhysicalMemorySize() const {
+    const size_t norm_size = m_page_table.GetNormalMemorySize();
+    const size_t other_size = m_code_size + m_main_thread_stack_size;
+
+    return norm_size + other_size;
+}
+
+size_t KProcess::GetTotalNonSystemUserPhysicalMemorySize() const {
+    // Get the amount of free and used size.
+    const size_t free_size =
+        m_resource_limit->GetFreeValue(Svc::LimitableResource::PhysicalMemoryMax);
+    const size_t max_size = m_max_process_memory;
+
+    // Determine used size.
+    // NOTE: This does *not* check this->IsDefaultApplicationSystemResource(), unlike
+    // GetUsedUserPhysicalMemorySize().
+    const size_t norm_size = m_page_table.GetNormalMemorySize();
+    const size_t other_size = m_code_size + m_main_thread_stack_size;
+    const size_t sec_size = this->GetRequiredSecureMemorySize();
+    const size_t used_size = norm_size + other_size + sec_size;
+
+    // NOTE: These function calls will recalculate, introducing a race...it is unclear why Nintendo
+    // does it this way.
+    if (used_size + free_size > max_size) {
+        return max_size - this->GetRequiredSecureMemorySizeNonDefault();
+    } else {
+        return free_size + this->GetUsedNonSystemUserPhysicalMemorySize();
+    }
 }
 
-Result KProcess::AllocateMainThreadStack(std::size_t stack_size) {
+Result KProcess::Run(s32 priority, size_t stack_size) {
+    // Lock ourselves, to prevent concurrent access.
+    KScopedLightLock lk(m_state_lock);
+
+    // Validate that we're in a state where we can initialize.
+    const auto state = m_state;
+    R_UNLESS(state == State::Created || state == State::CreatedAttached, ResultInvalidState);
+
+    // Place a tentative reservation of a thread for this process.
+    KScopedResourceReservation thread_reservation(this, Svc::LimitableResource::ThreadCountMax);
+    R_UNLESS(thread_reservation.Succeeded(), ResultLimitReached);
+
     // Ensure that we haven't already allocated stack.
     ASSERT(m_main_thread_stack_size == 0);
 
     // Ensure that we're allocating a valid stack.
     stack_size = Common::AlignUp(stack_size, PageSize);
-    // R_UNLESS(stack_size + image_size <= m_max_process_memory, ResultOutOfMemory);
-    R_UNLESS(stack_size + m_image_size >= m_image_size, ResultOutOfMemory);
+    R_UNLESS(stack_size + m_code_size <= m_max_process_memory, ResultOutOfMemory);
+    R_UNLESS(stack_size + m_code_size >= m_code_size, ResultOutOfMemory);
 
     // Place a tentative reservation of memory for our new stack.
     KScopedResourceReservation mem_reservation(this, Svc::LimitableResource::PhysicalMemoryMax,
@@ -707,21 +909,359 @@ Result KProcess::AllocateMainThreadStack(std::size_t stack_size) {
     R_UNLESS(mem_reservation.Succeeded(), ResultLimitReached);
 
     // Allocate and map our stack.
+    KProcessAddress stack_top = 0;
     if (stack_size) {
         KProcessAddress stack_bottom;
         R_TRY(m_page_table.MapPages(std::addressof(stack_bottom), stack_size / PageSize,
                                     KMemoryState::Stack, KMemoryPermission::UserReadWrite));
 
-        m_main_thread_stack_top = stack_bottom + stack_size;
+        stack_top = stack_bottom + stack_size;
         m_main_thread_stack_size = stack_size;
     }
 
+    // Ensure our stack is safe to clean up on exit.
+    ON_RESULT_FAILURE {
+        if (m_main_thread_stack_size) {
+            ASSERT(R_SUCCEEDED(m_page_table.UnmapPages(stack_top - m_main_thread_stack_size,
+                                                       m_main_thread_stack_size / PageSize,
+                                                       KMemoryState::Stack)));
+            m_main_thread_stack_size = 0;
+        }
+    };
+
+    // Set our maximum heap size.
+    R_TRY(m_page_table.SetMaxHeapSize(m_max_process_memory -
+                                      (m_main_thread_stack_size + m_code_size)));
+
+    // Initialize our handle table.
+    R_TRY(this->InitializeHandleTable(m_capabilities.GetHandleTableSize()));
+    ON_RESULT_FAILURE_2 {
+        this->FinalizeHandleTable();
+    };
+
+    // Create a new thread for the process.
+    KThread* main_thread = KThread::Create(m_kernel);
+    R_UNLESS(main_thread != nullptr, ResultOutOfResource);
+    SCOPE_EXIT({ main_thread->Close(); });
+
+    // Initialize the thread.
+    R_TRY(KThread::InitializeUserThread(m_kernel.System(), main_thread, this->GetEntryPoint(), 0,
+                                        stack_top, priority, m_ideal_core_id, this));
+
+    // Register the thread, and commit our reservation.
+    KThread::Register(m_kernel, main_thread);
+    thread_reservation.Commit();
+
+    // Add the thread to our handle table.
+    Handle thread_handle;
+    R_TRY(m_handle_table.Add(std::addressof(thread_handle), main_thread));
+
+    // Set the thread arguments.
+    main_thread->GetContext32().cpu_registers[0] = 0;
+    main_thread->GetContext64().cpu_registers[0] = 0;
+    main_thread->GetContext32().cpu_registers[1] = thread_handle;
+    main_thread->GetContext64().cpu_registers[1] = thread_handle;
+
+    // Update our state.
+    this->ChangeState((state == State::Created) ? State::Running : State::RunningAttached);
+    ON_RESULT_FAILURE_2 {
+        this->ChangeState(state);
+    };
+
+    // Suspend for debug, if we should.
+    if (m_kernel.System().DebuggerEnabled()) {
+        main_thread->RequestSuspend(SuspendType::Debug);
+    }
+
+    // Run our thread.
+    R_TRY(main_thread->Run());
+
+    // Open a reference to represent that we're running.
+    this->Open();
+
     // We succeeded! Commit our memory reservation.
     mem_reservation.Commit();
 
     R_SUCCEED();
 }
 
+Result KProcess::Reset() {
+    // Lock the process and the scheduler.
+    KScopedLightLock lk(m_state_lock);
+    KScopedSchedulerLock sl(m_kernel);
+
+    // Validate that we're in a state that we can reset.
+    R_UNLESS(m_state != State::Terminated, ResultInvalidState);
+    R_UNLESS(m_is_signaled, ResultInvalidState);
+
+    // Clear signaled.
+    m_is_signaled = false;
+    R_SUCCEED();
+}
+
+Result KProcess::SetActivity(Svc::ProcessActivity activity) {
+    // Lock ourselves and the scheduler.
+    KScopedLightLock lk(m_state_lock);
+    KScopedLightLock list_lk(m_list_lock);
+    KScopedSchedulerLock sl(m_kernel);
+
+    // Validate our state.
+    R_UNLESS(m_state != State::Terminating, ResultInvalidState);
+    R_UNLESS(m_state != State::Terminated, ResultInvalidState);
+
+    // Either pause or resume.
+    if (activity == Svc::ProcessActivity::Paused) {
+        // Verify that we're not suspended.
+        R_UNLESS(!m_is_suspended, ResultInvalidState);
+
+        // Suspend all threads.
+        auto end = this->GetThreadList().end();
+        for (auto it = this->GetThreadList().begin(); it != end; ++it) {
+            it->RequestSuspend(SuspendType::Process);
+        }
+
+        // Set ourselves as suspended.
+        this->SetSuspended(true);
+    } else {
+        ASSERT(activity == Svc::ProcessActivity::Runnable);
+
+        // Verify that we're suspended.
+        R_UNLESS(m_is_suspended, ResultInvalidState);
+
+        // Resume all threads.
+        auto end = this->GetThreadList().end();
+        for (auto it = this->GetThreadList().begin(); it != end; ++it) {
+            it->Resume(SuspendType::Process);
+        }
+
+        // Set ourselves as resumed.
+        this->SetSuspended(false);
+    }
+
+    R_SUCCEED();
+}
+
+void KProcess::PinCurrentThread() {
+    ASSERT(KScheduler::IsSchedulerLockedByCurrentThread(m_kernel));
+
+    // Get the current thread.
+    const s32 core_id = GetCurrentCoreId(m_kernel);
+    KThread* cur_thread = GetCurrentThreadPointer(m_kernel);
+
+    // If the thread isn't terminated, pin it.
+    if (!cur_thread->IsTerminationRequested()) {
+        // Pin it.
+        this->PinThread(core_id, cur_thread);
+        cur_thread->Pin(core_id);
+
+        // An update is needed.
+        KScheduler::SetSchedulerUpdateNeeded(m_kernel);
+    }
+}
+
+void KProcess::UnpinCurrentThread() {
+    ASSERT(KScheduler::IsSchedulerLockedByCurrentThread(m_kernel));
+
+    // Get the current thread.
+    const s32 core_id = GetCurrentCoreId(m_kernel);
+    KThread* cur_thread = GetCurrentThreadPointer(m_kernel);
+
+    // Unpin it.
+    cur_thread->Unpin();
+    this->UnpinThread(core_id, cur_thread);
+
+    // An update is needed.
+    KScheduler::SetSchedulerUpdateNeeded(m_kernel);
+}
+
+void KProcess::UnpinThread(KThread* thread) {
+    ASSERT(KScheduler::IsSchedulerLockedByCurrentThread(m_kernel));
+
+    // Get the thread's core id.
+    const auto core_id = thread->GetActiveCore();
+
+    // Unpin it.
+    this->UnpinThread(core_id, thread);
+    thread->Unpin();
+
+    // An update is needed.
+    KScheduler::SetSchedulerUpdateNeeded(m_kernel);
+}
+
+Result KProcess::GetThreadList(s32* out_num_threads, KProcessAddress out_thread_ids,
+                               s32 max_out_count) {
+    // TODO: use current memory reference
+    auto& memory = m_kernel.System().ApplicationMemory();
+
+    // Lock the list.
+    KScopedLightLock lk(m_list_lock);
+
+    // Iterate over the list.
+    s32 count = 0;
+    auto end = this->GetThreadList().end();
+    for (auto it = this->GetThreadList().begin(); it != end; ++it) {
+        // If we're within array bounds, write the id.
+        if (count < max_out_count) {
+            // Get the thread id.
+            KThread* thread = std::addressof(*it);
+            const u64 id = thread->GetId();
+
+            // Copy the id to userland.
+            memory.Write64(out_thread_ids + count * sizeof(u64), id);
+        }
+
+        // Increment the count.
+        ++count;
+    }
+
+    // We successfully iterated the list.
+    *out_num_threads = count;
+    R_SUCCEED();
+}
+
+void KProcess::Switch(KProcess* cur_process, KProcess* next_process) {}
+
+KProcess::KProcess(KernelCore& kernel)
+    : KAutoObjectWithSlabHeapAndContainer(kernel), m_page_table{kernel.System()},
+      m_state_lock{kernel}, m_list_lock{kernel}, m_cond_var{kernel.System()},
+      m_address_arbiter{kernel.System()}, m_handle_table{kernel} {}
+KProcess::~KProcess() = default;
+
+Result KProcess::LoadFromMetadata(const FileSys::ProgramMetadata& metadata, std::size_t code_size,
+                                  bool is_hbl) {
+    // Create a resource limit for the process.
+    const auto physical_memory_size =
+        m_kernel.MemoryManager().GetSize(Kernel::KMemoryManager::Pool::Application);
+    auto* res_limit =
+        Kernel::CreateResourceLimitForProcess(m_kernel.System(), physical_memory_size);
+
+    // Ensure we maintain a clean state on exit.
+    SCOPE_EXIT({ res_limit->Close(); });
+
+    // Declare flags and code address.
+    Svc::CreateProcessFlag flag{};
+    u64 code_address{};
+
+    // We are an application.
+    flag |= Svc::CreateProcessFlag::IsApplication;
+
+    // If we are 64-bit, create as such.
+    if (metadata.Is64BitProgram()) {
+        flag |= Svc::CreateProcessFlag::Is64Bit;
+    }
+
+    // Set the address space type and code address.
+    switch (metadata.GetAddressSpaceType()) {
+    case FileSys::ProgramAddressSpaceType::Is39Bit:
+        flag |= Svc::CreateProcessFlag::AddressSpace64Bit;
+
+        // For 39-bit processes, the ASLR region starts at 0x800'0000 and is ~512GiB large.
+        // However, some (buggy) programs/libraries like skyline incorrectly depend on the
+        // existence of ASLR pages before the entry point, so we will adjust the load address
+        // to point to about 2GiB into the ASLR region.
+        code_address = 0x8000'0000;
+        break;
+    case FileSys::ProgramAddressSpaceType::Is36Bit:
+        flag |= Svc::CreateProcessFlag::AddressSpace64BitDeprecated;
+        code_address = 0x800'0000;
+        break;
+    case FileSys::ProgramAddressSpaceType::Is32Bit:
+        flag |= Svc::CreateProcessFlag::AddressSpace32Bit;
+        code_address = 0x20'0000;
+        break;
+    case FileSys::ProgramAddressSpaceType::Is32BitNoMap:
+        flag |= Svc::CreateProcessFlag::AddressSpace32BitWithoutAlias;
+        code_address = 0x20'0000;
+        break;
+    }
+
+    Svc::CreateProcessParameter params{
+        .name = {},
+        .version = {},
+        .program_id = metadata.GetTitleID(),
+        .code_address = code_address,
+        .code_num_pages = static_cast<s32>(code_size / PageSize),
+        .flags = flag,
+        .reslimit = Svc::InvalidHandle,
+        .system_resource_num_pages = static_cast<s32>(metadata.GetSystemResourceSize() / PageSize),
+    };
+
+    // Set the process name.
+    const auto& name = metadata.GetName();
+    static_assert(sizeof(params.name) <= sizeof(name));
+    std::memcpy(params.name.data(), name.data(), sizeof(params.name));
+
+    // Initialize for application process.
+    R_TRY(this->Initialize(params, metadata.GetKernelCapabilities(), res_limit,
+                           KMemoryManager::Pool::Application));
+
+    // Assign remaining properties.
+    m_is_hbl = is_hbl;
+    m_ideal_core_id = metadata.GetMainThreadCore();
+
+    // We succeeded.
+    R_SUCCEED();
+}
+
+void KProcess::LoadModule(CodeSet code_set, KProcessAddress base_addr) {
+    const auto ReprotectSegment = [&](const CodeSet::Segment& segment,
+                                      Svc::MemoryPermission permission) {
+        m_page_table.SetProcessMemoryPermission(segment.addr + base_addr, segment.size, permission);
+    };
+
+    this->GetMemory().WriteBlock(base_addr, code_set.memory.data(), code_set.memory.size());
+
+    ReprotectSegment(code_set.CodeSegment(), Svc::MemoryPermission::ReadExecute);
+    ReprotectSegment(code_set.RODataSegment(), Svc::MemoryPermission::Read);
+    ReprotectSegment(code_set.DataSegment(), Svc::MemoryPermission::ReadWrite);
+}
+
+bool KProcess::InsertWatchpoint(KProcessAddress addr, u64 size, DebugWatchpointType type) {
+    const auto watch{std::find_if(m_watchpoints.begin(), m_watchpoints.end(), [&](const auto& wp) {
+        return wp.type == DebugWatchpointType::None;
+    })};
+
+    if (watch == m_watchpoints.end()) {
+        return false;
+    }
+
+    watch->start_address = addr;
+    watch->end_address = addr + size;
+    watch->type = type;
+
+    for (KProcessAddress page = Common::AlignDown(GetInteger(addr), PageSize); page < addr + size;
+         page += PageSize) {
+        m_debug_page_refcounts[page]++;
+        this->GetMemory().MarkRegionDebug(page, PageSize, true);
+    }
+
+    return true;
+}
+
+bool KProcess::RemoveWatchpoint(KProcessAddress addr, u64 size, DebugWatchpointType type) {
+    const auto watch{std::find_if(m_watchpoints.begin(), m_watchpoints.end(), [&](const auto& wp) {
+        return wp.start_address == addr && wp.end_address == addr + size && wp.type == type;
+    })};
+
+    if (watch == m_watchpoints.end()) {
+        return false;
+    }
+
+    watch->start_address = 0;
+    watch->end_address = 0;
+    watch->type = DebugWatchpointType::None;
+
+    for (KProcessAddress page = Common::AlignDown(GetInteger(addr), PageSize); page < addr + size;
+         page += PageSize) {
+        m_debug_page_refcounts[page]--;
+        if (!m_debug_page_refcounts[page]) {
+            this->GetMemory().MarkRegionDebug(page, PageSize, false);
+        }
+    }
+
+    return true;
+}
+
 Core::Memory::Memory& KProcess::GetMemory() const {
     // TODO: per-process memory
     return m_kernel.System().ApplicationMemory();
diff --git a/src/core/hle/kernel/k_process.h b/src/core/hle/kernel/k_process.h
index 146e07a57..f9f755afa 100644
--- a/src/core/hle/kernel/k_process.h
+++ b/src/core/hle/kernel/k_process.h
@@ -1,59 +1,23 @@
-// SPDX-FileCopyrightText: 2015 Citra Emulator Project
+// SPDX-FileCopyrightText: Copyright 2023 yuzu Emulator Project
 // SPDX-License-Identifier: GPL-2.0-or-later
 
 #pragma once
 
-#include <array>
-#include <cstddef>
-#include <list>
 #include <map>
-#include <string>
+
+#include "core/hle/kernel/code_set.h"
 #include "core/hle/kernel/k_address_arbiter.h"
-#include "core/hle/kernel/k_auto_object.h"
+#include "core/hle/kernel/k_capabilities.h"
 #include "core/hle/kernel/k_condition_variable.h"
 #include "core/hle/kernel/k_handle_table.h"
 #include "core/hle/kernel/k_page_table.h"
-#include "core/hle/kernel/k_synchronization_object.h"
+#include "core/hle/kernel/k_page_table_manager.h"
+#include "core/hle/kernel/k_system_resource.h"
+#include "core/hle/kernel/k_thread.h"
 #include "core/hle/kernel/k_thread_local_page.h"
-#include "core/hle/kernel/k_typed_address.h"
-#include "core/hle/kernel/k_worker_task.h"
-#include "core/hle/kernel/process_capability.h"
-#include "core/hle/kernel/slab_helpers.h"
-#include "core/hle/result.h"
-
-namespace Core {
-namespace Memory {
-class Memory;
-};
-
-class System;
-} // namespace Core
-
-namespace FileSys {
-class ProgramMetadata;
-}
 
 namespace Kernel {
 
-class KernelCore;
-class KResourceLimit;
-class KThread;
-class KSharedMemoryInfo;
-class TLSPage;
-
-struct CodeSet;
-
-enum class MemoryRegion : u16 {
-    APPLICATION = 1,
-    SYSTEM = 2,
-    BASE = 3,
-};
-
-enum class ProcessActivity : u32 {
-    Runnable,
-    Paused,
-};
-
 enum class DebugWatchpointType : u8 {
     None = 0,
     Read = 1 << 0,
@@ -72,9 +36,6 @@ class KProcess final : public KAutoObjectWithSlabHeapAndContainer<KProcess, KWor
     KERNEL_AUTOOBJECT_TRAITS(KProcess, KSynchronizationObject);
 
 public:
-    explicit KProcess(KernelCore& kernel);
-    ~KProcess() override;
-
     enum class State {
         Created = static_cast<u32>(Svc::ProcessState::Created),
         CreatedAttached = static_cast<u32>(Svc::ProcessState::CreatedAttached),
@@ -86,470 +47,493 @@ public:
         DebugBreak = static_cast<u32>(Svc::ProcessState::DebugBreak),
     };
 
-    enum : u64 {
-        /// Lowest allowed process ID for a kernel initial process.
-        InitialKIPIDMin = 1,
-        /// Highest allowed process ID for a kernel initial process.
-        InitialKIPIDMax = 80,
-
-        /// Lowest allowed process ID for a userland process.
-        ProcessIDMin = 81,
-        /// Highest allowed process ID for a userland process.
-        ProcessIDMax = 0xFFFFFFFFFFFFFFFF,
-    };
+    using ThreadList = Common::IntrusiveListMemberTraits<&KThread::m_process_list_node>::ListType;
 
-    // Used to determine how process IDs are assigned.
-    enum class ProcessType {
-        KernelInternal,
-        Userland,
-    };
+    static constexpr size_t AslrAlignment = 2_MiB;
 
-    static constexpr std::size_t RANDOM_ENTROPY_SIZE = 4;
+public:
+    static constexpr u64 InitialProcessIdMin = 1;
+    static constexpr u64 InitialProcessIdMax = 0x50;
 
-    static Result Initialize(KProcess* process, Core::System& system, std::string process_name,
-                             ProcessType type, KResourceLimit* res_limit);
+    static constexpr u64 ProcessIdMin = InitialProcessIdMax + 1;
+    static constexpr u64 ProcessIdMax = std::numeric_limits<u64>::max();
 
-    /// Gets a reference to the process' page table.
-    KPageTable& GetPageTable() {
-        return m_page_table;
-    }
+private:
+    using SharedMemoryInfoList = Common::IntrusiveListBaseTraits<KSharedMemoryInfo>::ListType;
+    using TLPTree =
+        Common::IntrusiveRedBlackTreeBaseTraits<KThreadLocalPage>::TreeType<KThreadLocalPage>;
+    using TLPIterator = TLPTree::iterator;
 
-    /// Gets const a reference to the process' page table.
-    const KPageTable& GetPageTable() const {
-        return m_page_table;
-    }
+private:
+    KPageTable m_page_table;
+    std::atomic<size_t> m_used_kernel_memory_size{};
+    TLPTree m_fully_used_tlp_tree{};
+    TLPTree m_partially_used_tlp_tree{};
+    s32 m_ideal_core_id{};
+    KResourceLimit* m_resource_limit{};
+    KSystemResource* m_system_resource{};
+    size_t m_memory_release_hint{};
+    State m_state{};
+    KLightLock m_state_lock;
+    KLightLock m_list_lock;
+    KConditionVariable m_cond_var;
+    KAddressArbiter m_address_arbiter;
+    std::array<u64, 4> m_entropy{};
+    bool m_is_signaled{};
+    bool m_is_initialized{};
+    bool m_is_application{};
+    bool m_is_default_application_system_resource{};
+    bool m_is_hbl{};
+    std::array<char, 13> m_name{};
+    std::atomic<u16> m_num_running_threads{};
+    Svc::CreateProcessFlag m_flags{};
+    KMemoryManager::Pool m_memory_pool{};
+    s64 m_schedule_count{};
+    KCapabilities m_capabilities{};
+    u64 m_program_id{};
+    u64 m_process_id{};
+    KProcessAddress m_code_address{};
+    size_t m_code_size{};
+    size_t m_main_thread_stack_size{};
+    size_t m_max_process_memory{};
+    u32 m_version{};
+    KHandleTable m_handle_table;
+    KProcessAddress m_plr_address{};
+    KThread* m_exception_thread{};
+    ThreadList m_thread_list{};
+    SharedMemoryInfoList m_shared_memory_list{};
+    bool m_is_suspended{};
+    bool m_is_immortal{};
+    bool m_is_handle_table_initialized{};
+    std::array<KThread*, Core::Hardware::NUM_CPU_CORES> m_running_threads{};
+    std::array<u64, Core::Hardware::NUM_CPU_CORES> m_running_thread_idle_counts{};
+    std::array<u64, Core::Hardware::NUM_CPU_CORES> m_running_thread_switch_counts{};
+    std::array<KThread*, Core::Hardware::NUM_CPU_CORES> m_pinned_threads{};
+    std::array<DebugWatchpoint, Core::Hardware::NUM_WATCHPOINTS> m_watchpoints{};
+    std::map<KProcessAddress, u64> m_debug_page_refcounts{};
+    std::atomic<s64> m_cpu_time{};
+    std::atomic<s64> m_num_process_switches{};
+    std::atomic<s64> m_num_thread_switches{};
+    std::atomic<s64> m_num_fpu_switches{};
+    std::atomic<s64> m_num_supervisor_calls{};
+    std::atomic<s64> m_num_ipc_messages{};
+    std::atomic<s64> m_num_ipc_replies{};
+    std::atomic<s64> m_num_ipc_receives{};
 
-    /// Gets a reference to the process' handle table.
-    KHandleTable& GetHandleTable() {
-        return m_handle_table;
-    }
+private:
+    Result StartTermination();
+    void FinishTermination();
 
-    /// Gets a const reference to the process' handle table.
-    const KHandleTable& GetHandleTable() const {
-        return m_handle_table;
+    void PinThread(s32 core_id, KThread* thread) {
+        ASSERT(0 <= core_id && core_id < static_cast<s32>(Core::Hardware::NUM_CPU_CORES));
+        ASSERT(thread != nullptr);
+        ASSERT(m_pinned_threads[core_id] == nullptr);
+        m_pinned_threads[core_id] = thread;
     }
 
-    /// Gets a reference to process's memory.
-    Core::Memory::Memory& GetMemory() const;
-
-    Result SignalToAddress(KProcessAddress address) {
-        return m_condition_var.SignalToAddress(address);
+    void UnpinThread(s32 core_id, KThread* thread) {
+        ASSERT(0 <= core_id && core_id < static_cast<s32>(Core::Hardware::NUM_CPU_CORES));
+        ASSERT(thread != nullptr);
+        ASSERT(m_pinned_threads[core_id] == thread);
+        m_pinned_threads[core_id] = nullptr;
     }
 
-    Result WaitForAddress(Handle handle, KProcessAddress address, u32 tag) {
-        return m_condition_var.WaitForAddress(handle, address, tag);
-    }
+public:
+    explicit KProcess(KernelCore& kernel);
+    ~KProcess() override;
 
-    void SignalConditionVariable(u64 cv_key, int32_t count) {
-        return m_condition_var.Signal(cv_key, count);
-    }
+    Result Initialize(const Svc::CreateProcessParameter& params, KResourceLimit* res_limit,
+                      bool is_real);
 
-    Result WaitConditionVariable(KProcessAddress address, u64 cv_key, u32 tag, s64 ns) {
-        R_RETURN(m_condition_var.Wait(address, cv_key, tag, ns));
-    }
+    Result Initialize(const Svc::CreateProcessParameter& params, const KPageGroup& pg,
+                      std::span<const u32> caps, KResourceLimit* res_limit,
+                      KMemoryManager::Pool pool, bool immortal);
+    Result Initialize(const Svc::CreateProcessParameter& params, std::span<const u32> user_caps,
+                      KResourceLimit* res_limit, KMemoryManager::Pool pool);
+    void Exit();
 
-    Result SignalAddressArbiter(uint64_t address, Svc::SignalType signal_type, s32 value,
-                                s32 count) {
-        R_RETURN(m_address_arbiter.SignalToAddress(address, signal_type, value, count));
+    const char* GetName() const {
+        return m_name.data();
     }
 
-    Result WaitAddressArbiter(uint64_t address, Svc::ArbitrationType arb_type, s32 value,
-                              s64 timeout) {
-        R_RETURN(m_address_arbiter.WaitForAddress(address, arb_type, value, timeout));
+    u64 GetProgramId() const {
+        return m_program_id;
     }
 
-    KProcessAddress GetProcessLocalRegionAddress() const {
-        return m_plr_address;
+    u64 GetProcessId() const {
+        return m_process_id;
     }
 
-    /// Gets the current status of the process
     State GetState() const {
         return m_state;
     }
 
-    /// Gets the unique ID that identifies this particular process.
-    u64 GetProcessId() const {
-        return m_process_id;
+    u64 GetCoreMask() const {
+        return m_capabilities.GetCoreMask();
+    }
+    u64 GetPhysicalCoreMask() const {
+        return m_capabilities.GetPhysicalCoreMask();
+    }
+    u64 GetPriorityMask() const {
+        return m_capabilities.GetPriorityMask();
     }
 
-    /// Gets the program ID corresponding to this process.
-    u64 GetProgramId() const {
-        return m_program_id;
+    s32 GetIdealCoreId() const {
+        return m_ideal_core_id;
+    }
+    void SetIdealCoreId(s32 core_id) {
+        m_ideal_core_id = core_id;
     }
 
-    KProcessAddress GetEntryPoint() const {
-        return m_code_address;
+    bool CheckThreadPriority(s32 prio) const {
+        return ((1ULL << prio) & this->GetPriorityMask()) != 0;
     }
 
-    /// Gets the resource limit descriptor for this process
-    KResourceLimit* GetResourceLimit() const;
+    u32 GetCreateProcessFlags() const {
+        return static_cast<u32>(m_flags);
+    }
 
-    /// Gets the ideal CPU core ID for this process
-    u8 GetIdealCoreId() const {
-        return m_ideal_core;
+    bool Is64Bit() const {
+        return True(m_flags & Svc::CreateProcessFlag::Is64Bit);
     }
 
-    /// Checks if the specified thread priority is valid.
-    bool CheckThreadPriority(s32 prio) const {
-        return ((1ULL << prio) & GetPriorityMask()) != 0;
+    KProcessAddress GetEntryPoint() const {
+        return m_code_address;
     }
 
-    /// Gets the bitmask of allowed cores that this process' threads can run on.
-    u64 GetCoreMask() const {
-        return m_capabilities.GetCoreMask();
+    size_t GetMainStackSize() const {
+        return m_main_thread_stack_size;
     }
 
-    /// Gets the bitmask of allowed thread priorities.
-    u64 GetPriorityMask() const {
-        return m_capabilities.GetPriorityMask();
+    KMemoryManager::Pool GetMemoryPool() const {
+        return m_memory_pool;
     }
 
-    /// Gets the amount of secure memory to allocate for memory management.
-    u32 GetSystemResourceSize() const {
-        return m_system_resource_size;
+    u64 GetRandomEntropy(size_t i) const {
+        return m_entropy[i];
     }
 
-    /// Gets the amount of secure memory currently in use for memory management.
-    u32 GetSystemResourceUsage() const {
-        // On hardware, this returns the amount of system resource memory that has
-        // been used by the kernel. This is problematic for Yuzu to emulate, because
-        // system resource memory is used for page tables -- and yuzu doesn't really
-        // have a way to calculate how much memory is required for page tables for
-        // the current process at any given time.
-        // TODO: Is this even worth implementing? Games may retrieve this value via
-        // an SDK function that gets used + available system resource size for debug
-        // or diagnostic purposes. However, it seems unlikely that a game would make
-        // decisions based on how much system memory is dedicated to its page tables.
-        // Is returning a value other than zero wise?
-        return 0;
+    bool IsApplication() const {
+        return m_is_application;
     }
 
-    /// Whether this process is an AArch64 or AArch32 process.
-    bool Is64BitProcess() const {
-        return m_is_64bit_process;
+    bool IsDefaultApplicationSystemResource() const {
+        return m_is_default_application_system_resource;
     }
 
     bool IsSuspended() const {
         return m_is_suspended;
     }
-
     void SetSuspended(bool suspended) {
         m_is_suspended = suspended;
     }
 
-    /// Gets the total running time of the process instance in ticks.
-    u64 GetCPUTimeTicks() const {
-        return m_total_process_running_time_ticks;
+    Result Terminate();
+
+    bool IsTerminated() const {
+        return m_state == State::Terminated;
     }
 
-    /// Updates the total running time, adding the given ticks to it.
-    void UpdateCPUTimeTicks(u64 ticks) {
-        m_total_process_running_time_ticks += ticks;
+    bool IsPermittedSvc(u32 svc_id) const {
+        return m_capabilities.IsPermittedSvc(svc_id);
     }
 
-    /// Gets the process schedule count, used for thread yielding
-    s64 GetScheduledCount() const {
-        return m_schedule_count;
+    bool IsPermittedInterrupt(s32 interrupt_id) const {
+        return m_capabilities.IsPermittedInterrupt(interrupt_id);
     }
 
-    /// Increments the process schedule count, used for thread yielding.
-    void IncrementScheduledCount() {
-        ++m_schedule_count;
+    bool IsPermittedDebug() const {
+        return m_capabilities.IsPermittedDebug();
     }
 
-    void IncrementRunningThreadCount();
-    void DecrementRunningThreadCount();
+    bool CanForceDebug() const {
+        return m_capabilities.CanForceDebug();
+    }
 
-    void SetRunningThread(s32 core, KThread* thread, u64 idle_count) {
-        m_running_threads[core] = thread;
-        m_running_thread_idle_counts[core] = idle_count;
+    bool IsHbl() const {
+        return m_is_hbl;
     }
 
-    void ClearRunningThread(KThread* thread) {
-        for (size_t i = 0; i < m_running_threads.size(); ++i) {
-            if (m_running_threads[i] == thread) {
-                m_running_threads[i] = nullptr;
-            }
-        }
+    Kernel::KMemoryManager::Direction GetAllocateOption() const {
+        // TODO: property of the KPageTableBase
+        return KMemoryManager::Direction::FromFront;
     }
 
-    [[nodiscard]] KThread* GetRunningThread(s32 core) const {
-        return m_running_threads[core];
+    ThreadList& GetThreadList() {
+        return m_thread_list;
+    }
+    const ThreadList& GetThreadList() const {
+        return m_thread_list;
     }
 
+    bool EnterUserException();
+    bool LeaveUserException();
     bool ReleaseUserException(KThread* thread);
 
-    [[nodiscard]] KThread* GetPinnedThread(s32 core_id) const {
+    KThread* GetPinnedThread(s32 core_id) const {
         ASSERT(0 <= core_id && core_id < static_cast<s32>(Core::Hardware::NUM_CPU_CORES));
         return m_pinned_threads[core_id];
     }
 
-    /// Gets 8 bytes of random data for svcGetInfo RandomEntropy
-    u64 GetRandomEntropy(std::size_t index) const {
-        return m_random_entropy.at(index);
+    const Svc::SvcAccessFlagSet& GetSvcPermissions() const {
+        return m_capabilities.GetSvcPermissions();
     }
 
-    /// Retrieves the total physical memory available to this process in bytes.
-    u64 GetTotalPhysicalMemoryAvailable();
-
-    /// Retrieves the total physical memory available to this process in bytes,
-    /// without the size of the personal system resource heap added to it.
-    u64 GetTotalPhysicalMemoryAvailableWithoutSystemResource();
-
-    /// Retrieves the total physical memory used by this process in bytes.
-    u64 GetTotalPhysicalMemoryUsed();
-
-    /// Retrieves the total physical memory used by this process in bytes,
-    /// without the size of the personal system resource heap added to it.
-    u64 GetTotalPhysicalMemoryUsedWithoutSystemResource();
-
-    /// Gets the list of all threads created with this process as their owner.
-    std::list<KThread*>& GetThreadList() {
-        return m_thread_list;
+    KResourceLimit* GetResourceLimit() const {
+        return m_resource_limit;
     }
 
-    /// Registers a thread as being created under this process,
-    /// adding it to this process' thread list.
-    void RegisterThread(KThread* thread);
+    bool ReserveResource(Svc::LimitableResource which, s64 value);
+    bool ReserveResource(Svc::LimitableResource which, s64 value, s64 timeout);
+    void ReleaseResource(Svc::LimitableResource which, s64 value);
+    void ReleaseResource(Svc::LimitableResource which, s64 value, s64 hint);
 
-    /// Unregisters a thread from this process, removing it
-    /// from this process' thread list.
-    void UnregisterThread(KThread* thread);
+    KLightLock& GetStateLock() {
+        return m_state_lock;
+    }
+    KLightLock& GetListLock() {
+        return m_list_lock;
+    }
 
-    /// Retrieves the number of available threads for this process.
-    u64 GetFreeThreadCount() const;
-
-    /// Clears the signaled state of the process if and only if it's signaled.
-    ///
-    /// @pre The process must not be already terminated. If this is called on a
-    ///      terminated process, then ResultInvalidState will be returned.
-    ///
-    /// @pre The process must be in a signaled state. If this is called on a
-    ///      process instance that is not signaled, ResultInvalidState will be
-    ///      returned.
-    Result Reset();
+    KPageTable& GetPageTable() {
+        return m_page_table;
+    }
+    const KPageTable& GetPageTable() const {
+        return m_page_table;
+    }
 
-    /**
-     * Loads process-specifics configuration info with metadata provided
-     * by an executable.
-     *
-     * @param metadata The provided metadata to load process specific info from.
-     *
-     * @returns ResultSuccess if all relevant metadata was able to be
-     *          loaded and parsed. Otherwise, an error code is returned.
-     */
-    Result LoadFromMetadata(const FileSys::ProgramMetadata& metadata, std::size_t code_size,
-                            bool is_hbl);
+    KHandleTable& GetHandleTable() {
+        return m_handle_table;
+    }
+    const KHandleTable& GetHandleTable() const {
+        return m_handle_table;
+    }
 
-    /**
-     * Starts the main application thread for this process.
-     *
-     * @param main_thread_priority The priority for the main thread.
-     * @param stack_size           The stack size for the main thread in bytes.
-     */
-    void Run(s32 main_thread_priority, u64 stack_size);
+    size_t GetUsedUserPhysicalMemorySize() const;
+    size_t GetTotalUserPhysicalMemorySize() const;
+    size_t GetUsedNonSystemUserPhysicalMemorySize() const;
+    size_t GetTotalNonSystemUserPhysicalMemorySize() const;
 
-    /**
-     * Prepares a process for termination by stopping all of its threads
-     * and clearing any other resources.
-     */
-    void PrepareForTermination();
+    Result AddSharedMemory(KSharedMemory* shmem, KProcessAddress address, size_t size);
+    void RemoveSharedMemory(KSharedMemory* shmem, KProcessAddress address, size_t size);
 
-    void LoadModule(CodeSet code_set, KProcessAddress base_addr);
+    Result CreateThreadLocalRegion(KProcessAddress* out);
+    Result DeleteThreadLocalRegion(KProcessAddress addr);
 
-    bool IsInitialized() const override {
-        return m_is_initialized;
+    KProcessAddress GetProcessLocalRegionAddress() const {
+        return m_plr_address;
     }
 
-    static void PostDestroy(uintptr_t arg) {}
-
-    void Finalize() override;
-
-    u64 GetId() const override {
-        return GetProcessId();
+    KThread* GetExceptionThread() const {
+        return m_exception_thread;
     }
 
-    bool IsHbl() const {
-        return m_is_hbl;
+    void AddCpuTime(s64 diff) {
+        m_cpu_time += diff;
+    }
+    s64 GetCpuTime() {
+        return m_cpu_time.load();
     }
 
-    bool IsSignaled() const override;
-
-    void DoWorkerTaskImpl();
+    s64 GetScheduledCount() const {
+        return m_schedule_count;
+    }
+    void IncrementScheduledCount() {
+        ++m_schedule_count;
+    }
 
-    Result SetActivity(ProcessActivity activity);
+    void IncrementRunningThreadCount();
+    void DecrementRunningThreadCount();
 
-    void PinCurrentThread(s32 core_id);
-    void UnpinCurrentThread(s32 core_id);
-    void UnpinThread(KThread* thread);
+    size_t GetRequiredSecureMemorySizeNonDefault() const {
+        if (!this->IsDefaultApplicationSystemResource() && m_system_resource->IsSecureResource()) {
+            auto* secure_system_resource = static_cast<KSecureSystemResource*>(m_system_resource);
+            return secure_system_resource->CalculateRequiredSecureMemorySize();
+        }
 
-    KLightLock& GetStateLock() {
-        return m_state_lock;
+        return 0;
     }
 
-    Result AddSharedMemory(KSharedMemory* shmem, KProcessAddress address, size_t size);
-    void RemoveSharedMemory(KSharedMemory* shmem, KProcessAddress address, size_t size);
-
-    ///////////////////////////////////////////////////////////////////////////////////////////////
-    // Thread-local storage management
-
-    // Marks the next available region as used and returns the address of the slot.
-    [[nodiscard]] Result CreateThreadLocalRegion(KProcessAddress* out);
+    size_t GetRequiredSecureMemorySize() const {
+        if (m_system_resource->IsSecureResource()) {
+            auto* secure_system_resource = static_cast<KSecureSystemResource*>(m_system_resource);
+            return secure_system_resource->CalculateRequiredSecureMemorySize();
+        }
 
-    // Frees a used TLS slot identified by the given address
-    Result DeleteThreadLocalRegion(KProcessAddress addr);
+        return 0;
+    }
 
-    ///////////////////////////////////////////////////////////////////////////////////////////////
-    // Debug watchpoint management
+    size_t GetTotalSystemResourceSize() const {
+        if (!this->IsDefaultApplicationSystemResource() && m_system_resource->IsSecureResource()) {
+            auto* secure_system_resource = static_cast<KSecureSystemResource*>(m_system_resource);
+            return secure_system_resource->GetSize();
+        }
 
-    // Attempts to insert a watchpoint into a free slot. Returns false if none are available.
-    bool InsertWatchpoint(KProcessAddress addr, u64 size, DebugWatchpointType type);
+        return 0;
+    }
 
-    // Attempts to remove the watchpoint specified by the given parameters.
-    bool RemoveWatchpoint(KProcessAddress addr, u64 size, DebugWatchpointType type);
+    size_t GetUsedSystemResourceSize() const {
+        if (!this->IsDefaultApplicationSystemResource() && m_system_resource->IsSecureResource()) {
+            auto* secure_system_resource = static_cast<KSecureSystemResource*>(m_system_resource);
+            return secure_system_resource->GetUsedSize();
+        }
 
-    const std::array<DebugWatchpoint, Core::Hardware::NUM_WATCHPOINTS>& GetWatchpoints() const {
-        return m_watchpoints;
+        return 0;
     }
 
-    const std::string& GetName() {
-        return name;
+    void SetRunningThread(s32 core, KThread* thread, u64 idle_count, u64 switch_count) {
+        m_running_threads[core] = thread;
+        m_running_thread_idle_counts[core] = idle_count;
+        m_running_thread_switch_counts[core] = switch_count;
     }
 
-private:
-    void PinThread(s32 core_id, KThread* thread) {
-        ASSERT(0 <= core_id && core_id < static_cast<s32>(Core::Hardware::NUM_CPU_CORES));
-        ASSERT(thread != nullptr);
-        ASSERT(m_pinned_threads[core_id] == nullptr);
-        m_pinned_threads[core_id] = thread;
+    void ClearRunningThread(KThread* thread) {
+        for (size_t i = 0; i < m_running_threads.size(); ++i) {
+            if (m_running_threads[i] == thread) {
+                m_running_threads[i] = nullptr;
+            }
+        }
     }
 
-    void UnpinThread(s32 core_id, KThread* thread) {
-        ASSERT(0 <= core_id && core_id < static_cast<s32>(Core::Hardware::NUM_CPU_CORES));
-        ASSERT(thread != nullptr);
-        ASSERT(m_pinned_threads[core_id] == thread);
-        m_pinned_threads[core_id] = nullptr;
+    const KSystemResource& GetSystemResource() const {
+        return *m_system_resource;
     }
 
-    void FinalizeHandleTable() {
-        // Finalize the table.
-        m_handle_table.Finalize();
-
-        // Note that the table is finalized.
-        m_is_handle_table_initialized = false;
+    const KMemoryBlockSlabManager& GetMemoryBlockSlabManager() const {
+        return m_system_resource->GetMemoryBlockSlabManager();
+    }
+    const KBlockInfoManager& GetBlockInfoManager() const {
+        return m_system_resource->GetBlockInfoManager();
+    }
+    const KPageTableManager& GetPageTableManager() const {
+        return m_system_resource->GetPageTableManager();
     }
 
-    void ChangeState(State new_state);
-
-    /// Allocates the main thread stack for the process, given the stack size in bytes.
-    Result AllocateMainThreadStack(std::size_t stack_size);
-
-    /// Memory manager for this process
-    KPageTable m_page_table;
-
-    /// Current status of the process
-    State m_state{};
+    KThread* GetRunningThread(s32 core) const {
+        return m_running_threads[core];
+    }
+    u64 GetRunningThreadIdleCount(s32 core) const {
+        return m_running_thread_idle_counts[core];
+    }
+    u64 GetRunningThreadSwitchCount(s32 core) const {
+        return m_running_thread_switch_counts[core];
+    }
 
-    /// The ID of this process
-    u64 m_process_id = 0;
+    void RegisterThread(KThread* thread);
+    void UnregisterThread(KThread* thread);
 
-    /// Title ID corresponding to the process
-    u64 m_program_id = 0;
+    Result Run(s32 priority, size_t stack_size);
 
-    /// Specifies additional memory to be reserved for the process's memory management by the
-    /// system. When this is non-zero, secure memory is allocated and used for page table allocation
-    /// instead of using the normal global page tables/memory block management.
-    u32 m_system_resource_size = 0;
+    Result Reset();
 
-    /// Resource limit descriptor for this process
-    KResourceLimit* m_resource_limit{};
+    void SetDebugBreak() {
+        if (m_state == State::RunningAttached) {
+            this->ChangeState(State::DebugBreak);
+        }
+    }
 
-    KVirtualAddress m_system_resource_address{};
+    void SetAttached() {
+        if (m_state == State::DebugBreak) {
+            this->ChangeState(State::RunningAttached);
+        }
+    }
 
-    /// The ideal CPU core for this process, threads are scheduled on this core by default.
-    u8 m_ideal_core = 0;
+    Result SetActivity(Svc::ProcessActivity activity);
 
-    /// Contains the parsed process capability descriptors.
-    ProcessCapabilities m_capabilities;
+    void PinCurrentThread();
+    void UnpinCurrentThread();
+    void UnpinThread(KThread* thread);
 
-    /// Whether or not this process is AArch64, or AArch32.
-    /// By default, we currently assume this is true, unless otherwise
-    /// specified by metadata provided to the process during loading.
-    bool m_is_64bit_process = true;
+    void SignalConditionVariable(uintptr_t cv_key, int32_t count) {
+        return m_cond_var.Signal(cv_key, count);
+    }
 
-    /// Total running time for the process in ticks.
-    std::atomic<u64> m_total_process_running_time_ticks = 0;
+    Result WaitConditionVariable(KProcessAddress address, uintptr_t cv_key, u32 tag, s64 ns) {
+        R_RETURN(m_cond_var.Wait(address, cv_key, tag, ns));
+    }
 
-    /// Per-process handle table for storing created object handles in.
-    KHandleTable m_handle_table;
+    Result SignalAddressArbiter(uintptr_t address, Svc::SignalType signal_type, s32 value,
+                                s32 count) {
+        R_RETURN(m_address_arbiter.SignalToAddress(address, signal_type, value, count));
+    }
 
-    /// Per-process address arbiter.
-    KAddressArbiter m_address_arbiter;
+    Result WaitAddressArbiter(uintptr_t address, Svc::ArbitrationType arb_type, s32 value,
+                              s64 timeout) {
+        R_RETURN(m_address_arbiter.WaitForAddress(address, arb_type, value, timeout));
+    }
 
-    /// The per-process mutex lock instance used for handling various
-    /// forms of services, such as lock arbitration, and condition
-    /// variable related facilities.
-    KConditionVariable m_condition_var;
+    Result GetThreadList(s32* out_num_threads, KProcessAddress out_thread_ids, s32 max_out_count);
 
-    /// Address indicating the location of the process' dedicated TLS region.
-    KProcessAddress m_plr_address = 0;
+    static void Switch(KProcess* cur_process, KProcess* next_process);
 
-    /// Address indicating the location of the process's entry point.
-    KProcessAddress m_code_address = 0;
+public:
+    // Attempts to insert a watchpoint into a free slot. Returns false if none are available.
+    bool InsertWatchpoint(KProcessAddress addr, u64 size, DebugWatchpointType type);
 
-    /// Random values for svcGetInfo RandomEntropy
-    std::array<u64, RANDOM_ENTROPY_SIZE> m_random_entropy{};
+    // Attempts to remove the watchpoint specified by the given parameters.
+    bool RemoveWatchpoint(KProcessAddress addr, u64 size, DebugWatchpointType type);
 
-    /// List of threads that are running with this process as their owner.
-    std::list<KThread*> m_thread_list;
+    const std::array<DebugWatchpoint, Core::Hardware::NUM_WATCHPOINTS>& GetWatchpoints() const {
+        return m_watchpoints;
+    }
 
-    /// List of shared memory that are running with this process as their owner.
-    std::list<KSharedMemoryInfo*> m_shared_memory_list;
+public:
+    Result LoadFromMetadata(const FileSys::ProgramMetadata& metadata, std::size_t code_size,
+                            bool is_hbl);
 
-    /// Address of the top of the main thread's stack
-    KProcessAddress m_main_thread_stack_top{};
+    void LoadModule(CodeSet code_set, KProcessAddress base_addr);
 
-    /// Size of the main thread's stack
-    std::size_t m_main_thread_stack_size{};
+    Core::Memory::Memory& GetMemory() const;
 
-    /// Memory usage capacity for the process
-    std::size_t m_memory_usage_capacity{};
+public:
+    // Overridden parent functions.
+    bool IsInitialized() const override {
+        return m_is_initialized;
+    }
 
-    /// Process total image size
-    std::size_t m_image_size{};
+    static void PostDestroy(uintptr_t arg) {}
 
-    /// Schedule count of this process
-    s64 m_schedule_count{};
+    void Finalize() override;
 
-    size_t m_memory_release_hint{};
+    u64 GetIdImpl() const {
+        return this->GetProcessId();
+    }
+    u64 GetId() const override {
+        return this->GetIdImpl();
+    }
 
-    std::string name{};
+    virtual bool IsSignaled() const override {
+        ASSERT(KScheduler::IsSchedulerLockedByCurrentThread(m_kernel));
+        return m_is_signaled;
+    }
 
-    bool m_is_signaled{};
-    bool m_is_suspended{};
-    bool m_is_immortal{};
-    bool m_is_handle_table_initialized{};
-    bool m_is_initialized{};
-    bool m_is_hbl{};
+    void DoWorkerTaskImpl();
 
-    std::atomic<u16> m_num_running_threads{};
+private:
+    void ChangeState(State new_state) {
+        if (m_state != new_state) {
+            m_state = new_state;
+            m_is_signaled = true;
+            this->NotifyAvailable();
+        }
+    }
 
-    std::array<KThread*, Core::Hardware::NUM_CPU_CORES> m_running_threads{};
-    std::array<u64, Core::Hardware::NUM_CPU_CORES> m_running_thread_idle_counts{};
-    std::array<KThread*, Core::Hardware::NUM_CPU_CORES> m_pinned_threads{};
-    std::array<DebugWatchpoint, Core::Hardware::NUM_WATCHPOINTS> m_watchpoints{};
-    std::map<KProcessAddress, u64> m_debug_page_refcounts;
+    Result InitializeHandleTable(s32 size) {
+        // Try to initialize the handle table.
+        R_TRY(m_handle_table.Initialize(size));
 
-    KThread* m_exception_thread{};
+        // We succeeded, so note that we did.
+        m_is_handle_table_initialized = true;
+        R_SUCCEED();
+    }
 
-    KLightLock m_state_lock;
-    KLightLock m_list_lock;
+    void FinalizeHandleTable() {
+        // Finalize the table.
+        m_handle_table.Finalize();
 
-    using TLPTree =
-        Common::IntrusiveRedBlackTreeBaseTraits<KThreadLocalPage>::TreeType<KThreadLocalPage>;
-    using TLPIterator = TLPTree::iterator;
-    TLPTree m_fully_used_tlp_tree;
-    TLPTree m_partially_used_tlp_tree;
+        // Note that the table is finalized.
+        m_is_handle_table_initialized = false;
+    }
 };
 
 } // namespace Kernel
diff --git a/src/core/hle/kernel/k_scheduler.cpp b/src/core/hle/kernel/k_scheduler.cpp
index d8143c650..1bce63a56 100644
--- a/src/core/hle/kernel/k_scheduler.cpp
+++ b/src/core/hle/kernel/k_scheduler.cpp
@@ -190,7 +190,7 @@ u64 KScheduler::UpdateHighestPriorityThread(KThread* highest_thread) {
         if (m_state.should_count_idle) {
             if (highest_thread != nullptr) [[likely]] {
                 if (KProcess* process = highest_thread->GetOwnerProcess(); process != nullptr) {
-                    process->SetRunningThread(m_core_id, highest_thread, m_state.idle_count);
+                    process->SetRunningThread(m_core_id, highest_thread, m_state.idle_count, 0);
                 }
             } else {
                 m_state.idle_count++;
@@ -356,7 +356,7 @@ void KScheduler::SwitchThread(KThread* next_thread) {
     const s64 tick_diff = cur_tick - prev_tick;
     cur_thread->AddCpuTime(m_core_id, tick_diff);
     if (cur_process != nullptr) {
-        cur_process->UpdateCPUTimeTicks(tick_diff);
+        cur_process->AddCpuTime(tick_diff);
     }
     m_last_context_switch_time = cur_tick;
 
diff --git a/src/core/hle/kernel/k_system_resource.cpp b/src/core/hle/kernel/k_system_resource.cpp
index e6c8d589a..07e92aa80 100644
--- a/src/core/hle/kernel/k_system_resource.cpp
+++ b/src/core/hle/kernel/k_system_resource.cpp
@@ -1,25 +1,100 @@
 // SPDX-FileCopyrightText: Copyright 2022 yuzu Emulator Project
 // SPDX-License-Identifier: GPL-2.0-or-later
 
+#include "core/core.h"
+#include "core/hle/kernel/k_scoped_resource_reservation.h"
 #include "core/hle/kernel/k_system_resource.h"
 
 namespace Kernel {
 
 Result KSecureSystemResource::Initialize(size_t size, KResourceLimit* resource_limit,
                                          KMemoryManager::Pool pool) {
-    // Unimplemented
-    UNREACHABLE();
+    // Set members.
+    m_resource_limit = resource_limit;
+    m_resource_size = size;
+    m_resource_pool = pool;
+
+    // Determine required size for our secure resource.
+    const size_t secure_size = this->CalculateRequiredSecureMemorySize();
+
+    // Reserve memory for our secure resource.
+    KScopedResourceReservation memory_reservation(
+        m_resource_limit, Svc::LimitableResource::PhysicalMemoryMax, secure_size);
+    R_UNLESS(memory_reservation.Succeeded(), ResultLimitReached);
+
+    // Allocate secure memory.
+    R_TRY(KSystemControl::AllocateSecureMemory(m_kernel, std::addressof(m_resource_address),
+                                               m_resource_size, static_cast<u32>(m_resource_pool)));
+    ASSERT(m_resource_address != 0);
+
+    // Ensure we clean up the secure memory, if we fail past this point.
+    ON_RESULT_FAILURE {
+        KSystemControl::FreeSecureMemory(m_kernel, m_resource_address, m_resource_size,
+                                         static_cast<u32>(m_resource_pool));
+    };
+
+    // Check that our allocation is bigger than the reference counts needed for it.
+    const size_t rc_size =
+        Common::AlignUp(KPageTableSlabHeap::CalculateReferenceCountSize(m_resource_size), PageSize);
+    R_UNLESS(m_resource_size > rc_size, ResultOutOfMemory);
+
+    // Get resource pointer.
+    KPhysicalAddress resource_paddr =
+        KPageTable::GetHeapPhysicalAddress(m_kernel.MemoryLayout(), m_resource_address);
+    auto* resource =
+        m_kernel.System().DeviceMemory().GetPointer<KPageTableManager::RefCount>(resource_paddr);
+
+    // Initialize slab heaps.
+    m_dynamic_page_manager.Initialize(m_resource_address + rc_size, m_resource_size - rc_size,
+                                      PageSize);
+    m_page_table_heap.Initialize(std::addressof(m_dynamic_page_manager), 0, resource);
+    m_memory_block_heap.Initialize(std::addressof(m_dynamic_page_manager), 0);
+    m_block_info_heap.Initialize(std::addressof(m_dynamic_page_manager), 0);
+
+    // Initialize managers.
+    m_page_table_manager.Initialize(std::addressof(m_dynamic_page_manager),
+                                    std::addressof(m_page_table_heap));
+    m_memory_block_slab_manager.Initialize(std::addressof(m_dynamic_page_manager),
+                                           std::addressof(m_memory_block_heap));
+    m_block_info_manager.Initialize(std::addressof(m_dynamic_page_manager),
+                                    std::addressof(m_block_info_heap));
+
+    // Set our managers.
+    this->SetManagers(m_memory_block_slab_manager, m_block_info_manager, m_page_table_manager);
+
+    // Commit the memory reservation.
+    memory_reservation.Commit();
+
+    // Open reference to our resource limit.
+    m_resource_limit->Open();
+
+    // Set ourselves as initialized.
+    m_is_initialized = true;
+
+    R_SUCCEED();
 }
 
 void KSecureSystemResource::Finalize() {
-    // Unimplemented
-    UNREACHABLE();
+    // Check that we have no outstanding allocations.
+    ASSERT(m_memory_block_slab_manager.GetUsed() == 0);
+    ASSERT(m_block_info_manager.GetUsed() == 0);
+    ASSERT(m_page_table_manager.GetUsed() == 0);
+
+    // Free our secure memory.
+    KSystemControl::FreeSecureMemory(m_kernel, m_resource_address, m_resource_size,
+                                     static_cast<u32>(m_resource_pool));
+
+    // Release the memory reservation.
+    m_resource_limit->Release(Svc::LimitableResource::PhysicalMemoryMax,
+                              this->CalculateRequiredSecureMemorySize());
+
+    // Close reference to our resource limit.
+    m_resource_limit->Close();
 }
 
 size_t KSecureSystemResource::CalculateRequiredSecureMemorySize(size_t size,
                                                                 KMemoryManager::Pool pool) {
-    // Unimplemented
-    UNREACHABLE();
+    return KSystemControl::CalculateRequiredSecureMemorySize(size, static_cast<u32>(pool));
 }
 
 } // namespace Kernel
diff --git a/src/core/hle/kernel/k_thread.cpp b/src/core/hle/kernel/k_thread.cpp
index 7df8fd7f7..a6deb50ec 100644
--- a/src/core/hle/kernel/k_thread.cpp
+++ b/src/core/hle/kernel/k_thread.cpp
@@ -122,16 +122,15 @@ Result KThread::Initialize(KThreadFunction func, uintptr_t arg, KProcessAddress
     case ThreadType::Main:
         ASSERT(arg == 0);
         [[fallthrough]];
-    case ThreadType::HighPriority:
-        [[fallthrough]];
-    case ThreadType::Dummy:
-        [[fallthrough]];
     case ThreadType::User:
         ASSERT(((owner == nullptr) ||
                 (owner->GetCoreMask() | (1ULL << virt_core)) == owner->GetCoreMask()));
         ASSERT(((owner == nullptr) || (prio > Svc::LowestThreadPriority) ||
                 (owner->GetPriorityMask() | (1ULL << prio)) == owner->GetPriorityMask()));
         break;
+    case ThreadType::HighPriority:
+    case ThreadType::Dummy:
+        break;
     case ThreadType::Kernel:
         UNIMPLEMENTED();
         break;
@@ -216,6 +215,7 @@ Result KThread::Initialize(KThreadFunction func, uintptr_t arg, KProcessAddress
         // Setup the TLS, if needed.
         if (type == ThreadType::User) {
             R_TRY(owner->CreateThreadLocalRegion(std::addressof(m_tls_address)));
+            owner->GetMemory().ZeroBlock(m_tls_address, Svc::ThreadLocalRegionSize);
         }
 
         m_parent = owner;
@@ -403,7 +403,7 @@ void KThread::StartTermination() {
     if (m_parent != nullptr) {
         m_parent->ReleaseUserException(this);
         if (m_parent->GetPinnedThread(GetCurrentCoreId(m_kernel)) == this) {
-            m_parent->UnpinCurrentThread(m_core_id);
+            m_parent->UnpinCurrentThread();
         }
     }
 
@@ -415,10 +415,6 @@ void KThread::StartTermination() {
         m_parent->ClearRunningThread(this);
     }
 
-    // Signal.
-    m_signaled = true;
-    KSynchronizationObject::NotifyAvailable();
-
     // Clear previous thread in KScheduler.
     KScheduler::ClearPreviousThread(m_kernel, this);
 
@@ -437,6 +433,13 @@ void KThread::FinishTermination() {
         }
     }
 
+    // Acquire the scheduler lock.
+    KScopedSchedulerLock sl{m_kernel};
+
+    // Signal.
+    m_signaled = true;
+    KSynchronizationObject::NotifyAvailable();
+
     // Close the thread.
     this->Close();
 }
@@ -820,7 +823,7 @@ void KThread::CloneFpuStatus() {
     ASSERT(this->GetOwnerProcess() != nullptr);
     ASSERT(this->GetOwnerProcess() == GetCurrentProcessPointer(m_kernel));
 
-    if (this->GetOwnerProcess()->Is64BitProcess()) {
+    if (this->GetOwnerProcess()->Is64Bit()) {
         // Clone FPSR and FPCR.
         ThreadContext64 cur_ctx{};
         m_kernel.System().CurrentArmInterface().SaveContext(cur_ctx);
@@ -923,7 +926,7 @@ Result KThread::GetThreadContext3(Common::ScratchBuffer<u8>& out) {
 
         // If we're not terminating, get the thread's user context.
         if (!this->IsTerminationRequested()) {
-            if (m_parent->Is64BitProcess()) {
+            if (m_parent->Is64Bit()) {
                 // Mask away mode bits, interrupt bits, IL bit, and other reserved bits.
                 auto context = GetContext64();
                 context.pstate &= 0xFF0FFE20;
@@ -1174,6 +1177,9 @@ Result KThread::Run() {
             owner->IncrementRunningThreadCount();
         }
 
+        // Open a reference, now that we're running.
+        this->Open();
+
         // Set our state and finish.
         this->SetState(ThreadState::Runnable);
 
diff --git a/src/core/hle/kernel/k_thread.h b/src/core/hle/kernel/k_thread.h
index d178c2453..e1f80b04f 100644
--- a/src/core/hle/kernel/k_thread.h
+++ b/src/core/hle/kernel/k_thread.h
@@ -721,6 +721,7 @@ private:
     // For core KThread implementation
     ThreadContext32 m_thread_context_32{};
     ThreadContext64 m_thread_context_64{};
+    Common::IntrusiveListNode m_process_list_node;
     Common::IntrusiveRedBlackTreeNode m_condvar_arbiter_tree_node{};
     s32 m_priority{};
     using ConditionVariableThreadTreeTraits =
diff --git a/src/core/hle/kernel/k_transfer_memory.cpp b/src/core/hle/kernel/k_transfer_memory.cpp
index 13d34125c..0e2e11743 100644
--- a/src/core/hle/kernel/k_transfer_memory.cpp
+++ b/src/core/hle/kernel/k_transfer_memory.cpp
@@ -1,6 +1,7 @@
 // SPDX-FileCopyrightText: Copyright 2021 yuzu Emulator Project
 // SPDX-License-Identifier: GPL-2.0-or-later
 
+#include "common/scope_exit.h"
 #include "core/hle/kernel/k_process.h"
 #include "core/hle/kernel/k_resource_limit.h"
 #include "core/hle/kernel/k_transfer_memory.h"
@@ -9,28 +10,50 @@
 namespace Kernel {
 
 KTransferMemory::KTransferMemory(KernelCore& kernel)
-    : KAutoObjectWithSlabHeapAndContainer{kernel} {}
+    : KAutoObjectWithSlabHeapAndContainer{kernel}, m_lock{kernel} {}
 
 KTransferMemory::~KTransferMemory() = default;
 
-Result KTransferMemory::Initialize(KProcessAddress address, std::size_t size,
-                                   Svc::MemoryPermission owner_perm) {
+Result KTransferMemory::Initialize(KProcessAddress addr, std::size_t size,
+                                   Svc::MemoryPermission own_perm) {
     // Set members.
     m_owner = GetCurrentProcessPointer(m_kernel);
 
-    // TODO(bunnei): Lock for transfer memory
+    // Get the owner page table.
+    auto& page_table = m_owner->GetPageTable();
+
+    // Construct the page group, guarding to make sure our state is valid on exit.
+    m_page_group.emplace(m_kernel, page_table.GetBlockInfoManager());
+    auto pg_guard = SCOPE_GUARD({ m_page_group.reset(); });
+
+    // Lock the memory.
+    R_TRY(page_table.LockForTransferMemory(std::addressof(*m_page_group), addr, size,
+                                           ConvertToKMemoryPermission(own_perm)));
 
     // Set remaining tracking members.
     m_owner->Open();
-    m_owner_perm = owner_perm;
-    m_address = address;
-    m_size = size;
+    m_owner_perm = own_perm;
+    m_address = addr;
     m_is_initialized = true;
+    m_is_mapped = false;
 
+    // We succeeded.
+    pg_guard.Cancel();
     R_SUCCEED();
 }
 
-void KTransferMemory::Finalize() {}
+void KTransferMemory::Finalize() {
+    // Unlock.
+    if (!m_is_mapped) {
+        const size_t size = m_page_group->GetNumPages() * PageSize;
+        ASSERT(R_SUCCEEDED(
+            m_owner->GetPageTable().UnlockForTransferMemory(m_address, size, *m_page_group)));
+    }
+
+    // Close the page group.
+    m_page_group->Close();
+    m_page_group->Finalize();
+}
 
 void KTransferMemory::PostDestroy(uintptr_t arg) {
     KProcess* owner = reinterpret_cast<KProcess*>(arg);
@@ -38,4 +61,54 @@ void KTransferMemory::PostDestroy(uintptr_t arg) {
     owner->Close();
 }
 
+Result KTransferMemory::Map(KProcessAddress address, size_t size, Svc::MemoryPermission map_perm) {
+    // Validate the size.
+    R_UNLESS(m_page_group->GetNumPages() == Common::DivideUp(size, PageSize), ResultInvalidSize);
+
+    // Validate the permission.
+    R_UNLESS(m_owner_perm == map_perm, ResultInvalidState);
+
+    // Lock ourselves.
+    KScopedLightLock lk(m_lock);
+
+    // Ensure we're not already mapped.
+    R_UNLESS(!m_is_mapped, ResultInvalidState);
+
+    // Map the memory.
+    const KMemoryState state = (m_owner_perm == Svc::MemoryPermission::None)
+                                   ? KMemoryState::Transfered
+                                   : KMemoryState::SharedTransfered;
+    R_TRY(GetCurrentProcess(m_kernel).GetPageTable().MapPageGroup(
+        address, *m_page_group, state, KMemoryPermission::UserReadWrite));
+
+    // Mark ourselves as mapped.
+    m_is_mapped = true;
+
+    R_SUCCEED();
+}
+
+Result KTransferMemory::Unmap(KProcessAddress address, size_t size) {
+    // Validate the size.
+    R_UNLESS(m_page_group->GetNumPages() == Common::DivideUp(size, PageSize), ResultInvalidSize);
+
+    // Lock ourselves.
+    KScopedLightLock lk(m_lock);
+
+    // Unmap the memory.
+    const KMemoryState state = (m_owner_perm == Svc::MemoryPermission::None)
+                                   ? KMemoryState::Transfered
+                                   : KMemoryState::SharedTransfered;
+    R_TRY(GetCurrentProcess(m_kernel).GetPageTable().UnmapPageGroup(address, *m_page_group, state));
+
+    // Mark ourselves as unmapped.
+    ASSERT(m_is_mapped);
+    m_is_mapped = false;
+
+    R_SUCCEED();
+}
+
+size_t KTransferMemory::GetSize() const {
+    return m_is_initialized ? m_page_group->GetNumPages() * PageSize : 0;
+}
+
 } // namespace Kernel
diff --git a/src/core/hle/kernel/k_transfer_memory.h b/src/core/hle/kernel/k_transfer_memory.h
index 54f97ccb4..8a0b08761 100644
--- a/src/core/hle/kernel/k_transfer_memory.h
+++ b/src/core/hle/kernel/k_transfer_memory.h
@@ -3,6 +3,9 @@
 
 #pragma once
 
+#include <optional>
+
+#include "core/hle/kernel/k_page_group.h"
 #include "core/hle/kernel/slab_helpers.h"
 #include "core/hle/kernel/svc_types.h"
 #include "core/hle/result.h"
@@ -48,16 +51,19 @@ public:
         return m_address;
     }
 
-    size_t GetSize() const {
-        return m_is_initialized ? m_size : 0;
-    }
+    size_t GetSize() const;
+
+    Result Map(KProcessAddress address, size_t size, Svc::MemoryPermission map_perm);
+    Result Unmap(KProcessAddress address, size_t size);
 
 private:
+    std::optional<KPageGroup> m_page_group{};
     KProcess* m_owner{};
     KProcessAddress m_address{};
+    KLightLock m_lock;
     Svc::MemoryPermission m_owner_perm{};
-    size_t m_size{};
     bool m_is_initialized{};
+    bool m_is_mapped{};
 };
 
 } // namespace Kernel
diff --git a/src/core/hle/kernel/kernel.cpp b/src/core/hle/kernel/kernel.cpp
index a1134b7e2..4a1559291 100644
--- a/src/core/hle/kernel/kernel.cpp
+++ b/src/core/hle/kernel/kernel.cpp
@@ -101,35 +101,31 @@ struct KernelCore::Impl {
 
     void InitializeCores() {
         for (u32 core_id = 0; core_id < Core::Hardware::NUM_CPU_CORES; core_id++) {
-            cores[core_id]->Initialize((*application_process).Is64BitProcess());
+            cores[core_id]->Initialize((*application_process).Is64Bit());
             system.ApplicationMemory().SetCurrentPageTable(*application_process, core_id);
         }
     }
 
-    void CloseApplicationProcess() {
-        KProcess* old_process = application_process.exchange(nullptr);
-        if (old_process == nullptr) {
-            return;
-        }
-
-        // old_process->Close();
-        // TODO: The process should be destroyed based on accurate ref counting after
-        // calling Close(). Adding a manual Destroy() call instead to avoid a memory leak.
-        old_process->Finalize();
-        old_process->Destroy();
+    void TerminateApplicationProcess() {
+        application_process.load()->Terminate();
     }
 
     void Shutdown() {
         is_shutting_down.store(true, std::memory_order_relaxed);
         SCOPE_EXIT({ is_shutting_down.store(false, std::memory_order_relaxed); });
 
-        process_list.clear();
-
         CloseServices();
 
+        auto* old_process = application_process.exchange(nullptr);
+        if (old_process) {
+            old_process->Close();
+        }
+
+        process_list.clear();
+
         next_object_id = 0;
-        next_kernel_process_id = KProcess::InitialKIPIDMin;
-        next_user_process_id = KProcess::ProcessIDMin;
+        next_kernel_process_id = KProcess::InitialProcessIdMin;
+        next_user_process_id = KProcess::ProcessIdMin;
         next_thread_id = 1;
 
         global_handle_table->Finalize();
@@ -176,8 +172,6 @@ struct KernelCore::Impl {
             }
         }
 
-        CloseApplicationProcess();
-
         // Track kernel objects that were not freed on shutdown
         {
             std::scoped_lock lk{registered_objects_lock};
@@ -344,6 +338,8 @@ struct KernelCore::Impl {
         // Create the system page table managers.
         app_system_resource = std::make_unique<KSystemResource>(kernel);
         sys_system_resource = std::make_unique<KSystemResource>(kernel);
+        KAutoObject::Create(std::addressof(*app_system_resource));
+        KAutoObject::Create(std::addressof(*sys_system_resource));
 
         // Set the managers for the system resources.
         app_system_resource->SetManagers(*app_memory_block_manager, *app_block_info_manager,
@@ -373,7 +369,7 @@ struct KernelCore::Impl {
     static inline thread_local u8 host_thread_id = UINT8_MAX;
 
     /// Sets the host thread ID for the caller.
-    u32 SetHostThreadId(std::size_t core_id) {
+    LTO_NOINLINE u32 SetHostThreadId(std::size_t core_id) {
         // This should only be called during core init.
         ASSERT(host_thread_id == UINT8_MAX);
 
@@ -384,13 +380,13 @@ struct KernelCore::Impl {
     }
 
     /// Gets the host thread ID for the caller
-    u32 GetHostThreadId() const {
+    LTO_NOINLINE u32 GetHostThreadId() const {
         return host_thread_id;
     }
 
     // Gets the dummy KThread for the caller, allocating a new one if this is the first time
-    KThread* GetHostDummyThread(KThread* existing_thread) {
-        const auto initialize{[](KThread* thread) {
+    LTO_NOINLINE KThread* GetHostDummyThread(KThread* existing_thread) {
+        const auto initialize{[](KThread* thread) LTO_NOINLINE {
             ASSERT(KThread::InitializeDummyThread(thread, nullptr).IsSuccess());
             return thread;
         }};
@@ -424,11 +420,11 @@ struct KernelCore::Impl {
 
     static inline thread_local bool is_phantom_mode_for_singlecore{false};
 
-    bool IsPhantomModeForSingleCore() const {
+    LTO_NOINLINE bool IsPhantomModeForSingleCore() const {
         return is_phantom_mode_for_singlecore;
     }
 
-    void SetIsPhantomModeForSingleCore(bool value) {
+    LTO_NOINLINE void SetIsPhantomModeForSingleCore(bool value) {
         ASSERT(!is_multicore);
         is_phantom_mode_for_singlecore = value;
     }
@@ -439,14 +435,14 @@ struct KernelCore::Impl {
 
     static inline thread_local KThread* current_thread{nullptr};
 
-    KThread* GetCurrentEmuThread() {
+    LTO_NOINLINE KThread* GetCurrentEmuThread() {
         if (!current_thread) {
             current_thread = GetHostDummyThread(nullptr);
         }
         return current_thread;
     }
 
-    void SetCurrentEmuThread(KThread* thread) {
+    LTO_NOINLINE void SetCurrentEmuThread(KThread* thread) {
         current_thread = thread;
     }
 
@@ -623,14 +619,33 @@ struct KernelCore::Impl {
         ASSERT(memory_layout->GetPhysicalMemoryRegionTree().Insert(
             GetInteger(slab_start_phys_addr), slab_region_size, KMemoryRegionType_DramKernelSlab));
 
+        // Insert a physical region for the secure applet memory.
+        const auto secure_applet_end_phys_addr =
+            slab_end_phys_addr + KSystemControl::SecureAppletMemorySize;
+        if constexpr (KSystemControl::SecureAppletMemorySize > 0) {
+            ASSERT(memory_layout->GetPhysicalMemoryRegionTree().Insert(
+                GetInteger(slab_end_phys_addr), KSystemControl::SecureAppletMemorySize,
+                KMemoryRegionType_DramKernelSecureAppletMemory));
+        }
+
+        // Insert a physical region for the unknown debug2 region.
+        constexpr size_t SecureUnknownRegionSize = 0;
+        const size_t secure_unknown_size = SecureUnknownRegionSize;
+        const auto secure_unknown_end_phys_addr = secure_applet_end_phys_addr + secure_unknown_size;
+        if constexpr (SecureUnknownRegionSize > 0) {
+            ASSERT(memory_layout->GetPhysicalMemoryRegionTree().Insert(
+                GetInteger(secure_applet_end_phys_addr), secure_unknown_size,
+                KMemoryRegionType_DramKernelSecureUnknown));
+        }
+
         // Determine size available for kernel page table heaps, requiring > 8 MB.
         const KPhysicalAddress resource_end_phys_addr = slab_start_phys_addr + resource_region_size;
-        const size_t page_table_heap_size = resource_end_phys_addr - slab_end_phys_addr;
+        const size_t page_table_heap_size = resource_end_phys_addr - secure_unknown_end_phys_addr;
         ASSERT(page_table_heap_size / 4_MiB > 2);
 
         // Insert a physical region for the kernel page table heap region
         ASSERT(memory_layout->GetPhysicalMemoryRegionTree().Insert(
-            GetInteger(slab_end_phys_addr), page_table_heap_size,
+            GetInteger(secure_unknown_end_phys_addr), page_table_heap_size,
             KMemoryRegionType_DramKernelPtHeap));
 
         // All DRAM regions that we haven't tagged by this point will be mapped under the linear
@@ -773,8 +788,8 @@ struct KernelCore::Impl {
     std::mutex registered_in_use_objects_lock;
 
     std::atomic<u32> next_object_id{0};
-    std::atomic<u64> next_kernel_process_id{KProcess::InitialKIPIDMin};
-    std::atomic<u64> next_user_process_id{KProcess::ProcessIDMin};
+    std::atomic<u64> next_kernel_process_id{KProcess::InitialProcessIdMin};
+    std::atomic<u64> next_user_process_id{KProcess::ProcessIdMin};
     std::atomic<u64> next_thread_id{1};
 
     // Lists all processes that exist in the current session.
@@ -905,10 +920,6 @@ const KProcess* KernelCore::ApplicationProcess() const {
     return impl->application_process;
 }
 
-void KernelCore::CloseApplicationProcess() {
-    impl->CloseApplicationProcess();
-}
-
 const std::vector<KProcess*>& KernelCore::GetProcessList() const {
     return impl->process_list;
 }
@@ -1109,8 +1120,8 @@ std::jthread KernelCore::RunOnHostCoreProcess(std::string&& process_name,
                                               std::function<void()> func) {
     // Make a new process.
     KProcess* process = KProcess::Create(*this);
-    ASSERT(R_SUCCEEDED(KProcess::Initialize(process, System(), "", KProcess::ProcessType::Userland,
-                                            GetSystemResourceLimit())));
+    ASSERT(R_SUCCEEDED(
+        process->Initialize(Svc::CreateProcessParameter{}, GetSystemResourceLimit(), false)));
 
     // Ensure that we don't hold onto any extra references.
     SCOPE_EXIT({ process->Close(); });
@@ -1137,8 +1148,8 @@ void KernelCore::RunOnGuestCoreProcess(std::string&& process_name, std::function
 
     // Make a new process.
     KProcess* process = KProcess::Create(*this);
-    ASSERT(R_SUCCEEDED(KProcess::Initialize(process, System(), "", KProcess::ProcessType::Userland,
-                                            GetSystemResourceLimit())));
+    ASSERT(R_SUCCEEDED(
+        process->Initialize(Svc::CreateProcessParameter{}, GetSystemResourceLimit(), false)));
 
     // Ensure that we don't hold onto any extra references.
     SCOPE_EXIT({ process->Close(); });
@@ -1247,7 +1258,8 @@ const Kernel::KSharedMemory& KernelCore::GetHidBusSharedMem() const {
 
 void KernelCore::SuspendApplication(bool suspended) {
     const bool should_suspend{exception_exited || suspended};
-    const auto activity = should_suspend ? ProcessActivity::Paused : ProcessActivity::Runnable;
+    const auto activity =
+        should_suspend ? Svc::ProcessActivity::Paused : Svc::ProcessActivity::Runnable;
 
     // Get the application process.
     KScopedAutoObject<KProcess> process = ApplicationProcess();
@@ -1281,6 +1293,8 @@ void KernelCore::SuspendApplication(bool suspended) {
 }
 
 void KernelCore::ShutdownCores() {
+    impl->TerminateApplicationProcess();
+
     KScopedSchedulerLock lk{*this};
 
     for (auto* thread : impl->shutdown_threads) {
diff --git a/src/core/hle/kernel/kernel.h b/src/core/hle/kernel/kernel.h
index d5b08eeb5..d8086c0ea 100644
--- a/src/core/hle/kernel/kernel.h
+++ b/src/core/hle/kernel/kernel.h
@@ -134,9 +134,6 @@ public:
     /// Retrieves a const pointer to the application process.
     const KProcess* ApplicationProcess() const;
 
-    /// Closes the application process.
-    void CloseApplicationProcess();
-
     /// Retrieves the list of processes.
     const std::vector<KProcess*>& GetProcessList() const;
 
diff --git a/src/core/hle/kernel/svc.cpp b/src/core/hle/kernel/svc.cpp
index 871d541d4..b76683969 100644
--- a/src/core/hle/kernel/svc.cpp
+++ b/src/core/hle/kernel/svc.cpp
@@ -4426,7 +4426,7 @@ void Call(Core::System& system, u32 imm) {
     auto& kernel = system.Kernel();
     kernel.EnterSVCProfile();
 
-    if (GetCurrentProcess(system.Kernel()).Is64BitProcess()) {
+    if (GetCurrentProcess(system.Kernel()).Is64Bit()) {
         Call64(system, imm);
     } else {
         Call32(system, imm);
diff --git a/src/core/hle/kernel/svc/svc_info.cpp b/src/core/hle/kernel/svc/svc_info.cpp
index f99964028..ada998772 100644
--- a/src/core/hle/kernel/svc/svc_info.cpp
+++ b/src/core/hle/kernel/svc/svc_info.cpp
@@ -86,20 +86,19 @@ Result GetInfo(Core::System& system, u64* result, InfoType info_id_type, Handle
             R_SUCCEED();
 
         case InfoType::TotalMemorySize:
-            *result = process->GetTotalPhysicalMemoryAvailable();
+            *result = process->GetTotalUserPhysicalMemorySize();
             R_SUCCEED();
 
         case InfoType::UsedMemorySize:
-            *result = process->GetTotalPhysicalMemoryUsed();
+            *result = process->GetUsedUserPhysicalMemorySize();
             R_SUCCEED();
 
         case InfoType::SystemResourceSizeTotal:
-            *result = process->GetSystemResourceSize();
+            *result = process->GetTotalSystemResourceSize();
             R_SUCCEED();
 
         case InfoType::SystemResourceSizeUsed:
-            LOG_WARNING(Kernel_SVC, "(STUBBED) Attempted to query system resource usage");
-            *result = process->GetSystemResourceUsage();
+            *result = process->GetUsedSystemResourceSize();
             R_SUCCEED();
 
         case InfoType::ProgramId:
@@ -111,20 +110,29 @@ Result GetInfo(Core::System& system, u64* result, InfoType info_id_type, Handle
             R_SUCCEED();
 
         case InfoType::TotalNonSystemMemorySize:
-            *result = process->GetTotalPhysicalMemoryAvailableWithoutSystemResource();
+            *result = process->GetTotalNonSystemUserPhysicalMemorySize();
             R_SUCCEED();
 
         case InfoType::UsedNonSystemMemorySize:
-            *result = process->GetTotalPhysicalMemoryUsedWithoutSystemResource();
+            *result = process->GetUsedNonSystemUserPhysicalMemorySize();
             R_SUCCEED();
 
         case InfoType::IsApplication:
             LOG_WARNING(Kernel_SVC, "(STUBBED) Assuming process is application");
-            *result = true;
+            *result = process->IsApplication();
             R_SUCCEED();
 
         case InfoType::FreeThreadCount:
-            *result = process->GetFreeThreadCount();
+            if (KResourceLimit* resource_limit = process->GetResourceLimit();
+                resource_limit != nullptr) {
+                const auto current_value =
+                    resource_limit->GetCurrentValue(Svc::LimitableResource::ThreadCountMax);
+                const auto limit_value =
+                    resource_limit->GetLimitValue(Svc::LimitableResource::ThreadCountMax);
+                *result = limit_value - current_value;
+            } else {
+                *result = 0;
+            }
             R_SUCCEED();
 
         default:
@@ -161,7 +169,7 @@ Result GetInfo(Core::System& system, u64* result, InfoType info_id_type, Handle
 
     case InfoType::RandomEntropy:
         R_UNLESS(handle == 0, ResultInvalidHandle);
-        R_UNLESS(info_sub_id < KProcess::RANDOM_ENTROPY_SIZE, ResultInvalidCombination);
+        R_UNLESS(info_sub_id < 4, ResultInvalidCombination);
 
         *result = GetCurrentProcess(system.Kernel()).GetRandomEntropy(info_sub_id);
         R_SUCCEED();
diff --git a/src/core/hle/kernel/svc/svc_lock.cpp b/src/core/hle/kernel/svc/svc_lock.cpp
index 1d7bc4246..5f0833fcb 100644
--- a/src/core/hle/kernel/svc/svc_lock.cpp
+++ b/src/core/hle/kernel/svc/svc_lock.cpp
@@ -17,7 +17,7 @@ Result ArbitrateLock(Core::System& system, Handle thread_handle, u64 address, u3
     R_UNLESS(!IsKernelAddress(address), ResultInvalidCurrentMemory);
     R_UNLESS(Common::IsAligned(address, sizeof(u32)), ResultInvalidAddress);
 
-    R_RETURN(GetCurrentProcess(system.Kernel()).WaitForAddress(thread_handle, address, tag));
+    R_RETURN(KConditionVariable::WaitForAddress(system.Kernel(), thread_handle, address, tag));
 }
 
 /// Unlock a mutex
@@ -28,7 +28,7 @@ Result ArbitrateUnlock(Core::System& system, u64 address) {
     R_UNLESS(!IsKernelAddress(address), ResultInvalidCurrentMemory);
     R_UNLESS(Common::IsAligned(address, sizeof(u32)), ResultInvalidAddress);
 
-    R_RETURN(GetCurrentProcess(system.Kernel()).SignalToAddress(address));
+    R_RETURN(KConditionVariable::SignalToAddress(system.Kernel(), address));
 }
 
 Result ArbitrateLock64(Core::System& system, Handle thread_handle, uint64_t address, uint32_t tag) {
diff --git a/src/core/hle/kernel/svc/svc_memory.cpp b/src/core/hle/kernel/svc/svc_memory.cpp
index 2cab74127..97f1210de 100644
--- a/src/core/hle/kernel/svc/svc_memory.cpp
+++ b/src/core/hle/kernel/svc/svc_memory.cpp
@@ -76,7 +76,7 @@ Result MapUnmapMemorySanityChecks(const KPageTable& manager, u64 dst_addr, u64 s
 } // namespace
 
 Result SetMemoryPermission(Core::System& system, u64 address, u64 size, MemoryPermission perm) {
-    LOG_DEBUG(Kernel_SVC, "called, address=0x{:016X}, size=0x{:X}, perm=0x{:08X", address, size,
+    LOG_DEBUG(Kernel_SVC, "called, address=0x{:016X}, size=0x{:X}, perm=0x{:08X}", address, size,
               perm);
 
     // Validate address / size.
@@ -108,10 +108,16 @@ Result SetMemoryAttribute(Core::System& system, u64 address, u64 size, u32 mask,
     R_UNLESS((address < address + size), ResultInvalidCurrentMemory);
 
     // Validate the attribute and mask.
-    constexpr u32 SupportedMask = static_cast<u32>(MemoryAttribute::Uncached);
+    constexpr u32 SupportedMask =
+        static_cast<u32>(MemoryAttribute::Uncached | MemoryAttribute::PermissionLocked);
     R_UNLESS((mask | attr) == mask, ResultInvalidCombination);
     R_UNLESS((mask | attr | SupportedMask) == SupportedMask, ResultInvalidCombination);
 
+    // Check that permission locked is either being set or not masked.
+    R_UNLESS((static_cast<Svc::MemoryAttribute>(mask) & Svc::MemoryAttribute::PermissionLocked) ==
+                 (static_cast<Svc::MemoryAttribute>(attr) & Svc::MemoryAttribute::PermissionLocked),
+             ResultInvalidCombination);
+
     // Validate that the region is in range for the current process.
     auto& page_table{GetCurrentProcess(system.Kernel()).GetPageTable()};
     R_UNLESS(page_table.Contains(address, size), ResultInvalidCurrentMemory);
diff --git a/src/core/hle/kernel/svc/svc_physical_memory.cpp b/src/core/hle/kernel/svc/svc_physical_memory.cpp
index d3545f232..99330d02a 100644
--- a/src/core/hle/kernel/svc/svc_physical_memory.cpp
+++ b/src/core/hle/kernel/svc/svc_physical_memory.cpp
@@ -46,7 +46,7 @@ Result MapPhysicalMemory(Core::System& system, u64 addr, u64 size) {
     KProcess* const current_process{GetCurrentProcessPointer(system.Kernel())};
     auto& page_table{current_process->GetPageTable()};
 
-    if (current_process->GetSystemResourceSize() == 0) {
+    if (current_process->GetTotalSystemResourceSize() == 0) {
         LOG_ERROR(Kernel_SVC, "System Resource Size is zero");
         R_THROW(ResultInvalidState);
     }
@@ -95,7 +95,7 @@ Result UnmapPhysicalMemory(Core::System& system, u64 addr, u64 size) {
     KProcess* const current_process{GetCurrentProcessPointer(system.Kernel())};
     auto& page_table{current_process->GetPageTable()};
 
-    if (current_process->GetSystemResourceSize() == 0) {
+    if (current_process->GetTotalSystemResourceSize() == 0) {
         LOG_ERROR(Kernel_SVC, "System Resource Size is zero");
         R_THROW(ResultInvalidState);
     }
diff --git a/src/core/hle/kernel/svc/svc_synchronization.cpp b/src/core/hle/kernel/svc/svc_synchronization.cpp
index 8ebc1bd1c..6c79cfd8d 100644
--- a/src/core/hle/kernel/svc/svc_synchronization.cpp
+++ b/src/core/hle/kernel/svc/svc_synchronization.cpp
@@ -132,7 +132,7 @@ void SynchronizePreemptionState(Core::System& system) {
         GetCurrentThread(kernel).ClearInterruptFlag();
 
         // Unpin the current thread.
-        cur_process->UnpinCurrentThread(core_id);
+        cur_process->UnpinCurrentThread();
     }
 }
 
diff --git a/src/core/hle/kernel/svc/svc_thread.cpp b/src/core/hle/kernel/svc/svc_thread.cpp
index 933b82e30..755fd62b5 100644
--- a/src/core/hle/kernel/svc/svc_thread.cpp
+++ b/src/core/hle/kernel/svc/svc_thread.cpp
@@ -85,10 +85,6 @@ Result StartThread(Core::System& system, Handle thread_handle) {
     // Try to start the thread.
     R_TRY(thread->Run());
 
-    // If we succeeded, persist a reference to the thread.
-    thread->Open();
-    system.Kernel().RegisterInUseObject(thread.GetPointerUnsafe());
-
     R_SUCCEED();
 }
 
@@ -99,7 +95,6 @@ void ExitThread(Core::System& system) {
     auto* const current_thread = GetCurrentThreadPointer(system.Kernel());
     system.GlobalSchedulerContext().RemoveThread(current_thread);
     current_thread->Exit();
-    system.Kernel().UnregisterInUseObject(current_thread);
 }
 
 /// Sleep the current thread
@@ -260,7 +255,7 @@ Result GetThreadList(Core::System& system, s32* out_num_threads, u64 out_thread_
 
     auto list_iter = thread_list.cbegin();
     for (std::size_t i = 0; i < copy_amount; ++i, ++list_iter) {
-        memory.Write64(out_thread_ids, (*list_iter)->GetThreadId());
+        memory.Write64(out_thread_ids, list_iter->GetThreadId());
         out_thread_ids += sizeof(u64);
     }
 
diff --git a/src/core/hle/kernel/svc/svc_transfer_memory.cpp b/src/core/hle/kernel/svc/svc_transfer_memory.cpp
index 7d94e7f09..1f97121b3 100644
--- a/src/core/hle/kernel/svc/svc_transfer_memory.cpp
+++ b/src/core/hle/kernel/svc/svc_transfer_memory.cpp
@@ -71,15 +71,59 @@ Result CreateTransferMemory(Core::System& system, Handle* out, u64 address, u64
 }
 
 Result MapTransferMemory(Core::System& system, Handle trmem_handle, uint64_t address, uint64_t size,
-                         MemoryPermission owner_perm) {
-    UNIMPLEMENTED();
-    R_THROW(ResultNotImplemented);
+                         MemoryPermission map_perm) {
+    // Validate the address/size.
+    R_UNLESS(Common::IsAligned(address, PageSize), ResultInvalidAddress);
+    R_UNLESS(Common::IsAligned(size, PageSize), ResultInvalidSize);
+    R_UNLESS(size > 0, ResultInvalidSize);
+    R_UNLESS((address < address + size), ResultInvalidCurrentMemory);
+
+    // Validate the permission.
+    R_UNLESS(IsValidTransferMemoryPermission(map_perm), ResultInvalidState);
+
+    // Get the transfer memory.
+    KScopedAutoObject trmem = GetCurrentProcess(system.Kernel())
+                                  .GetHandleTable()
+                                  .GetObject<KTransferMemory>(trmem_handle);
+    R_UNLESS(trmem.IsNotNull(), ResultInvalidHandle);
+
+    // Verify that the mapping is in range.
+    R_UNLESS(GetCurrentProcess(system.Kernel())
+                 .GetPageTable()
+                 .CanContain(address, size, KMemoryState::Transfered),
+             ResultInvalidMemoryRegion);
+
+    // Map the transfer memory.
+    R_TRY(trmem->Map(address, size, map_perm));
+
+    // We succeeded.
+    R_SUCCEED();
 }
 
 Result UnmapTransferMemory(Core::System& system, Handle trmem_handle, uint64_t address,
                            uint64_t size) {
-    UNIMPLEMENTED();
-    R_THROW(ResultNotImplemented);
+    // Validate the address/size.
+    R_UNLESS(Common::IsAligned(address, PageSize), ResultInvalidAddress);
+    R_UNLESS(Common::IsAligned(size, PageSize), ResultInvalidSize);
+    R_UNLESS(size > 0, ResultInvalidSize);
+    R_UNLESS((address < address + size), ResultInvalidCurrentMemory);
+
+    // Get the transfer memory.
+    KScopedAutoObject trmem = GetCurrentProcess(system.Kernel())
+                                  .GetHandleTable()
+                                  .GetObject<KTransferMemory>(trmem_handle);
+    R_UNLESS(trmem.IsNotNull(), ResultInvalidHandle);
+
+    // Verify that the mapping is in range.
+    R_UNLESS(GetCurrentProcess(system.Kernel())
+                 .GetPageTable()
+                 .CanContain(address, size, KMemoryState::Transfered),
+             ResultInvalidMemoryRegion);
+
+    // Unmap the transfer memory.
+    R_TRY(trmem->Unmap(address, size));
+
+    R_SUCCEED();
 }
 
 Result MapTransferMemory64(Core::System& system, Handle trmem_handle, uint64_t address,
diff --git a/src/core/hle/kernel/svc_generator.py b/src/core/hle/kernel/svc_generator.py
index 7fcbb1ba1..5531faac6 100644
--- a/src/core/hle/kernel/svc_generator.py
+++ b/src/core/hle/kernel/svc_generator.py
@@ -592,7 +592,7 @@ void Call(Core::System& system, u32 imm) {
     auto& kernel = system.Kernel();
     kernel.EnterSVCProfile();
 
-    if (GetCurrentProcess(system.Kernel()).Is64BitProcess()) {
+    if (GetCurrentProcess(system.Kernel()).Is64Bit()) {
         Call64(system, imm);
     } else {
         Call32(system, imm);
diff --git a/src/core/hle/kernel/svc_types.h b/src/core/hle/kernel/svc_types.h
index 7f380ca4f..50de02e36 100644
--- a/src/core/hle/kernel/svc_types.h
+++ b/src/core/hle/kernel/svc_types.h
@@ -46,6 +46,7 @@ enum class MemoryAttribute : u32 {
     IpcLocked = (1 << 1),
     DeviceShared = (1 << 2),
     Uncached = (1 << 3),
+    PermissionLocked = (1 << 4),
 };
 DECLARE_ENUM_FLAG_OPERATORS(MemoryAttribute);
 
@@ -603,13 +604,57 @@ enum class ProcessActivity : u32 {
     Paused,
 };
 
+enum class CreateProcessFlag : u32 {
+    // Is 64 bit?
+    Is64Bit = (1 << 0),
+
+    // What kind of address space?
+    AddressSpaceShift = 1,
+    AddressSpaceMask = (7 << AddressSpaceShift),
+    AddressSpace32Bit = (0 << AddressSpaceShift),
+    AddressSpace64BitDeprecated = (1 << AddressSpaceShift),
+    AddressSpace32BitWithoutAlias = (2 << AddressSpaceShift),
+    AddressSpace64Bit = (3 << AddressSpaceShift),
+
+    // Should JIT debug be done on crash?
+    EnableDebug = (1 << 4),
+
+    // Should ASLR be enabled for the process?
+    EnableAslr = (1 << 5),
+
+    // Is the process an application?
+    IsApplication = (1 << 6),
+
+    // 4.x deprecated: Should use secure memory?
+    DeprecatedUseSecureMemory = (1 << 7),
+
+    // 5.x+ Pool partition type.
+    PoolPartitionShift = 7,
+    PoolPartitionMask = (0xF << PoolPartitionShift),
+    PoolPartitionApplication = (0 << PoolPartitionShift),
+    PoolPartitionApplet = (1 << PoolPartitionShift),
+    PoolPartitionSystem = (2 << PoolPartitionShift),
+    PoolPartitionSystemNonSecure = (3 << PoolPartitionShift),
+
+    // 7.x+ Should memory allocation be optimized? This requires IsApplication.
+    OptimizeMemoryAllocation = (1 << 11),
+
+    // 11.x+ DisableDeviceAddressSpaceMerge.
+    DisableDeviceAddressSpaceMerge = (1 << 12),
+
+    // Mask of all flags.
+    All = Is64Bit | AddressSpaceMask | EnableDebug | EnableAslr | IsApplication |
+          PoolPartitionMask | OptimizeMemoryAllocation | DisableDeviceAddressSpaceMerge,
+};
+DECLARE_ENUM_FLAG_OPERATORS(CreateProcessFlag);
+
 struct CreateProcessParameter {
     std::array<char, 12> name;
     u32 version;
     u64 program_id;
     u64 code_address;
     s32 code_num_pages;
-    u32 flags;
+    CreateProcessFlag flags;
     Handle reslimit;
     s32 system_resource_num_pages;
 };
diff --git a/src/core/hle/service/acc/acc.cpp b/src/core/hle/service/acc/acc.cpp
index b971401e6..1b1c8190e 100644
--- a/src/core/hle/service/acc/acc.cpp
+++ b/src/core/hle/service/acc/acc.cpp
@@ -49,7 +49,7 @@ public:
         : ServiceFramework{system_, "IManagerForSystemService"} {
         // clang-format off
         static const FunctionInfo functions[] = {
-            {0, nullptr, "CheckAvailability"},
+            {0, &IManagerForSystemService::CheckAvailability, "CheckAvailability"},
             {1, nullptr, "GetAccountId"},
             {2, nullptr, "EnsureIdTokenCacheAsync"},
             {3, nullptr, "LoadIdTokenCache"},
@@ -78,6 +78,13 @@ public:
 
         RegisterHandlers(functions);
     }
+
+private:
+    void CheckAvailability(HLERequestContext& ctx) {
+        LOG_WARNING(Service_ACC, "(STUBBED) called");
+        IPC::ResponseBuilder rb{ctx, 2};
+        rb.Push(ResultSuccess);
+    }
 };
 
 // 3.0.0+
@@ -400,13 +407,13 @@ protected:
         IPC::RequestParser rp{ctx};
         const auto base = rp.PopRaw<ProfileBase>();
 
-        const auto user_data = ctx.ReadBuffer();
-        const auto image_data = ctx.ReadBuffer(1);
+        const auto image_data = ctx.ReadBufferA(0);
+        const auto user_data = ctx.ReadBufferX(0);
 
-        LOG_DEBUG(Service_ACC, "called, username='{}', timestamp={:016X}, uuid=0x{}",
-                  Common::StringFromFixedZeroTerminatedBuffer(
-                      reinterpret_cast<const char*>(base.username.data()), base.username.size()),
-                  base.timestamp, base.user_uuid.RawString());
+        LOG_INFO(Service_ACC, "called, username='{}', timestamp={:016X}, uuid=0x{}",
+                 Common::StringFromFixedZeroTerminatedBuffer(
+                     reinterpret_cast<const char*>(base.username.data()), base.username.size()),
+                 base.timestamp, base.user_uuid.RawString());
 
         if (user_data.size() < sizeof(UserData)) {
             LOG_ERROR(Service_ACC, "UserData buffer too small!");
@@ -837,6 +844,29 @@ void Module::Interface::InitializeApplicationInfoV2(HLERequestContext& ctx) {
     rb.Push(ResultSuccess);
 }
 
+void Module::Interface::BeginUserRegistration(HLERequestContext& ctx) {
+    const auto user_id = Common::UUID::MakeRandom();
+    profile_manager->CreateNewUser(user_id, "yuzu");
+
+    LOG_INFO(Service_ACC, "called, uuid={}", user_id.FormattedString());
+
+    IPC::ResponseBuilder rb{ctx, 6};
+    rb.Push(ResultSuccess);
+    rb.PushRaw(user_id);
+}
+
+void Module::Interface::CompleteUserRegistration(HLERequestContext& ctx) {
+    IPC::RequestParser rp{ctx};
+    Common::UUID user_id = rp.PopRaw<Common::UUID>();
+
+    LOG_INFO(Service_ACC, "called, uuid={}", user_id.FormattedString());
+
+    profile_manager->WriteUserSaveFile();
+
+    IPC::ResponseBuilder rb{ctx, 2};
+    rb.Push(ResultSuccess);
+}
+
 void Module::Interface::GetProfileEditor(HLERequestContext& ctx) {
     IPC::RequestParser rp{ctx};
     Common::UUID user_id = rp.PopRaw<Common::UUID>();
@@ -880,6 +910,17 @@ void Module::Interface::StoreSaveDataThumbnailApplication(HLERequestContext& ctx
     StoreSaveDataThumbnail(ctx, uuid, tid);
 }
 
+void Module::Interface::GetBaasAccountManagerForSystemService(HLERequestContext& ctx) {
+    IPC::RequestParser rp{ctx};
+    const auto uuid = rp.PopRaw<Common::UUID>();
+
+    LOG_INFO(Service_ACC, "called, uuid=0x{}", uuid.RawString());
+
+    IPC::ResponseBuilder rb{ctx, 2, 0, 1};
+    rb.Push(ResultSuccess);
+    rb.PushIpcInterface<IManagerForSystemService>(system, uuid);
+}
+
 void Module::Interface::StoreSaveDataThumbnailSystem(HLERequestContext& ctx) {
     IPC::RequestParser rp{ctx};
     const auto uuid = rp.PopRaw<Common::UUID>();
diff --git a/src/core/hle/service/acc/acc.h b/src/core/hle/service/acc/acc.h
index 6b4735c2f..0395229b4 100644
--- a/src/core/hle/service/acc/acc.h
+++ b/src/core/hle/service/acc/acc.h
@@ -33,10 +33,13 @@ public:
         void TrySelectUserWithoutInteraction(HLERequestContext& ctx);
         void IsUserAccountSwitchLocked(HLERequestContext& ctx);
         void InitializeApplicationInfoV2(HLERequestContext& ctx);
+        void BeginUserRegistration(HLERequestContext& ctx);
+        void CompleteUserRegistration(HLERequestContext& ctx);
         void GetProfileEditor(HLERequestContext& ctx);
         void ListQualifiedUsers(HLERequestContext& ctx);
         void ListOpenContextStoredUsers(HLERequestContext& ctx);
         void StoreSaveDataThumbnailApplication(HLERequestContext& ctx);
+        void GetBaasAccountManagerForSystemService(HLERequestContext& ctx);
         void StoreSaveDataThumbnailSystem(HLERequestContext& ctx);
 
     private:
diff --git a/src/core/hle/service/acc/acc_su.cpp b/src/core/hle/service/acc/acc_su.cpp
index d9882ecd3..770d13ec5 100644
--- a/src/core/hle/service/acc/acc_su.cpp
+++ b/src/core/hle/service/acc/acc_su.cpp
@@ -23,7 +23,7 @@ ACC_SU::ACC_SU(std::shared_ptr<Module> module_, std::shared_ptr<ProfileManager>
         {99, nullptr, "DebugActivateOpenContextRetention"},
         {100, nullptr, "GetUserRegistrationNotifier"},
         {101, nullptr, "GetUserStateChangeNotifier"},
-        {102, nullptr, "GetBaasAccountManagerForSystemService"},
+        {102, &ACC_SU::GetBaasAccountManagerForSystemService, "GetBaasAccountManagerForSystemService"},
         {103, nullptr, "GetBaasUserAvailabilityChangeNotifier"},
         {104, nullptr, "GetProfileUpdateNotifier"},
         {105, nullptr, "CheckNetworkServiceAvailabilityAsync"},
@@ -40,8 +40,8 @@ ACC_SU::ACC_SU(std::shared_ptr<Module> module_, std::shared_ptr<ProfileManager>
         {152, nullptr, "LoadSignedDeviceIdentifierCacheForNintendoAccount"},
         {190, nullptr, "GetUserLastOpenedApplication"},
         {191, nullptr, "ActivateOpenContextHolder"},
-        {200, nullptr, "BeginUserRegistration"},
-        {201, nullptr, "CompleteUserRegistration"},
+        {200, &ACC_SU::BeginUserRegistration, "BeginUserRegistration"},
+        {201, &ACC_SU::CompleteUserRegistration, "CompleteUserRegistration"},
         {202, nullptr, "CancelUserRegistration"},
         {203, nullptr, "DeleteUser"},
         {204, nullptr, "SetUserPosition"},
diff --git a/src/core/hle/service/acc/profile_manager.h b/src/core/hle/service/acc/profile_manager.h
index 993a5a57a..900e32200 100644
--- a/src/core/hle/service/acc/profile_manager.h
+++ b/src/core/hle/service/acc/profile_manager.h
@@ -96,9 +96,10 @@ public:
     bool SetProfileBaseAndData(Common::UUID uuid, const ProfileBase& profile_new,
                                const UserData& data_new);
 
+    void WriteUserSaveFile();
+
 private:
     void ParseUserSaveFile();
-    void WriteUserSaveFile();
     std::optional<std::size_t> AddToProfiles(const ProfileInfo& profile);
     bool RemoveProfileAtIndex(std::size_t index);
 
diff --git a/src/core/hle/service/am/am.cpp b/src/core/hle/service/am/am.cpp
index 819dea6a7..ff067c8d9 100644
--- a/src/core/hle/service/am/am.cpp
+++ b/src/core/hle/service/am/am.cpp
@@ -31,7 +31,7 @@
 #include "core/hle/service/apm/apm_controller.h"
 #include "core/hle/service/apm/apm_interface.h"
 #include "core/hle/service/bcat/backend/backend.h"
-#include "core/hle/service/caps/caps.h"
+#include "core/hle/service/caps/caps_types.h"
 #include "core/hle/service/filesystem/filesystem.h"
 #include "core/hle/service/ipc_helpers.h"
 #include "core/hle/service/ns/ns.h"
@@ -210,8 +210,8 @@ IDisplayController::IDisplayController(Core::System& system_)
         {21, nullptr, "ClearAppletTransitionBuffer"},
         {22, nullptr, "AcquireLastApplicationCaptureSharedBuffer"},
         {23, nullptr, "ReleaseLastApplicationCaptureSharedBuffer"},
-        {24, nullptr, "AcquireLastForegroundCaptureSharedBuffer"},
-        {25, nullptr, "ReleaseLastForegroundCaptureSharedBuffer"},
+        {24, &IDisplayController::AcquireLastForegroundCaptureSharedBuffer, "AcquireLastForegroundCaptureSharedBuffer"},
+        {25, &IDisplayController::ReleaseLastForegroundCaptureSharedBuffer, "ReleaseLastForegroundCaptureSharedBuffer"},
         {26, &IDisplayController::AcquireCallerAppletCaptureSharedBuffer, "AcquireCallerAppletCaptureSharedBuffer"},
         {27, &IDisplayController::ReleaseCallerAppletCaptureSharedBuffer, "ReleaseCallerAppletCaptureSharedBuffer"},
         {28, nullptr, "TakeScreenShotOfOwnLayerEx"},
@@ -239,6 +239,22 @@ void IDisplayController::TakeScreenShotOfOwnLayer(HLERequestContext& ctx) {
     rb.Push(ResultSuccess);
 }
 
+void IDisplayController::AcquireLastForegroundCaptureSharedBuffer(HLERequestContext& ctx) {
+    LOG_WARNING(Service_AM, "(STUBBED) called");
+
+    IPC::ResponseBuilder rb{ctx, 4};
+    rb.Push(ResultSuccess);
+    rb.Push(1U);
+    rb.Push(0);
+}
+
+void IDisplayController::ReleaseLastForegroundCaptureSharedBuffer(HLERequestContext& ctx) {
+    LOG_WARNING(Service_AM, "(STUBBED) called");
+
+    IPC::ResponseBuilder rb{ctx, 2};
+    rb.Push(ResultSuccess);
+}
+
 void IDisplayController::AcquireCallerAppletCaptureSharedBuffer(HLERequestContext& ctx) {
     LOG_WARNING(Service_AM, "(STUBBED) called");
 
@@ -764,6 +780,68 @@ void AppletMessageQueue::OperationModeChanged() {
     on_operation_mode_changed->Signal();
 }
 
+ILockAccessor::ILockAccessor(Core::System& system_)
+    : ServiceFramework{system_, "ILockAccessor"}, service_context{system_, "ILockAccessor"} {
+    // clang-format off
+        static const FunctionInfo functions[] = {
+            {1, &ILockAccessor::TryLock, "TryLock"},
+            {2, &ILockAccessor::Unlock, "Unlock"},
+            {3, &ILockAccessor::GetEvent, "GetEvent"},
+            {4,&ILockAccessor::IsLocked, "IsLocked"},
+        };
+    // clang-format on
+
+    RegisterHandlers(functions);
+
+    lock_event = service_context.CreateEvent("ILockAccessor::LockEvent");
+}
+
+ILockAccessor::~ILockAccessor() {
+    service_context.CloseEvent(lock_event);
+};
+
+void ILockAccessor::TryLock(HLERequestContext& ctx) {
+    IPC::RequestParser rp{ctx};
+    const auto return_handle = rp.Pop<bool>();
+
+    LOG_WARNING(Service_AM, "(STUBBED) called, return_handle={}", return_handle);
+
+    // TODO: When return_handle is true this function should return the lock handle
+
+    is_locked = true;
+
+    IPC::ResponseBuilder rb{ctx, 3};
+    rb.Push(ResultSuccess);
+    rb.Push<u8>(is_locked);
+}
+
+void ILockAccessor::Unlock(HLERequestContext& ctx) {
+    LOG_INFO(Service_AM, "called");
+
+    is_locked = false;
+
+    IPC::ResponseBuilder rb{ctx, 2};
+    rb.Push(ResultSuccess);
+}
+
+void ILockAccessor::GetEvent(HLERequestContext& ctx) {
+    LOG_INFO(Service_AM, "called");
+
+    lock_event->Signal();
+
+    IPC::ResponseBuilder rb{ctx, 2, 1};
+    rb.Push(ResultSuccess);
+    rb.PushCopyObjects(lock_event->GetReadableEvent());
+}
+
+void ILockAccessor::IsLocked(HLERequestContext& ctx) {
+    LOG_INFO(Service_AM, "called");
+
+    IPC::ResponseBuilder rb{ctx, 2};
+    rb.Push(ResultSuccess);
+    rb.Push<u8>(is_locked);
+}
+
 ICommonStateGetter::ICommonStateGetter(Core::System& system_,
                                        std::shared_ptr<AppletMessageQueue> msg_queue_)
     : ServiceFramework{system_, "ICommonStateGetter"}, msg_queue{std::move(msg_queue_)},
@@ -787,7 +865,7 @@ ICommonStateGetter::ICommonStateGetter(Core::System& system_,
         {14, nullptr, "GetWakeupCount"},
         {20, nullptr, "PushToGeneralChannel"},
         {30, nullptr, "GetHomeButtonReaderLockAccessor"},
-        {31, nullptr, "GetReaderLockAccessorEx"},
+        {31, &ICommonStateGetter::GetReaderLockAccessorEx, "GetReaderLockAccessorEx"},
         {32, nullptr, "GetWriterLockAccessorEx"},
         {40, nullptr, "GetCradleFwVersion"},
         {50, &ICommonStateGetter::IsVrModeEnabled, "IsVrModeEnabled"},
@@ -805,7 +883,7 @@ ICommonStateGetter::ICommonStateGetter(Core::System& system_,
         {65, nullptr, "GetApplicationIdByContentActionName"},
         {66, &ICommonStateGetter::SetCpuBoostMode, "SetCpuBoostMode"},
         {67, nullptr, "CancelCpuBoostMode"},
-        {68, nullptr, "GetBuiltInDisplayType"},
+        {68, &ICommonStateGetter::GetBuiltInDisplayType, "GetBuiltInDisplayType"},
         {80, &ICommonStateGetter::PerformSystemButtonPressingIfInFocus, "PerformSystemButtonPressingIfInFocus"},
         {90, nullptr, "SetPerformanceConfigurationChangedNotification"},
         {91, nullptr, "GetCurrentPerformanceConfiguration"},
@@ -833,7 +911,9 @@ ICommonStateGetter::ICommonStateGetter(Core::System& system_,
     msg_queue->PushMessage(AppletMessageQueue::AppletMessage::ChangeIntoForeground);
 }
 
-ICommonStateGetter::~ICommonStateGetter() = default;
+ICommonStateGetter::~ICommonStateGetter() {
+    service_context.CloseEvent(sleep_lock_event);
+};
 
 void ICommonStateGetter::GetBootMode(HLERequestContext& ctx) {
     LOG_DEBUG(Service_AM, "called");
@@ -886,6 +966,18 @@ void ICommonStateGetter::RequestToAcquireSleepLock(HLERequestContext& ctx) {
     rb.Push(ResultSuccess);
 }
 
+void ICommonStateGetter::GetReaderLockAccessorEx(HLERequestContext& ctx) {
+    IPC::RequestParser rp{ctx};
+    const auto unknown = rp.Pop<u32>();
+
+    LOG_INFO(Service_AM, "called, unknown={}", unknown);
+
+    IPC::ResponseBuilder rb{ctx, 2, 0, 1};
+
+    rb.Push(ResultSuccess);
+    rb.PushIpcInterface<ILockAccessor>(system);
+}
+
 void ICommonStateGetter::GetAcquiredSleepLockEvent(HLERequestContext& ctx) {
     LOG_WARNING(Service_AM, "called");
 
@@ -970,6 +1062,14 @@ void ICommonStateGetter::SetCpuBoostMode(HLERequestContext& ctx) {
     apm_sys->SetCpuBoostMode(ctx);
 }
 
+void ICommonStateGetter::GetBuiltInDisplayType(HLERequestContext& ctx) {
+    LOG_WARNING(Service_AM, "(STUBBED) called");
+
+    IPC::ResponseBuilder rb{ctx, 3};
+    rb.Push(ResultSuccess);
+    rb.Push(0);
+}
+
 void ICommonStateGetter::PerformSystemButtonPressingIfInFocus(HLERequestContext& ctx) {
     IPC::RequestParser rp{ctx};
     const auto system_button{rp.PopEnum<SystemButtonType>()};
@@ -1477,7 +1577,7 @@ ILibraryAppletSelfAccessor::ILibraryAppletSelfAccessor(Core::System& system_)
         {100, nullptr, "CreateGameMovieTrimmer"},
         {101, nullptr, "ReserveResourceForMovieOperation"},
         {102, nullptr, "UnreserveResourceForMovieOperation"},
-        {110, nullptr, "GetMainAppletAvailableUsers"},
+        {110, &ILibraryAppletSelfAccessor::GetMainAppletAvailableUsers, "GetMainAppletAvailableUsers"},
         {120, nullptr, "GetLaunchStorageInfoForDebug"},
         {130, nullptr, "GetGpuErrorDetectedSystemEvent"},
         {140, nullptr, "SetApplicationMemoryReservation"},
@@ -1493,6 +1593,9 @@ ILibraryAppletSelfAccessor::ILibraryAppletSelfAccessor(Core::System& system_)
     case Applets::AppletId::MiiEdit:
         PushInShowMiiEditData();
         break;
+    case Applets::AppletId::PhotoViewer:
+        PushInShowAlbum();
+        break;
     default:
         break;
     }
@@ -1569,6 +1672,42 @@ void ILibraryAppletSelfAccessor::GetCallerAppletIdentityInfo(HLERequestContext&
     rb.PushRaw(applet_info);
 }
 
+void ILibraryAppletSelfAccessor::GetMainAppletAvailableUsers(HLERequestContext& ctx) {
+    const Service::Account::ProfileManager manager{};
+    bool is_empty{true};
+    s32 user_count{-1};
+
+    LOG_INFO(Service_AM, "called");
+
+    if (manager.GetUserCount() > 0) {
+        is_empty = false;
+        user_count = static_cast<s32>(manager.GetUserCount());
+        ctx.WriteBuffer(manager.GetAllUsers());
+    }
+
+    IPC::ResponseBuilder rb{ctx, 4};
+    rb.Push(ResultSuccess);
+    rb.Push<u8>(is_empty);
+    rb.Push(user_count);
+}
+
+void ILibraryAppletSelfAccessor::PushInShowAlbum() {
+    const Applets::CommonArguments arguments{
+        .arguments_version = Applets::CommonArgumentVersion::Version3,
+        .size = Applets::CommonArgumentSize::Version3,
+        .library_version = 1,
+        .theme_color = Applets::ThemeColor::BasicBlack,
+        .play_startup_sound = true,
+        .system_tick = system.CoreTiming().GetClockTicks(),
+    };
+
+    std::vector<u8> argument_data(sizeof(arguments));
+    std::vector<u8> settings_data{2};
+    std::memcpy(argument_data.data(), &arguments, sizeof(arguments));
+    queue_data.emplace_back(std::move(argument_data));
+    queue_data.emplace_back(std::move(settings_data));
+}
+
 void ILibraryAppletSelfAccessor::PushInShowCabinetData() {
     const Applets::CommonArguments arguments{
         .arguments_version = Applets::CommonArgumentVersion::Version3,
diff --git a/src/core/hle/service/am/am.h b/src/core/hle/service/am/am.h
index 349482dcc..64b3f3fe2 100644
--- a/src/core/hle/service/am/am.h
+++ b/src/core/hle/service/am/am.h
@@ -124,6 +124,8 @@ public:
 private:
     void GetCallerAppletCaptureImageEx(HLERequestContext& ctx);
     void TakeScreenShotOfOwnLayer(HLERequestContext& ctx);
+    void AcquireLastForegroundCaptureSharedBuffer(HLERequestContext& ctx);
+    void ReleaseLastForegroundCaptureSharedBuffer(HLERequestContext& ctx);
     void AcquireCallerAppletCaptureSharedBuffer(HLERequestContext& ctx);
     void ReleaseCallerAppletCaptureSharedBuffer(HLERequestContext& ctx);
 };
@@ -195,6 +197,23 @@ private:
     ScreenshotPermission screenshot_permission = ScreenshotPermission::Inherit;
 };
 
+class ILockAccessor final : public ServiceFramework<ILockAccessor> {
+public:
+    explicit ILockAccessor(Core::System& system_);
+    ~ILockAccessor() override;
+
+private:
+    void TryLock(HLERequestContext& ctx);
+    void Unlock(HLERequestContext& ctx);
+    void GetEvent(HLERequestContext& ctx);
+    void IsLocked(HLERequestContext& ctx);
+
+    bool is_locked{};
+
+    Kernel::KEvent* lock_event;
+    KernelHelpers::ServiceContext service_context;
+};
+
 class ICommonStateGetter final : public ServiceFramework<ICommonStateGetter> {
 public:
     explicit ICommonStateGetter(Core::System& system_,
@@ -237,6 +256,7 @@ private:
     void GetCurrentFocusState(HLERequestContext& ctx);
     void RequestToAcquireSleepLock(HLERequestContext& ctx);
     void GetAcquiredSleepLockEvent(HLERequestContext& ctx);
+    void GetReaderLockAccessorEx(HLERequestContext& ctx);
     void GetDefaultDisplayResolutionChangeEvent(HLERequestContext& ctx);
     void GetOperationMode(HLERequestContext& ctx);
     void GetPerformanceMode(HLERequestContext& ctx);
@@ -248,6 +268,7 @@ private:
     void EndVrModeEx(HLERequestContext& ctx);
     void GetDefaultDisplayResolution(HLERequestContext& ctx);
     void SetCpuBoostMode(HLERequestContext& ctx);
+    void GetBuiltInDisplayType(HLERequestContext& ctx);
     void PerformSystemButtonPressingIfInFocus(HLERequestContext& ctx);
     void GetSettingsPlatformRegion(HLERequestContext& ctx);
     void SetRequestExitToLibraryAppletAtExecuteNextProgramEnabled(HLERequestContext& ctx);
@@ -326,7 +347,9 @@ private:
     void GetLibraryAppletInfo(HLERequestContext& ctx);
     void ExitProcessAndReturn(HLERequestContext& ctx);
     void GetCallerAppletIdentityInfo(HLERequestContext& ctx);
+    void GetMainAppletAvailableUsers(HLERequestContext& ctx);
 
+    void PushInShowAlbum();
     void PushInShowCabinetData();
     void PushInShowMiiEditData();
 
diff --git a/src/core/hle/service/am/applet_ae.cpp b/src/core/hle/service/am/applet_ae.cpp
index eb12312cc..e30e6478a 100644
--- a/src/core/hle/service/am/applet_ae.cpp
+++ b/src/core/hle/service/am/applet_ae.cpp
@@ -28,8 +28,8 @@ public:
             {11, &ILibraryAppletProxy::GetLibraryAppletCreator, "GetLibraryAppletCreator"},
             {20, &ILibraryAppletProxy::OpenLibraryAppletSelfAccessor, "OpenLibraryAppletSelfAccessor"},
             {21, &ILibraryAppletProxy::GetAppletCommonFunctions, "GetAppletCommonFunctions"},
-            {22, nullptr, "GetHomeMenuFunctions"},
-            {23, nullptr, "GetGlobalStateController"},
+            {22, &ILibraryAppletProxy::GetHomeMenuFunctions, "GetHomeMenuFunctions"},
+            {23, &ILibraryAppletProxy::GetGlobalStateController, "GetGlobalStateController"},
             {1000, &ILibraryAppletProxy::GetDebugFunctions, "GetDebugFunctions"},
         };
         // clang-format on
@@ -110,6 +110,22 @@ private:
         rb.PushIpcInterface<IAppletCommonFunctions>(system);
     }
 
+    void GetHomeMenuFunctions(HLERequestContext& ctx) {
+        LOG_DEBUG(Service_AM, "called");
+
+        IPC::ResponseBuilder rb{ctx, 2, 0, 1};
+        rb.Push(ResultSuccess);
+        rb.PushIpcInterface<IHomeMenuFunctions>(system);
+    }
+
+    void GetGlobalStateController(HLERequestContext& ctx) {
+        LOG_DEBUG(Service_AM, "called");
+
+        IPC::ResponseBuilder rb{ctx, 2, 0, 1};
+        rb.Push(ResultSuccess);
+        rb.PushIpcInterface<IGlobalStateController>(system);
+    }
+
     void GetDebugFunctions(HLERequestContext& ctx) {
         LOG_DEBUG(Service_AM, "called");
 
diff --git a/src/core/hle/service/am/applets/applet_cabinet.cpp b/src/core/hle/service/am/applets/applet_cabinet.cpp
index 19ed184e8..b379dadeb 100644
--- a/src/core/hle/service/am/applets/applet_cabinet.cpp
+++ b/src/core/hle/service/am/applets/applet_cabinet.cpp
@@ -25,7 +25,9 @@ Cabinet::Cabinet(Core::System& system_, LibraryAppletMode applet_mode_,
         service_context.CreateEvent("CabinetApplet:AvailabilityChangeEvent");
 }
 
-Cabinet::~Cabinet() = default;
+Cabinet::~Cabinet() {
+    service_context.CloseEvent(availability_change_event);
+};
 
 void Cabinet::Initialize() {
     Applet::Initialize();
diff --git a/src/core/hle/service/am/applets/applet_error.cpp b/src/core/hle/service/am/applets/applet_error.cpp
index b46ea840c..5d17c353f 100644
--- a/src/core/hle/service/am/applets/applet_error.cpp
+++ b/src/core/hle/service/am/applets/applet_error.cpp
@@ -138,6 +138,10 @@ void Error::Initialize() {
         CopyArgumentData(data, args->application_error);
         error_code = Result(args->application_error.error_code);
         break;
+    case ErrorAppletMode::ShowErrorPctl:
+        CopyArgumentData(data, args->error_record);
+        error_code = Decode64BitError(args->error_record.error_code_64);
+        break;
     case ErrorAppletMode::ShowErrorRecord:
         CopyArgumentData(data, args->error_record);
         error_code = Decode64BitError(args->error_record.error_code_64);
@@ -191,6 +195,7 @@ void Error::Execute() {
         frontend.ShowCustomErrorText(error_code, main_text_string, detail_text_string, callback);
         break;
     }
+    case ErrorAppletMode::ShowErrorPctl:
     case ErrorAppletMode::ShowErrorRecord:
         reporter.SaveErrorReport(title_id, error_code,
                                  fmt::format("{:016X}", args->error_record.posix_time));
diff --git a/src/core/hle/service/caps/caps.cpp b/src/core/hle/service/caps/caps.cpp
index 610fe9940..cd1dfe993 100644
--- a/src/core/hle/service/caps/caps.cpp
+++ b/src/core/hle/service/caps/caps.cpp
@@ -4,6 +4,7 @@
 #include "core/hle/service/caps/caps.h"
 #include "core/hle/service/caps/caps_a.h"
 #include "core/hle/service/caps/caps_c.h"
+#include "core/hle/service/caps/caps_manager.h"
 #include "core/hle/service/caps/caps_sc.h"
 #include "core/hle/service/caps/caps_ss.h"
 #include "core/hle/service/caps/caps_su.h"
@@ -15,13 +16,22 @@ namespace Service::Capture {
 
 void LoopProcess(Core::System& system) {
     auto server_manager = std::make_unique<ServerManager>(system);
+    auto album_manager = std::make_shared<AlbumManager>(system);
+
+    server_manager->RegisterNamedService(
+        "caps:a", std::make_shared<IAlbumAccessorService>(system, album_manager));
+    server_manager->RegisterNamedService(
+        "caps:c", std::make_shared<IAlbumControlService>(system, album_manager));
+    server_manager->RegisterNamedService(
+        "caps:u", std::make_shared<IAlbumApplicationService>(system, album_manager));
+
+    server_manager->RegisterNamedService(
+        "caps:ss", std::make_shared<IScreenShotService>(system, album_manager));
+    server_manager->RegisterNamedService("caps:sc",
+                                         std::make_shared<IScreenShotControlService>(system));
+    server_manager->RegisterNamedService(
+        "caps:su", std::make_shared<IScreenShotApplicationService>(system, album_manager));
 
-    server_manager->RegisterNamedService("caps:a", std::make_shared<CAPS_A>(system));
-    server_manager->RegisterNamedService("caps:c", std::make_shared<CAPS_C>(system));
-    server_manager->RegisterNamedService("caps:u", std::make_shared<CAPS_U>(system));
-    server_manager->RegisterNamedService("caps:sc", std::make_shared<CAPS_SC>(system));
-    server_manager->RegisterNamedService("caps:ss", std::make_shared<CAPS_SS>(system));
-    server_manager->RegisterNamedService("caps:su", std::make_shared<CAPS_SU>(system));
     ServerManager::RunServer(std::move(server_manager));
 }
 
diff --git a/src/core/hle/service/caps/caps.h b/src/core/hle/service/caps/caps.h
index 15f0ecfaa..58e9725b8 100644
--- a/src/core/hle/service/caps/caps.h
+++ b/src/core/hle/service/caps/caps.h
@@ -3,93 +3,12 @@
 
 #pragma once
 
-#include "common/common_funcs.h"
-#include "common/common_types.h"
-
 namespace Core {
 class System;
 }
 
-namespace Service::SM {
-class ServiceManager;
-}
-
 namespace Service::Capture {
 
-enum class AlbumImageOrientation {
-    Orientation0 = 0,
-    Orientation1 = 1,
-    Orientation2 = 2,
-    Orientation3 = 3,
-};
-
-enum class AlbumReportOption : s32 {
-    Disable = 0,
-    Enable = 1,
-};
-
-enum class ContentType : u8 {
-    Screenshot = 0,
-    Movie = 1,
-    ExtraMovie = 3,
-};
-
-enum class AlbumStorage : u8 {
-    NAND = 0,
-    SD = 1,
-};
-
-struct AlbumFileDateTime {
-    s16 year{};
-    s8 month{};
-    s8 day{};
-    s8 hour{};
-    s8 minute{};
-    s8 second{};
-    s8 uid{};
-};
-static_assert(sizeof(AlbumFileDateTime) == 0x8, "AlbumFileDateTime has incorrect size.");
-
-struct AlbumEntry {
-    u64 size{};
-    u64 application_id{};
-    AlbumFileDateTime datetime{};
-    AlbumStorage storage{};
-    ContentType content{};
-    INSERT_PADDING_BYTES(6);
-};
-static_assert(sizeof(AlbumEntry) == 0x20, "AlbumEntry has incorrect size.");
-
-struct AlbumFileEntry {
-    u64 size{}; // Size of the entry
-    u64 hash{}; // AES256 with hardcoded key over AlbumEntry
-    AlbumFileDateTime datetime{};
-    AlbumStorage storage{};
-    ContentType content{};
-    INSERT_PADDING_BYTES(5);
-    u8 unknown{1}; // Set to 1 on official SW
-};
-static_assert(sizeof(AlbumFileEntry) == 0x20, "AlbumFileEntry has incorrect size.");
-
-struct ApplicationAlbumEntry {
-    u64 size{}; // Size of the entry
-    u64 hash{}; // AES256 with hardcoded key over AlbumEntry
-    AlbumFileDateTime datetime{};
-    AlbumStorage storage{};
-    ContentType content{};
-    INSERT_PADDING_BYTES(5);
-    u8 unknown{1}; // Set to 1 on official SW
-};
-static_assert(sizeof(ApplicationAlbumEntry) == 0x20, "ApplicationAlbumEntry has incorrect size.");
-
-struct ApplicationAlbumFileEntry {
-    ApplicationAlbumEntry entry{};
-    AlbumFileDateTime datetime{};
-    u64 unknown{};
-};
-static_assert(sizeof(ApplicationAlbumFileEntry) == 0x30,
-              "ApplicationAlbumFileEntry has incorrect size.");
-
 void LoopProcess(Core::System& system);
 
 } // namespace Service::Capture
diff --git a/src/core/hle/service/caps/caps_a.cpp b/src/core/hle/service/caps/caps_a.cpp
index 44267b284..9925720a3 100644
--- a/src/core/hle/service/caps/caps_a.cpp
+++ b/src/core/hle/service/caps/caps_a.cpp
@@ -1,40 +1,26 @@
 // SPDX-FileCopyrightText: Copyright 2020 yuzu Emulator Project
 // SPDX-License-Identifier: GPL-2.0-or-later
 
+#include "common/logging/log.h"
 #include "core/hle/service/caps/caps_a.h"
+#include "core/hle/service/caps/caps_manager.h"
+#include "core/hle/service/caps/caps_result.h"
+#include "core/hle/service/caps/caps_types.h"
+#include "core/hle/service/ipc_helpers.h"
 
 namespace Service::Capture {
 
-class IAlbumAccessorSession final : public ServiceFramework<IAlbumAccessorSession> {
-public:
-    explicit IAlbumAccessorSession(Core::System& system_)
-        : ServiceFramework{system_, "IAlbumAccessorSession"} {
-        // clang-format off
-        static const FunctionInfo functions[] = {
-            {2001, nullptr, "OpenAlbumMovieReadStream"},
-            {2002, nullptr, "CloseAlbumMovieReadStream"},
-            {2003, nullptr, "GetAlbumMovieReadStreamMovieDataSize"},
-            {2004, nullptr, "ReadMovieDataFromAlbumMovieReadStream"},
-            {2005, nullptr, "GetAlbumMovieReadStreamBrokenReason"},
-            {2006, nullptr, "GetAlbumMovieReadStreamImageDataSize"},
-            {2007, nullptr, "ReadImageDataFromAlbumMovieReadStream"},
-            {2008, nullptr, "ReadFileAttributeFromAlbumMovieReadStream"},
-        };
-        // clang-format on
-
-        RegisterHandlers(functions);
-    }
-};
-
-CAPS_A::CAPS_A(Core::System& system_) : ServiceFramework{system_, "caps:a"} {
+IAlbumAccessorService::IAlbumAccessorService(Core::System& system_,
+                                             std::shared_ptr<AlbumManager> album_manager)
+    : ServiceFramework{system_, "caps:a"}, manager{album_manager} {
     // clang-format off
     static const FunctionInfo functions[] = {
         {0, nullptr, "GetAlbumFileCount"},
         {1, nullptr, "GetAlbumFileList"},
         {2, nullptr, "LoadAlbumFile"},
-        {3, nullptr, "DeleteAlbumFile"},
+        {3, &IAlbumAccessorService::DeleteAlbumFile, "DeleteAlbumFile"},
         {4, nullptr, "StorageCopyAlbumFile"},
-        {5, nullptr, "IsAlbumMounted"},
+        {5, &IAlbumAccessorService::IsAlbumMounted, "IsAlbumMounted"},
         {6, nullptr, "GetAlbumUsage"},
         {7, nullptr, "GetAlbumFileSize"},
         {8, nullptr, "LoadAlbumFileThumbnail"},
@@ -47,18 +33,18 @@ CAPS_A::CAPS_A(Core::System& system_) : ServiceFramework{system_, "caps:a"} {
         {15, nullptr, "GetAlbumUsage3"},
         {16, nullptr, "GetAlbumMountResult"},
         {17, nullptr, "GetAlbumUsage16"},
-        {18, nullptr, "Unknown18"},
+        {18, &IAlbumAccessorService::Unknown18, "Unknown18"},
         {19, nullptr, "Unknown19"},
         {100, nullptr, "GetAlbumFileCountEx0"},
-        {101, nullptr, "GetAlbumFileListEx0"},
+        {101, &IAlbumAccessorService::GetAlbumFileListEx0, "GetAlbumFileListEx0"},
         {202, nullptr, "SaveEditedScreenShot"},
         {301, nullptr, "GetLastThumbnail"},
         {302, nullptr, "GetLastOverlayMovieThumbnail"},
-        {401, nullptr, "GetAutoSavingStorage"},
+        {401,  &IAlbumAccessorService::GetAutoSavingStorage, "GetAutoSavingStorage"},
         {501, nullptr, "GetRequiredStorageSpaceSizeToCopyAll"},
         {1001, nullptr, "LoadAlbumScreenShotThumbnailImageEx0"},
-        {1002, nullptr, "LoadAlbumScreenShotImageEx1"},
-        {1003, nullptr, "LoadAlbumScreenShotThumbnailImageEx1"},
+        {1002, &IAlbumAccessorService::LoadAlbumScreenShotImageEx1, "LoadAlbumScreenShotImageEx1"},
+        {1003, &IAlbumAccessorService::LoadAlbumScreenShotThumbnailImageEx1, "LoadAlbumScreenShotThumbnailImageEx1"},
         {8001, nullptr, "ForceAlbumUnmounted"},
         {8002, nullptr, "ResetAlbumMountStatus"},
         {8011, nullptr, "RefreshAlbumCache"},
@@ -74,6 +60,199 @@ CAPS_A::CAPS_A(Core::System& system_) : ServiceFramework{system_, "caps:a"} {
     RegisterHandlers(functions);
 }
 
-CAPS_A::~CAPS_A() = default;
+IAlbumAccessorService::~IAlbumAccessorService() = default;
+
+void IAlbumAccessorService::DeleteAlbumFile(HLERequestContext& ctx) {
+    IPC::RequestParser rp{ctx};
+    const auto file_id{rp.PopRaw<AlbumFileId>()};
+
+    LOG_INFO(Service_Capture, "called, application_id=0x{:0x}, storage={}, type={}",
+             file_id.application_id, file_id.storage, file_id.type);
+
+    Result result = manager->DeleteAlbumFile(file_id);
+    result = TranslateResult(result);
+
+    IPC::ResponseBuilder rb{ctx, 2};
+    rb.Push(result);
+}
+
+void IAlbumAccessorService::IsAlbumMounted(HLERequestContext& ctx) {
+    IPC::RequestParser rp{ctx};
+    const auto storage{rp.PopEnum<AlbumStorage>()};
+
+    LOG_INFO(Service_Capture, "called, storage={}", storage);
+
+    Result result = manager->IsAlbumMounted(storage);
+    const bool is_mounted = result.IsSuccess();
+    result = TranslateResult(result);
+
+    IPC::ResponseBuilder rb{ctx, 3};
+    rb.Push(result);
+    rb.Push<u8>(is_mounted);
+}
+
+void IAlbumAccessorService::Unknown18(HLERequestContext& ctx) {
+    struct UnknownBuffer {
+        INSERT_PADDING_BYTES(0x10);
+    };
+    static_assert(sizeof(UnknownBuffer) == 0x10, "UnknownBuffer is an invalid size");
+
+    LOG_WARNING(Service_Capture, "(STUBBED) called");
+
+    std::vector<UnknownBuffer> buffer{};
+
+    if (!buffer.empty()) {
+        ctx.WriteBuffer(buffer);
+    }
+
+    IPC::ResponseBuilder rb{ctx, 3};
+    rb.Push(ResultSuccess);
+    rb.Push(static_cast<u32>(buffer.size()));
+}
+
+void IAlbumAccessorService::GetAlbumFileListEx0(HLERequestContext& ctx) {
+    IPC::RequestParser rp{ctx};
+    const auto storage{rp.PopEnum<AlbumStorage>()};
+    const auto flags{rp.Pop<u8>()};
+    const auto album_entry_size{ctx.GetWriteBufferNumElements<AlbumEntry>()};
+
+    LOG_INFO(Service_Capture, "called, storage={}, flags={}", storage, flags);
+
+    std::vector<AlbumEntry> entries;
+    Result result = manager->GetAlbumFileList(entries, storage, flags);
+    result = TranslateResult(result);
+
+    entries.resize(std::min(album_entry_size, entries.size()));
+
+    if (!entries.empty()) {
+        ctx.WriteBuffer(entries);
+    }
+
+    IPC::ResponseBuilder rb{ctx, 4};
+    rb.Push(result);
+    rb.Push<u64>(entries.size());
+}
+
+void IAlbumAccessorService::GetAutoSavingStorage(HLERequestContext& ctx) {
+    LOG_WARNING(Service_Capture, "(STUBBED) called");
+
+    bool is_autosaving{};
+    Result result = manager->GetAutoSavingStorage(is_autosaving);
+    result = TranslateResult(result);
+
+    IPC::ResponseBuilder rb{ctx, 3};
+    rb.Push(result);
+    rb.Push<u8>(is_autosaving);
+}
+
+void IAlbumAccessorService::LoadAlbumScreenShotImageEx1(HLERequestContext& ctx) {
+    IPC::RequestParser rp{ctx};
+    const auto file_id{rp.PopRaw<AlbumFileId>()};
+    const auto decoder_options{rp.PopRaw<ScreenShotDecodeOption>()};
+    const auto image_buffer_size{ctx.GetWriteBufferSize(1)};
+
+    LOG_INFO(Service_Capture, "called, application_id=0x{:0x}, storage={}, type={}, flags={}",
+             file_id.application_id, file_id.storage, file_id.type, decoder_options.flags);
+
+    std::vector<u8> image;
+    LoadAlbumScreenShotImageOutput image_output;
+    Result result =
+        manager->LoadAlbumScreenShotImage(image_output, image, file_id, decoder_options);
+    result = TranslateResult(result);
+
+    if (image.size() > image_buffer_size) {
+        result = ResultWorkMemoryError;
+    }
+
+    if (result.IsSuccess()) {
+        ctx.WriteBuffer(image_output, 0);
+        ctx.WriteBuffer(image, 1);
+    }
+
+    IPC::ResponseBuilder rb{ctx, 2};
+    rb.Push(result);
+}
+
+void IAlbumAccessorService::LoadAlbumScreenShotThumbnailImageEx1(HLERequestContext& ctx) {
+    IPC::RequestParser rp{ctx};
+    const auto file_id{rp.PopRaw<AlbumFileId>()};
+    const auto decoder_options{rp.PopRaw<ScreenShotDecodeOption>()};
+
+    LOG_INFO(Service_Capture, "called, application_id=0x{:0x}, storage={}, type={}, flags={}",
+             file_id.application_id, file_id.storage, file_id.type, decoder_options.flags);
+
+    std::vector<u8> image(ctx.GetWriteBufferSize(1));
+    LoadAlbumScreenShotImageOutput image_output;
+    Result result =
+        manager->LoadAlbumScreenShotThumbnail(image_output, image, file_id, decoder_options);
+    result = TranslateResult(result);
+
+    if (result.IsSuccess()) {
+        ctx.WriteBuffer(image_output, 0);
+        ctx.WriteBuffer(image, 1);
+    }
+
+    IPC::ResponseBuilder rb{ctx, 2};
+    rb.Push(result);
+}
+
+Result IAlbumAccessorService::TranslateResult(Result in_result) {
+    if (in_result.IsSuccess()) {
+        return in_result;
+    }
+
+    if ((in_result.raw & 0x3801ff) == ResultUnknown1024.raw) {
+        if (in_result.description - 0x514 < 100) {
+            return ResultInvalidFileData;
+        }
+        if (in_result.description - 0x5dc < 100) {
+            return ResultInvalidFileData;
+        }
+
+        if (in_result.description - 0x578 < 100) {
+            if (in_result == ResultFileCountLimit) {
+                return ResultUnknown22;
+            }
+            return ResultUnknown25;
+        }
+
+        if (in_result.raw < ResultUnknown1801.raw) {
+            if (in_result == ResultUnknown1202) {
+                return ResultUnknown810;
+            }
+            if (in_result == ResultUnknown1203) {
+                return ResultUnknown810;
+            }
+            if (in_result == ResultUnknown1701) {
+                return ResultUnknown5;
+            }
+        } else if (in_result.raw < ResultUnknown1803.raw) {
+            if (in_result == ResultUnknown1801) {
+                return ResultUnknown5;
+            }
+            if (in_result == ResultUnknown1802) {
+                return ResultUnknown6;
+            }
+        } else {
+            if (in_result == ResultUnknown1803) {
+                return ResultUnknown7;
+            }
+            if (in_result == ResultUnknown1804) {
+                return ResultOutOfRange;
+            }
+        }
+        return ResultUnknown1024;
+    }
+
+    if (in_result.module == ErrorModule::FS) {
+        if ((in_result.description >> 0xc < 0x7d) || (in_result.description - 1000 < 2000) ||
+            (((in_result.description - 3000) >> 3) < 0x271)) {
+            // TODO: Translate FS error
+            return in_result;
+        }
+    }
+
+    return in_result;
+}
 
 } // namespace Service::Capture
diff --git a/src/core/hle/service/caps/caps_a.h b/src/core/hle/service/caps/caps_a.h
index 98a21a5ad..c90cff71e 100644
--- a/src/core/hle/service/caps/caps_a.h
+++ b/src/core/hle/service/caps/caps_a.h
@@ -10,11 +10,26 @@ class System;
 }
 
 namespace Service::Capture {
+class AlbumManager;
 
-class CAPS_A final : public ServiceFramework<CAPS_A> {
+class IAlbumAccessorService final : public ServiceFramework<IAlbumAccessorService> {
 public:
-    explicit CAPS_A(Core::System& system_);
-    ~CAPS_A() override;
+    explicit IAlbumAccessorService(Core::System& system_,
+                                   std::shared_ptr<AlbumManager> album_manager);
+    ~IAlbumAccessorService() override;
+
+private:
+    void DeleteAlbumFile(HLERequestContext& ctx);
+    void IsAlbumMounted(HLERequestContext& ctx);
+    void Unknown18(HLERequestContext& ctx);
+    void GetAlbumFileListEx0(HLERequestContext& ctx);
+    void GetAutoSavingStorage(HLERequestContext& ctx);
+    void LoadAlbumScreenShotImageEx1(HLERequestContext& ctx);
+    void LoadAlbumScreenShotThumbnailImageEx1(HLERequestContext& ctx);
+
+    Result TranslateResult(Result in_result);
+
+    std::shared_ptr<AlbumManager> manager = nullptr;
 };
 
 } // namespace Service::Capture
diff --git a/src/core/hle/service/caps/caps_c.cpp b/src/core/hle/service/caps/caps_c.cpp
index fc77e35cd..1e7fe6474 100644
--- a/src/core/hle/service/caps/caps_c.cpp
+++ b/src/core/hle/service/caps/caps_c.cpp
@@ -3,53 +3,21 @@
 
 #include "common/logging/log.h"
 #include "core/hle/service/caps/caps_c.h"
+#include "core/hle/service/caps/caps_manager.h"
+#include "core/hle/service/caps/caps_result.h"
+#include "core/hle/service/caps/caps_types.h"
 #include "core/hle/service/ipc_helpers.h"
 
 namespace Service::Capture {
 
-class IAlbumControlSession final : public ServiceFramework<IAlbumControlSession> {
-public:
-    explicit IAlbumControlSession(Core::System& system_)
-        : ServiceFramework{system_, "IAlbumControlSession"} {
-        // clang-format off
-        static const FunctionInfo functions[] = {
-            {2001, nullptr, "OpenAlbumMovieReadStream"},
-            {2002, nullptr, "CloseAlbumMovieReadStream"},
-            {2003, nullptr, "GetAlbumMovieReadStreamMovieDataSize"},
-            {2004, nullptr, "ReadMovieDataFromAlbumMovieReadStream"},
-            {2005, nullptr, "GetAlbumMovieReadStreamBrokenReason"},
-            {2006, nullptr, "GetAlbumMovieReadStreamImageDataSize"},
-            {2007, nullptr, "ReadImageDataFromAlbumMovieReadStream"},
-            {2008, nullptr, "ReadFileAttributeFromAlbumMovieReadStream"},
-            {2401, nullptr, "OpenAlbumMovieWriteStream"},
-            {2402, nullptr, "FinishAlbumMovieWriteStream"},
-            {2403, nullptr, "CommitAlbumMovieWriteStream"},
-            {2404, nullptr, "DiscardAlbumMovieWriteStream"},
-            {2405, nullptr, "DiscardAlbumMovieWriteStreamNoDelete"},
-            {2406, nullptr, "CommitAlbumMovieWriteStreamEx"},
-            {2411, nullptr, "StartAlbumMovieWriteStreamDataSection"},
-            {2412, nullptr, "EndAlbumMovieWriteStreamDataSection"},
-            {2413, nullptr, "StartAlbumMovieWriteStreamMetaSection"},
-            {2414, nullptr, "EndAlbumMovieWriteStreamMetaSection"},
-            {2421, nullptr, "ReadDataFromAlbumMovieWriteStream"},
-            {2422, nullptr, "WriteDataToAlbumMovieWriteStream"},
-            {2424, nullptr, "WriteMetaToAlbumMovieWriteStream"},
-            {2431, nullptr, "GetAlbumMovieWriteStreamBrokenReason"},
-            {2433, nullptr, "GetAlbumMovieWriteStreamDataSize"},
-            {2434, nullptr, "SetAlbumMovieWriteStreamDataSize"},
-        };
-        // clang-format on
-
-        RegisterHandlers(functions);
-    }
-};
-
-CAPS_C::CAPS_C(Core::System& system_) : ServiceFramework{system_, "caps:c"} {
+IAlbumControlService::IAlbumControlService(Core::System& system_,
+                                           std::shared_ptr<AlbumManager> album_manager)
+    : ServiceFramework{system_, "caps:c"}, manager{album_manager} {
     // clang-format off
     static const FunctionInfo functions[] = {
         {1, nullptr, "CaptureRawImage"},
         {2, nullptr, "CaptureRawImageWithTimeout"},
-        {33, &CAPS_C::SetShimLibraryVersion, "SetShimLibraryVersion"},
+        {33, &IAlbumControlService::SetShimLibraryVersion, "SetShimLibraryVersion"},
         {1001, nullptr, "RequestTakingScreenShot"},
         {1002, nullptr, "RequestTakingScreenShotWithTimeout"},
         {1011, nullptr, "NotifyTakingScreenShotRefused"},
@@ -72,9 +40,9 @@ CAPS_C::CAPS_C(Core::System& system_) : ServiceFramework{system_, "caps:c"} {
     RegisterHandlers(functions);
 }
 
-CAPS_C::~CAPS_C() = default;
+IAlbumControlService::~IAlbumControlService() = default;
 
-void CAPS_C::SetShimLibraryVersion(HLERequestContext& ctx) {
+void IAlbumControlService::SetShimLibraryVersion(HLERequestContext& ctx) {
     IPC::RequestParser rp{ctx};
     const auto library_version{rp.Pop<u64>()};
     const auto applet_resource_user_id{rp.Pop<u64>()};
diff --git a/src/core/hle/service/caps/caps_c.h b/src/core/hle/service/caps/caps_c.h
index 537b3a2e3..92ba242db 100644
--- a/src/core/hle/service/caps/caps_c.h
+++ b/src/core/hle/service/caps/caps_c.h
@@ -10,14 +10,18 @@ class System;
 }
 
 namespace Service::Capture {
+class AlbumManager;
 
-class CAPS_C final : public ServiceFramework<CAPS_C> {
+class IAlbumControlService final : public ServiceFramework<IAlbumControlService> {
 public:
-    explicit CAPS_C(Core::System& system_);
-    ~CAPS_C() override;
+    explicit IAlbumControlService(Core::System& system_,
+                                  std::shared_ptr<AlbumManager> album_manager);
+    ~IAlbumControlService() override;
 
 private:
     void SetShimLibraryVersion(HLERequestContext& ctx);
+
+    std::shared_ptr<AlbumManager> manager = nullptr;
 };
 
 } // namespace Service::Capture
diff --git a/src/core/hle/service/caps/caps_manager.cpp b/src/core/hle/service/caps/caps_manager.cpp
new file mode 100644
index 000000000..7d733eb54
--- /dev/null
+++ b/src/core/hle/service/caps/caps_manager.cpp
@@ -0,0 +1,469 @@
+// SPDX-FileCopyrightText: Copyright 2023 yuzu Emulator Project
+// SPDX-License-Identifier: GPL-2.0-or-later
+
+#include <sstream>
+
+#include "common/fs/file.h"
+#include "common/fs/path_util.h"
+#include "common/logging/log.h"
+#include "common/stb.h"
+#include "core/core.h"
+#include "core/hle/service/caps/caps_manager.h"
+#include "core/hle/service/caps/caps_result.h"
+#include "core/hle/service/time/time_manager.h"
+#include "core/hle/service/time/time_zone_content_manager.h"
+
+namespace Service::Capture {
+
+AlbumManager::AlbumManager(Core::System& system_) : system{system_} {}
+
+AlbumManager::~AlbumManager() = default;
+
+Result AlbumManager::DeleteAlbumFile(const AlbumFileId& file_id) {
+    if (file_id.storage > AlbumStorage::Sd) {
+        return ResultInvalidStorage;
+    }
+
+    if (!is_mounted) {
+        return ResultIsNotMounted;
+    }
+
+    std::filesystem::path path;
+    const auto result = GetFile(path, file_id);
+
+    if (result.IsError()) {
+        return result;
+    }
+
+    if (!Common::FS::RemoveFile(path)) {
+        return ResultFileNotFound;
+    }
+
+    return ResultSuccess;
+}
+
+Result AlbumManager::IsAlbumMounted(AlbumStorage storage) {
+    if (storage > AlbumStorage::Sd) {
+        return ResultInvalidStorage;
+    }
+
+    is_mounted = true;
+
+    if (storage == AlbumStorage::Sd) {
+        FindScreenshots();
+    }
+
+    return is_mounted ? ResultSuccess : ResultIsNotMounted;
+}
+
+Result AlbumManager::GetAlbumFileList(std::vector<AlbumEntry>& out_entries, AlbumStorage storage,
+                                      u8 flags) const {
+    if (storage > AlbumStorage::Sd) {
+        return ResultInvalidStorage;
+    }
+
+    if (!is_mounted) {
+        return ResultIsNotMounted;
+    }
+
+    for (auto& [file_id, path] : album_files) {
+        if (file_id.storage != storage) {
+            continue;
+        }
+        if (out_entries.size() >= SdAlbumFileLimit) {
+            break;
+        }
+
+        const auto entry_size = Common::FS::GetSize(path);
+        out_entries.push_back({
+            .entry_size = entry_size,
+            .file_id = file_id,
+        });
+    }
+
+    return ResultSuccess;
+}
+
+Result AlbumManager::GetAlbumFileList(std::vector<ApplicationAlbumFileEntry>& out_entries,
+                                      ContentType contex_type, s64 start_posix_time,
+                                      s64 end_posix_time, u64 aruid) const {
+    if (!is_mounted) {
+        return ResultIsNotMounted;
+    }
+
+    std::vector<ApplicationAlbumEntry> album_entries;
+    const auto start_date = ConvertToAlbumDateTime(start_posix_time);
+    const auto end_date = ConvertToAlbumDateTime(end_posix_time);
+    const auto result = GetAlbumFileList(album_entries, contex_type, start_date, end_date, aruid);
+
+    if (result.IsError()) {
+        return result;
+    }
+
+    for (const auto& album_entry : album_entries) {
+        ApplicationAlbumFileEntry entry{
+            .entry = album_entry,
+            .datetime = album_entry.datetime,
+            .unknown = {},
+        };
+        out_entries.push_back(entry);
+    }
+
+    return ResultSuccess;
+}
+
+Result AlbumManager::GetAlbumFileList(std::vector<ApplicationAlbumEntry>& out_entries,
+                                      ContentType contex_type, AlbumFileDateTime start_date,
+                                      AlbumFileDateTime end_date, u64 aruid) const {
+    if (!is_mounted) {
+        return ResultIsNotMounted;
+    }
+
+    for (auto& [file_id, path] : album_files) {
+        if (file_id.type != contex_type) {
+            continue;
+        }
+        if (file_id.date > start_date) {
+            continue;
+        }
+        if (file_id.date < end_date) {
+            continue;
+        }
+        if (out_entries.size() >= SdAlbumFileLimit) {
+            break;
+        }
+
+        const auto entry_size = Common::FS::GetSize(path);
+        ApplicationAlbumEntry entry{
+            .size = entry_size,
+            .hash{},
+            .datetime = file_id.date,
+            .storage = file_id.storage,
+            .content = contex_type,
+            .unknown = 1,
+        };
+        out_entries.push_back(entry);
+    }
+
+    return ResultSuccess;
+}
+
+Result AlbumManager::GetAutoSavingStorage(bool& out_is_autosaving) const {
+    out_is_autosaving = false;
+    return ResultSuccess;
+}
+
+Result AlbumManager::LoadAlbumScreenShotImage(LoadAlbumScreenShotImageOutput& out_image_output,
+                                              std::vector<u8>& out_image,
+                                              const AlbumFileId& file_id,
+                                              const ScreenShotDecodeOption& decoder_options) const {
+    if (file_id.storage > AlbumStorage::Sd) {
+        return ResultInvalidStorage;
+    }
+
+    if (!is_mounted) {
+        return ResultIsNotMounted;
+    }
+
+    out_image_output = {
+        .width = 1280,
+        .height = 720,
+        .attribute =
+            {
+                .unknown_0{},
+                .orientation = AlbumImageOrientation::None,
+                .unknown_1{},
+                .unknown_2{},
+            },
+    };
+
+    std::filesystem::path path;
+    const auto result = GetFile(path, file_id);
+
+    if (result.IsError()) {
+        return result;
+    }
+
+    out_image.resize(out_image_output.height * out_image_output.width * STBI_rgb_alpha);
+
+    return LoadImage(out_image, path, static_cast<int>(out_image_output.width),
+                     +static_cast<int>(out_image_output.height), decoder_options.flags);
+}
+
+Result AlbumManager::LoadAlbumScreenShotThumbnail(
+    LoadAlbumScreenShotImageOutput& out_image_output, std::vector<u8>& out_image,
+    const AlbumFileId& file_id, const ScreenShotDecodeOption& decoder_options) const {
+    if (file_id.storage > AlbumStorage::Sd) {
+        return ResultInvalidStorage;
+    }
+
+    if (!is_mounted) {
+        return ResultIsNotMounted;
+    }
+
+    out_image_output = {
+        .width = 320,
+        .height = 180,
+        .attribute =
+            {
+                .unknown_0{},
+                .orientation = AlbumImageOrientation::None,
+                .unknown_1{},
+                .unknown_2{},
+            },
+    };
+
+    std::filesystem::path path;
+    const auto result = GetFile(path, file_id);
+
+    if (result.IsError()) {
+        return result;
+    }
+
+    out_image.resize(out_image_output.height * out_image_output.width * STBI_rgb_alpha);
+
+    return LoadImage(out_image, path, static_cast<int>(out_image_output.width),
+                     +static_cast<int>(out_image_output.height), decoder_options.flags);
+}
+
+Result AlbumManager::SaveScreenShot(ApplicationAlbumEntry& out_entry,
+                                    const ScreenShotAttribute& attribute,
+                                    std::span<const u8> image_data, u64 aruid) {
+    return SaveScreenShot(out_entry, attribute, {}, image_data, aruid);
+}
+
+Result AlbumManager::SaveScreenShot(ApplicationAlbumEntry& out_entry,
+                                    const ScreenShotAttribute& attribute,
+                                    const ApplicationData& app_data, std::span<const u8> image_data,
+                                    u64 aruid) {
+    const u64 title_id = system.GetApplicationProcessProgramID();
+    const auto& user_clock = system.GetTimeManager().GetStandardUserSystemClockCore();
+
+    s64 posix_time{};
+    Result result = user_clock.GetCurrentTime(system, posix_time);
+
+    if (result.IsError()) {
+        return result;
+    }
+
+    const auto date = ConvertToAlbumDateTime(posix_time);
+
+    return SaveImage(out_entry, image_data, title_id, date);
+}
+
+Result AlbumManager::SaveEditedScreenShot(ApplicationAlbumEntry& out_entry,
+                                          const ScreenShotAttribute& attribute,
+                                          const AlbumFileId& file_id,
+                                          std::span<const u8> image_data) {
+    const auto& user_clock = system.GetTimeManager().GetStandardUserSystemClockCore();
+
+    s64 posix_time{};
+    Result result = user_clock.GetCurrentTime(system, posix_time);
+
+    if (result.IsError()) {
+        return result;
+    }
+
+    const auto date = ConvertToAlbumDateTime(posix_time);
+
+    return SaveImage(out_entry, image_data, file_id.application_id, date);
+}
+
+Result AlbumManager::GetFile(std::filesystem::path& out_path, const AlbumFileId& file_id) const {
+    const auto file = album_files.find(file_id);
+
+    if (file == album_files.end()) {
+        return ResultFileNotFound;
+    }
+
+    out_path = file->second;
+    return ResultSuccess;
+}
+
+void AlbumManager::FindScreenshots() {
+    is_mounted = false;
+    album_files.clear();
+
+    // TODO: Swap this with a blocking operation.
+    const auto screenshots_dir = Common::FS::GetYuzuPath(Common::FS::YuzuPath::ScreenshotsDir);
+    Common::FS::IterateDirEntries(
+        screenshots_dir,
+        [this](const std::filesystem::path& full_path) {
+            AlbumEntry entry;
+            if (GetAlbumEntry(entry, full_path).IsError()) {
+                return true;
+            }
+            while (album_files.contains(entry.file_id)) {
+                if (++entry.file_id.date.unique_id == 0) {
+                    break;
+                }
+            }
+            album_files[entry.file_id] = full_path;
+            return true;
+        },
+        Common::FS::DirEntryFilter::File);
+
+    is_mounted = true;
+}
+
+Result AlbumManager::GetAlbumEntry(AlbumEntry& out_entry, const std::filesystem::path& path) const {
+    std::istringstream line_stream(path.filename().string());
+    std::string date;
+    std::string application;
+    std::string time;
+
+    // Parse filename to obtain entry properties
+    std::getline(line_stream, application, '_');
+    std::getline(line_stream, date, '_');
+    std::getline(line_stream, time, '_');
+
+    std::istringstream date_stream(date);
+    std::istringstream time_stream(time);
+    std::string year;
+    std::string month;
+    std::string day;
+    std::string hour;
+    std::string minute;
+    std::string second;
+
+    std::getline(date_stream, year, '-');
+    std::getline(date_stream, month, '-');
+    std::getline(date_stream, day, '-');
+
+    std::getline(time_stream, hour, '-');
+    std::getline(time_stream, minute, '-');
+    std::getline(time_stream, second, '-');
+
+    try {
+        out_entry = {
+            .entry_size = 1,
+            .file_id{
+                .application_id = static_cast<u64>(std::stoll(application, 0, 16)),
+                .date =
+                    {
+                        .year = static_cast<s16>(std::stoi(year)),
+                        .month = static_cast<s8>(std::stoi(month)),
+                        .day = static_cast<s8>(std::stoi(day)),
+                        .hour = static_cast<s8>(std::stoi(hour)),
+                        .minute = static_cast<s8>(std::stoi(minute)),
+                        .second = static_cast<s8>(std::stoi(second)),
+                        .unique_id = 0,
+                    },
+                .storage = AlbumStorage::Sd,
+                .type = ContentType::Screenshot,
+                .unknown = 1,
+            },
+        };
+    } catch (const std::invalid_argument&) {
+        return ResultUnknown;
+    } catch (const std::out_of_range&) {
+        return ResultUnknown;
+    } catch (const std::exception&) {
+        return ResultUnknown;
+    }
+
+    return ResultSuccess;
+}
+
+Result AlbumManager::LoadImage(std::span<u8> out_image, const std::filesystem::path& path,
+                               int width, int height, ScreenShotDecoderFlag flag) const {
+    if (out_image.size() != static_cast<std::size_t>(width * height * STBI_rgb_alpha)) {
+        return ResultUnknown;
+    }
+
+    const Common::FS::IOFile db_file{path, Common::FS::FileAccessMode::Read,
+                                     Common::FS::FileType::BinaryFile};
+
+    std::vector<u8> raw_file(db_file.GetSize());
+    if (db_file.Read(raw_file) != raw_file.size()) {
+        return ResultUnknown;
+    }
+
+    int filter_flag = STBIR_FILTER_DEFAULT;
+    int original_width, original_height, color_channels;
+    const auto dbi_image =
+        stbi_load_from_memory(raw_file.data(), static_cast<int>(raw_file.size()), &original_width,
+                              &original_height, &color_channels, STBI_rgb_alpha);
+
+    if (dbi_image == nullptr) {
+        return ResultUnknown;
+    }
+
+    switch (flag) {
+    case ScreenShotDecoderFlag::EnableFancyUpsampling:
+        filter_flag = STBIR_FILTER_TRIANGLE;
+        break;
+    case ScreenShotDecoderFlag::EnableBlockSmoothing:
+        filter_flag = STBIR_FILTER_BOX;
+        break;
+    default:
+        filter_flag = STBIR_FILTER_DEFAULT;
+        break;
+    }
+
+    stbir_resize_uint8_srgb(dbi_image, original_width, original_height, 0, out_image.data(), width,
+                            height, 0, STBI_rgb_alpha, 3, filter_flag);
+
+    return ResultSuccess;
+}
+
+static void PNGToMemory(void* context, void* png, int len) {
+    std::vector<u8>* png_image = static_cast<std::vector<u8>*>(context);
+    png_image->reserve(len);
+    std::memcpy(png_image->data(), png, len);
+}
+
+Result AlbumManager::SaveImage(ApplicationAlbumEntry& out_entry, std::span<const u8> image,
+                               u64 title_id, const AlbumFileDateTime& date) const {
+    const auto screenshot_path =
+        Common::FS::GetYuzuPathString(Common::FS::YuzuPath::ScreenshotsDir);
+    const std::string formatted_date =
+        fmt::format("{:04}-{:02}-{:02}_{:02}-{:02}-{:02}-{:03}", date.year, date.month, date.day,
+                    date.hour, date.minute, date.second, 0);
+    const std::string file_path =
+        fmt::format("{}/{:016x}_{}.png", screenshot_path, title_id, formatted_date);
+
+    const Common::FS::IOFile db_file{file_path, Common::FS::FileAccessMode::Write,
+                                     Common::FS::FileType::BinaryFile};
+
+    std::vector<u8> png_image;
+    if (!stbi_write_png_to_func(PNGToMemory, &png_image, 1280, 720, STBI_rgb_alpha, image.data(),
+                                0)) {
+        return ResultFileCountLimit;
+    }
+
+    if (db_file.Write(png_image) != png_image.size()) {
+        return ResultFileCountLimit;
+    }
+
+    out_entry = {
+        .size = png_image.size(),
+        .hash = {},
+        .datetime = date,
+        .storage = AlbumStorage::Sd,
+        .content = ContentType::Screenshot,
+        .unknown = 1,
+    };
+
+    return ResultSuccess;
+}
+
+AlbumFileDateTime AlbumManager::ConvertToAlbumDateTime(u64 posix_time) const {
+    Time::TimeZone::CalendarInfo calendar_date{};
+    const auto& time_zone_manager =
+        system.GetTimeManager().GetTimeZoneContentManager().GetTimeZoneManager();
+
+    time_zone_manager.ToCalendarTimeWithMyRules(posix_time, calendar_date);
+
+    return {
+        .year = calendar_date.time.year,
+        .month = calendar_date.time.month,
+        .day = calendar_date.time.day,
+        .hour = calendar_date.time.hour,
+        .minute = calendar_date.time.minute,
+        .second = calendar_date.time.second,
+        .unique_id = 0,
+    };
+}
+
+} // namespace Service::Capture
diff --git a/src/core/hle/service/caps/caps_manager.h b/src/core/hle/service/caps/caps_manager.h
new file mode 100644
index 000000000..44d85117f
--- /dev/null
+++ b/src/core/hle/service/caps/caps_manager.h
@@ -0,0 +1,90 @@
+// SPDX-FileCopyrightText: Copyright 2023 yuzu Emulator Project
+// SPDX-License-Identifier: GPL-2.0-or-later
+
+#pragma once
+
+#include <unordered_map>
+
+#include "common/fs/fs.h"
+#include "core/hle/result.h"
+#include "core/hle/service/caps/caps_types.h"
+
+namespace Core {
+class System;
+}
+
+namespace std {
+// Hash used to create lists from AlbumFileId data
+template <>
+struct hash<Service::Capture::AlbumFileId> {
+    size_t operator()(const Service::Capture::AlbumFileId& pad_id) const noexcept {
+        u64 hash_value = (static_cast<u64>(pad_id.date.year) << 8);
+        hash_value ^= (static_cast<u64>(pad_id.date.month) << 7);
+        hash_value ^= (static_cast<u64>(pad_id.date.day) << 6);
+        hash_value ^= (static_cast<u64>(pad_id.date.hour) << 5);
+        hash_value ^= (static_cast<u64>(pad_id.date.minute) << 4);
+        hash_value ^= (static_cast<u64>(pad_id.date.second) << 3);
+        hash_value ^= (static_cast<u64>(pad_id.date.unique_id) << 2);
+        hash_value ^= (static_cast<u64>(pad_id.storage) << 1);
+        hash_value ^= static_cast<u64>(pad_id.type);
+        return static_cast<size_t>(hash_value);
+    }
+};
+
+} // namespace std
+
+namespace Service::Capture {
+
+class AlbumManager {
+public:
+    explicit AlbumManager(Core::System& system_);
+    ~AlbumManager();
+
+    Result DeleteAlbumFile(const AlbumFileId& file_id);
+    Result IsAlbumMounted(AlbumStorage storage);
+    Result GetAlbumFileList(std::vector<AlbumEntry>& out_entries, AlbumStorage storage,
+                            u8 flags) const;
+    Result GetAlbumFileList(std::vector<ApplicationAlbumFileEntry>& out_entries,
+                            ContentType contex_type, s64 start_posix_time, s64 end_posix_time,
+                            u64 aruid) const;
+    Result GetAlbumFileList(std::vector<ApplicationAlbumEntry>& out_entries,
+                            ContentType contex_type, AlbumFileDateTime start_date,
+                            AlbumFileDateTime end_date, u64 aruid) const;
+    Result GetAutoSavingStorage(bool& out_is_autosaving) const;
+    Result LoadAlbumScreenShotImage(LoadAlbumScreenShotImageOutput& out_image_output,
+                                    std::vector<u8>& out_image, const AlbumFileId& file_id,
+                                    const ScreenShotDecodeOption& decoder_options) const;
+    Result LoadAlbumScreenShotThumbnail(LoadAlbumScreenShotImageOutput& out_image_output,
+                                        std::vector<u8>& out_image, const AlbumFileId& file_id,
+                                        const ScreenShotDecodeOption& decoder_options) const;
+
+    Result SaveScreenShot(ApplicationAlbumEntry& out_entry, const ScreenShotAttribute& attribute,
+                          std::span<const u8> image_data, u64 aruid);
+    Result SaveScreenShot(ApplicationAlbumEntry& out_entry, const ScreenShotAttribute& attribute,
+                          const ApplicationData& app_data, std::span<const u8> image_data,
+                          u64 aruid);
+    Result SaveEditedScreenShot(ApplicationAlbumEntry& out_entry,
+                                const ScreenShotAttribute& attribute, const AlbumFileId& file_id,
+                                std::span<const u8> image_data);
+
+private:
+    static constexpr std::size_t NandAlbumFileLimit = 1000;
+    static constexpr std::size_t SdAlbumFileLimit = 10000;
+
+    void FindScreenshots();
+    Result GetFile(std::filesystem::path& out_path, const AlbumFileId& file_id) const;
+    Result GetAlbumEntry(AlbumEntry& out_entry, const std::filesystem::path& path) const;
+    Result LoadImage(std::span<u8> out_image, const std::filesystem::path& path, int width,
+                     int height, ScreenShotDecoderFlag flag) const;
+    Result SaveImage(ApplicationAlbumEntry& out_entry, std::span<const u8> image, u64 title_id,
+                     const AlbumFileDateTime& date) const;
+
+    AlbumFileDateTime ConvertToAlbumDateTime(u64 posix_time) const;
+
+    bool is_mounted{};
+    std::unordered_map<AlbumFileId, std::filesystem::path> album_files;
+
+    Core::System& system;
+};
+
+} // namespace Service::Capture
diff --git a/src/core/hle/service/caps/caps_result.h b/src/core/hle/service/caps/caps_result.h
new file mode 100644
index 000000000..c65e5fb9a
--- /dev/null
+++ b/src/core/hle/service/caps/caps_result.h
@@ -0,0 +1,35 @@
+// SPDX-FileCopyrightText: Copyright 2023 yuzu Emulator Project
+// SPDX-License-Identifier: GPL-2.0-or-later
+
+#pragma once
+
+#include "core/hle/result.h"
+
+namespace Service::Capture {
+
+constexpr Result ResultWorkMemoryError(ErrorModule::Capture, 3);
+constexpr Result ResultUnknown5(ErrorModule::Capture, 5);
+constexpr Result ResultUnknown6(ErrorModule::Capture, 6);
+constexpr Result ResultUnknown7(ErrorModule::Capture, 7);
+constexpr Result ResultOutOfRange(ErrorModule::Capture, 8);
+constexpr Result ResulInvalidTimestamp(ErrorModule::Capture, 12);
+constexpr Result ResultInvalidStorage(ErrorModule::Capture, 13);
+constexpr Result ResultInvalidFileContents(ErrorModule::Capture, 14);
+constexpr Result ResultIsNotMounted(ErrorModule::Capture, 21);
+constexpr Result ResultUnknown22(ErrorModule::Capture, 22);
+constexpr Result ResultFileNotFound(ErrorModule::Capture, 23);
+constexpr Result ResultInvalidFileData(ErrorModule::Capture, 24);
+constexpr Result ResultUnknown25(ErrorModule::Capture, 25);
+constexpr Result ResultReadBufferShortage(ErrorModule::Capture, 30);
+constexpr Result ResultUnknown810(ErrorModule::Capture, 810);
+constexpr Result ResultUnknown1024(ErrorModule::Capture, 1024);
+constexpr Result ResultUnknown1202(ErrorModule::Capture, 1202);
+constexpr Result ResultUnknown1203(ErrorModule::Capture, 1203);
+constexpr Result ResultFileCountLimit(ErrorModule::Capture, 1401);
+constexpr Result ResultUnknown1701(ErrorModule::Capture, 1701);
+constexpr Result ResultUnknown1801(ErrorModule::Capture, 1801);
+constexpr Result ResultUnknown1802(ErrorModule::Capture, 1802);
+constexpr Result ResultUnknown1803(ErrorModule::Capture, 1803);
+constexpr Result ResultUnknown1804(ErrorModule::Capture, 1804);
+
+} // namespace Service::Capture
diff --git a/src/core/hle/service/caps/caps_sc.cpp b/src/core/hle/service/caps/caps_sc.cpp
index 395b13da7..6117cb7c6 100644
--- a/src/core/hle/service/caps/caps_sc.cpp
+++ b/src/core/hle/service/caps/caps_sc.cpp
@@ -5,7 +5,8 @@
 
 namespace Service::Capture {
 
-CAPS_SC::CAPS_SC(Core::System& system_) : ServiceFramework{system_, "caps:sc"} {
+IScreenShotControlService::IScreenShotControlService(Core::System& system_)
+    : ServiceFramework{system_, "caps:sc"} {
     // clang-format off
     static const FunctionInfo functions[] = {
         {1, nullptr, "CaptureRawImage"},
@@ -34,6 +35,6 @@ CAPS_SC::CAPS_SC(Core::System& system_) : ServiceFramework{system_, "caps:sc"} {
     RegisterHandlers(functions);
 }
 
-CAPS_SC::~CAPS_SC() = default;
+IScreenShotControlService::~IScreenShotControlService() = default;
 
 } // namespace Service::Capture
diff --git a/src/core/hle/service/caps/caps_sc.h b/src/core/hle/service/caps/caps_sc.h
index e5600f6d7..d555f4979 100644
--- a/src/core/hle/service/caps/caps_sc.h
+++ b/src/core/hle/service/caps/caps_sc.h
@@ -11,10 +11,10 @@ class System;
 
 namespace Service::Capture {
 
-class CAPS_SC final : public ServiceFramework<CAPS_SC> {
+class IScreenShotControlService final : public ServiceFramework<IScreenShotControlService> {
 public:
-    explicit CAPS_SC(Core::System& system_);
-    ~CAPS_SC() override;
+    explicit IScreenShotControlService(Core::System& system_);
+    ~IScreenShotControlService() override;
 };
 
 } // namespace Service::Capture
diff --git a/src/core/hle/service/caps/caps_ss.cpp b/src/core/hle/service/caps/caps_ss.cpp
index 62b9edd41..1ba2b7972 100644
--- a/src/core/hle/service/caps/caps_ss.cpp
+++ b/src/core/hle/service/caps/caps_ss.cpp
@@ -1,18 +1,25 @@
 // SPDX-FileCopyrightText: Copyright 2020 yuzu Emulator Project
 // SPDX-License-Identifier: GPL-2.0-or-later
 
+#include "common/logging/log.h"
+#include "core/hle/service/caps/caps_manager.h"
+#include "core/hle/service/caps/caps_types.h"
+#include "core/hle/service/ipc_helpers.h"
+
 #include "core/hle/service/caps/caps_ss.h"
 
 namespace Service::Capture {
 
-CAPS_SS::CAPS_SS(Core::System& system_) : ServiceFramework{system_, "caps:ss"} {
+IScreenShotService::IScreenShotService(Core::System& system_,
+                                       std::shared_ptr<AlbumManager> album_manager)
+    : ServiceFramework{system_, "caps:ss"}, manager{album_manager} {
     // clang-format off
     static const FunctionInfo functions[] = {
         {201, nullptr, "SaveScreenShot"},
         {202, nullptr, "SaveEditedScreenShot"},
-        {203, nullptr, "SaveScreenShotEx0"},
+        {203, &IScreenShotService::SaveScreenShotEx0, "SaveScreenShotEx0"},
         {204, nullptr, "SaveEditedScreenShotEx0"},
-        {206, nullptr, "Unknown206"},
+        {206, &IScreenShotService::SaveEditedScreenShotEx1, "SaveEditedScreenShotEx1"},
         {208, nullptr, "SaveScreenShotOfMovieEx1"},
         {1000, nullptr, "Unknown1000"},
     };
@@ -21,6 +28,67 @@ CAPS_SS::CAPS_SS(Core::System& system_) : ServiceFramework{system_, "caps:ss"} {
     RegisterHandlers(functions);
 }
 
-CAPS_SS::~CAPS_SS() = default;
+IScreenShotService::~IScreenShotService() = default;
+
+void IScreenShotService::SaveScreenShotEx0(HLERequestContext& ctx) {
+    IPC::RequestParser rp{ctx};
+    struct Parameters {
+        ScreenShotAttribute attribute{};
+        u32 report_option{};
+        INSERT_PADDING_BYTES(0x4);
+        u64 applet_resource_user_id{};
+    };
+    static_assert(sizeof(Parameters) == 0x50, "Parameters has incorrect size.");
+
+    const auto parameters{rp.PopRaw<Parameters>()};
+    const auto image_data_buffer = ctx.ReadBuffer();
+
+    LOG_INFO(Service_Capture,
+             "called, report_option={}, image_data_buffer_size={}, applet_resource_user_id={}",
+             parameters.report_option, image_data_buffer.size(),
+             parameters.applet_resource_user_id);
+
+    ApplicationAlbumEntry entry{};
+    const auto result = manager->SaveScreenShot(entry, parameters.attribute, image_data_buffer,
+                                                parameters.applet_resource_user_id);
+
+    IPC::ResponseBuilder rb{ctx, 10};
+    rb.Push(result);
+    rb.PushRaw(entry);
+}
+void IScreenShotService::SaveEditedScreenShotEx1(HLERequestContext& ctx) {
+    IPC::RequestParser rp{ctx};
+    struct Parameters {
+        ScreenShotAttribute attribute;
+        u64 width;
+        u64 height;
+        u64 thumbnail_width;
+        u64 thumbnail_height;
+        AlbumFileId file_id;
+    };
+    static_assert(sizeof(Parameters) == 0x78, "Parameters has incorrect size.");
+
+    const auto parameters{rp.PopRaw<Parameters>()};
+    const auto application_data_buffer = ctx.ReadBuffer(0);
+    const auto image_data_buffer = ctx.ReadBuffer(1);
+    const auto thumbnail_image_data_buffer = ctx.ReadBuffer(2);
+
+    LOG_INFO(Service_Capture,
+             "called, width={}, height={}, thumbnail_width={}, thumbnail_height={}, "
+             "application_id={:016x},  storage={},  type={}, app_data_buffer_size={}, "
+             "image_data_buffer_size={}, thumbnail_image_buffer_size={}",
+             parameters.width, parameters.height, parameters.thumbnail_width,
+             parameters.thumbnail_height, parameters.file_id.application_id,
+             parameters.file_id.storage, parameters.file_id.type, application_data_buffer.size(),
+             image_data_buffer.size(), thumbnail_image_data_buffer.size());
+
+    ApplicationAlbumEntry entry{};
+    const auto result = manager->SaveEditedScreenShot(entry, parameters.attribute,
+                                                      parameters.file_id, image_data_buffer);
+
+    IPC::ResponseBuilder rb{ctx, 10};
+    rb.Push(result);
+    rb.PushRaw(entry);
+}
 
 } // namespace Service::Capture
diff --git a/src/core/hle/service/caps/caps_ss.h b/src/core/hle/service/caps/caps_ss.h
index 718ade485..a7e9972ab 100644
--- a/src/core/hle/service/caps/caps_ss.h
+++ b/src/core/hle/service/caps/caps_ss.h
@@ -11,10 +11,16 @@ class System;
 
 namespace Service::Capture {
 
-class CAPS_SS final : public ServiceFramework<CAPS_SS> {
+class IScreenShotService final : public ServiceFramework<IScreenShotService> {
 public:
-    explicit CAPS_SS(Core::System& system_);
-    ~CAPS_SS() override;
+    explicit IScreenShotService(Core::System& system_, std::shared_ptr<AlbumManager> album_manager);
+    ~IScreenShotService() override;
+
+private:
+    void SaveScreenShotEx0(HLERequestContext& ctx);
+    void SaveEditedScreenShotEx1(HLERequestContext& ctx);
+
+    std::shared_ptr<AlbumManager> manager;
 };
 
 } // namespace Service::Capture
diff --git a/src/core/hle/service/caps/caps_su.cpp b/src/core/hle/service/caps/caps_su.cpp
index 3b11cc95c..e85625ee4 100644
--- a/src/core/hle/service/caps/caps_su.cpp
+++ b/src/core/hle/service/caps/caps_su.cpp
@@ -2,18 +2,22 @@
 // SPDX-License-Identifier: GPL-2.0-or-later
 
 #include "common/logging/log.h"
+#include "core/hle/service/caps/caps_manager.h"
 #include "core/hle/service/caps/caps_su.h"
+#include "core/hle/service/caps/caps_types.h"
 #include "core/hle/service/ipc_helpers.h"
 
 namespace Service::Capture {
 
-CAPS_SU::CAPS_SU(Core::System& system_) : ServiceFramework{system_, "caps:su"} {
+IScreenShotApplicationService::IScreenShotApplicationService(
+    Core::System& system_, std::shared_ptr<AlbumManager> album_manager)
+    : ServiceFramework{system_, "caps:su"}, manager{album_manager} {
     // clang-format off
     static const FunctionInfo functions[] = {
-        {32, &CAPS_SU::SetShimLibraryVersion, "SetShimLibraryVersion"},
+        {32, &IScreenShotApplicationService::SetShimLibraryVersion, "SetShimLibraryVersion"},
         {201, nullptr, "SaveScreenShot"},
-        {203, nullptr, "SaveScreenShotEx0"},
-        {205, nullptr, "SaveScreenShotEx1"},
+        {203, &IScreenShotApplicationService::SaveScreenShotEx0, "SaveScreenShotEx0"},
+        {205, &IScreenShotApplicationService::SaveScreenShotEx1, "SaveScreenShotEx1"},
         {210, nullptr, "SaveScreenShotEx2"},
     };
     // clang-format on
@@ -21,9 +25,9 @@ CAPS_SU::CAPS_SU(Core::System& system_) : ServiceFramework{system_, "caps:su"} {
     RegisterHandlers(functions);
 }
 
-CAPS_SU::~CAPS_SU() = default;
+IScreenShotApplicationService::~IScreenShotApplicationService() = default;
 
-void CAPS_SU::SetShimLibraryVersion(HLERequestContext& ctx) {
+void IScreenShotApplicationService::SetShimLibraryVersion(HLERequestContext& ctx) {
     IPC::RequestParser rp{ctx};
     const auto library_version{rp.Pop<u64>()};
     const auto applet_resource_user_id{rp.Pop<u64>()};
@@ -35,4 +39,62 @@ void CAPS_SU::SetShimLibraryVersion(HLERequestContext& ctx) {
     rb.Push(ResultSuccess);
 }
 
+void IScreenShotApplicationService::SaveScreenShotEx0(HLERequestContext& ctx) {
+    IPC::RequestParser rp{ctx};
+    struct Parameters {
+        ScreenShotAttribute attribute{};
+        AlbumReportOption report_option{};
+        INSERT_PADDING_BYTES(0x4);
+        u64 applet_resource_user_id{};
+    };
+    static_assert(sizeof(Parameters) == 0x50, "Parameters has incorrect size.");
+
+    const auto parameters{rp.PopRaw<Parameters>()};
+    const auto image_data_buffer = ctx.ReadBuffer();
+
+    LOG_INFO(Service_Capture,
+             "called, report_option={}, image_data_buffer_size={}, applet_resource_user_id={}",
+             parameters.report_option, image_data_buffer.size(),
+             parameters.applet_resource_user_id);
+
+    ApplicationAlbumEntry entry{};
+    const auto result = manager->SaveScreenShot(entry, parameters.attribute, image_data_buffer,
+                                                parameters.applet_resource_user_id);
+
+    IPC::ResponseBuilder rb{ctx, 10};
+    rb.Push(result);
+    rb.PushRaw(entry);
+}
+
+void IScreenShotApplicationService::SaveScreenShotEx1(HLERequestContext& ctx) {
+    IPC::RequestParser rp{ctx};
+    struct Parameters {
+        ScreenShotAttribute attribute{};
+        AlbumReportOption report_option{};
+        INSERT_PADDING_BYTES(0x4);
+        u64 applet_resource_user_id{};
+    };
+    static_assert(sizeof(Parameters) == 0x50, "Parameters has incorrect size.");
+
+    const auto parameters{rp.PopRaw<Parameters>()};
+    const auto app_data_buffer = ctx.ReadBuffer(0);
+    const auto image_data_buffer = ctx.ReadBuffer(1);
+
+    LOG_INFO(Service_Capture,
+             "called, report_option={}, image_data_buffer_size={}, applet_resource_user_id={}",
+             parameters.report_option, image_data_buffer.size(),
+             parameters.applet_resource_user_id);
+
+    ApplicationAlbumEntry entry{};
+    ApplicationData app_data{};
+    std::memcpy(&app_data, app_data_buffer.data(), sizeof(ApplicationData));
+    const auto result =
+        manager->SaveScreenShot(entry, parameters.attribute, app_data, image_data_buffer,
+                                parameters.applet_resource_user_id);
+
+    IPC::ResponseBuilder rb{ctx, 10};
+    rb.Push(result);
+    rb.PushRaw(entry);
+}
+
 } // namespace Service::Capture
diff --git a/src/core/hle/service/caps/caps_su.h b/src/core/hle/service/caps/caps_su.h
index c6398858d..89e71f506 100644
--- a/src/core/hle/service/caps/caps_su.h
+++ b/src/core/hle/service/caps/caps_su.h
@@ -10,14 +10,20 @@ class System;
 }
 
 namespace Service::Capture {
+class AlbumManager;
 
-class CAPS_SU final : public ServiceFramework<CAPS_SU> {
+class IScreenShotApplicationService final : public ServiceFramework<IScreenShotApplicationService> {
 public:
-    explicit CAPS_SU(Core::System& system_);
-    ~CAPS_SU() override;
+    explicit IScreenShotApplicationService(Core::System& system_,
+                                           std::shared_ptr<AlbumManager> album_manager);
+    ~IScreenShotApplicationService() override;
 
 private:
     void SetShimLibraryVersion(HLERequestContext& ctx);
+    void SaveScreenShotEx0(HLERequestContext& ctx);
+    void SaveScreenShotEx1(HLERequestContext& ctx);
+
+    std::shared_ptr<AlbumManager> manager;
 };
 
 } // namespace Service::Capture
diff --git a/src/core/hle/service/caps/caps_types.h b/src/core/hle/service/caps/caps_types.h
new file mode 100644
index 000000000..589ac28d3
--- /dev/null
+++ b/src/core/hle/service/caps/caps_types.h
@@ -0,0 +1,186 @@
+// SPDX-FileCopyrightText: Copyright 2018 yuzu Emulator Project
+// SPDX-License-Identifier: GPL-2.0-or-later
+
+#pragma once
+
+#include "common/common_funcs.h"
+#include "common/common_types.h"
+
+namespace Service::Capture {
+
+// This is nn::album::ImageOrientation
+enum class AlbumImageOrientation {
+    None,
+    Rotate90,
+    Rotate180,
+    Rotate270,
+};
+
+// This is nn::album::AlbumReportOption
+enum class AlbumReportOption : s32 {
+    Disable,
+    Enable,
+    Unknown2,
+    Unknown3,
+};
+
+enum class ContentType : u8 {
+    Screenshot = 0,
+    Movie = 1,
+    ExtraMovie = 3,
+};
+
+enum class AlbumStorage : u8 {
+    Nand,
+    Sd,
+};
+
+enum class ScreenShotDecoderFlag : u64 {
+    None = 0,
+    EnableFancyUpsampling = 1 << 0,
+    EnableBlockSmoothing = 1 << 1,
+};
+
+// This is nn::capsrv::AlbumFileDateTime
+struct AlbumFileDateTime {
+    s16 year{};
+    s8 month{};
+    s8 day{};
+    s8 hour{};
+    s8 minute{};
+    s8 second{};
+    s8 unique_id{};
+
+    friend constexpr bool operator==(const AlbumFileDateTime&, const AlbumFileDateTime&) = default;
+    friend constexpr bool operator>(const AlbumFileDateTime& a, const AlbumFileDateTime& b) {
+        if (a.year > b.year) {
+            return true;
+        }
+        if (a.month > b.month) {
+            return true;
+        }
+        if (a.day > b.day) {
+            return true;
+        }
+        if (a.hour > b.hour) {
+            return true;
+        }
+        if (a.minute > b.minute) {
+            return true;
+        }
+        return a.second > b.second;
+    };
+    friend constexpr bool operator<(const AlbumFileDateTime& a, const AlbumFileDateTime& b) {
+        if (a.year < b.year) {
+            return true;
+        }
+        if (a.month < b.month) {
+            return true;
+        }
+        if (a.day < b.day) {
+            return true;
+        }
+        if (a.hour < b.hour) {
+            return true;
+        }
+        if (a.minute < b.minute) {
+            return true;
+        }
+        return a.second < b.second;
+    };
+};
+static_assert(sizeof(AlbumFileDateTime) == 0x8, "AlbumFileDateTime has incorrect size.");
+
+// This is nn::album::AlbumEntry
+struct AlbumFileEntry {
+    u64 size{}; // Size of the entry
+    u64 hash{}; // AES256 with hardcoded key over AlbumEntry
+    AlbumFileDateTime datetime{};
+    AlbumStorage storage{};
+    ContentType content{};
+    INSERT_PADDING_BYTES(5);
+    u8 unknown{}; // Set to 1 on official SW
+};
+static_assert(sizeof(AlbumFileEntry) == 0x20, "AlbumFileEntry has incorrect size.");
+
+struct AlbumFileId {
+    u64 application_id{};
+    AlbumFileDateTime date{};
+    AlbumStorage storage{};
+    ContentType type{};
+    INSERT_PADDING_BYTES(0x5);
+    u8 unknown{};
+
+    friend constexpr bool operator==(const AlbumFileId&, const AlbumFileId&) = default;
+};
+static_assert(sizeof(AlbumFileId) == 0x18, "AlbumFileId is an invalid size");
+
+// This is nn::capsrv::AlbumEntry
+struct AlbumEntry {
+    u64 entry_size{};
+    AlbumFileId file_id{};
+};
+static_assert(sizeof(AlbumEntry) == 0x20, "AlbumEntry has incorrect size.");
+
+// This is nn::capsrv::ApplicationAlbumEntry
+struct ApplicationAlbumEntry {
+    u64 size{}; // Size of the entry
+    u64 hash{}; // AES256 with hardcoded key over AlbumEntry
+    AlbumFileDateTime datetime{};
+    AlbumStorage storage{};
+    ContentType content{};
+    INSERT_PADDING_BYTES(5);
+    u8 unknown{1}; // Set to 1 on official SW
+};
+static_assert(sizeof(ApplicationAlbumEntry) == 0x20, "ApplicationAlbumEntry has incorrect size.");
+
+// This is nn::capsrv::ApplicationAlbumFileEntry
+struct ApplicationAlbumFileEntry {
+    ApplicationAlbumEntry entry{};
+    AlbumFileDateTime datetime{};
+    u64 unknown{};
+};
+static_assert(sizeof(ApplicationAlbumFileEntry) == 0x30,
+              "ApplicationAlbumFileEntry has incorrect size.");
+
+struct ApplicationData {
+    std::array<u8, 0x400> data{};
+    u32 data_size{};
+};
+static_assert(sizeof(ApplicationData) == 0x404, "ApplicationData is an invalid size");
+
+struct ScreenShotAttribute {
+    u32 unknown_0{};
+    AlbumImageOrientation orientation{};
+    u32 unknown_1{};
+    u32 unknown_2{};
+    INSERT_PADDING_BYTES(0x30);
+};
+static_assert(sizeof(ScreenShotAttribute) == 0x40, "ScreenShotAttribute is an invalid size");
+
+struct ScreenShotDecodeOption {
+    ScreenShotDecoderFlag flags{};
+    INSERT_PADDING_BYTES(0x18);
+};
+static_assert(sizeof(ScreenShotDecodeOption) == 0x20, "ScreenShotDecodeOption is an invalid size");
+
+struct LoadAlbumScreenShotImageOutput {
+    s64 width{};
+    s64 height{};
+    ScreenShotAttribute attribute{};
+    INSERT_PADDING_BYTES(0x400);
+};
+static_assert(sizeof(LoadAlbumScreenShotImageOutput) == 0x450,
+              "LoadAlbumScreenShotImageOutput is an invalid size");
+
+struct LoadAlbumScreenShotImageOutputForApplication {
+    s64 width{};
+    s64 height{};
+    ScreenShotAttribute attribute{};
+    ApplicationData data{};
+    INSERT_PADDING_BYTES(0xAC);
+};
+static_assert(sizeof(LoadAlbumScreenShotImageOutputForApplication) == 0x500,
+              "LoadAlbumScreenShotImageOutput is an invalid size");
+
+} // namespace Service::Capture
diff --git a/src/core/hle/service/caps/caps_u.cpp b/src/core/hle/service/caps/caps_u.cpp
index bffe0f8d0..b6b33fb2f 100644
--- a/src/core/hle/service/caps/caps_u.cpp
+++ b/src/core/hle/service/caps/caps_u.cpp
@@ -2,45 +2,29 @@
 // SPDX-License-Identifier: GPL-2.0-or-later
 
 #include "common/logging/log.h"
-#include "core/hle/service/caps/caps.h"
+#include "core/hle/service/caps/caps_manager.h"
+#include "core/hle/service/caps/caps_types.h"
 #include "core/hle/service/caps/caps_u.h"
 #include "core/hle/service/ipc_helpers.h"
 
 namespace Service::Capture {
 
-class IAlbumAccessorApplicationSession final
-    : public ServiceFramework<IAlbumAccessorApplicationSession> {
-public:
-    explicit IAlbumAccessorApplicationSession(Core::System& system_)
-        : ServiceFramework{system_, "IAlbumAccessorApplicationSession"} {
-        // clang-format off
-        static const FunctionInfo functions[] = {
-            {2001, nullptr, "OpenAlbumMovieReadStream"},
-            {2002, nullptr, "CloseAlbumMovieReadStream"},
-            {2003, nullptr, "GetAlbumMovieReadStreamMovieDataSize"},
-            {2004, nullptr, "ReadMovieDataFromAlbumMovieReadStream"},
-            {2005, nullptr, "GetAlbumMovieReadStreamBrokenReason"},
-        };
-        // clang-format on
-
-        RegisterHandlers(functions);
-    }
-};
-
-CAPS_U::CAPS_U(Core::System& system_) : ServiceFramework{system_, "caps:u"} {
+IAlbumApplicationService::IAlbumApplicationService(Core::System& system_,
+                                                   std::shared_ptr<AlbumManager> album_manager)
+    : ServiceFramework{system_, "caps:u"}, manager{album_manager} {
     // clang-format off
     static const FunctionInfo functions[] = {
-        {32, &CAPS_U::SetShimLibraryVersion, "SetShimLibraryVersion"},
-        {102, &CAPS_U::GetAlbumContentsFileListForApplication, "GetAlbumContentsFileListForApplication"},
-        {103, nullptr, "DeleteAlbumContentsFileForApplication"},
-        {104, nullptr, "GetAlbumContentsFileSizeForApplication"},
+        {32, &IAlbumApplicationService::SetShimLibraryVersion, "SetShimLibraryVersion"},
+        {102, &IAlbumApplicationService::GetAlbumFileList0AafeAruidDeprecated, "GetAlbumFileList0AafeAruidDeprecated"},
+        {103, nullptr, "DeleteAlbumFileByAruid"},
+        {104, nullptr, "GetAlbumFileSizeByAruid"},
         {105, nullptr, "DeleteAlbumFileByAruidForDebug"},
-        {110, nullptr, "LoadAlbumContentsFileScreenShotImageForApplication"},
-        {120, nullptr, "LoadAlbumContentsFileThumbnailImageForApplication"},
-        {130, nullptr, "PrecheckToCreateContentsForApplication"},
+        {110, nullptr, "LoadAlbumScreenShotImageByAruid"},
+        {120, nullptr, "LoadAlbumScreenShotThumbnailImageByAruid"},
+        {130, nullptr, "PrecheckToCreateContentsByAruid"},
         {140, nullptr, "GetAlbumFileList1AafeAruidDeprecated"},
         {141, nullptr, "GetAlbumFileList2AafeUidAruidDeprecated"},
-        {142, &CAPS_U::GetAlbumFileList3AaeAruid, "GetAlbumFileList3AaeAruid"},
+        {142, &IAlbumApplicationService::GetAlbumFileList3AaeAruid, "GetAlbumFileList3AaeAruid"},
         {143, nullptr, "GetAlbumFileList4AaeUidAruid"},
         {144, nullptr, "GetAllAlbumFileList3AaeAruid"},
         {60002, nullptr, "OpenAccessorSessionForApplication"},
@@ -50,9 +34,9 @@ CAPS_U::CAPS_U(Core::System& system_) : ServiceFramework{system_, "caps:u"} {
     RegisterHandlers(functions);
 }
 
-CAPS_U::~CAPS_U() = default;
+IAlbumApplicationService::~IAlbumApplicationService() = default;
 
-void CAPS_U::SetShimLibraryVersion(HLERequestContext& ctx) {
+void IAlbumApplicationService::SetShimLibraryVersion(HLERequestContext& ctx) {
     IPC::RequestParser rp{ctx};
     const auto library_version{rp.Pop<u64>()};
     const auto applet_resource_user_id{rp.Pop<u64>()};
@@ -64,37 +48,89 @@ void CAPS_U::SetShimLibraryVersion(HLERequestContext& ctx) {
     rb.Push(ResultSuccess);
 }
 
-void CAPS_U::GetAlbumContentsFileListForApplication(HLERequestContext& ctx) {
-    // Takes a type-0x6 output buffer containing an array of ApplicationAlbumFileEntry, a PID, an
-    // u8 ContentType, two s64s, and an u64 AppletResourceUserId. Returns an output u64 for total
-    // output entries (which is copied to a s32 by official SW).
+void IAlbumApplicationService::GetAlbumFileList0AafeAruidDeprecated(HLERequestContext& ctx) {
     IPC::RequestParser rp{ctx};
-    const auto pid{rp.Pop<s32>()};
-    const auto content_type{rp.PopEnum<ContentType>()};
-    const auto start_posix_time{rp.Pop<s64>()};
-    const auto end_posix_time{rp.Pop<s64>()};
-    const auto applet_resource_user_id{rp.Pop<u64>()};
+    struct Parameters {
+        ContentType content_type;
+        INSERT_PADDING_BYTES(7);
+        s64 start_posix_time;
+        s64 end_posix_time;
+        u64 applet_resource_user_id;
+    };
+    static_assert(sizeof(Parameters) == 0x20, "Parameters has incorrect size.");
 
-    // TODO: Update this when we implement the album.
-    // Currently we do not have a method of accessing album entries, set this to 0 for now.
-    constexpr u32 total_entries_1{};
-    constexpr u32 total_entries_2{};
+    const auto parameters{rp.PopRaw<Parameters>()};
 
-    LOG_WARNING(
-        Service_Capture,
-        "(STUBBED) called. pid={}, content_type={}, start_posix_time={}, "
-        "end_posix_time={}, applet_resource_user_id={}, total_entries_1={}, total_entries_2={}",
-        pid, content_type, start_posix_time, end_posix_time, applet_resource_user_id,
-        total_entries_1, total_entries_2);
+    LOG_WARNING(Service_Capture,
+                "(STUBBED) called. content_type={}, start_posix_time={}, end_posix_time={}, "
+                "applet_resource_user_id={}",
+                parameters.content_type, parameters.start_posix_time, parameters.end_posix_time,
+                parameters.applet_resource_user_id);
+
+    Result result = ResultSuccess;
+
+    if (result.IsSuccess()) {
+        result = manager->IsAlbumMounted(AlbumStorage::Sd);
+    }
+
+    std::vector<ApplicationAlbumFileEntry> entries;
+    if (result.IsSuccess()) {
+        result = manager->GetAlbumFileList(entries, parameters.content_type,
+                                           parameters.start_posix_time, parameters.end_posix_time,
+                                           parameters.applet_resource_user_id);
+    }
+
+    if (!entries.empty()) {
+        ctx.WriteBuffer(entries);
+    }
 
     IPC::ResponseBuilder rb{ctx, 4};
-    rb.Push(ResultSuccess);
-    rb.Push(total_entries_1);
-    rb.Push(total_entries_2);
+    rb.Push(result);
+    rb.Push<u64>(entries.size());
 }
 
-void CAPS_U::GetAlbumFileList3AaeAruid(HLERequestContext& ctx) {
-    GetAlbumContentsFileListForApplication(ctx);
+void IAlbumApplicationService::GetAlbumFileList3AaeAruid(HLERequestContext& ctx) {
+    IPC::RequestParser rp{ctx};
+    struct Parameters {
+        ContentType content_type;
+        INSERT_PADDING_BYTES(1);
+        AlbumFileDateTime start_date_time;
+        AlbumFileDateTime end_date_time;
+        INSERT_PADDING_BYTES(6);
+        u64 applet_resource_user_id;
+    };
+    static_assert(sizeof(Parameters) == 0x20, "Parameters has incorrect size.");
+
+    const auto parameters{rp.PopRaw<Parameters>()};
+
+    LOG_WARNING(Service_Capture,
+                "(STUBBED) called. content_type={}, start_date={}/{}/{}, "
+                "end_date={}/{}/{}, applet_resource_user_id={}",
+                parameters.content_type, parameters.start_date_time.year,
+                parameters.start_date_time.month, parameters.start_date_time.day,
+                parameters.end_date_time.year, parameters.end_date_time.month,
+                parameters.end_date_time.day, parameters.applet_resource_user_id);
+
+    Result result = ResultSuccess;
+
+    if (result.IsSuccess()) {
+        result = manager->IsAlbumMounted(AlbumStorage::Sd);
+    }
+
+    std::vector<ApplicationAlbumEntry> entries;
+    if (result.IsSuccess()) {
+        result =
+            manager->GetAlbumFileList(entries, parameters.content_type, parameters.start_date_time,
+                                      parameters.end_date_time, parameters.applet_resource_user_id);
+    }
+
+    if (!entries.empty()) {
+        ctx.WriteBuffer(entries);
+    }
+
+    IPC::ResponseBuilder rb{ctx, 4};
+    rb.Push(result);
+    rb.Push<u64>(entries.size());
 }
 
 } // namespace Service::Capture
diff --git a/src/core/hle/service/caps/caps_u.h b/src/core/hle/service/caps/caps_u.h
index e8dd037d7..9458c128e 100644
--- a/src/core/hle/service/caps/caps_u.h
+++ b/src/core/hle/service/caps/caps_u.h
@@ -10,16 +10,20 @@ class System;
 }
 
 namespace Service::Capture {
+class AlbumManager;
 
-class CAPS_U final : public ServiceFramework<CAPS_U> {
+class IAlbumApplicationService final : public ServiceFramework<IAlbumApplicationService> {
 public:
-    explicit CAPS_U(Core::System& system_);
-    ~CAPS_U() override;
+    explicit IAlbumApplicationService(Core::System& system_,
+                                      std::shared_ptr<AlbumManager> album_manager);
+    ~IAlbumApplicationService() override;
 
 private:
     void SetShimLibraryVersion(HLERequestContext& ctx);
-    void GetAlbumContentsFileListForApplication(HLERequestContext& ctx);
+    void GetAlbumFileList0AafeAruidDeprecated(HLERequestContext& ctx);
     void GetAlbumFileList3AaeAruid(HLERequestContext& ctx);
+
+    std::shared_ptr<AlbumManager> manager = nullptr;
 };
 
 } // namespace Service::Capture
diff --git a/src/core/hle/service/hid/controllers/palma.cpp b/src/core/hle/service/hid/controllers/palma.cpp
index 14c67e454..73a2a2b91 100644
--- a/src/core/hle/service/hid/controllers/palma.cpp
+++ b/src/core/hle/service/hid/controllers/palma.cpp
@@ -19,7 +19,9 @@ Controller_Palma::Controller_Palma(Core::HID::HIDCore& hid_core_, u8* raw_shared
     operation_complete_event = service_context.CreateEvent("hid:PalmaOperationCompleteEvent");
 }
 
-Controller_Palma::~Controller_Palma() = default;
+Controller_Palma::~Controller_Palma() {
+    service_context.CloseEvent(operation_complete_event);
+};
 
 void Controller_Palma::OnInit() {}
 
diff --git a/src/core/hle/service/hid/hid.cpp b/src/core/hle/service/hid/hid.cpp
index 4d70006c1..929dd5f03 100644
--- a/src/core/hle/service/hid/hid.cpp
+++ b/src/core/hle/service/hid/hid.cpp
@@ -2757,6 +2757,10 @@ public:
         joy_detach_event = service_context.CreateEvent("HidSys::JoyDetachEvent");
     }
 
+    ~HidSys() {
+        service_context.CloseEvent(joy_detach_event);
+    };
+
 private:
     void ApplyNpadSystemCommonPolicy(HLERequestContext& ctx) {
         LOG_WARNING(Service_HID, "called");
diff --git a/src/core/hle/service/hid/hidbus/hidbus_base.cpp b/src/core/hle/service/hid/hidbus/hidbus_base.cpp
index ee522c36e..8c44f93e8 100644
--- a/src/core/hle/service/hid/hidbus/hidbus_base.cpp
+++ b/src/core/hle/service/hid/hidbus/hidbus_base.cpp
@@ -13,7 +13,10 @@ HidbusBase::HidbusBase(Core::System& system_, KernelHelpers::ServiceContext& ser
     : system(system_), service_context(service_context_) {
     send_command_async_event = service_context.CreateEvent("hidbus:SendCommandAsyncEvent");
 }
-HidbusBase::~HidbusBase() = default;
+
+HidbusBase::~HidbusBase() {
+    service_context.CloseEvent(send_command_async_event);
+};
 
 void HidbusBase::ActivateDevice() {
     if (is_activated) {
diff --git a/src/core/hle/service/hle_ipc.cpp b/src/core/hle/service/hle_ipc.cpp
index f6a1e54f2..ff374ae39 100644
--- a/src/core/hle/service/hle_ipc.cpp
+++ b/src/core/hle/service/hle_ipc.cpp
@@ -23,6 +23,19 @@
 #include "core/hle/service/ipc_helpers.h"
 #include "core/memory.h"
 
+namespace {
+static thread_local std::array read_buffer_data_a{
+    Common::ScratchBuffer<u8>(),
+    Common::ScratchBuffer<u8>(),
+    Common::ScratchBuffer<u8>(),
+};
+static thread_local std::array read_buffer_data_x{
+    Common::ScratchBuffer<u8>(),
+    Common::ScratchBuffer<u8>(),
+    Common::ScratchBuffer<u8>(),
+};
+} // Anonymous namespace
+
 namespace Service {
 
 SessionRequestHandler::SessionRequestHandler(Kernel::KernelCore& kernel_, const char* service_name_)
@@ -328,26 +341,61 @@ std::vector<u8> HLERequestContext::ReadBufferCopy(std::size_t buffer_index) cons
     }
 }
 
-std::span<const u8> HLERequestContext::ReadBuffer(std::size_t buffer_index) const {
+std::span<const u8> HLERequestContext::ReadBufferA(std::size_t buffer_index) const {
     static thread_local std::array read_buffer_a{
         Core::Memory::CpuGuestMemory<u8, Core::Memory::GuestMemoryFlags::SafeRead>(memory, 0, 0),
         Core::Memory::CpuGuestMemory<u8, Core::Memory::GuestMemoryFlags::SafeRead>(memory, 0, 0),
+        Core::Memory::CpuGuestMemory<u8, Core::Memory::GuestMemoryFlags::SafeRead>(memory, 0, 0),
     };
-    static thread_local std::array read_buffer_data_a{
-        Common::ScratchBuffer<u8>(),
-        Common::ScratchBuffer<u8>(),
-    };
+
+    ASSERT_OR_EXECUTE_MSG(
+        BufferDescriptorA().size() > buffer_index, { return {}; },
+        "BufferDescriptorA invalid buffer_index {}", buffer_index);
+    auto& read_buffer = read_buffer_a[buffer_index];
+    return read_buffer.Read(BufferDescriptorA()[buffer_index].Address(),
+                            BufferDescriptorA()[buffer_index].Size(),
+                            &read_buffer_data_a[buffer_index]);
+}
+
+std::span<const u8> HLERequestContext::ReadBufferX(std::size_t buffer_index) const {
     static thread_local std::array read_buffer_x{
         Core::Memory::CpuGuestMemory<u8, Core::Memory::GuestMemoryFlags::SafeRead>(memory, 0, 0),
         Core::Memory::CpuGuestMemory<u8, Core::Memory::GuestMemoryFlags::SafeRead>(memory, 0, 0),
+        Core::Memory::CpuGuestMemory<u8, Core::Memory::GuestMemoryFlags::SafeRead>(memory, 0, 0),
+    };
+
+    ASSERT_OR_EXECUTE_MSG(
+        BufferDescriptorX().size() > buffer_index, { return {}; },
+        "BufferDescriptorX invalid buffer_index {}", buffer_index);
+    auto& read_buffer = read_buffer_x[buffer_index];
+    return read_buffer.Read(BufferDescriptorX()[buffer_index].Address(),
+                            BufferDescriptorX()[buffer_index].Size(),
+                            &read_buffer_data_x[buffer_index]);
+}
+
+std::span<const u8> HLERequestContext::ReadBuffer(std::size_t buffer_index) const {
+    static thread_local std::array read_buffer_a{
+        Core::Memory::CpuGuestMemory<u8, Core::Memory::GuestMemoryFlags::SafeRead>(memory, 0, 0),
+        Core::Memory::CpuGuestMemory<u8, Core::Memory::GuestMemoryFlags::SafeRead>(memory, 0, 0),
+        Core::Memory::CpuGuestMemory<u8, Core::Memory::GuestMemoryFlags::SafeRead>(memory, 0, 0),
     };
-    static thread_local std::array read_buffer_data_x{
-        Common::ScratchBuffer<u8>(),
-        Common::ScratchBuffer<u8>(),
+    static thread_local std::array read_buffer_x{
+        Core::Memory::CpuGuestMemory<u8, Core::Memory::GuestMemoryFlags::SafeRead>(memory, 0, 0),
+        Core::Memory::CpuGuestMemory<u8, Core::Memory::GuestMemoryFlags::SafeRead>(memory, 0, 0),
+        Core::Memory::CpuGuestMemory<u8, Core::Memory::GuestMemoryFlags::SafeRead>(memory, 0, 0),
     };
 
     const bool is_buffer_a{BufferDescriptorA().size() > buffer_index &&
                            BufferDescriptorA()[buffer_index].Size()};
+    const bool is_buffer_x{BufferDescriptorX().size() > buffer_index &&
+                           BufferDescriptorX()[buffer_index].Size()};
+
+    if (is_buffer_a && is_buffer_x) {
+        LOG_WARNING(Input, "Both buffer descriptors are available a.size={}, x.size={}",
+                    BufferDescriptorA()[buffer_index].Size(),
+                    BufferDescriptorX()[buffer_index].Size());
+    }
+
     if (is_buffer_a) {
         ASSERT_OR_EXECUTE_MSG(
             BufferDescriptorA().size() > buffer_index, { return {}; },
diff --git a/src/core/hle/service/hle_ipc.h b/src/core/hle/service/hle_ipc.h
index 4bd24c899..ad5259a5c 100644
--- a/src/core/hle/service/hle_ipc.h
+++ b/src/core/hle/service/hle_ipc.h
@@ -253,6 +253,12 @@ public:
         return domain_message_header.has_value();
     }
 
+    /// Helper function to get a span of a buffer using the buffer descriptor A
+    [[nodiscard]] std::span<const u8> ReadBufferA(std::size_t buffer_index = 0) const;
+
+    /// Helper function to get a span of a buffer using the buffer descriptor X
+    [[nodiscard]] std::span<const u8> ReadBufferX(std::size_t buffer_index = 0) const;
+
     /// Helper function to get a span of a buffer using the appropriate buffer descriptor
     [[nodiscard]] std::span<const u8> ReadBuffer(std::size_t buffer_index = 0) const;
 
diff --git a/src/core/hle/service/jit/jit_context.cpp b/src/core/hle/service/jit/jit_context.cpp
index 4ed3f02e2..0090e8568 100644
--- a/src/core/hle/service/jit/jit_context.cpp
+++ b/src/core/hle/service/jit/jit_context.cpp
@@ -156,6 +156,8 @@ public:
 
     bool LoadNRO(std::span<const u8> data) {
         local_memory.clear();
+
+        relocbase = local_memory.size();
         local_memory.insert(local_memory.end(), data.begin(), data.end());
 
         if (FixupRelocations()) {
@@ -181,8 +183,8 @@ public:
         // https://refspecs.linuxbase.org/elf/gabi4+/ch5.dynamic.html
         // https://refspecs.linuxbase.org/elf/gabi4+/ch4.reloc.html
         VAddr dynamic_offset{mod_offset + callbacks->MemoryRead32(mod_offset + 4)};
-        VAddr rela_dyn = 0;
-        size_t num_rela = 0;
+        VAddr rela_dyn = 0, relr_dyn = 0;
+        size_t num_rela = 0, num_relr = 0;
         while (true) {
             const auto dyn{callbacks->ReadMemory<Elf64_Dyn>(dynamic_offset)};
             dynamic_offset += sizeof(Elf64_Dyn);
@@ -196,6 +198,12 @@ public:
             if (dyn.d_tag == ElfDtRelasz) {
                 num_rela = dyn.d_un.d_val / sizeof(Elf64_Rela);
             }
+            if (dyn.d_tag == ElfDtRelr) {
+                relr_dyn = dyn.d_un.d_ptr;
+            }
+            if (dyn.d_tag == ElfDtRelrsz) {
+                num_relr = dyn.d_un.d_val / sizeof(Elf64_Relr);
+            }
         }
 
         for (size_t i = 0; i < num_rela; i++) {
@@ -207,6 +215,29 @@ public:
             callbacks->MemoryWrite64(rela.r_offset, contents + rela.r_addend);
         }
 
+        VAddr relr_where = 0;
+        for (size_t i = 0; i < num_relr; i++) {
+            const auto relr{callbacks->ReadMemory<Elf64_Relr>(relr_dyn + i * sizeof(Elf64_Relr))};
+            const auto incr{[&](VAddr where) {
+                callbacks->MemoryWrite64(where, callbacks->MemoryRead64(where) + relocbase);
+            }};
+
+            if ((relr & 1) == 0) {
+                // where pointer
+                relr_where = relocbase + relr;
+                incr(relr_where);
+                relr_where += sizeof(Elf64_Addr);
+            } else {
+                // bitmap
+                for (int bit = 1; bit < 64; bit++) {
+                    if ((relr & (1ULL << bit)) != 0) {
+                        incr(relr_where + i * sizeof(Elf64_Addr));
+                    }
+                }
+                relr_where += 63 * sizeof(Elf64_Addr);
+            }
+        }
+
         return true;
     }
 
@@ -313,6 +344,7 @@ public:
     Core::Memory::Memory& memory;
     VAddr top_of_stack;
     VAddr heap_pointer;
+    VAddr relocbase;
 };
 
 void DynarmicCallbacks64::CallSVC(u32 swi) {
diff --git a/src/core/hle/service/kernel_helpers.cpp b/src/core/hle/service/kernel_helpers.cpp
index 6a313a03b..f51e63564 100644
--- a/src/core/hle/service/kernel_helpers.cpp
+++ b/src/core/hle/service/kernel_helpers.cpp
@@ -21,10 +21,8 @@ ServiceContext::ServiceContext(Core::System& system_, std::string name_)
 
     // Create the process.
     process = Kernel::KProcess::Create(kernel);
-    ASSERT(Kernel::KProcess::Initialize(process, system_, std::move(name_),
-                                        Kernel::KProcess::ProcessType::KernelInternal,
-                                        kernel.GetSystemResourceLimit())
-               .IsSuccess());
+    ASSERT(R_SUCCEEDED(process->Initialize(Kernel::Svc::CreateProcessParameter{},
+                                           kernel.GetSystemResourceLimit(), false)));
 
     // Register the process.
     Kernel::KProcess::Register(kernel, process);
diff --git a/src/core/hle/service/mii/types/core_data.cpp b/src/core/hle/service/mii/types/core_data.cpp
index 970c748ca..ba1da76ba 100644
--- a/src/core/hle/service/mii/types/core_data.cpp
+++ b/src/core/hle/service/mii/types/core_data.cpp
@@ -41,6 +41,7 @@ void CoreData::BuildRandom(Age age, Gender gender, Race race) {
         }
     }
 
+    SetDefault();
     SetGender(gender);
     SetFavoriteColor(MiiUtil::GetRandomValue(FavoriteColor::Max));
     SetRegionMove(0);
diff --git a/src/core/hle/service/nifm/nifm.cpp b/src/core/hle/service/nifm/nifm.cpp
index 21b06d10b..22dc55a6d 100644
--- a/src/core/hle/service/nifm/nifm.cpp
+++ b/src/core/hle/service/nifm/nifm.cpp
@@ -545,6 +545,16 @@ void IGeneralService::IsAnyInternetRequestAccepted(HLERequestContext& ctx) {
     }
 }
 
+void IGeneralService::IsAnyForegroundRequestAccepted(HLERequestContext& ctx) {
+    const bool is_accepted{};
+
+    LOG_WARNING(Service_NIFM, "(STUBBED) called, is_accepted={}", is_accepted);
+
+    IPC::ResponseBuilder rb{ctx, 3};
+    rb.Push(ResultSuccess);
+    rb.Push<u8>(is_accepted);
+}
+
 IGeneralService::IGeneralService(Core::System& system_)
     : ServiceFramework{system_, "IGeneralService"}, network{system_.GetRoomNetwork()} {
     // clang-format off
@@ -569,7 +579,7 @@ IGeneralService::IGeneralService(Core::System& system_)
         {19, nullptr, "SetEthernetCommunicationEnabled"},
         {20, &IGeneralService::IsEthernetCommunicationEnabled, "IsEthernetCommunicationEnabled"},
         {21, &IGeneralService::IsAnyInternetRequestAccepted, "IsAnyInternetRequestAccepted"},
-        {22, nullptr, "IsAnyForegroundRequestAccepted"},
+        {22, &IGeneralService::IsAnyForegroundRequestAccepted, "IsAnyForegroundRequestAccepted"},
         {23, nullptr, "PutToSleep"},
         {24, nullptr, "WakeUp"},
         {25, nullptr, "GetSsidListVersion"},
diff --git a/src/core/hle/service/nifm/nifm.h b/src/core/hle/service/nifm/nifm.h
index ae99c4695..b74b66438 100644
--- a/src/core/hle/service/nifm/nifm.h
+++ b/src/core/hle/service/nifm/nifm.h
@@ -35,6 +35,7 @@ private:
     void GetInternetConnectionStatus(HLERequestContext& ctx);
     void IsEthernetCommunicationEnabled(HLERequestContext& ctx);
     void IsAnyInternetRequestAccepted(HLERequestContext& ctx);
+    void IsAnyForegroundRequestAccepted(HLERequestContext& ctx);
 
     Network::RoomNetwork& network;
 };
diff --git a/src/core/hle/service/ns/ns.cpp b/src/core/hle/service/ns/ns.cpp
index 6e0baf0be..f9e0e272d 100644
--- a/src/core/hle/service/ns/ns.cpp
+++ b/src/core/hle/service/ns/ns.cpp
@@ -7,6 +7,7 @@
 #include "core/file_sys/control_metadata.h"
 #include "core/file_sys/patch_manager.h"
 #include "core/file_sys/vfs.h"
+#include "core/hle/service/filesystem/filesystem.h"
 #include "core/hle/service/glue/glue_manager.h"
 #include "core/hle/service/ipc_helpers.h"
 #include "core/hle/service/ns/errors.h"
@@ -502,8 +503,8 @@ IContentManagementInterface::IContentManagementInterface(Core::System& system_)
     static const FunctionInfo functions[] = {
         {11, nullptr, "CalculateApplicationOccupiedSize"},
         {43, nullptr, "CheckSdCardMountStatus"},
-        {47, nullptr, "GetTotalSpaceSize"},
-        {48, nullptr, "GetFreeSpaceSize"},
+        {47, &IContentManagementInterface::GetTotalSpaceSize, "GetTotalSpaceSize"},
+        {48, &IContentManagementInterface::GetFreeSpaceSize, "GetFreeSpaceSize"},
         {600, nullptr, "CountApplicationContentMeta"},
         {601, nullptr, "ListApplicationContentMetaStatus"},
         {605, nullptr, "ListApplicationContentMetaStatusWithRightsCheck"},
@@ -516,6 +517,28 @@ IContentManagementInterface::IContentManagementInterface(Core::System& system_)
 
 IContentManagementInterface::~IContentManagementInterface() = default;
 
+void IContentManagementInterface::GetTotalSpaceSize(HLERequestContext& ctx) {
+    IPC::RequestParser rp{ctx};
+    const auto storage{rp.PopEnum<FileSys::StorageId>()};
+
+    LOG_INFO(Service_Capture, "called, storage={}", storage);
+
+    IPC::ResponseBuilder rb{ctx, 4};
+    rb.Push(ResultSuccess);
+    rb.Push<u64>(system.GetFileSystemController().GetTotalSpaceSize(storage));
+}
+
+void IContentManagementInterface::GetFreeSpaceSize(HLERequestContext& ctx) {
+    IPC::RequestParser rp{ctx};
+    const auto storage{rp.PopEnum<FileSys::StorageId>()};
+
+    LOG_INFO(Service_Capture, "called, storage={}", storage);
+
+    IPC::ResponseBuilder rb{ctx, 4};
+    rb.Push(ResultSuccess);
+    rb.Push<u64>(system.GetFileSystemController().GetFreeSpaceSize(storage));
+}
+
 IDocumentInterface::IDocumentInterface(Core::System& system_)
     : ServiceFramework{system_, "IDocumentInterface"} {
     // clang-format off
diff --git a/src/core/hle/service/ns/ns.h b/src/core/hle/service/ns/ns.h
index 175dad780..34d2a45dc 100644
--- a/src/core/hle/service/ns/ns.h
+++ b/src/core/hle/service/ns/ns.h
@@ -48,6 +48,10 @@ class IContentManagementInterface final : public ServiceFramework<IContentManage
 public:
     explicit IContentManagementInterface(Core::System& system_);
     ~IContentManagementInterface() override;
+
+private:
+    void GetTotalSpaceSize(HLERequestContext& ctx);
+    void GetFreeSpaceSize(HLERequestContext& ctx);
 };
 
 class IDocumentInterface final : public ServiceFramework<IDocumentInterface> {
diff --git a/src/core/hle/service/nvnflinger/buffer_queue_core.cpp b/src/core/hle/service/nvnflinger/buffer_queue_core.cpp
index 2dbe29616..ed66f6f5b 100644
--- a/src/core/hle/service/nvnflinger/buffer_queue_core.cpp
+++ b/src/core/hle/service/nvnflinger/buffer_queue_core.cpp
@@ -41,7 +41,7 @@ bool BufferQueueCore::WaitForDequeueCondition(std::unique_lock<std::mutex>& lk)
 s32 BufferQueueCore::GetMinUndequeuedBufferCountLocked(bool async) const {
     // If DequeueBuffer is allowed to error out, we don't have to add an extra buffer.
     if (!use_async_buffer) {
-        return max_acquired_buffer_count;
+        return 0;
     }
 
     if (dequeue_buffer_cannot_block || async) {
@@ -52,7 +52,7 @@ s32 BufferQueueCore::GetMinUndequeuedBufferCountLocked(bool async) const {
 }
 
 s32 BufferQueueCore::GetMinMaxBufferCountLocked(bool async) const {
-    return GetMinUndequeuedBufferCountLocked(async) + 1;
+    return GetMinUndequeuedBufferCountLocked(async);
 }
 
 s32 BufferQueueCore::GetMaxBufferCountLocked(bool async) const {
@@ -61,7 +61,7 @@ s32 BufferQueueCore::GetMaxBufferCountLocked(bool async) const {
 
     if (override_max_buffer_count != 0) {
         ASSERT(override_max_buffer_count >= min_buffer_count);
-        max_buffer_count = override_max_buffer_count;
+        return override_max_buffer_count;
     }
 
     // Any buffers that are dequeued by the producer or sitting in the queue waiting to be consumed
diff --git a/src/core/hle/service/nvnflinger/buffer_queue_producer.cpp b/src/core/hle/service/nvnflinger/buffer_queue_producer.cpp
index dc6917d5d..6e7a49658 100644
--- a/src/core/hle/service/nvnflinger/buffer_queue_producer.cpp
+++ b/src/core/hle/service/nvnflinger/buffer_queue_producer.cpp
@@ -134,7 +134,7 @@ Status BufferQueueProducer::WaitForFreeSlotThenRelock(bool async, s32* found, St
         const s32 max_buffer_count = core->GetMaxBufferCountLocked(async);
         if (async && core->override_max_buffer_count) {
             if (core->override_max_buffer_count < max_buffer_count) {
-                LOG_ERROR(Service_Nvnflinger, "async mode is invalid with buffer count override");
+                *found = BufferQueueCore::INVALID_BUFFER_SLOT;
                 return Status::BadValue;
             }
         }
@@ -142,7 +142,8 @@ Status BufferQueueProducer::WaitForFreeSlotThenRelock(bool async, s32* found, St
         // Free up any buffers that are in slots beyond the max buffer count
         for (s32 s = max_buffer_count; s < BufferQueueDefs::NUM_BUFFER_SLOTS; ++s) {
             ASSERT(slots[s].buffer_state == BufferState::Free);
-            if (slots[s].graphic_buffer != nullptr) {
+            if (slots[s].graphic_buffer != nullptr && slots[s].buffer_state == BufferState::Free &&
+                !slots[s].is_preallocated) {
                 core->FreeBufferLocked(s);
                 *return_flags |= Status::ReleaseAllBuffers;
             }
diff --git a/src/core/hle/service/nvnflinger/fb_share_buffer_manager.cpp b/src/core/hle/service/nvnflinger/fb_share_buffer_manager.cpp
index 469a53244..2e29bc848 100644
--- a/src/core/hle/service/nvnflinger/fb_share_buffer_manager.cpp
+++ b/src/core/hle/service/nvnflinger/fb_share_buffer_manager.cpp
@@ -46,7 +46,7 @@ Result AllocateIoForProcessAddressSpace(Common::ProcessAddress* out_map_address,
     // Get bounds of where mapping is possible.
     const VAddr alias_code_begin = GetInteger(page_table.GetAliasCodeRegionStart());
     const VAddr alias_code_size = page_table.GetAliasCodeRegionSize() / YUZU_PAGESIZE;
-    const auto state = Kernel::KMemoryState::Io;
+    const auto state = Kernel::KMemoryState::IoMemory;
     const auto perm = Kernel::KMemoryPermission::UserReadWrite;
     std::mt19937_64 rng{process->GetRandomEntropy(0)};
 
diff --git a/src/core/hle/service/nvnflinger/nvnflinger.cpp b/src/core/hle/service/nvnflinger/nvnflinger.cpp
index a07c621d9..bebb45eae 100644
--- a/src/core/hle/service/nvnflinger/nvnflinger.cpp
+++ b/src/core/hle/service/nvnflinger/nvnflinger.cpp
@@ -66,7 +66,6 @@ Nvnflinger::Nvnflinger(Core::System& system_, HosBinderDriverServer& hos_binder_
         "ScreenComposition",
         [this](std::uintptr_t, s64 time,
                std::chrono::nanoseconds ns_late) -> std::optional<std::chrono::nanoseconds> {
-            { const auto lock_guard = Lock(); }
             vsync_signal.Set();
             return std::chrono::nanoseconds(GetNextTicks());
         });
@@ -99,6 +98,7 @@ Nvnflinger::~Nvnflinger() {
     }
 
     ShutdownLayers();
+    vsync_thread = {};
 
     if (nvdrv) {
         nvdrv->Close(disp_fd);
@@ -106,6 +106,7 @@ Nvnflinger::~Nvnflinger() {
 }
 
 void Nvnflinger::ShutdownLayers() {
+    const auto lock_guard = Lock();
     for (auto& display : displays) {
         for (size_t layer = 0; layer < display.GetNumLayers(); ++layer) {
             display.GetLayer(layer).Core().NotifyShutdown();
@@ -229,16 +230,6 @@ VI::Layer* Nvnflinger::FindLayer(u64 display_id, u64 layer_id) {
     return display->FindLayer(layer_id);
 }
 
-const VI::Layer* Nvnflinger::FindLayer(u64 display_id, u64 layer_id) const {
-    const auto* const display = FindDisplay(display_id);
-
-    if (display == nullptr) {
-        return nullptr;
-    }
-
-    return display->FindLayer(layer_id);
-}
-
 VI::Layer* Nvnflinger::FindOrCreateLayer(u64 display_id, u64 layer_id) {
     auto* const display = FindDisplay(display_id);
 
@@ -288,7 +279,6 @@ void Nvnflinger::Compose() {
         auto nvdisp = nvdrv->GetDevice<Nvidia::Devices::nvdisp_disp0>(disp_fd);
         ASSERT(nvdisp);
 
-        guard->unlock();
         Common::Rectangle<int> crop_rect{
             static_cast<int>(buffer.crop.Left()), static_cast<int>(buffer.crop.Top()),
             static_cast<int>(buffer.crop.Right()), static_cast<int>(buffer.crop.Bottom())};
@@ -299,7 +289,6 @@ void Nvnflinger::Compose() {
                      buffer.fence.fences, buffer.fence.num_fences);
 
         MicroProfileFlip();
-        guard->lock();
 
         swap_interval = buffer.swap_interval;
 
diff --git a/src/core/hle/service/nvnflinger/nvnflinger.h b/src/core/hle/service/nvnflinger/nvnflinger.h
index 14c783582..959d8b46b 100644
--- a/src/core/hle/service/nvnflinger/nvnflinger.h
+++ b/src/core/hle/service/nvnflinger/nvnflinger.h
@@ -117,9 +117,6 @@ private:
     /// Finds the layer identified by the specified ID in the desired display.
     [[nodiscard]] VI::Layer* FindLayer(u64 display_id, u64 layer_id);
 
-    /// Finds the layer identified by the specified ID in the desired display.
-    [[nodiscard]] const VI::Layer* FindLayer(u64 display_id, u64 layer_id) const;
-
     /// Finds the layer identified by the specified ID in the desired display,
     /// or creates the layer if it is not found.
     /// To be used when the system expects the specified ID to already exist.
diff --git a/src/core/hle/service/pctl/pctl_module.cpp b/src/core/hle/service/pctl/pctl_module.cpp
index 5db1703d1..6a7fd72bc 100644
--- a/src/core/hle/service/pctl/pctl_module.cpp
+++ b/src/core/hle/service/pctl/pctl_module.cpp
@@ -33,7 +33,7 @@ public:
             {1001, &IParentalControlService::CheckFreeCommunicationPermission, "CheckFreeCommunicationPermission"},
             {1002, nullptr, "ConfirmLaunchApplicationPermission"},
             {1003, nullptr, "ConfirmResumeApplicationPermission"},
-            {1004, nullptr, "ConfirmSnsPostPermission"},
+            {1004, &IParentalControlService::ConfirmSnsPostPermission, "ConfirmSnsPostPermission"},
             {1005, nullptr, "ConfirmSystemSettingsPermission"},
             {1006, &IParentalControlService::IsRestrictionTemporaryUnlocked, "IsRestrictionTemporaryUnlocked"},
             {1007, nullptr, "RevertRestrictionTemporaryUnlocked"},
@@ -141,6 +141,12 @@ public:
             service_context.CreateEvent("IParentalControlService::RequestSuspensionEvent");
     }
 
+    ~IParentalControlService() {
+        service_context.CloseEvent(synchronization_event);
+        service_context.CloseEvent(unlinked_event);
+        service_context.CloseEvent(request_suspension_event);
+    };
+
 private:
     bool CheckFreeCommunicationPermissionImpl() const {
         if (states.temporary_unlocked) {
@@ -236,6 +242,13 @@ private:
         states.free_communication = true;
     }
 
+    void ConfirmSnsPostPermission(HLERequestContext& ctx) {
+        LOG_WARNING(Service_PCTL, "(STUBBED) called");
+
+        IPC::ResponseBuilder rb{ctx, 2};
+        rb.Push(Error::ResultNoFreeCommunication);
+    }
+
     void IsRestrictionTemporaryUnlocked(HLERequestContext& ctx) {
         const bool is_temporary_unlocked = false;
 
diff --git a/src/core/hle/service/pm/pm.cpp b/src/core/hle/service/pm/pm.cpp
index f9cf2dda3..d92499f05 100644
--- a/src/core/hle/service/pm/pm.cpp
+++ b/src/core/hle/service/pm/pm.cpp
@@ -37,7 +37,7 @@ std::optional<Kernel::KProcess*> SearchProcessList(
 void GetApplicationPidGeneric(HLERequestContext& ctx,
                               const std::vector<Kernel::KProcess*>& process_list) {
     const auto process = SearchProcessList(process_list, [](const auto& proc) {
-        return proc->GetProcessId() == Kernel::KProcess::ProcessIDMin;
+        return proc->GetProcessId() == Kernel::KProcess::ProcessIdMin;
     });
 
     IPC::ResponseBuilder rb{ctx, 4};
diff --git a/src/core/hle/service/prepo/prepo.cpp b/src/core/hle/service/prepo/prepo.cpp
index ec4a84989..14e8df63a 100644
--- a/src/core/hle/service/prepo/prepo.cpp
+++ b/src/core/hle/service/prepo/prepo.cpp
@@ -58,14 +58,8 @@ private:
         IPC::RequestParser rp{ctx};
         const auto process_id = rp.PopRaw<u64>();
 
-        const auto data1 = ctx.ReadBuffer(0);
-        const auto data2 = [&ctx] {
-            if (ctx.CanReadBuffer(1)) {
-                return ctx.ReadBuffer(1);
-            }
-
-            return std::span<const u8>{};
-        }();
+        const auto data1 = ctx.ReadBufferA(0);
+        const auto data2 = ctx.ReadBufferX(0);
 
         LOG_DEBUG(Service_PREPO,
                   "called, type={:02X}, process_id={:016X}, data1_size={:016X}, data2_size={:016X}",
@@ -85,14 +79,8 @@ private:
         const auto user_id = rp.PopRaw<u128>();
         const auto process_id = rp.PopRaw<u64>();
 
-        const auto data1 = ctx.ReadBuffer(0);
-        const auto data2 = [&ctx] {
-            if (ctx.CanReadBuffer(1)) {
-                return ctx.ReadBuffer(1);
-            }
-
-            return std::span<const u8>{};
-        }();
+        const auto data1 = ctx.ReadBufferA(0);
+        const auto data2 = ctx.ReadBufferX(0);
 
         LOG_DEBUG(Service_PREPO,
                   "called, type={:02X}, user_id={:016X}{:016X}, process_id={:016X}, "
@@ -137,14 +125,8 @@ private:
         IPC::RequestParser rp{ctx};
         const auto title_id = rp.PopRaw<u64>();
 
-        const auto data1 = ctx.ReadBuffer(0);
-        const auto data2 = [&ctx] {
-            if (ctx.CanReadBuffer(1)) {
-                return ctx.ReadBuffer(1);
-            }
-
-            return std::span<const u8>{};
-        }();
+        const auto data1 = ctx.ReadBufferA(0);
+        const auto data2 = ctx.ReadBufferX(0);
 
         LOG_DEBUG(Service_PREPO, "called, title_id={:016X}, data1_size={:016X}, data2_size={:016X}",
                   title_id, data1.size(), data2.size());
@@ -161,14 +143,8 @@ private:
         const auto user_id = rp.PopRaw<u128>();
         const auto title_id = rp.PopRaw<u64>();
 
-        const auto data1 = ctx.ReadBuffer(0);
-        const auto data2 = [&ctx] {
-            if (ctx.CanReadBuffer(1)) {
-                return ctx.ReadBuffer(1);
-            }
-
-            return std::span<const u8>{};
-        }();
+        const auto data1 = ctx.ReadBufferA(0);
+        const auto data2 = ctx.ReadBufferX(0);
 
         LOG_DEBUG(Service_PREPO,
                   "called, user_id={:016X}{:016X}, title_id={:016X}, data1_size={:016X}, "
diff --git a/src/core/hle/service/ptm/ts.cpp b/src/core/hle/service/ptm/ts.cpp
index ca064dd90..652f38b97 100644
--- a/src/core/hle/service/ptm/ts.cpp
+++ b/src/core/hle/service/ptm/ts.cpp
@@ -9,6 +9,35 @@
 
 namespace Service::PTM {
 
+enum class Location : u8 {
+    Internal,
+    External,
+};
+
+class ISession : public ServiceFramework<ISession> {
+public:
+    explicit ISession(Core::System& system_) : ServiceFramework{system_, "ISession"} {
+        // clang-format off
+        static const FunctionInfo functions[] = {
+            {0, nullptr, "GetTemperatureRange"},
+            {2, nullptr, "SetMeasurementMode"},
+            {4, &ISession::GetTemperature, "GetTemperature"},
+        };
+        // clang-format on
+
+        RegisterHandlers(functions);
+    }
+
+private:
+    void GetTemperature(HLERequestContext& ctx) {
+        constexpr f32 temperature = 35;
+
+        IPC::ResponseBuilder rb{ctx, 3};
+        rb.Push(ResultSuccess);
+        rb.Push(temperature);
+    }
+};
+
 TS::TS(Core::System& system_) : ServiceFramework{system_, "ts"} {
     // clang-format off
     static const FunctionInfo functions[] = {
@@ -16,7 +45,7 @@ TS::TS(Core::System& system_) : ServiceFramework{system_, "ts"} {
         {1, &TS::GetTemperature, "GetTemperature"},
         {2, nullptr, "SetMeasurementMode"},
         {3, &TS::GetTemperatureMilliC, "GetTemperatureMilliC"},
-        {4, nullptr, "OpenSession"},
+        {4, &TS::OpenSession, "OpenSession"},
     };
     // clang-format on
 
@@ -47,4 +76,13 @@ void TS::GetTemperatureMilliC(HLERequestContext& ctx) {
     rb.Push(temperature);
 }
 
+void TS::OpenSession(HLERequestContext& ctx) {
+    IPC::RequestParser rp{ctx};
+    [[maybe_unused]] const u32 device_code = rp.Pop<u32>();
+
+    IPC::ResponseBuilder rb{ctx, 2, 0, 1};
+    rb.Push(ResultSuccess);
+    rb.PushIpcInterface<ISession>(system);
+}
+
 } // namespace Service::PTM
diff --git a/src/core/hle/service/ptm/ts.h b/src/core/hle/service/ptm/ts.h
index c3f43d5a3..a10a91a64 100644
--- a/src/core/hle/service/ptm/ts.h
+++ b/src/core/hle/service/ptm/ts.h
@@ -14,13 +14,9 @@ public:
     ~TS() override;
 
 private:
-    enum class Location : u8 {
-        Internal,
-        External,
-    };
-
     void GetTemperature(HLERequestContext& ctx);
     void GetTemperatureMilliC(HLERequestContext& ctx);
+    void OpenSession(HLERequestContext& ctx);
 };
 
 } // namespace Service::PTM
diff --git a/src/core/hle/service/set/set_sys.cpp b/src/core/hle/service/set/set_sys.cpp
index 165b97dad..ec3af80af 100644
--- a/src/core/hle/service/set/set_sys.cpp
+++ b/src/core/hle/service/set/set_sys.cpp
@@ -5,8 +5,13 @@
 #include "common/logging/log.h"
 #include "common/settings.h"
 #include "common/string_util.h"
+#include "core/core.h"
+#include "core/file_sys/content_archive.h"
 #include "core/file_sys/errors.h"
-#include "core/file_sys/system_archive/system_version.h"
+#include "core/file_sys/nca_metadata.h"
+#include "core/file_sys/registered_cache.h"
+#include "core/file_sys/romfs.h"
+#include "core/file_sys/system_archive/system_archive.h"
 #include "core/hle/service/filesystem/filesystem.h"
 #include "core/hle/service/ipc_helpers.h"
 #include "core/hle/service/set/set.h"
@@ -22,18 +27,30 @@ enum class GetFirmwareVersionType {
     Version2,
 };
 
-void GetFirmwareVersionImpl(HLERequestContext& ctx, GetFirmwareVersionType type) {
-    LOG_WARNING(Service_SET, "called - Using hardcoded firmware version '{}'",
-                FileSys::SystemArchive::GetLongDisplayVersion());
-
+void GetFirmwareVersionImpl(Core::System& system, HLERequestContext& ctx,
+                            GetFirmwareVersionType type) {
     ASSERT_MSG(ctx.GetWriteBufferSize() == 0x100,
                "FirmwareVersion output buffer must be 0x100 bytes in size!");
 
-    // Instead of using the normal procedure of checking for the real system archive and if it
-    // doesn't exist, synthesizing one, I feel that that would lead to strange bugs because a
-    // used is using a really old or really new SystemVersion title. The synthesized one ensures
-    // consistence (currently reports as 5.1.0-0.0)
-    const auto archive = FileSys::SystemArchive::SystemVersion();
+    constexpr u64 FirmwareVersionSystemDataId = 0x0100000000000809;
+    auto& fsc = system.GetFileSystemController();
+
+    // Attempt to load version data from disk
+    const FileSys::RegisteredCache* bis_system{};
+    std::unique_ptr<FileSys::NCA> nca{};
+    FileSys::VirtualDir romfs{};
+
+    bis_system = fsc.GetSystemNANDContents();
+    if (bis_system) {
+        nca = bis_system->GetEntry(FirmwareVersionSystemDataId, FileSys::ContentRecordType::Data);
+    }
+    if (nca) {
+        romfs = FileSys::ExtractRomFS(nca->GetRomFS());
+    }
+    if (!romfs) {
+        romfs = FileSys::ExtractRomFS(
+            FileSys::SystemArchive::SynthesizeSystemArchive(FirmwareVersionSystemDataId));
+    }
 
     const auto early_exit_failure = [&ctx](std::string_view desc, Result code) {
         LOG_ERROR(Service_SET, "General failure while attempting to resolve firmware version ({}).",
@@ -42,13 +59,7 @@ void GetFirmwareVersionImpl(HLERequestContext& ctx, GetFirmwareVersionType type)
         rb.Push(code);
     };
 
-    if (archive == nullptr) {
-        early_exit_failure("The system version archive couldn't be synthesized.",
-                           FileSys::ERROR_FAILED_MOUNT_ARCHIVE);
-        return;
-    }
-
-    const auto ver_file = archive->GetFile("file");
+    const auto ver_file = romfs->GetFile("file");
     if (ver_file == nullptr) {
         early_exit_failure("The system version archive didn't contain the file 'file'.",
                            FileSys::ERROR_INVALID_ARGUMENT);
@@ -87,12 +98,12 @@ void SET_SYS::SetLanguageCode(HLERequestContext& ctx) {
 
 void SET_SYS::GetFirmwareVersion(HLERequestContext& ctx) {
     LOG_DEBUG(Service_SET, "called");
-    GetFirmwareVersionImpl(ctx, GetFirmwareVersionType::Version1);
+    GetFirmwareVersionImpl(system, ctx, GetFirmwareVersionType::Version1);
 }
 
 void SET_SYS::GetFirmwareVersion2(HLERequestContext& ctx) {
     LOG_DEBUG(Service_SET, "called");
-    GetFirmwareVersionImpl(ctx, GetFirmwareVersionType::Version2);
+    GetFirmwareVersionImpl(system, ctx, GetFirmwareVersionType::Version2);
 }
 
 void SET_SYS::GetAccountSettings(HLERequestContext& ctx) {
diff --git a/src/core/loader/deconstructed_rom_directory.cpp b/src/core/loader/deconstructed_rom_directory.cpp
index 5a42dea48..5c36b71e5 100644
--- a/src/core/loader/deconstructed_rom_directory.cpp
+++ b/src/core/loader/deconstructed_rom_directory.cpp
@@ -118,7 +118,7 @@ AppLoader_DeconstructedRomDirectory::LoadResult AppLoader_DeconstructedRomDirect
         return {ResultStatus::ErrorMissingNPDM, {}};
     }
 
-    const ResultStatus result2 = metadata.Load(npdm);
+    const ResultStatus result2 = metadata.Reload(npdm);
     if (result2 != ResultStatus::Success) {
         return {result2, {}};
     }
diff --git a/src/core/memory/cheat_engine.cpp b/src/core/memory/cheat_engine.cpp
index a06e99166..53a89cc8f 100644
--- a/src/core/memory/cheat_engine.cpp
+++ b/src/core/memory/cheat_engine.cpp
@@ -19,16 +19,23 @@ namespace Core::Memory {
 namespace {
 constexpr auto CHEAT_ENGINE_NS = std::chrono::nanoseconds{1000000000 / 12};
 
-std::string_view ExtractName(std::string_view data, std::size_t start_index, char match) {
+std::string_view ExtractName(std::size_t& out_name_size, std::string_view data,
+                             std::size_t start_index, char match) {
     auto end_index = start_index;
     while (data[end_index] != match) {
         ++end_index;
-        if (end_index > data.size() ||
-            (end_index - start_index - 1) > sizeof(CheatDefinition::readable_name)) {
+        if (end_index > data.size()) {
             return {};
         }
     }
 
+    out_name_size = end_index - start_index;
+
+    // Clamp name if it's too big
+    if (out_name_size > sizeof(CheatDefinition::readable_name)) {
+        end_index = start_index + sizeof(CheatDefinition::readable_name);
+    }
+
     return data.substr(start_index, end_index - start_index);
 }
 } // Anonymous namespace
@@ -113,7 +120,8 @@ std::vector<CheatEntry> TextCheatParser::Parse(std::string_view data) const {
                 return {};
             }
 
-            const auto name = ExtractName(data, i + 1, '}');
+            std::size_t name_size{};
+            const auto name = ExtractName(name_size, data, i + 1, '}');
             if (name.empty()) {
                 return {};
             }
@@ -125,12 +133,13 @@ std::vector<CheatEntry> TextCheatParser::Parse(std::string_view data) const {
                 .definition.readable_name[out[*current_entry].definition.readable_name.size() - 1] =
                 '\0';
 
-            i += name.length() + 1;
+            i += name_size + 1;
         } else if (data[i] == '[') {
             current_entry = out.size();
             out.emplace_back();
 
-            const auto name = ExtractName(data, i + 1, ']');
+            std::size_t name_size{};
+            const auto name = ExtractName(name_size, data, i + 1, ']');
             if (name.empty()) {
                 return {};
             }
@@ -142,7 +151,7 @@ std::vector<CheatEntry> TextCheatParser::Parse(std::string_view data) const {
                 .definition.readable_name[out[*current_entry].definition.readable_name.size() - 1] =
                 '\0';
 
-            i += name.length() + 1;
+            i += name_size + 1;
         } else if (::isxdigit(data[i])) {
             if (!current_entry || out[*current_entry].definition.num_opcodes >=
                                       out[*current_entry].definition.opcodes.size()) {
diff --git a/src/core/reporter.cpp b/src/core/reporter.cpp
index ed875d444..5d168cbc1 100644
--- a/src/core/reporter.cpp
+++ b/src/core/reporter.cpp
@@ -116,7 +116,7 @@ json GetProcessorStateDataAuto(Core::System& system) {
     Core::ARM_Interface::ThreadContext64 context{};
     arm.SaveContext(context);
 
-    return GetProcessorStateData(process->Is64BitProcess() ? "AArch64" : "AArch32",
+    return GetProcessorStateData(process->Is64Bit() ? "AArch64" : "AArch32",
                                  GetInteger(process->GetEntryPoint()), context.sp, context.pc,
                                  context.pstate, context.cpu_registers);
 }
diff --git a/src/input_common/drivers/udp_client.cpp b/src/input_common/drivers/udp_client.cpp
index 808b21069..77db60e92 100644
--- a/src/input_common/drivers/udp_client.cpp
+++ b/src/input_common/drivers/udp_client.cpp
@@ -338,6 +338,7 @@ void UDPClient::StartCommunication(std::size_t client, const std::string& host,
     for (std::size_t index = 0; index < PADS_PER_CLIENT; ++index) {
         const PadIdentifier identifier = GetPadIdentifier(client * PADS_PER_CLIENT + index);
         PreSetController(identifier);
+        PreSetMotion(identifier, 0);
     }
 }
 
diff --git a/src/input_common/helpers/joycon_driver.cpp b/src/input_common/helpers/joycon_driver.cpp
index cf51f3481..c9f903213 100644
--- a/src/input_common/helpers/joycon_driver.cpp
+++ b/src/input_common/helpers/joycon_driver.cpp
@@ -139,7 +139,7 @@ void JoyconDriver::InputThread(std::stop_token stop_token) {
     input_thread_running = true;
 
     // Max update rate is 5ms, ensure we are always able to read a bit faster
-    constexpr int ThreadDelay = 2;
+    constexpr int ThreadDelay = 3;
     std::vector<u8> buffer(MaxBufferSize);
 
     while (!stop_token.stop_requested()) {
@@ -163,6 +163,17 @@ void JoyconDriver::InputThread(std::stop_token stop_token) {
             OnNewData(buffer);
         }
 
+        if (!vibration_queue.Empty()) {
+            VibrationValue vibration_value;
+            vibration_queue.Pop(vibration_value);
+            last_vibration_result = rumble_protocol->SendVibration(vibration_value);
+        }
+
+        // We can't keep up with vibrations. Start skipping.
+        while (vibration_queue.Size() > 6) {
+            vibration_queue.Pop();
+        }
+
         std::this_thread::yield();
     }
 
@@ -402,7 +413,8 @@ Common::Input::DriverResult JoyconDriver::SetVibration(const VibrationValue& vib
     if (disable_input_thread) {
         return Common::Input::DriverResult::HandleInUse;
     }
-    return rumble_protocol->SendVibration(vibration);
+    vibration_queue.Push(vibration);
+    return last_vibration_result;
 }
 
 Common::Input::DriverResult JoyconDriver::SetLedConfig(u8 led_pattern) {
diff --git a/src/input_common/helpers/joycon_driver.h b/src/input_common/helpers/joycon_driver.h
index 335e12cc3..5355780fb 100644
--- a/src/input_common/helpers/joycon_driver.h
+++ b/src/input_common/helpers/joycon_driver.h
@@ -9,6 +9,7 @@
 #include <span>
 #include <thread>
 
+#include "common/threadsafe_queue.h"
 #include "input_common/helpers/joycon_protocol/joycon_types.h"
 
 namespace Common::Input {
@@ -152,6 +153,10 @@ private:
     SerialNumber handle_serial_number{}; // Serial number type reported by hidapi
     SupportedFeatures supported_features{};
 
+    /// Queue of vibration request to controllers
+    Common::Input::DriverResult last_vibration_result{Common::Input::DriverResult::Success};
+    Common::SPSCQueue<VibrationValue> vibration_queue;
+
     // Thread related
     mutable std::mutex mutex;
     std::jthread input_thread;
diff --git a/src/shader_recompiler/backend/glsl/emit_glsl.cpp b/src/shader_recompiler/backend/glsl/emit_glsl.cpp
index d91e04446..66ecfc9f7 100644
--- a/src/shader_recompiler/backend/glsl/emit_glsl.cpp
+++ b/src/shader_recompiler/backend/glsl/emit_glsl.cpp
@@ -242,6 +242,7 @@ std::string EmitGLSL(const Profile& profile, const RuntimeInfo& runtime_info, IR
     }
     if (program.info.uses_subgroup_shuffles) {
         ctx.header += "bool shfl_in_bounds;";
+        ctx.header += "uint shfl_result;";
     }
     ctx.code.insert(0, ctx.header);
     ctx.code += '}';
diff --git a/src/shader_recompiler/backend/glsl/emit_glsl_warp.cpp b/src/shader_recompiler/backend/glsl/emit_glsl_warp.cpp
index 1245c9429..f9be5de1c 100644
--- a/src/shader_recompiler/backend/glsl/emit_glsl_warp.cpp
+++ b/src/shader_recompiler/backend/glsl/emit_glsl_warp.cpp
@@ -141,7 +141,8 @@ void EmitShuffleIndex(EmitContext& ctx, IR::Inst& inst, std::string_view value,
     const auto src_thread_id{fmt::format("({})|({})", lhs, min_thread_id)};
     ctx.Add("shfl_in_bounds=int({})<=int({});", src_thread_id, max_thread_id);
     SetInBoundsFlag(ctx, inst);
-    ctx.AddU32("{}=shfl_in_bounds?readInvocationARB({},{}):{};", inst, value, src_thread_id, value);
+    ctx.Add("shfl_result=readInvocationARB({},{});", value, src_thread_id);
+    ctx.AddU32("{}=shfl_in_bounds?shfl_result:{};", inst, value);
 }
 
 void EmitShuffleUp(EmitContext& ctx, IR::Inst& inst, std::string_view value, std::string_view index,
@@ -158,7 +159,8 @@ void EmitShuffleUp(EmitContext& ctx, IR::Inst& inst, std::string_view value, std
     const auto src_thread_id{fmt::format("({}-{})", THREAD_ID, index)};
     ctx.Add("shfl_in_bounds=int({})>=int({});", src_thread_id, max_thread_id);
     SetInBoundsFlag(ctx, inst);
-    ctx.AddU32("{}=shfl_in_bounds?readInvocationARB({},{}):{};", inst, value, src_thread_id, value);
+    ctx.Add("shfl_result=readInvocationARB({},{});", value, src_thread_id);
+    ctx.AddU32("{}=shfl_in_bounds?shfl_result:{};", inst, value);
 }
 
 void EmitShuffleDown(EmitContext& ctx, IR::Inst& inst, std::string_view value,
@@ -175,7 +177,8 @@ void EmitShuffleDown(EmitContext& ctx, IR::Inst& inst, std::string_view value,
     const auto src_thread_id{fmt::format("({}+{})", THREAD_ID, index)};
     ctx.Add("shfl_in_bounds=int({})<=int({});", src_thread_id, max_thread_id);
     SetInBoundsFlag(ctx, inst);
-    ctx.AddU32("{}=shfl_in_bounds?readInvocationARB({},{}):{};", inst, value, src_thread_id, value);
+    ctx.Add("shfl_result=readInvocationARB({},{});", value, src_thread_id);
+    ctx.AddU32("{}=shfl_in_bounds?shfl_result:{};", inst, value);
 }
 
 void EmitShuffleButterfly(EmitContext& ctx, IR::Inst& inst, std::string_view value,
@@ -193,7 +196,8 @@ void EmitShuffleButterfly(EmitContext& ctx, IR::Inst& inst, std::string_view val
     const auto src_thread_id{fmt::format("({}^{})", THREAD_ID, index)};
     ctx.Add("shfl_in_bounds=int({})<=int({});", src_thread_id, max_thread_id);
     SetInBoundsFlag(ctx, inst);
-    ctx.AddU32("{}=shfl_in_bounds?readInvocationARB({},{}):{};", inst, value, src_thread_id, value);
+    ctx.Add("shfl_result=readInvocationARB({},{});", value, src_thread_id);
+    ctx.AddU32("{}=shfl_in_bounds?shfl_result:{};", inst, value);
 }
 
 void EmitFSwizzleAdd(EmitContext& ctx, IR::Inst& inst, std::string_view op_a, std::string_view op_b,
diff --git a/src/shader_recompiler/backend/spirv/emit_spirv_context_get_set.cpp b/src/shader_recompiler/backend/spirv/emit_spirv_context_get_set.cpp
index 2868fc57d..1d77426e0 100644
--- a/src/shader_recompiler/backend/spirv/emit_spirv_context_get_set.cpp
+++ b/src/shader_recompiler/backend/spirv/emit_spirv_context_get_set.cpp
@@ -111,16 +111,33 @@ Id GetCbuf(EmitContext& ctx, Id result_type, Id UniformDefinitions::*member_ptr,
     } else if (element_size > 1) {
         const u32 log2_element_size{static_cast<u32>(std::countr_zero(element_size))};
         const Id shift{ctx.Const(log2_element_size)};
-        buffer_offset = ctx.OpShiftRightArithmetic(ctx.U32[1], ctx.Def(offset), shift);
+        buffer_offset = ctx.OpShiftRightLogical(ctx.U32[1], ctx.Def(offset), shift);
     } else {
         buffer_offset = ctx.Def(offset);
     }
     if (!binding.IsImmediate()) {
         return ctx.OpFunctionCall(result_type, indirect_func, ctx.Def(binding), buffer_offset);
     }
+
     const Id cbuf{ctx.cbufs[binding.U32()].*member_ptr};
     const Id access_chain{ctx.OpAccessChain(uniform_type, cbuf, ctx.u32_zero_value, buffer_offset)};
-    return ctx.OpLoad(result_type, access_chain);
+    const Id val = ctx.OpLoad(result_type, access_chain);
+
+    if (offset.IsImmediate() || !ctx.profile.has_broken_robust) {
+        return val;
+    }
+
+    const auto is_float = UniformDefinitions::IsFloat(member_ptr);
+    const auto num_elements = UniformDefinitions::NumElements(member_ptr);
+    const std::array zero_vec{
+        is_float ? ctx.Const(0.0f) : ctx.Const(0u),
+        is_float ? ctx.Const(0.0f) : ctx.Const(0u),
+        is_float ? ctx.Const(0.0f) : ctx.Const(0u),
+        is_float ? ctx.Const(0.0f) : ctx.Const(0u),
+    };
+    const Id cond = ctx.OpULessThanEqual(ctx.TypeBool(), buffer_offset, ctx.Const(0xFFFFu));
+    const Id zero = ctx.OpCompositeConstruct(result_type, std::span(zero_vec.data(), num_elements));
+    return ctx.OpSelect(result_type, cond, val, zero);
 }
 
 Id GetCbufU32(EmitContext& ctx, const IR::Value& binding, const IR::Value& offset) {
@@ -138,7 +155,7 @@ Id GetCbufElement(EmitContext& ctx, Id vector, const IR::Value& offset, u32 inde
         const u32 element{(offset.U32() / 4) % 4 + index_offset};
         return ctx.OpCompositeExtract(ctx.U32[1], vector, element);
     }
-    const Id shift{ctx.OpShiftRightArithmetic(ctx.U32[1], ctx.Def(offset), ctx.Const(2u))};
+    const Id shift{ctx.OpShiftRightLogical(ctx.U32[1], ctx.Def(offset), ctx.Const(2u))};
     Id element{ctx.OpBitwiseAnd(ctx.U32[1], shift, ctx.Const(3u))};
     if (index_offset > 0) {
         element = ctx.OpIAdd(ctx.U32[1], element, ctx.Const(index_offset));
diff --git a/src/shader_recompiler/backend/spirv/spirv_emit_context.h b/src/shader_recompiler/backend/spirv/spirv_emit_context.h
index 7c49fd504..1aa79863d 100644
--- a/src/shader_recompiler/backend/spirv/spirv_emit_context.h
+++ b/src/shader_recompiler/backend/spirv/spirv_emit_context.h
@@ -64,6 +64,42 @@ struct UniformDefinitions {
     Id F32{};
     Id U32x2{};
     Id U32x4{};
+
+    constexpr static size_t NumElements(Id UniformDefinitions::*member_ptr) {
+        if (member_ptr == &UniformDefinitions::U8) {
+            return 1;
+        }
+        if (member_ptr == &UniformDefinitions::S8) {
+            return 1;
+        }
+        if (member_ptr == &UniformDefinitions::U16) {
+            return 1;
+        }
+        if (member_ptr == &UniformDefinitions::S16) {
+            return 1;
+        }
+        if (member_ptr == &UniformDefinitions::U32) {
+            return 1;
+        }
+        if (member_ptr == &UniformDefinitions::F32) {
+            return 1;
+        }
+        if (member_ptr == &UniformDefinitions::U32x2) {
+            return 2;
+        }
+        if (member_ptr == &UniformDefinitions::U32x4) {
+            return 4;
+        }
+        ASSERT(false);
+        return 1;
+    }
+
+    constexpr static bool IsFloat(Id UniformDefinitions::*member_ptr) {
+        if (member_ptr == &UniformDefinitions::F32) {
+            return true;
+        }
+        return false;
+    }
 };
 
 struct StorageTypeDefinition {
diff --git a/src/shader_recompiler/profile.h b/src/shader_recompiler/profile.h
index 9ca97f6a4..38d820db2 100644
--- a/src/shader_recompiler/profile.h
+++ b/src/shader_recompiler/profile.h
@@ -9,7 +9,6 @@ namespace Shader {
 
 struct Profile {
     u32 supported_spirv{0x00010000};
-
     bool unified_descriptor_binding{};
     bool support_descriptor_aliasing{};
     bool support_int8{};
@@ -82,6 +81,9 @@ struct Profile {
     bool has_broken_spirv_subgroup_mask_vector_extract_dynamic{};
 
     u32 gl_max_compute_smem_size{};
+
+    /// Maxwell and earlier nVidia architectures have broken robust support
+    bool has_broken_robust{};
 };
 
 } // namespace Shader
diff --git a/src/tests/common/unique_function.cpp b/src/tests/common/unique_function.cpp
index f7a23e876..42e6ade09 100644
--- a/src/tests/common/unique_function.cpp
+++ b/src/tests/common/unique_function.cpp
@@ -46,8 +46,8 @@ TEST_CASE("UniqueFunction", "[common]") {
         Noisy noisy;
         REQUIRE(noisy.state == "Default constructed");
 
-        Common::UniqueFunction<void> func = [noisy = std::move(noisy)] {
-            REQUIRE(noisy.state == "Move constructed");
+        Common::UniqueFunction<void> func = [noisy_inner = std::move(noisy)] {
+            REQUIRE(noisy_inner.state == "Move constructed");
         };
         REQUIRE(noisy.state == "Moved away");
         func();
@@ -101,7 +101,7 @@ TEST_CASE("UniqueFunction", "[common]") {
         };
         Foo object{&num_destroyed};
         {
-            Common::UniqueFunction<void> func = [object = std::move(object)] {};
+            Common::UniqueFunction<void> func = [object_inner = std::move(object)] {};
             REQUIRE(num_destroyed == 0);
         }
         REQUIRE(num_destroyed == 1);
diff --git a/src/video_core/buffer_cache/buffer_cache.h b/src/video_core/buffer_cache/buffer_cache.h
index 9e90c587c..081a574e8 100644
--- a/src/video_core/buffer_cache/buffer_cache.h
+++ b/src/video_core/buffer_cache/buffer_cache.h
@@ -544,7 +544,7 @@ void BufferCache<P>::CommitAsyncFlushesHigh() {
         it++;
     }
 
-    boost::container::small_vector<std::pair<BufferCopy, BufferId>, 1> downloads;
+    boost::container::small_vector<std::pair<BufferCopy, BufferId>, 16> downloads;
     u64 total_size_bytes = 0;
     u64 largest_copy = 0;
     for (const IntervalSet& intervals : committed_ranges) {
@@ -914,6 +914,11 @@ void BufferCache<P>::BindHostGraphicsStorageBuffers(size_t stage) {
 
         const u32 offset = buffer.Offset(binding.cpu_addr);
         const bool is_written = ((channel_state->written_storage_buffers[stage] >> index) & 1) != 0;
+
+        if (is_written) {
+            MarkWrittenBuffer(binding.buffer_id, binding.cpu_addr, size);
+        }
+
         if constexpr (NEEDS_BIND_STORAGE_INDEX) {
             runtime.BindStorageBuffer(stage, binding_index, buffer, offset, size, is_written);
             ++binding_index;
@@ -931,6 +936,11 @@ void BufferCache<P>::BindHostGraphicsTextureBuffers(size_t stage) {
         const u32 size = binding.size;
         SynchronizeBuffer(buffer, binding.cpu_addr, size);
 
+        const bool is_written = ((channel_state->written_texture_buffers[stage] >> index) & 1) != 0;
+        if (is_written) {
+            MarkWrittenBuffer(binding.buffer_id, binding.cpu_addr, size);
+        }
+
         const u32 offset = buffer.Offset(binding.cpu_addr);
         const PixelFormat format = binding.format;
         if constexpr (SEPARATE_IMAGE_BUFFERS_BINDINGS) {
@@ -962,6 +972,8 @@ void BufferCache<P>::BindHostTransformFeedbackBuffers() {
         const u32 size = binding.size;
         SynchronizeBuffer(buffer, binding.cpu_addr, size);
 
+        MarkWrittenBuffer(binding.buffer_id, binding.cpu_addr, size);
+
         const u32 offset = buffer.Offset(binding.cpu_addr);
         host_bindings.buffers.push_back(&buffer);
         host_bindings.offsets.push_back(offset);
@@ -1011,6 +1023,11 @@ void BufferCache<P>::BindHostComputeStorageBuffers() {
         const u32 offset = buffer.Offset(binding.cpu_addr);
         const bool is_written =
             ((channel_state->written_compute_storage_buffers >> index) & 1) != 0;
+
+        if (is_written) {
+            MarkWrittenBuffer(binding.buffer_id, binding.cpu_addr, size);
+        }
+
         if constexpr (NEEDS_BIND_STORAGE_INDEX) {
             runtime.BindComputeStorageBuffer(binding_index, buffer, offset, size, is_written);
             ++binding_index;
@@ -1028,6 +1045,12 @@ void BufferCache<P>::BindHostComputeTextureBuffers() {
         const u32 size = binding.size;
         SynchronizeBuffer(buffer, binding.cpu_addr, size);
 
+        const bool is_written =
+            ((channel_state->written_compute_texture_buffers >> index) & 1) != 0;
+        if (is_written) {
+            MarkWrittenBuffer(binding.buffer_id, binding.cpu_addr, size);
+        }
+
         const u32 offset = buffer.Offset(binding.cpu_addr);
         const PixelFormat format = binding.format;
         if constexpr (SEPARATE_IMAGE_BUFFERS_BINDINGS) {
@@ -1044,8 +1067,7 @@ void BufferCache<P>::BindHostComputeTextureBuffers() {
 
 template <class P>
 void BufferCache<P>::DoUpdateGraphicsBuffers(bool is_indexed) {
-    do {
-        channel_state->has_deleted_buffers = false;
+    BufferOperations([&]() {
         if (is_indexed) {
             UpdateIndexBuffer();
         }
@@ -1059,14 +1081,16 @@ void BufferCache<P>::DoUpdateGraphicsBuffers(bool is_indexed) {
         if (current_draw_indirect) {
             UpdateDrawIndirect();
         }
-    } while (channel_state->has_deleted_buffers);
+    });
 }
 
 template <class P>
 void BufferCache<P>::DoUpdateComputeBuffers() {
-    UpdateComputeUniformBuffers();
-    UpdateComputeStorageBuffers();
-    UpdateComputeTextureBuffers();
+    BufferOperations([&]() {
+        UpdateComputeUniformBuffers();
+        UpdateComputeStorageBuffers();
+        UpdateComputeTextureBuffers();
+    });
 }
 
 template <class P>
@@ -1201,16 +1225,11 @@ void BufferCache<P>::UpdateUniformBuffers(size_t stage) {
 
 template <class P>
 void BufferCache<P>::UpdateStorageBuffers(size_t stage) {
-    const u32 written_mask = channel_state->written_storage_buffers[stage];
     ForEachEnabledBit(channel_state->enabled_storage_buffers[stage], [&](u32 index) {
         // Resolve buffer
         Binding& binding = channel_state->storage_buffers[stage][index];
         const BufferId buffer_id = FindBuffer(binding.cpu_addr, binding.size);
         binding.buffer_id = buffer_id;
-        // Mark buffer as written if needed
-        if (((written_mask >> index) & 1) != 0) {
-            MarkWrittenBuffer(buffer_id, binding.cpu_addr, binding.size);
-        }
     });
 }
 
@@ -1219,10 +1238,6 @@ void BufferCache<P>::UpdateTextureBuffers(size_t stage) {
     ForEachEnabledBit(channel_state->enabled_texture_buffers[stage], [&](u32 index) {
         Binding& binding = channel_state->texture_buffers[stage][index];
         binding.buffer_id = FindBuffer(binding.cpu_addr, binding.size);
-        // Mark buffer as written if needed
-        if (((channel_state->written_texture_buffers[stage] >> index) & 1) != 0) {
-            MarkWrittenBuffer(binding.buffer_id, binding.cpu_addr, binding.size);
-        }
     });
 }
 
@@ -1252,7 +1267,6 @@ void BufferCache<P>::UpdateTransformFeedbackBuffer(u32 index) {
         .size = size,
         .buffer_id = buffer_id,
     };
-    MarkWrittenBuffer(buffer_id, *cpu_addr, size);
 }
 
 template <class P>
@@ -1279,10 +1293,6 @@ void BufferCache<P>::UpdateComputeStorageBuffers() {
         // Resolve buffer
         Binding& binding = channel_state->compute_storage_buffers[index];
         binding.buffer_id = FindBuffer(binding.cpu_addr, binding.size);
-        // Mark as written if needed
-        if (((channel_state->written_compute_storage_buffers >> index) & 1) != 0) {
-            MarkWrittenBuffer(binding.buffer_id, binding.cpu_addr, binding.size);
-        }
     });
 }
 
@@ -1291,18 +1301,11 @@ void BufferCache<P>::UpdateComputeTextureBuffers() {
     ForEachEnabledBit(channel_state->enabled_compute_texture_buffers, [&](u32 index) {
         Binding& binding = channel_state->compute_texture_buffers[index];
         binding.buffer_id = FindBuffer(binding.cpu_addr, binding.size);
-        // Mark as written if needed
-        if (((channel_state->written_compute_texture_buffers >> index) & 1) != 0) {
-            MarkWrittenBuffer(binding.buffer_id, binding.cpu_addr, binding.size);
-        }
     });
 }
 
 template <class P>
 void BufferCache<P>::MarkWrittenBuffer(BufferId buffer_id, VAddr cpu_addr, u32 size) {
-    if (memory_tracker.IsRegionCpuModified(cpu_addr, size)) {
-        SynchronizeBuffer(slot_buffers[buffer_id], cpu_addr, size);
-    }
     memory_tracker.MarkRegionAsGpuModified(cpu_addr, size);
 
     const IntervalType base_interval{cpu_addr, cpu_addr + size};
diff --git a/src/video_core/buffer_cache/buffer_cache_base.h b/src/video_core/buffer_cache/buffer_cache_base.h
index c4f6e8d12..eed267361 100644
--- a/src/video_core/buffer_cache/buffer_cache_base.h
+++ b/src/video_core/buffer_cache/buffer_cache_base.h
@@ -62,7 +62,11 @@ using BufferId = SlotId;
 using VideoCore::Surface::PixelFormat;
 using namespace Common::Literals;
 
+#ifdef __APPLE__
+constexpr u32 NUM_VERTEX_BUFFERS = 16;
+#else
 constexpr u32 NUM_VERTEX_BUFFERS = 32;
+#endif
 constexpr u32 NUM_TRANSFORM_FEEDBACK_BUFFERS = 4;
 constexpr u32 NUM_GRAPHICS_UNIFORM_BUFFERS = 18;
 constexpr u32 NUM_COMPUTE_UNIFORM_BUFFERS = 8;
diff --git a/src/video_core/engines/draw_manager.cpp b/src/video_core/engines/draw_manager.cpp
index f34090791..d77ff455b 100644
--- a/src/video_core/engines/draw_manager.cpp
+++ b/src/video_core/engines/draw_manager.cpp
@@ -48,8 +48,14 @@ void DrawManager::ProcessMethodCall(u32 method, u32 argument) {
         SetInlineIndexBuffer(regs.inline_index_4x8.index3);
         break;
     case MAXWELL3D_REG_INDEX(vertex_array_instance_first):
+        DrawArrayInstanced(regs.vertex_array_instance_first.topology.Value(),
+                           regs.vertex_array_instance_first.start.Value(),
+                           regs.vertex_array_instance_first.count.Value(), false);
+        break;
     case MAXWELL3D_REG_INDEX(vertex_array_instance_subsequent): {
-        LOG_WARNING(HW_GPU, "(STUBBED) called");
+        DrawArrayInstanced(regs.vertex_array_instance_subsequent.topology.Value(),
+                           regs.vertex_array_instance_subsequent.start.Value(),
+                           regs.vertex_array_instance_subsequent.count.Value(), true);
         break;
     }
     case MAXWELL3D_REG_INDEX(draw_texture.src_y0): {
@@ -84,6 +90,22 @@ void DrawManager::DrawArray(PrimitiveTopology topology, u32 vertex_first, u32 ve
     ProcessDraw(false, num_instances);
 }
 
+void DrawManager::DrawArrayInstanced(PrimitiveTopology topology, u32 vertex_first, u32 vertex_count,
+                                     bool subsequent) {
+    draw_state.topology = topology;
+    draw_state.vertex_buffer.first = vertex_first;
+    draw_state.vertex_buffer.count = vertex_count;
+
+    if (!subsequent) {
+        draw_state.instance_count = 1;
+    }
+
+    draw_state.base_instance = draw_state.instance_count - 1;
+    draw_state.draw_mode = DrawMode::Instance;
+    draw_state.instance_count++;
+    ProcessDraw(false, 1);
+}
+
 void DrawManager::DrawIndex(PrimitiveTopology topology, u32 index_first, u32 index_count,
                             u32 base_index, u32 base_instance, u32 num_instances) {
     const auto& regs{maxwell3d->regs};
diff --git a/src/video_core/engines/draw_manager.h b/src/video_core/engines/draw_manager.h
index 18d959143..cfc8127fc 100644
--- a/src/video_core/engines/draw_manager.h
+++ b/src/video_core/engines/draw_manager.h
@@ -66,6 +66,8 @@ public:
 
     void DrawArray(PrimitiveTopology topology, u32 vertex_first, u32 vertex_count,
                    u32 base_instance, u32 num_instances);
+    void DrawArrayInstanced(PrimitiveTopology topology, u32 vertex_first, u32 vertex_count,
+                            bool subsequent);
 
     void DrawIndex(PrimitiveTopology topology, u32 index_first, u32 index_count, u32 base_index,
                    u32 base_instance, u32 num_instances);
diff --git a/src/video_core/fence_manager.h b/src/video_core/fence_manager.h
index 805a89900..c0e6471fe 100644
--- a/src/video_core/fence_manager.h
+++ b/src/video_core/fence_manager.h
@@ -86,7 +86,10 @@ public:
             uncommitted_operations.emplace_back(std::move(func));
         }
         pending_operations.emplace_back(std::move(uncommitted_operations));
-        QueueFence(new_fence);
+        {
+            std::scoped_lock lock{buffer_cache.mutex, texture_cache.mutex};
+            QueueFence(new_fence);
+        }
         if (!delay_fence) {
             func();
         }
diff --git a/src/video_core/host1x/codecs/codec.cpp b/src/video_core/host1x/codecs/codec.cpp
index 8d7da50fc..dbcf508e5 100644
--- a/src/video_core/host1x/codecs/codec.cpp
+++ b/src/video_core/host1x/codecs/codec.cpp
@@ -137,16 +137,6 @@ bool Codec::CreateGpuAvDevice() {
                 break;
             }
             if ((config->methods & HW_CONFIG_METHOD) != 0 && config->device_type == type) {
-#if defined(__unix__)
-                // Some linux decoding backends are reported to crash with this config method
-                // TODO(ameerj): Properly support this method
-                if ((config->methods & AV_CODEC_HW_CONFIG_METHOD_HW_FRAMES_CTX) != 0) {
-                    // skip zero-copy decoders, we don't currently support them
-                    LOG_DEBUG(Service_NVDRV, "Skipping decoder {} with unsupported capability {}.",
-                              av_hwdevice_get_type_name(type), config->methods);
-                    continue;
-                }
-#endif
                 LOG_INFO(Service_NVDRV, "Using {} GPU decoder", av_hwdevice_get_type_name(type));
                 av_codec_ctx->pix_fmt = config->pix_fmt;
                 return true;
diff --git a/src/video_core/host_shaders/CMakeLists.txt b/src/video_core/host_shaders/CMakeLists.txt
index 6b912027f..cd2549232 100644
--- a/src/video_core/host_shaders/CMakeLists.txt
+++ b/src/video_core/host_shaders/CMakeLists.txt
@@ -19,6 +19,8 @@ set(SHADER_FILES
     block_linear_unswizzle_2d.comp
     block_linear_unswizzle_3d.comp
     convert_abgr8_to_d24s8.frag
+    convert_abgr8_to_d32f.frag
+    convert_d32f_to_abgr8.frag
     convert_d24s8_to_abgr8.frag
     convert_depth_to_float.frag
     convert_float_to_depth.frag
diff --git a/src/video_core/host_shaders/convert_abgr8_to_d32f.frag b/src/video_core/host_shaders/convert_abgr8_to_d32f.frag
new file mode 100644
index 000000000..095b910c2
--- /dev/null
+++ b/src/video_core/host_shaders/convert_abgr8_to_d32f.frag
@@ -0,0 +1,15 @@
+// SPDX-FileCopyrightText: Copyright 2023 yuzu Emulator Project
+// SPDX-License-Identifier: GPL-2.0-or-later
+
+#version 450
+
+layout(binding = 0) uniform sampler2D color_texture;
+
+void main() {
+    ivec2 coord = ivec2(gl_FragCoord.xy);
+    vec4 color = texelFetch(color_texture, coord, 0).abgr;
+
+    float value = color.a * (color.r + color.g + color.b) / 3.0f;
+
+    gl_FragDepth = value;
+}
diff --git a/src/video_core/host_shaders/convert_d24s8_to_abgr8.frag b/src/video_core/host_shaders/convert_d24s8_to_abgr8.frag
index d33131d7c..b81a54056 100644
--- a/src/video_core/host_shaders/convert_d24s8_to_abgr8.frag
+++ b/src/video_core/host_shaders/convert_d24s8_to_abgr8.frag
@@ -3,16 +3,16 @@
 
 #version 450
 
+precision mediump int;
+precision highp float;
+
 layout(binding = 0) uniform sampler2D depth_tex;
-layout(binding = 1) uniform isampler2D stencil_tex;
+layout(binding = 1) uniform usampler2D stencil_tex;
 
 layout(location = 0) out vec4 color;
 
 void main() {
     ivec2 coord = ivec2(gl_FragCoord.xy);
-    uint depth = uint(textureLod(depth_tex, coord, 0).r * (exp2(24.0) - 1.0f));
-    uint stencil = uint(textureLod(stencil_tex, coord, 0).r);
-
     highp uint depth_val =
         uint(textureLod(depth_tex, coord, 0).r * (exp2(32.0) - 1.0));
     lowp uint stencil_val = textureLod(stencil_tex, coord, 0).r;
diff --git a/src/video_core/host_shaders/convert_d32f_to_abgr8.frag b/src/video_core/host_shaders/convert_d32f_to_abgr8.frag
new file mode 100644
index 000000000..4e5a9f955
--- /dev/null
+++ b/src/video_core/host_shaders/convert_d32f_to_abgr8.frag
@@ -0,0 +1,14 @@
+// SPDX-FileCopyrightText: Copyright 2021 yuzu Emulator Project
+// SPDX-License-Identifier: GPL-2.0-or-later
+
+#version 450
+
+layout(binding = 0) uniform sampler2D depth_tex;
+
+layout(location = 0) out vec4 color;
+
+void main() {
+    ivec2 coord = ivec2(gl_FragCoord.xy);
+    float depth = texelFetch(depth_tex, coord, 0).r;
+    color = vec4(depth, depth, depth, 1.0);
+}
diff --git a/src/video_core/host_shaders/convert_s8d24_to_abgr8.frag b/src/video_core/host_shaders/convert_s8d24_to_abgr8.frag
index 31db7d426..6a457981d 100644
--- a/src/video_core/host_shaders/convert_s8d24_to_abgr8.frag
+++ b/src/video_core/host_shaders/convert_s8d24_to_abgr8.frag
@@ -3,16 +3,16 @@
 
 #version 450
 
+precision mediump int;
+precision highp float;
+
 layout(binding = 0) uniform sampler2D depth_tex;
-layout(binding = 1) uniform isampler2D stencil_tex;
+layout(binding = 1) uniform usampler2D stencil_tex;
 
 layout(location = 0) out vec4 color;
 
 void main() {
     ivec2 coord = ivec2(gl_FragCoord.xy);
-    uint depth = uint(textureLod(depth_tex, coord, 0).r * (exp2(24.0) - 1.0f));
-    uint stencil = uint(textureLod(stencil_tex, coord, 0).r);
-
     highp uint depth_val =
         uint(textureLod(depth_tex, coord, 0).r * (exp2(32.0) - 1.0));
     lowp uint stencil_val = textureLod(stencil_tex, coord, 0).r;
diff --git a/src/video_core/renderer_base.h b/src/video_core/renderer_base.h
index 3e12a8813..78ea5208b 100644
--- a/src/video_core/renderer_base.h
+++ b/src/video_core/renderer_base.h
@@ -89,9 +89,6 @@ public:
     void RequestScreenshot(void* data, std::function<void(bool)> callback,
                            const Layout::FramebufferLayout& layout);
 
-    /// This is called to notify the rendering backend of a surface change
-    virtual void NotifySurfaceChanged() {}
-
 protected:
     Core::Frontend::EmuWindow& render_window; ///< Reference to the render window handle.
     std::unique_ptr<Core::Frontend::GraphicsContext> context;
diff --git a/src/video_core/renderer_opengl/gl_texture_cache.cpp b/src/video_core/renderer_opengl/gl_texture_cache.cpp
index 9cafd2983..512eef575 100644
--- a/src/video_core/renderer_opengl/gl_texture_cache.cpp
+++ b/src/video_core/renderer_opengl/gl_texture_cache.cpp
@@ -1048,6 +1048,10 @@ void Image::Scale(bool up_scale) {
 }
 
 bool Image::ScaleUp(bool ignore) {
+    const auto& resolution = runtime->resolution;
+    if (!resolution.active) {
+        return false;
+    }
     if (True(flags & ImageFlagBits::Rescaled)) {
         return false;
     }
@@ -1060,9 +1064,6 @@ bool Image::ScaleUp(bool ignore) {
         return false;
     }
     flags |= ImageFlagBits::Rescaled;
-    if (!runtime->resolution.active) {
-        return false;
-    }
     has_scaled = true;
     if (ignore) {
         current_texture = upscaled_backup.handle;
@@ -1073,13 +1074,14 @@ bool Image::ScaleUp(bool ignore) {
 }
 
 bool Image::ScaleDown(bool ignore) {
-    if (False(flags & ImageFlagBits::Rescaled)) {
+    const auto& resolution = runtime->resolution;
+    if (!resolution.active) {
         return false;
     }
-    flags &= ~ImageFlagBits::Rescaled;
-    if (!runtime->resolution.active) {
+    if (False(flags & ImageFlagBits::Rescaled)) {
         return false;
     }
+    flags &= ~ImageFlagBits::Rescaled;
     if (ignore) {
         current_texture = texture.handle;
         return true;
diff --git a/src/video_core/renderer_opengl/gl_texture_cache.h b/src/video_core/renderer_opengl/gl_texture_cache.h
index 3676eaaa9..e71b87e99 100644
--- a/src/video_core/renderer_opengl/gl_texture_cache.h
+++ b/src/video_core/renderer_opengl/gl_texture_cache.h
@@ -118,6 +118,8 @@ public:
 
     void InsertUploadMemoryBarrier();
 
+    void TransitionImageLayout(Image& image) {}
+
     FormatProperties FormatInfo(VideoCommon::ImageType type, GLenum internal_format) const;
 
     bool HasNativeBgr() const noexcept {
diff --git a/src/video_core/renderer_opengl/maxwell_to_gl.h b/src/video_core/renderer_opengl/maxwell_to_gl.h
index c7dc7e0a1..5ea9e2378 100644
--- a/src/video_core/renderer_opengl/maxwell_to_gl.h
+++ b/src/video_core/renderer_opengl/maxwell_to_gl.h
@@ -116,6 +116,7 @@ constexpr std::array<FormatTuple, VideoCore::Surface::MaxPixelFormat> FORMAT_TAB
     {GL_RGB9_E5, GL_RGB, GL_UNSIGNED_INT_5_9_9_9_REV},                // E5B9G9R9_FLOAT
     {GL_DEPTH_COMPONENT32F, GL_DEPTH_COMPONENT, GL_FLOAT},            // D32_FLOAT
     {GL_DEPTH_COMPONENT16, GL_DEPTH_COMPONENT, GL_UNSIGNED_SHORT},    // D16_UNORM
+    {GL_DEPTH_COMPONENT24, GL_DEPTH_COMPONENT, GL_UNSIGNED_INT_24_8}, // X8_D24_UNORM
     {GL_STENCIL_INDEX8, GL_STENCIL, GL_UNSIGNED_BYTE},                // S8_UINT
     {GL_DEPTH24_STENCIL8, GL_DEPTH_STENCIL, GL_UNSIGNED_INT_24_8},    // D24_UNORM_S8_UINT
     {GL_DEPTH24_STENCIL8, GL_DEPTH_STENCIL, GL_UNSIGNED_INT_24_8},    // S8_UINT_D24_UNORM
diff --git a/src/video_core/renderer_vulkan/blit_image.cpp b/src/video_core/renderer_vulkan/blit_image.cpp
index 1032c9d12..c3db09424 100644
--- a/src/video_core/renderer_vulkan/blit_image.cpp
+++ b/src/video_core/renderer_vulkan/blit_image.cpp
@@ -8,7 +8,9 @@
 #include "common/settings.h"
 #include "video_core/host_shaders/blit_color_float_frag_spv.h"
 #include "video_core/host_shaders/convert_abgr8_to_d24s8_frag_spv.h"
+#include "video_core/host_shaders/convert_abgr8_to_d32f_frag_spv.h"
 #include "video_core/host_shaders/convert_d24s8_to_abgr8_frag_spv.h"
+#include "video_core/host_shaders/convert_d32f_to_abgr8_frag_spv.h"
 #include "video_core/host_shaders/convert_depth_to_float_frag_spv.h"
 #include "video_core/host_shaders/convert_float_to_depth_frag_spv.h"
 #include "video_core/host_shaders/convert_s8d24_to_abgr8_frag_spv.h"
@@ -433,6 +435,8 @@ BlitImageHelper::BlitImageHelper(const Device& device_, Scheduler& scheduler_,
       convert_depth_to_float_frag(BuildShader(device, CONVERT_DEPTH_TO_FLOAT_FRAG_SPV)),
       convert_float_to_depth_frag(BuildShader(device, CONVERT_FLOAT_TO_DEPTH_FRAG_SPV)),
       convert_abgr8_to_d24s8_frag(BuildShader(device, CONVERT_ABGR8_TO_D24S8_FRAG_SPV)),
+      convert_abgr8_to_d32f_frag(BuildShader(device, CONVERT_ABGR8_TO_D32F_FRAG_SPV)),
+      convert_d32f_to_abgr8_frag(BuildShader(device, CONVERT_D32F_TO_ABGR8_FRAG_SPV)),
       convert_d24s8_to_abgr8_frag(BuildShader(device, CONVERT_D24S8_TO_ABGR8_FRAG_SPV)),
       convert_s8d24_to_abgr8_frag(BuildShader(device, CONVERT_S8D24_TO_ABGR8_FRAG_SPV)),
       linear_sampler(device.GetLogical().CreateSampler(SAMPLER_CREATE_INFO<VK_FILTER_LINEAR>)),
@@ -557,6 +561,20 @@ void BlitImageHelper::ConvertABGR8ToD24S8(const Framebuffer* dst_framebuffer,
     Convert(*convert_abgr8_to_d24s8_pipeline, dst_framebuffer, src_image_view);
 }
 
+void BlitImageHelper::ConvertABGR8ToD32F(const Framebuffer* dst_framebuffer,
+                                         const ImageView& src_image_view) {
+    ConvertPipelineDepthTargetEx(convert_abgr8_to_d32f_pipeline, dst_framebuffer->RenderPass(),
+                                 convert_abgr8_to_d32f_frag);
+    Convert(*convert_abgr8_to_d32f_pipeline, dst_framebuffer, src_image_view);
+}
+
+void BlitImageHelper::ConvertD32FToABGR8(const Framebuffer* dst_framebuffer,
+                                         ImageView& src_image_view) {
+    ConvertPipelineColorTargetEx(convert_d32f_to_abgr8_pipeline, dst_framebuffer->RenderPass(),
+                                 convert_d32f_to_abgr8_frag);
+    ConvertDepthStencil(*convert_d32f_to_abgr8_pipeline, dst_framebuffer, src_image_view);
+}
+
 void BlitImageHelper::ConvertD24S8ToABGR8(const Framebuffer* dst_framebuffer,
                                           ImageView& src_image_view) {
     ConvertPipelineColorTargetEx(convert_d24s8_to_abgr8_pipeline, dst_framebuffer->RenderPass(),
@@ -609,6 +627,8 @@ void BlitImageHelper::ClearDepthStencil(const Framebuffer* dst_framebuffer, bool
     const VkPipelineLayout layout = *clear_color_pipeline_layout;
     scheduler.RequestRenderpass(dst_framebuffer);
     scheduler.Record([pipeline, layout, clear_depth, dst_region](vk::CommandBuffer cmdbuf) {
+        constexpr std::array blend_constants{0.0f, 0.0f, 0.0f, 0.0f};
+        cmdbuf.SetBlendConstants(blend_constants.data());
         cmdbuf.BindPipeline(VK_PIPELINE_BIND_POINT_GRAPHICS, pipeline);
         BindBlitState(cmdbuf, dst_region);
         cmdbuf.PushConstants(layout, VK_SHADER_STAGE_FRAGMENT_BIT, clear_depth);
@@ -865,7 +885,7 @@ VkPipeline BlitImageHelper::FindOrEmplaceClearStencilPipeline(
         .sType = VK_STRUCTURE_TYPE_PIPELINE_DEPTH_STENCIL_STATE_CREATE_INFO,
         .pNext = nullptr,
         .flags = 0,
-        .depthTestEnable = VK_FALSE,
+        .depthTestEnable = key.depth_clear,
         .depthWriteEnable = key.depth_clear,
         .depthCompareOp = VK_COMPARE_OP_ALWAYS,
         .depthBoundsTestEnable = VK_FALSE,
diff --git a/src/video_core/renderer_vulkan/blit_image.h b/src/video_core/renderer_vulkan/blit_image.h
index dcfe217aa..b2104a59e 100644
--- a/src/video_core/renderer_vulkan/blit_image.h
+++ b/src/video_core/renderer_vulkan/blit_image.h
@@ -67,6 +67,10 @@ public:
 
     void ConvertABGR8ToD24S8(const Framebuffer* dst_framebuffer, const ImageView& src_image_view);
 
+    void ConvertABGR8ToD32F(const Framebuffer* dst_framebuffer, const ImageView& src_image_view);
+
+    void ConvertD32FToABGR8(const Framebuffer* dst_framebuffer, ImageView& src_image_view);
+
     void ConvertD24S8ToABGR8(const Framebuffer* dst_framebuffer, ImageView& src_image_view);
 
     void ConvertS8D24ToABGR8(const Framebuffer* dst_framebuffer, ImageView& src_image_view);
@@ -128,6 +132,8 @@ private:
     vk::ShaderModule convert_depth_to_float_frag;
     vk::ShaderModule convert_float_to_depth_frag;
     vk::ShaderModule convert_abgr8_to_d24s8_frag;
+    vk::ShaderModule convert_abgr8_to_d32f_frag;
+    vk::ShaderModule convert_d32f_to_abgr8_frag;
     vk::ShaderModule convert_d24s8_to_abgr8_frag;
     vk::ShaderModule convert_s8d24_to_abgr8_frag;
     vk::Sampler linear_sampler;
@@ -146,6 +152,8 @@ private:
     vk::Pipeline convert_d16_to_r16_pipeline;
     vk::Pipeline convert_r16_to_d16_pipeline;
     vk::Pipeline convert_abgr8_to_d24s8_pipeline;
+    vk::Pipeline convert_abgr8_to_d32f_pipeline;
+    vk::Pipeline convert_d32f_to_abgr8_pipeline;
     vk::Pipeline convert_d24s8_to_abgr8_pipeline;
     vk::Pipeline convert_s8d24_to_abgr8_pipeline;
 };
diff --git a/src/video_core/renderer_vulkan/maxwell_to_vk.cpp b/src/video_core/renderer_vulkan/maxwell_to_vk.cpp
index 208e88533..a08f2f67f 100644
--- a/src/video_core/renderer_vulkan/maxwell_to_vk.cpp
+++ b/src/video_core/renderer_vulkan/maxwell_to_vk.cpp
@@ -214,8 +214,9 @@ struct FormatTuple {
     {VK_FORMAT_E5B9G9R9_UFLOAT_PACK32},                        // E5B9G9R9_FLOAT
 
     // Depth formats
-    {VK_FORMAT_D32_SFLOAT, Attachable}, // D32_FLOAT
-    {VK_FORMAT_D16_UNORM, Attachable},  // D16_UNORM
+    {VK_FORMAT_D32_SFLOAT, Attachable},          // D32_FLOAT
+    {VK_FORMAT_D16_UNORM, Attachable},           // D16_UNORM
+    {VK_FORMAT_X8_D24_UNORM_PACK32, Attachable}, // X8_D24_UNORM
 
     // Stencil formats
     {VK_FORMAT_S8_UINT, Attachable}, // S8_UINT
diff --git a/src/video_core/renderer_vulkan/renderer_vulkan.cpp b/src/video_core/renderer_vulkan/renderer_vulkan.cpp
index c4c30d807..7e7a80740 100644
--- a/src/video_core/renderer_vulkan/renderer_vulkan.cpp
+++ b/src/video_core/renderer_vulkan/renderer_vulkan.cpp
@@ -132,12 +132,16 @@ void RendererVulkan::SwapBuffers(const Tegra::FramebufferConfig* framebuffer) {
     const bool use_accelerated =
         rasterizer.AccelerateDisplay(*framebuffer, framebuffer_addr, framebuffer->stride);
     const bool is_srgb = use_accelerated && screen_info.is_srgb;
-    RenderScreenshot(*framebuffer, use_accelerated);
 
-    Frame* frame = present_manager.GetRenderFrame();
-    blit_screen.DrawToSwapchain(frame, *framebuffer, use_accelerated, is_srgb);
-    scheduler.Flush(*frame->render_ready);
-    present_manager.Present(frame);
+    {
+        std::scoped_lock lock{rasterizer.LockCaches()};
+        RenderScreenshot(*framebuffer, use_accelerated);
+
+        Frame* frame = present_manager.GetRenderFrame();
+        blit_screen.DrawToSwapchain(frame, *framebuffer, use_accelerated, is_srgb);
+        scheduler.Flush(*frame->render_ready);
+        present_manager.Present(frame);
+    }
 
     gpu.RendererFrameEndNotify();
     rasterizer.TickFrame();
diff --git a/src/video_core/renderer_vulkan/renderer_vulkan.h b/src/video_core/renderer_vulkan/renderer_vulkan.h
index 590bc1c64..14e257cf7 100644
--- a/src/video_core/renderer_vulkan/renderer_vulkan.h
+++ b/src/video_core/renderer_vulkan/renderer_vulkan.h
@@ -56,10 +56,6 @@ public:
         return device.GetDriverName();
     }
 
-    void NotifySurfaceChanged() override {
-        present_manager.NotifySurfaceChanged();
-    }
-
 private:
     void Report() const;
 
diff --git a/src/video_core/renderer_vulkan/vk_blit_screen.cpp b/src/video_core/renderer_vulkan/vk_blit_screen.cpp
index 31928bb94..52fc142d1 100644
--- a/src/video_core/renderer_vulkan/vk_blit_screen.cpp
+++ b/src/video_core/renderer_vulkan/vk_blit_screen.cpp
@@ -96,6 +96,7 @@ std::size_t GetSizeInBytes(const Tegra::FramebufferConfig& framebuffer) {
 VkFormat GetFormat(const Tegra::FramebufferConfig& framebuffer) {
     switch (framebuffer.pixel_format) {
     case Service::android::PixelFormat::Rgba8888:
+    case Service::android::PixelFormat::Rgbx8888:
         return VK_FORMAT_A8B8G8R8_UNORM_PACK32;
     case Service::android::PixelFormat::Rgb565:
         return VK_FORMAT_R5G6B5_UNORM_PACK16;
diff --git a/src/video_core/renderer_vulkan/vk_pipeline_cache.cpp b/src/video_core/renderer_vulkan/vk_pipeline_cache.cpp
index a1ec1a100..22bf8cc77 100644
--- a/src/video_core/renderer_vulkan/vk_pipeline_cache.cpp
+++ b/src/video_core/renderer_vulkan/vk_pipeline_cache.cpp
@@ -356,7 +356,11 @@ PipelineCache::PipelineCache(RasterizerVulkan& rasterizer_, const Device& device
         .has_broken_fp16_float_controls = driver_id == VK_DRIVER_ID_NVIDIA_PROPRIETARY,
         .ignore_nan_fp_comparisons = false,
         .has_broken_spirv_subgroup_mask_vector_extract_dynamic =
-            driver_id == VK_DRIVER_ID_QUALCOMM_PROPRIETARY};
+            driver_id == VK_DRIVER_ID_QUALCOMM_PROPRIETARY,
+        .has_broken_robust =
+            device.IsNvidia() && device.GetNvidiaArch() <= NvidiaArchitecture::Arch_Pascal,
+    };
+
     host_info = Shader::HostTranslateInfo{
         .support_float64 = device.IsFloat64Supported(),
         .support_float16 = device.IsFloat16Supported(),
diff --git a/src/video_core/renderer_vulkan/vk_present_manager.cpp b/src/video_core/renderer_vulkan/vk_present_manager.cpp
index d681bd22a..2ef36583b 100644
--- a/src/video_core/renderer_vulkan/vk_present_manager.cpp
+++ b/src/video_core/renderer_vulkan/vk_present_manager.cpp
@@ -103,8 +103,7 @@ PresentManager::PresentManager(const vk::Instance& instance_,
       surface{surface_}, blit_supported{CanBlitToSwapchain(device.GetPhysical(),
                                                            swapchain.GetImageViewFormat())},
       use_present_thread{Settings::values.async_presentation.GetValue()},
-      image_count{swapchain.GetImageCount()}, last_render_surface{
-                                                  render_window_.GetWindowInfo().render_surface} {
+      image_count{swapchain.GetImageCount()} {
 
     auto& dld = device.GetLogical();
     cmdpool = dld.CreateCommandPool({
@@ -289,44 +288,36 @@ void PresentManager::PresentThread(std::stop_token token) {
     }
 }
 
-void PresentManager::NotifySurfaceChanged() {
-#ifdef ANDROID
-    std::scoped_lock lock{recreate_surface_mutex};
-    recreate_surface_cv.notify_one();
-#endif
+void PresentManager::RecreateSwapchain(Frame* frame) {
+    swapchain.Create(*surface, frame->width, frame->height, frame->is_srgb);
+    image_count = swapchain.GetImageCount();
 }
 
 void PresentManager::CopyToSwapchain(Frame* frame) {
-    MICROPROFILE_SCOPE(Vulkan_CopyToSwapchain);
-
-    const auto recreate_swapchain = [&] {
-        swapchain.Create(*surface, frame->width, frame->height, frame->is_srgb);
-        image_count = swapchain.GetImageCount();
-    };
-
-#ifdef ANDROID
-    std::unique_lock lock{recreate_surface_mutex};
-
-    const auto needs_recreation = [&] {
-        if (last_render_surface != render_window.GetWindowInfo().render_surface) {
-            return true;
-        }
-        if (swapchain.NeedsRecreation(frame->is_srgb)) {
-            return true;
+    bool requires_recreation = false;
+
+    while (true) {
+        try {
+            // Recreate surface and swapchain if needed.
+            if (requires_recreation) {
+                surface = CreateSurface(instance, render_window.GetWindowInfo());
+                RecreateSwapchain(frame);
+            }
+
+            // Draw to swapchain.
+            return CopyToSwapchainImpl(frame);
+        } catch (const vk::Exception& except) {
+            if (except.GetResult() != VK_ERROR_SURFACE_LOST_KHR) {
+                throw;
+            }
+
+            requires_recreation = true;
         }
-        return false;
-    };
-
-    recreate_surface_cv.wait_for(lock, std::chrono::milliseconds(400),
-                                 [&]() { return !needs_recreation(); });
-
-    // If the frontend recreated the surface, recreate the renderer surface and swapchain.
-    if (last_render_surface != render_window.GetWindowInfo().render_surface) {
-        last_render_surface = render_window.GetWindowInfo().render_surface;
-        surface = CreateSurface(instance, render_window.GetWindowInfo());
-        recreate_swapchain();
     }
-#endif
+}
+
+void PresentManager::CopyToSwapchainImpl(Frame* frame) {
+    MICROPROFILE_SCOPE(Vulkan_CopyToSwapchain);
 
     // If the size or colorspace of the incoming frames has changed, recreate the swapchain
     // to account for that.
@@ -334,11 +325,11 @@ void PresentManager::CopyToSwapchain(Frame* frame) {
     const bool size_changed =
         swapchain.GetWidth() != frame->width || swapchain.GetHeight() != frame->height;
     if (srgb_changed || size_changed) {
-        recreate_swapchain();
+        RecreateSwapchain(frame);
     }
 
     while (swapchain.AcquireNextImage()) {
-        recreate_swapchain();
+        RecreateSwapchain(frame);
     }
 
     const vk::CommandBuffer cmdbuf{frame->cmdbuf};
@@ -488,4 +479,4 @@ void PresentManager::CopyToSwapchain(Frame* frame) {
     swapchain.Present(render_semaphore);
 }
 
-} // namespace Vulkan
\ No newline at end of file
+} // namespace Vulkan
diff --git a/src/video_core/renderer_vulkan/vk_present_manager.h b/src/video_core/renderer_vulkan/vk_present_manager.h
index 83e859416..a3d825fe6 100644
--- a/src/video_core/renderer_vulkan/vk_present_manager.h
+++ b/src/video_core/renderer_vulkan/vk_present_manager.h
@@ -54,14 +54,15 @@ public:
     /// Waits for the present thread to finish presenting all queued frames.
     void WaitPresent();
 
-    /// This is called to notify the rendering backend of a surface change
-    void NotifySurfaceChanged();
-
 private:
     void PresentThread(std::stop_token token);
 
     void CopyToSwapchain(Frame* frame);
 
+    void CopyToSwapchainImpl(Frame* frame);
+
+    void RecreateSwapchain(Frame* frame);
+
 private:
     const vk::Instance& instance;
     Core::Frontend::EmuWindow& render_window;
@@ -76,16 +77,13 @@ private:
     std::queue<Frame*> free_queue;
     std::condition_variable_any frame_cv;
     std::condition_variable free_cv;
-    std::condition_variable recreate_surface_cv;
     std::mutex swapchain_mutex;
-    std::mutex recreate_surface_mutex;
     std::mutex queue_mutex;
     std::mutex free_mutex;
     std::jthread present_thread;
     bool blit_supported;
     bool use_present_thread;
     std::size_t image_count{};
-    void* last_render_surface{};
 };
 
 } // namespace Vulkan
diff --git a/src/video_core/renderer_vulkan/vk_query_cache.cpp b/src/video_core/renderer_vulkan/vk_query_cache.cpp
index 2edaafa7e..66c03bf17 100644
--- a/src/video_core/renderer_vulkan/vk_query_cache.cpp
+++ b/src/video_core/renderer_vulkan/vk_query_cache.cpp
@@ -1436,6 +1436,7 @@ void QueryCacheRuntime::Barriers(bool is_prebarrier) {
         .srcAccessMask = VK_ACCESS_TRANSFER_WRITE_BIT,
         .dstAccessMask = VK_ACCESS_MEMORY_READ_BIT | VK_ACCESS_MEMORY_WRITE_BIT,
     };
+    impl->scheduler.RequestOutsideRenderPassOperationContext();
     if (is_prebarrier) {
         impl->scheduler.Record([](vk::CommandBuffer cmdbuf) {
             cmdbuf.PipelineBarrier(VK_PIPELINE_STAGE_ALL_COMMANDS_BIT,
diff --git a/src/video_core/renderer_vulkan/vk_rasterizer.cpp b/src/video_core/renderer_vulkan/vk_rasterizer.cpp
index 1628d76d6..059b7cb40 100644
--- a/src/video_core/renderer_vulkan/vk_rasterizer.cpp
+++ b/src/video_core/renderer_vulkan/vk_rasterizer.cpp
@@ -13,6 +13,7 @@
 #include "common/microprofile.h"
 #include "common/scope_exit.h"
 #include "common/settings.h"
+#include "video_core/buffer_cache/buffer_cache.h"
 #include "video_core/control/channel_state.h"
 #include "video_core/engines/draw_manager.h"
 #include "video_core/engines/kepler_compute.h"
@@ -198,7 +199,7 @@ void RasterizerVulkan::PrepareDraw(bool is_indexed, Func&& draw_func) {
     if (!pipeline) {
         return;
     }
-    std::scoped_lock lock{buffer_cache.mutex, texture_cache.mutex};
+    std::scoped_lock lock{LockCaches()};
     // update engine as channel may be different.
     pipeline->SetEngine(maxwell3d, gpu_memory);
     pipeline->Configure(is_indexed);
@@ -285,6 +286,7 @@ void RasterizerVulkan::DrawTexture() {
 
     query_cache.NotifySegment(true);
 
+    std::scoped_lock l{texture_cache.mutex};
     texture_cache.SynchronizeGraphicsDescriptors();
     texture_cache.UpdateRenderTargets(false);
 
@@ -422,7 +424,8 @@ void RasterizerVulkan::Clear(u32 layer_count) {
         return;
     }
 
-    if (use_stencil && regs.stencil_front_mask != 0xFF && regs.stencil_front_mask != 0) {
+    if (use_stencil && framebuffer->HasAspectStencilBit() && regs.stencil_front_mask != 0xFF &&
+        regs.stencil_front_mask != 0) {
         Region2D dst_region = {
             Offset2D{.x = clear_rect.rect.offset.x, .y = clear_rect.rect.offset.y},
             Offset2D{.x = clear_rect.rect.offset.x + static_cast<s32>(clear_rect.rect.extent.width),
@@ -707,6 +710,7 @@ void RasterizerVulkan::TiledCacheBarrier() {
 }
 
 void RasterizerVulkan::FlushCommands() {
+    std::scoped_lock lock{LockCaches()};
     if (draw_counter == 0) {
         return;
     }
@@ -804,6 +808,7 @@ void RasterizerVulkan::FlushWork() {
     if ((++draw_counter & 7) != 7) {
         return;
     }
+    std::scoped_lock lock{LockCaches()};
     if (draw_counter < DRAWS_TO_DISPATCH) {
         // Send recorded tasks to the worker thread
         scheduler.DispatchWork();
@@ -974,6 +979,19 @@ void RasterizerVulkan::UpdateScissorsState(Tegra::Engines::Maxwell3D::Regs& regs
     if (!state_tracker.TouchScissors()) {
         return;
     }
+    if (!regs.viewport_scale_offset_enabled) {
+        const auto x = static_cast<float>(regs.surface_clip.x);
+        const auto y = static_cast<float>(regs.surface_clip.y);
+        const auto width = static_cast<float>(regs.surface_clip.width);
+        const auto height = static_cast<float>(regs.surface_clip.height);
+        VkRect2D scissor;
+        scissor.offset.x = static_cast<u32>(x);
+        scissor.offset.y = static_cast<u32>(y);
+        scissor.extent.width = static_cast<u32>(width != 0.0f ? width : 1.0f);
+        scissor.extent.height = static_cast<u32>(height != 0.0f ? height : 1.0f);
+        scheduler.Record([scissor](vk::CommandBuffer cmdbuf) { cmdbuf.SetScissor(0, scissor); });
+        return;
+    }
     u32 up_scale = 1;
     u32 down_shift = 0;
     if (texture_cache.IsRescaling()) {
@@ -1485,7 +1503,7 @@ void RasterizerVulkan::UpdateVertexInput(Tegra::Engines::Maxwell3D::Regs& regs)
 void RasterizerVulkan::InitializeChannel(Tegra::Control::ChannelState& channel) {
     CreateChannel(channel);
     {
-        std::scoped_lock lock{buffer_cache.mutex, texture_cache.mutex};
+        std::scoped_lock lock{LockCaches()};
         texture_cache.CreateChannel(channel);
         buffer_cache.CreateChannel(channel);
     }
@@ -1498,7 +1516,7 @@ void RasterizerVulkan::BindChannel(Tegra::Control::ChannelState& channel) {
     const s32 channel_id = channel.bind_id;
     BindToChannel(channel_id);
     {
-        std::scoped_lock lock{buffer_cache.mutex, texture_cache.mutex};
+        std::scoped_lock lock{LockCaches()};
         texture_cache.BindToChannel(channel_id);
         buffer_cache.BindToChannel(channel_id);
     }
@@ -1511,7 +1529,7 @@ void RasterizerVulkan::BindChannel(Tegra::Control::ChannelState& channel) {
 void RasterizerVulkan::ReleaseChannel(s32 channel_id) {
     EraseChannel(channel_id);
     {
-        std::scoped_lock lock{buffer_cache.mutex, texture_cache.mutex};
+        std::scoped_lock lock{LockCaches()};
         texture_cache.EraseChannel(channel_id);
         buffer_cache.EraseChannel(channel_id);
     }
diff --git a/src/video_core/renderer_vulkan/vk_rasterizer.h b/src/video_core/renderer_vulkan/vk_rasterizer.h
index ad069556c..ce3dfbaab 100644
--- a/src/video_core/renderer_vulkan/vk_rasterizer.h
+++ b/src/video_core/renderer_vulkan/vk_rasterizer.h
@@ -133,6 +133,10 @@ public:
 
     void ReleaseChannel(s32 channel_id) override;
 
+    std::scoped_lock<std::recursive_mutex, std::recursive_mutex> LockCaches() {
+        return std::scoped_lock{buffer_cache.mutex, texture_cache.mutex};
+    }
+
 private:
     static constexpr size_t MAX_TEXTURES = 192;
     static constexpr size_t MAX_IMAGES = 48;
diff --git a/src/video_core/renderer_vulkan/vk_render_pass_cache.cpp b/src/video_core/renderer_vulkan/vk_render_pass_cache.cpp
index ae9f1de64..7746a88d3 100644
--- a/src/video_core/renderer_vulkan/vk_render_pass_cache.cpp
+++ b/src/video_core/renderer_vulkan/vk_render_pass_cache.cpp
@@ -19,7 +19,7 @@ VkAttachmentDescription AttachmentDescription(const Device& device, PixelFormat
                                               VkSampleCountFlagBits samples) {
     using MaxwellToVK::SurfaceFormat;
     return {
-        .flags = VK_ATTACHMENT_DESCRIPTION_MAY_ALIAS_BIT,
+        .flags = {},
         .format = SurfaceFormat(device, FormatType::Optimal, true, format).format,
         .samples = samples,
         .loadOp = VK_ATTACHMENT_LOAD_OP_LOAD,
diff --git a/src/video_core/renderer_vulkan/vk_staging_buffer_pool.cpp b/src/video_core/renderer_vulkan/vk_staging_buffer_pool.cpp
index ce92f66ab..b278614e6 100644
--- a/src/video_core/renderer_vulkan/vk_staging_buffer_pool.cpp
+++ b/src/video_core/renderer_vulkan/vk_staging_buffer_pool.cpp
@@ -24,25 +24,38 @@ using namespace Common::Literals;
 
 // Maximum potential alignment of a Vulkan buffer
 constexpr VkDeviceSize MAX_ALIGNMENT = 256;
-// Maximum size to put elements in the stream buffer
-constexpr VkDeviceSize MAX_STREAM_BUFFER_REQUEST_SIZE = 8_MiB;
 // Stream buffer size in bytes
-constexpr VkDeviceSize STREAM_BUFFER_SIZE = 128_MiB;
-constexpr VkDeviceSize REGION_SIZE = STREAM_BUFFER_SIZE / StagingBufferPool::NUM_SYNCS;
+constexpr VkDeviceSize MAX_STREAM_BUFFER_SIZE = 128_MiB;
 
-size_t Region(size_t iterator) noexcept {
-    return iterator / REGION_SIZE;
+size_t GetStreamBufferSize(const Device& device) {
+    VkDeviceSize size{0};
+    if (device.HasDebuggingToolAttached()) {
+        ForEachDeviceLocalHostVisibleHeap(device, [&size](size_t index, VkMemoryHeap& heap) {
+            size = std::max(size, heap.size);
+        });
+        // If rebar is not supported, cut the max heap size to 40%. This will allow 2 captures to be
+        // loaded at the same time in RenderDoc. If rebar is supported, this shouldn't be an issue
+        // as the heap will be much larger.
+        if (size <= 256_MiB) {
+            size = size * 40 / 100;
+        }
+    } else {
+        size = MAX_STREAM_BUFFER_SIZE;
+    }
+    return std::min(Common::AlignUp(size, MAX_ALIGNMENT), MAX_STREAM_BUFFER_SIZE);
 }
 } // Anonymous namespace
 
 StagingBufferPool::StagingBufferPool(const Device& device_, MemoryAllocator& memory_allocator_,
                                      Scheduler& scheduler_)
-    : device{device_}, memory_allocator{memory_allocator_}, scheduler{scheduler_} {
+    : device{device_}, memory_allocator{memory_allocator_}, scheduler{scheduler_},
+      stream_buffer_size{GetStreamBufferSize(device)}, region_size{stream_buffer_size /
+                                                                   StagingBufferPool::NUM_SYNCS} {
     VkBufferCreateInfo stream_ci = {
         .sType = VK_STRUCTURE_TYPE_BUFFER_CREATE_INFO,
         .pNext = nullptr,
         .flags = 0,
-        .size = STREAM_BUFFER_SIZE,
+        .size = stream_buffer_size,
         .usage = VK_BUFFER_USAGE_TRANSFER_SRC_BIT | VK_BUFFER_USAGE_UNIFORM_BUFFER_BIT |
                  VK_BUFFER_USAGE_INDEX_BUFFER_BIT | VK_BUFFER_USAGE_STORAGE_BUFFER_BIT,
         .sharingMode = VK_SHARING_MODE_EXCLUSIVE,
@@ -63,7 +76,7 @@ StagingBufferPool::StagingBufferPool(const Device& device_, MemoryAllocator& mem
 StagingBufferPool::~StagingBufferPool() = default;
 
 StagingBufferRef StagingBufferPool::Request(size_t size, MemoryUsage usage, bool deferred) {
-    if (!deferred && usage == MemoryUsage::Upload && size <= MAX_STREAM_BUFFER_REQUEST_SIZE) {
+    if (!deferred && usage == MemoryUsage::Upload && size <= region_size) {
         return GetStreamBuffer(size);
     }
     return GetStagingBuffer(size, usage, deferred);
@@ -101,7 +114,7 @@ StagingBufferRef StagingBufferPool::GetStreamBuffer(size_t size) {
     used_iterator = iterator;
     free_iterator = std::max(free_iterator, iterator + size);
 
-    if (iterator + size >= STREAM_BUFFER_SIZE) {
+    if (iterator + size >= stream_buffer_size) {
         std::fill(sync_ticks.begin() + Region(used_iterator), sync_ticks.begin() + NUM_SYNCS,
                   current_tick);
         used_iterator = 0;
diff --git a/src/video_core/renderer_vulkan/vk_staging_buffer_pool.h b/src/video_core/renderer_vulkan/vk_staging_buffer_pool.h
index 5f69f08b1..d3deb9072 100644
--- a/src/video_core/renderer_vulkan/vk_staging_buffer_pool.h
+++ b/src/video_core/renderer_vulkan/vk_staging_buffer_pool.h
@@ -90,6 +90,9 @@ private:
     void ReleaseCache(MemoryUsage usage);
 
     void ReleaseLevel(StagingBuffersCache& cache, size_t log2);
+    size_t Region(size_t iter) const noexcept {
+        return iter / region_size;
+    }
 
     const Device& device;
     MemoryAllocator& memory_allocator;
@@ -97,6 +100,8 @@ private:
 
     vk::Buffer stream_buffer;
     std::span<u8> stream_pointer;
+    VkDeviceSize stream_buffer_size;
+    VkDeviceSize region_size;
 
     size_t iterator = 0;
     size_t used_iterator = 0;
diff --git a/src/video_core/renderer_vulkan/vk_state_tracker.cpp b/src/video_core/renderer_vulkan/vk_state_tracker.cpp
index d56558a83..daaea2979 100644
--- a/src/video_core/renderer_vulkan/vk_state_tracker.cpp
+++ b/src/video_core/renderer_vulkan/vk_state_tracker.cpp
@@ -190,7 +190,7 @@ void SetupDirtySpecialOps(Tables& tables) {
 void SetupDirtyViewportSwizzles(Tables& tables) {
     static constexpr size_t swizzle_offset = 6;
     for (size_t index = 0; index < Regs::NumViewports; ++index) {
-        tables[0][OFF(viewport_transform) + index * NUM(viewport_transform[0]) + swizzle_offset] =
+        tables[1][OFF(viewport_transform) + index * NUM(viewport_transform[0]) + swizzle_offset] =
             ViewportSwizzles;
     }
 }
diff --git a/src/video_core/renderer_vulkan/vk_texture_cache.cpp b/src/video_core/renderer_vulkan/vk_texture_cache.cpp
index 71fdec809..de34f6d49 100644
--- a/src/video_core/renderer_vulkan/vk_texture_cache.cpp
+++ b/src/video_core/renderer_vulkan/vk_texture_cache.cpp
@@ -238,6 +238,7 @@ constexpr VkBorderColor ConvertBorderColor(const std::array<float, 4>& color) {
         return any_r ? VK_IMAGE_ASPECT_STENCIL_BIT : VK_IMAGE_ASPECT_DEPTH_BIT;
     case PixelFormat::D16_UNORM:
     case PixelFormat::D32_FLOAT:
+    case PixelFormat::X8_D24_UNORM:
         return VK_IMAGE_ASPECT_DEPTH_BIT;
     case PixelFormat::S8_UINT:
         return VK_IMAGE_ASPECT_STENCIL_BIT;
@@ -1193,6 +1194,11 @@ void TextureCacheRuntime::ConvertImage(Framebuffer* dst, ImageView& dst_view, Im
             return blit_image_helper.ConvertD16ToR16(dst, src_view);
         }
         break;
+    case PixelFormat::A8B8G8R8_SRGB:
+        if (src_view.format == PixelFormat::D32_FLOAT) {
+            return blit_image_helper.ConvertD32FToABGR8(dst, src_view);
+        }
+        break;
     case PixelFormat::A8B8G8R8_UNORM:
         if (src_view.format == PixelFormat::S8_UINT_D24_UNORM) {
             return blit_image_helper.ConvertD24S8ToABGR8(dst, src_view);
@@ -1200,6 +1206,19 @@ void TextureCacheRuntime::ConvertImage(Framebuffer* dst, ImageView& dst_view, Im
         if (src_view.format == PixelFormat::D24_UNORM_S8_UINT) {
             return blit_image_helper.ConvertS8D24ToABGR8(dst, src_view);
         }
+        if (src_view.format == PixelFormat::D32_FLOAT) {
+            return blit_image_helper.ConvertD32FToABGR8(dst, src_view);
+        }
+        break;
+    case PixelFormat::B8G8R8A8_SRGB:
+        if (src_view.format == PixelFormat::D32_FLOAT) {
+            return blit_image_helper.ConvertD32FToABGR8(dst, src_view);
+        }
+        break;
+    case PixelFormat::B8G8R8A8_UNORM:
+        if (src_view.format == PixelFormat::D32_FLOAT) {
+            return blit_image_helper.ConvertD32FToABGR8(dst, src_view);
+        }
         break;
     case PixelFormat::R32_FLOAT:
         if (src_view.format == PixelFormat::D32_FLOAT) {
@@ -1218,6 +1237,12 @@ void TextureCacheRuntime::ConvertImage(Framebuffer* dst, ImageView& dst_view, Im
         }
         break;
     case PixelFormat::D32_FLOAT:
+        if (src_view.format == PixelFormat::A8B8G8R8_UNORM ||
+            src_view.format == PixelFormat::B8G8R8A8_UNORM ||
+            src_view.format == PixelFormat::A8B8G8R8_SRGB ||
+            src_view.format == PixelFormat::B8G8R8A8_SRGB) {
+            return blit_image_helper.ConvertABGR8ToD32F(dst, src_view);
+        }
         if (src_view.format == PixelFormat::R32_FLOAT) {
             return blit_image_helper.ConvertR32ToD32(dst, src_view);
         }
@@ -1526,15 +1551,15 @@ bool Image::IsRescaled() const noexcept {
 }
 
 bool Image::ScaleUp(bool ignore) {
+    const auto& resolution = runtime->resolution;
+    if (!resolution.active) {
+        return false;
+    }
     if (True(flags & ImageFlagBits::Rescaled)) {
         return false;
     }
     ASSERT(info.type != ImageType::Linear);
     flags |= ImageFlagBits::Rescaled;
-    const auto& resolution = runtime->resolution;
-    if (!resolution.active) {
-        return false;
-    }
     has_scaled = true;
     if (!scaled_image) {
         const bool is_2d = info.type == ImageType::e2D;
@@ -1563,15 +1588,15 @@ bool Image::ScaleUp(bool ignore) {
 }
 
 bool Image::ScaleDown(bool ignore) {
+    const auto& resolution = runtime->resolution;
+    if (!resolution.active) {
+        return false;
+    }
     if (False(flags & ImageFlagBits::Rescaled)) {
         return false;
     }
     ASSERT(info.type != ImageType::Linear);
     flags &= ~ImageFlagBits::Rescaled;
-    const auto& resolution = runtime->resolution;
-    if (!resolution.active) {
-        return false;
-    }
     current_image = *original_image;
     if (ignore) {
         return true;
@@ -2009,4 +2034,32 @@ void TextureCacheRuntime::AccelerateImageUpload(
     ASSERT(false);
 }
 
+void TextureCacheRuntime::TransitionImageLayout(Image& image) {
+    if (!image.ExchangeInitialization()) {
+        VkImageMemoryBarrier barrier{
+            .sType = VK_STRUCTURE_TYPE_IMAGE_MEMORY_BARRIER,
+            .pNext = nullptr,
+            .srcAccessMask = VK_ACCESS_NONE,
+            .dstAccessMask = VK_ACCESS_MEMORY_READ_BIT | VK_ACCESS_MEMORY_WRITE_BIT,
+            .oldLayout = VK_IMAGE_LAYOUT_UNDEFINED,
+            .newLayout = VK_IMAGE_LAYOUT_GENERAL,
+            .srcQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED,
+            .dstQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED,
+            .image = image.Handle(),
+            .subresourceRange{
+                .aspectMask = image.AspectMask(),
+                .baseMipLevel = 0,
+                .levelCount = VK_REMAINING_MIP_LEVELS,
+                .baseArrayLayer = 0,
+                .layerCount = VK_REMAINING_ARRAY_LAYERS,
+            },
+        };
+        scheduler.RequestOutsideRenderPassOperationContext();
+        scheduler.Record([barrier](vk::CommandBuffer cmdbuf) {
+            cmdbuf.PipelineBarrier(VK_PIPELINE_STAGE_ALL_COMMANDS_BIT,
+                                   VK_PIPELINE_STAGE_ALL_COMMANDS_BIT, 0, barrier);
+        });
+    }
+}
+
 } // namespace Vulkan
diff --git a/src/video_core/renderer_vulkan/vk_texture_cache.h b/src/video_core/renderer_vulkan/vk_texture_cache.h
index d6c5a15cc..7a0807709 100644
--- a/src/video_core/renderer_vulkan/vk_texture_cache.h
+++ b/src/video_core/renderer_vulkan/vk_texture_cache.h
@@ -92,6 +92,8 @@ public:
 
     void InsertUploadMemoryBarrier() {}
 
+    void TransitionImageLayout(Image& image);
+
     bool HasBrokenTextureViewFormats() const noexcept {
         // No known Vulkan driver has broken image views
         return false;
diff --git a/src/video_core/surface.cpp b/src/video_core/surface.cpp
index e16cd5e73..5b3c7aa5a 100644
--- a/src/video_core/surface.cpp
+++ b/src/video_core/surface.cpp
@@ -85,6 +85,8 @@ PixelFormat PixelFormatFromDepthFormat(Tegra::DepthFormat format) {
         return PixelFormat::S8_UINT;
     case Tegra::DepthFormat::Z32_FLOAT_X24S8_UINT:
         return PixelFormat::D32_FLOAT_S8_UINT;
+    case Tegra::DepthFormat::X8Z24_UNORM:
+        return PixelFormat::X8_D24_UNORM;
     default:
         UNIMPLEMENTED_MSG("Unimplemented format={}", format);
         return PixelFormat::S8_UINT_D24_UNORM;
@@ -202,6 +204,7 @@ PixelFormat PixelFormatFromRenderTargetFormat(Tegra::RenderTargetFormat format)
 PixelFormat PixelFormatFromGPUPixelFormat(Service::android::PixelFormat format) {
     switch (format) {
     case Service::android::PixelFormat::Rgba8888:
+    case Service::android::PixelFormat::Rgbx8888:
         return PixelFormat::A8B8G8R8_UNORM;
     case Service::android::PixelFormat::Rgb565:
         return PixelFormat::R5G6B5_UNORM;
diff --git a/src/video_core/surface.h b/src/video_core/surface.h
index 9b9c4d9bc..a5e8e2f62 100644
--- a/src/video_core/surface.h
+++ b/src/video_core/surface.h
@@ -115,6 +115,7 @@ enum class PixelFormat {
     // Depth formats
     D32_FLOAT = MaxColorFormat,
     D16_UNORM,
+    X8_D24_UNORM,
 
     MaxDepthFormat,
 
@@ -251,6 +252,7 @@ constexpr std::array<u8, MaxPixelFormat> BLOCK_WIDTH_TABLE = {{
     1,  // E5B9G9R9_FLOAT
     1,  // D32_FLOAT
     1,  // D16_UNORM
+    1,  // X8_D24_UNORM
     1,  // S8_UINT
     1,  // D24_UNORM_S8_UINT
     1,  // S8_UINT_D24_UNORM
@@ -360,6 +362,7 @@ constexpr std::array<u8, MaxPixelFormat> BLOCK_HEIGHT_TABLE = {{
     1,  // E5B9G9R9_FLOAT
     1,  // D32_FLOAT
     1,  // D16_UNORM
+    1,  // X8_D24_UNORM
     1,  // S8_UINT
     1,  // D24_UNORM_S8_UINT
     1,  // S8_UINT_D24_UNORM
@@ -469,6 +472,7 @@ constexpr std::array<u8, MaxPixelFormat> BITS_PER_BLOCK_TABLE = {{
     32,  // E5B9G9R9_FLOAT
     32,  // D32_FLOAT
     16,  // D16_UNORM
+    32,  // X8_D24_UNORM
     8,   // S8_UINT
     32,  // D24_UNORM_S8_UINT
     32,  // S8_UINT_D24_UNORM
diff --git a/src/video_core/texture_cache/format_lookup_table.cpp b/src/video_core/texture_cache/format_lookup_table.cpp
index 56307d030..8c774f512 100644
--- a/src/video_core/texture_cache/format_lookup_table.cpp
+++ b/src/video_core/texture_cache/format_lookup_table.cpp
@@ -138,10 +138,16 @@ PixelFormat PixelFormatFromTextureInfo(TextureFormat format, ComponentType red,
         return PixelFormat::E5B9G9R9_FLOAT;
     case Hash(TextureFormat::Z32, FLOAT):
         return PixelFormat::D32_FLOAT;
+    case Hash(TextureFormat::Z32, FLOAT, UINT, UINT, UINT, LINEAR):
+        return PixelFormat::D32_FLOAT;
     case Hash(TextureFormat::Z16, UNORM):
         return PixelFormat::D16_UNORM;
     case Hash(TextureFormat::Z16, UNORM, UINT, UINT, UINT, LINEAR):
         return PixelFormat::D16_UNORM;
+    case Hash(TextureFormat::X8Z24, UNORM):
+        return PixelFormat::X8_D24_UNORM;
+    case Hash(TextureFormat::X8Z24, UNORM, UINT, UINT, UINT, LINEAR):
+        return PixelFormat::X8_D24_UNORM;
     case Hash(TextureFormat::Z24S8, UINT, UNORM, UNORM, UNORM, LINEAR):
         return PixelFormat::S8_UINT_D24_UNORM;
     case Hash(TextureFormat::Z24S8, UINT, UNORM, UINT, UINT, LINEAR):
diff --git a/src/video_core/texture_cache/formatter.cpp b/src/video_core/texture_cache/formatter.cpp
index 6279d8e9e..2b7e0df72 100644
--- a/src/video_core/texture_cache/formatter.cpp
+++ b/src/video_core/texture_cache/formatter.cpp
@@ -10,19 +10,23 @@
 #include "video_core/texture_cache/image_info.h"
 #include "video_core/texture_cache/image_view_base.h"
 #include "video_core/texture_cache/render_targets.h"
+#include "video_core/texture_cache/samples_helper.h"
 
 namespace VideoCommon {
 
 std::string Name(const ImageBase& image) {
     const GPUVAddr gpu_addr = image.gpu_addr;
     const ImageInfo& info = image.info;
-    const u32 width = info.size.width;
-    const u32 height = info.size.height;
+    u32 width = info.size.width;
+    u32 height = info.size.height;
     const u32 depth = info.size.depth;
     const u32 num_layers = image.info.resources.layers;
     const u32 num_levels = image.info.resources.levels;
     std::string resource;
     if (image.info.num_samples > 1) {
+        const auto [samples_x, samples_y] = VideoCommon::SamplesLog2(image.info.num_samples);
+        width >>= samples_x;
+        height >>= samples_y;
         resource += fmt::format(":{}xMSAA", image.info.num_samples);
     }
     if (num_layers > 1) {
diff --git a/src/video_core/texture_cache/formatter.h b/src/video_core/texture_cache/formatter.h
index 9ee57a076..cabbfcb2d 100644
--- a/src/video_core/texture_cache/formatter.h
+++ b/src/video_core/texture_cache/formatter.h
@@ -211,6 +211,8 @@ struct fmt::formatter<VideoCore::Surface::PixelFormat> : fmt::formatter<fmt::str
                 return "D32_FLOAT";
             case PixelFormat::D16_UNORM:
                 return "D16_UNORM";
+            case PixelFormat::X8_D24_UNORM:
+                return "X8_D24_UNORM";
             case PixelFormat::S8_UINT:
                 return "S8_UINT";
             case PixelFormat::D24_UNORM_S8_UINT:
diff --git a/src/video_core/texture_cache/image_view_base.cpp b/src/video_core/texture_cache/image_view_base.cpp
index 0c5f4450d..18b9250f9 100644
--- a/src/video_core/texture_cache/image_view_base.cpp
+++ b/src/video_core/texture_cache/image_view_base.cpp
@@ -85,6 +85,7 @@ bool ImageViewBase::SupportsAnisotropy() const noexcept {
     // Depth formats
     case PixelFormat::D32_FLOAT:
     case PixelFormat::D16_UNORM:
+    case PixelFormat::X8_D24_UNORM:
     // Stencil formats
     case PixelFormat::S8_UINT:
     // DepthStencil formats
diff --git a/src/video_core/texture_cache/samples_helper.h b/src/video_core/texture_cache/samples_helper.h
index 203ac1b11..2ee2f8312 100644
--- a/src/video_core/texture_cache/samples_helper.h
+++ b/src/video_core/texture_cache/samples_helper.h
@@ -24,7 +24,7 @@ namespace VideoCommon {
         return {2, 2};
     }
     ASSERT_MSG(false, "Invalid number of samples={}", num_samples);
-    return {1, 1};
+    return {0, 0};
 }
 
 [[nodiscard]] inline int NumSamples(Tegra::Texture::MsaaMode msaa_mode) {
diff --git a/src/video_core/texture_cache/texture_cache.h b/src/video_core/texture_cache/texture_cache.h
index 1bdb0def5..d575c57ca 100644
--- a/src/video_core/texture_cache/texture_cache.h
+++ b/src/video_core/texture_cache/texture_cache.h
@@ -1016,6 +1016,7 @@ void TextureCache<P>::RefreshContents(Image& image, ImageId image_id) {
 
     if (image.info.num_samples > 1 && !runtime.CanUploadMSAA()) {
         LOG_WARNING(HW_GPU, "MSAA image uploads are not implemented");
+        runtime.TransitionImageLayout(image);
         return;
     }
     if (True(image.flags & ImageFlagBits::AsynchronousDecode)) {
diff --git a/src/video_core/texture_cache/util.cpp b/src/video_core/texture_cache/util.cpp
index 0a86ce139..15596c925 100644
--- a/src/video_core/texture_cache/util.cpp
+++ b/src/video_core/texture_cache/util.cpp
@@ -68,6 +68,7 @@ struct LevelInfo {
     Extent2D tile_size;
     u32 bpp_log2;
     u32 tile_width_spacing;
+    u32 num_levels;
 };
 
 [[nodiscard]] constexpr u32 AdjustTileSize(u32 shift, u32 unit_factor, u32 dimension) {
@@ -118,11 +119,11 @@ template <u32 GOB_EXTENT>
 }
 
 [[nodiscard]] constexpr Extent3D AdjustMipBlockSize(Extent3D num_tiles, Extent3D block_size,
-                                                    u32 level) {
+                                                    u32 level, u32 num_levels) {
     return {
         .width = AdjustMipBlockSize<GOB_SIZE_X>(num_tiles.width, block_size.width, level),
         .height = AdjustMipBlockSize<GOB_SIZE_Y>(num_tiles.height, block_size.height, level),
-        .depth = level == 0
+        .depth = level == 0 && num_levels == 1
                      ? block_size.depth
                      : AdjustMipBlockSize<GOB_SIZE_Z>(num_tiles.depth, block_size.depth, level),
     };
@@ -166,7 +167,7 @@ template <u32 GOB_EXTENT>
 }
 
 [[nodiscard]] constexpr Extent3D TileShift(const LevelInfo& info, u32 level) {
-    if (level == 0) {
+    if (level == 0 && info.num_levels == 1) {
         return Extent3D{
             .width = info.block.width,
             .height = info.block.height,
@@ -257,7 +258,7 @@ template <u32 GOB_EXTENT>
 }
 
 [[nodiscard]] constexpr LevelInfo MakeLevelInfo(PixelFormat format, Extent3D size, Extent3D block,
-                                                u32 tile_width_spacing) {
+                                                u32 tile_width_spacing, u32 num_levels) {
     const u32 bytes_per_block = BytesPerBlock(format);
     return {
         .size =
@@ -270,16 +271,18 @@ template <u32 GOB_EXTENT>
         .tile_size = DefaultBlockSize(format),
         .bpp_log2 = BytesPerBlockLog2(bytes_per_block),
         .tile_width_spacing = tile_width_spacing,
+        .num_levels = num_levels,
     };
 }
 
 [[nodiscard]] constexpr LevelInfo MakeLevelInfo(const ImageInfo& info) {
-    return MakeLevelInfo(info.format, info.size, info.block, info.tile_width_spacing);
+    return MakeLevelInfo(info.format, info.size, info.block, info.tile_width_spacing,
+                         info.resources.levels);
 }
 
 [[nodiscard]] constexpr u32 CalculateLevelOffset(PixelFormat format, Extent3D size, Extent3D block,
                                                  u32 tile_width_spacing, u32 level) {
-    const LevelInfo info = MakeLevelInfo(format, size, block, tile_width_spacing);
+    const LevelInfo info = MakeLevelInfo(format, size, block, tile_width_spacing, level);
     u32 offset = 0;
     for (u32 current_level = 0; current_level < level; ++current_level) {
         offset += CalculateLevelSize(info, current_level);
@@ -466,7 +469,7 @@ template <u32 GOB_EXTENT>
     };
     const u32 bpp_log2 = BytesPerBlockLog2(info.format);
     const u32 alignment = StrideAlignment(num_tiles, info.block, bpp_log2, info.tile_width_spacing);
-    const Extent3D mip_block = AdjustMipBlockSize(num_tiles, info.block, 0);
+    const Extent3D mip_block = AdjustMipBlockSize(num_tiles, info.block, 0, info.resources.levels);
     return Extent3D{
         .width = Common::AlignUpLog2(num_tiles.width, alignment),
         .height = Common::AlignUpLog2(num_tiles.height, GOB_SIZE_Y_SHIFT + mip_block.height),
@@ -533,7 +536,8 @@ void SwizzleBlockLinearImage(Tegra::MemoryManager& gpu_memory, GPUVAddr gpu_addr
     UNIMPLEMENTED_IF(copy.image_extent != level_size);
 
     const Extent3D num_tiles = AdjustTileSize(level_size, tile_size);
-    const Extent3D block = AdjustMipBlockSize(num_tiles, level_info.block, level);
+    const Extent3D block =
+        AdjustMipBlockSize(num_tiles, level_info.block, level, level_info.num_levels);
 
     size_t host_offset = copy.buffer_offset;
 
@@ -698,7 +702,7 @@ u32 CalculateLevelStrideAlignment(const ImageInfo& info, u32 level) {
     const Extent2D tile_size = DefaultBlockSize(info.format);
     const Extent3D level_size = AdjustMipSize(info.size, level);
     const Extent3D num_tiles = AdjustTileSize(level_size, tile_size);
-    const Extent3D block = AdjustMipBlockSize(num_tiles, info.block, level);
+    const Extent3D block = AdjustMipBlockSize(num_tiles, info.block, level, info.resources.levels);
     const u32 bpp_log2 = BytesPerBlockLog2(info.format);
     return StrideAlignment(num_tiles, block, bpp_log2, info.tile_width_spacing);
 }
@@ -887,7 +891,8 @@ boost::container::small_vector<BufferImageCopy, 16> UnswizzleImage(Tegra::Memory
             .image_extent = level_size,
         };
         const Extent3D num_tiles = AdjustTileSize(level_size, tile_size);
-        const Extent3D block = AdjustMipBlockSize(num_tiles, level_info.block, level);
+        const Extent3D block =
+            AdjustMipBlockSize(num_tiles, level_info.block, level, level_info.num_levels);
         const u32 stride_alignment = StrideAlignment(num_tiles, info.block, gob, bpp_log2);
         size_t guest_layer_offset = 0;
 
@@ -1041,7 +1046,7 @@ Extent3D MipBlockSize(const ImageInfo& info, u32 level) {
     const Extent2D tile_size = DefaultBlockSize(info.format);
     const Extent3D level_size = AdjustMipSize(info.size, level);
     const Extent3D num_tiles = AdjustTileSize(level_size, tile_size);
-    return AdjustMipBlockSize(num_tiles, level_info.block, level);
+    return AdjustMipBlockSize(num_tiles, level_info.block, level, level_info.num_levels);
 }
 
 boost::container::small_vector<SwizzleParameters, 16> FullUploadSwizzles(const ImageInfo& info) {
@@ -1063,7 +1068,8 @@ boost::container::small_vector<SwizzleParameters, 16> FullUploadSwizzles(const I
     for (s32 level = 0; level < num_levels; ++level) {
         const Extent3D level_size = AdjustMipSize(size, level);
         const Extent3D num_tiles = AdjustTileSize(level_size, tile_size);
-        const Extent3D block = AdjustMipBlockSize(num_tiles, level_info.block, level);
+        const Extent3D block =
+            AdjustMipBlockSize(num_tiles, level_info.block, level, level_info.num_levels);
         params[level] = SwizzleParameters{
             .num_tiles = num_tiles,
             .block = block,
@@ -1292,11 +1298,11 @@ u32 MapSizeBytes(const ImageBase& image) {
     }
 }
 
-static_assert(CalculateLevelSize(LevelInfo{{1920, 1080, 1}, {0, 2, 0}, {1, 1}, 2, 0}, 0) ==
+static_assert(CalculateLevelSize(LevelInfo{{1920, 1080, 1}, {0, 2, 0}, {1, 1}, 2, 0, 1}, 0) ==
               0x7f8000);
-static_assert(CalculateLevelSize(LevelInfo{{32, 32, 1}, {0, 0, 4}, {1, 1}, 4, 0}, 0) == 0x40000);
+static_assert(CalculateLevelSize(LevelInfo{{32, 32, 1}, {0, 0, 4}, {1, 1}, 4, 0, 1}, 0) == 0x40000);
 
-static_assert(CalculateLevelSize(LevelInfo{{128, 8, 1}, {0, 4, 0}, {1, 1}, 4, 0}, 0) == 0x40000);
+static_assert(CalculateLevelSize(LevelInfo{{128, 8, 1}, {0, 4, 0}, {1, 1}, 4, 0, 1}, 0) == 0x40000);
 
 static_assert(CalculateLevelOffset(PixelFormat::R8_SINT, {1920, 1080, 1}, {0, 2, 0}, 0, 7) ==
               0x2afc00);
diff --git a/src/video_core/vulkan_common/vulkan_device.cpp b/src/video_core/vulkan_common/vulkan_device.cpp
index 3960b135a..e518756d2 100644
--- a/src/video_core/vulkan_common/vulkan_device.cpp
+++ b/src/video_core/vulkan_common/vulkan_device.cpp
@@ -83,12 +83,6 @@ constexpr std::array VK_FORMAT_A4B4G4R4_UNORM_PACK16{
 
 } // namespace Alternatives
 
-enum class NvidiaArchitecture {
-    AmpereOrNewer,
-    Turing,
-    VoltaOrOlder,
-};
-
 template <typename T>
 void SetNext(void**& next, T& data) {
     *next = &data;
@@ -200,6 +194,7 @@ std::unordered_map<VkFormat, VkFormatProperties> GetFormatProperties(vk::Physica
         VK_FORMAT_BC7_UNORM_BLOCK,
         VK_FORMAT_D16_UNORM,
         VK_FORMAT_D16_UNORM_S8_UINT,
+        VK_FORMAT_X8_D24_UNORM_PACK32,
         VK_FORMAT_D24_UNORM_S8_UINT,
         VK_FORMAT_D32_SFLOAT,
         VK_FORMAT_D32_SFLOAT_S8_UINT,
@@ -321,13 +316,38 @@ NvidiaArchitecture GetNvidiaArchitecture(vk::PhysicalDevice physical,
         physical.GetProperties2(physical_properties);
         if (shading_rate_props.primitiveFragmentShadingRateWithMultipleViewports) {
             // Only Ampere and newer support this feature
-            return NvidiaArchitecture::AmpereOrNewer;
+            // TODO: Find a way to differentiate Ampere and Ada
+            return NvidiaArchitecture::Arch_AmpereOrNewer;
         }
+        return NvidiaArchitecture::Arch_Turing;
     }
-    if (exts.contains(VK_NV_SHADING_RATE_IMAGE_EXTENSION_NAME)) {
-        return NvidiaArchitecture::Turing;
+
+    if (exts.contains(VK_EXT_BLEND_OPERATION_ADVANCED_EXTENSION_NAME)) {
+        VkPhysicalDeviceBlendOperationAdvancedPropertiesEXT advanced_blending_props{};
+        advanced_blending_props.sType =
+            VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_BLEND_OPERATION_ADVANCED_PROPERTIES_EXT;
+        VkPhysicalDeviceProperties2 physical_properties{};
+        physical_properties.sType = VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_PROPERTIES_2;
+        physical_properties.pNext = &advanced_blending_props;
+        physical.GetProperties2(physical_properties);
+        if (advanced_blending_props.advancedBlendMaxColorAttachments == 1) {
+            return NvidiaArchitecture::Arch_Maxwell;
+        }
+
+        if (exts.contains(VK_EXT_CONSERVATIVE_RASTERIZATION_EXTENSION_NAME)) {
+            VkPhysicalDeviceConservativeRasterizationPropertiesEXT conservative_raster_props{};
+            conservative_raster_props.sType =
+                VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_CONSERVATIVE_RASTERIZATION_PROPERTIES_EXT;
+            physical_properties.pNext = &conservative_raster_props;
+            physical.GetProperties2(physical_properties);
+            if (conservative_raster_props.degenerateLinesRasterized) {
+                return NvidiaArchitecture::Arch_Volta;
+            }
+            return NvidiaArchitecture::Arch_Pascal;
+        }
     }
-    return NvidiaArchitecture::VoltaOrOlder;
+
+    return NvidiaArchitecture::Arch_KeplerOrOlder;
 }
 
 std::vector<const char*> ExtensionListForVulkan(
@@ -407,6 +427,10 @@ Device::Device(VkInstance instance_, vk::PhysicalDevice physical_, VkSurfaceKHR
         throw vk::Exception(VK_ERROR_INCOMPATIBLE_DRIVER);
     }
 
+    if (is_nvidia) {
+        nvidia_arch = GetNvidiaArchitecture(physical, supported_extensions);
+    }
+
     SetupFamilies(surface);
     const auto queue_cis = GetDeviceQueueCreateInfos();
 
@@ -503,20 +527,15 @@ Device::Device(VkInstance instance_, vk::PhysicalDevice physical_, VkSurfaceKHR
 
     if (is_nvidia) {
         const u32 nv_major_version = (properties.properties.driverVersion >> 22) & 0x3ff;
-        const auto arch = GetNvidiaArchitecture(physical, supported_extensions);
-        switch (arch) {
-        case NvidiaArchitecture::AmpereOrNewer:
+        const auto arch = GetNvidiaArch();
+        if (arch >= NvidiaArchitecture::Arch_AmpereOrNewer) {
             LOG_WARNING(Render_Vulkan, "Ampere and newer have broken float16 math");
             features.shader_float16_int8.shaderFloat16 = false;
-            break;
-        case NvidiaArchitecture::Turing:
-            break;
-        case NvidiaArchitecture::VoltaOrOlder:
+        } else if (arch <= NvidiaArchitecture::Arch_Volta) {
             if (nv_major_version < 527) {
                 LOG_WARNING(Render_Vulkan, "Volta and older have broken VK_KHR_push_descriptor");
                 RemoveExtension(extensions.push_descriptor, VK_KHR_PUSH_DESCRIPTOR_EXTENSION_NAME);
             }
-            break;
         }
         if (nv_major_version >= 510) {
             LOG_WARNING(Render_Vulkan, "NVIDIA Drivers >= 510 do not support MSAA image blits");
@@ -661,7 +680,15 @@ Device::Device(VkInstance instance_, vk::PhysicalDevice physical_, VkSurfaceKHR
                         "ANV drivers 22.3.0 to 23.1.0 have broken VK_KHR_push_descriptor");
             RemoveExtension(extensions.push_descriptor, VK_KHR_PUSH_DESCRIPTOR_EXTENSION_NAME);
         }
+    } else if (extensions.push_descriptor && is_nvidia) {
+        const auto arch = GetNvidiaArch();
+        if (arch <= NvidiaArchitecture::Arch_Pascal) {
+            LOG_WARNING(Render_Vulkan,
+                        "Pascal and older architectures have broken VK_KHR_push_descriptor");
+            RemoveExtension(extensions.push_descriptor, VK_KHR_PUSH_DESCRIPTOR_EXTENSION_NAME);
+        }
     }
+
     if (is_mvk) {
         LOG_WARNING(Render_Vulkan,
                     "MVK driver breaks when using more than 16 vertex attributes/bindings");
diff --git a/src/video_core/vulkan_common/vulkan_device.h b/src/video_core/vulkan_common/vulkan_device.h
index 282a2925d..b213ed7dd 100644
--- a/src/video_core/vulkan_common/vulkan_device.h
+++ b/src/video_core/vulkan_common/vulkan_device.h
@@ -177,6 +177,15 @@ enum class FormatType { Linear, Optimal, Buffer };
 /// Subgroup size of the guest emulated hardware (Nvidia has 32 threads per subgroup).
 const u32 GuestWarpSize = 32;
 
+enum class NvidiaArchitecture {
+    Arch_KeplerOrOlder,
+    Arch_Maxwell,
+    Arch_Pascal,
+    Arch_Volta,
+    Arch_Turing,
+    Arch_AmpereOrNewer,
+};
+
 /// Handles data specific to a physical device.
 class Device {
 public:
@@ -670,6 +679,14 @@ public:
         return false;
     }
 
+    bool IsNvidia() const noexcept {
+        return properties.driver.driverID == VK_DRIVER_ID_NVIDIA_PROPRIETARY;
+    }
+
+    NvidiaArchitecture GetNvidiaArch() const noexcept {
+        return nvidia_arch;
+    }
+
 private:
     /// Checks if the physical device is suitable and configures the object state
     /// with all necessary info about its properties.
@@ -788,6 +805,7 @@ private:
     bool supports_conditional_barriers{};      ///< Allows barriers in conditional control flow.
     u64 device_access_memory{};                ///< Total size of device local memory in bytes.
     u32 sets_per_pool{};                       ///< Sets per Description Pool
+    NvidiaArchitecture nvidia_arch{NvidiaArchitecture::Arch_AmpereOrNewer};
 
     // Telemetry parameters
     std::set<std::string, std::less<>> supported_extensions; ///< Reported Vulkan extensions.
diff --git a/src/video_core/vulkan_common/vulkan_memory_allocator.cpp b/src/video_core/vulkan_common/vulkan_memory_allocator.cpp
index 3ef381a38..8dd1667f3 100644
--- a/src/video_core/vulkan_common/vulkan_memory_allocator.cpp
+++ b/src/video_core/vulkan_common/vulkan_memory_allocator.cpp
@@ -9,6 +9,7 @@
 #include "common/alignment.h"
 #include "common/assert.h"
 #include "common/common_types.h"
+#include "common/literals.h"
 #include "common/logging/log.h"
 #include "common/polyfill_ranges.h"
 #include "video_core/vulkan_common/vma.h"
@@ -65,12 +66,12 @@ struct Range {
     switch (usage) {
     case MemoryUsage::Upload:
     case MemoryUsage::Stream:
-        return VMA_ALLOCATION_CREATE_HOST_ACCESS_SEQUENTIAL_WRITE_BIT;
+        return VMA_ALLOCATION_CREATE_MAPPED_BIT |
+               VMA_ALLOCATION_CREATE_HOST_ACCESS_SEQUENTIAL_WRITE_BIT;
     case MemoryUsage::Download:
-        return VMA_ALLOCATION_CREATE_HOST_ACCESS_RANDOM_BIT;
+        return VMA_ALLOCATION_CREATE_MAPPED_BIT | VMA_ALLOCATION_CREATE_HOST_ACCESS_RANDOM_BIT;
     case MemoryUsage::DeviceLocal:
-        return VMA_ALLOCATION_CREATE_HOST_ACCESS_SEQUENTIAL_WRITE_BIT |
-               VMA_ALLOCATION_CREATE_HOST_ACCESS_ALLOW_TRANSFER_INSTEAD_BIT;
+        return {};
     }
     return {};
 }
@@ -212,7 +213,20 @@ MemoryAllocator::MemoryAllocator(const Device& device_)
     : device{device_}, allocator{device.GetAllocator()},
       properties{device_.GetPhysical().GetMemoryProperties().memoryProperties},
       buffer_image_granularity{
-          device_.GetPhysical().GetProperties().limits.bufferImageGranularity} {}
+          device_.GetPhysical().GetProperties().limits.bufferImageGranularity} {
+    // GPUs not supporting rebar may only have a region with less than 256MB host visible/device
+    // local memory. In that case, opening 2 RenderDoc captures side-by-side is not possible due to
+    // the heap running out of memory. With RenderDoc attached and only a small host/device region,
+    // only allow the stream buffer in this memory heap.
+    if (device.HasDebuggingToolAttached()) {
+        using namespace Common::Literals;
+        ForEachDeviceLocalHostVisibleHeap(device, [this](size_t index, VkMemoryHeap& heap) {
+            if (heap.size <= 256_MiB) {
+                valid_memory_types &= ~(1u << index);
+            }
+        });
+    }
+}
 
 MemoryAllocator::~MemoryAllocator() = default;
 
@@ -239,12 +253,11 @@ vk::Image MemoryAllocator::CreateImage(const VkImageCreateInfo& ci) const {
 
 vk::Buffer MemoryAllocator::CreateBuffer(const VkBufferCreateInfo& ci, MemoryUsage usage) const {
     const VmaAllocationCreateInfo alloc_ci = {
-        .flags = VMA_ALLOCATION_CREATE_WITHIN_BUDGET_BIT | VMA_ALLOCATION_CREATE_MAPPED_BIT |
-                 MemoryUsageVmaFlags(usage),
+        .flags = VMA_ALLOCATION_CREATE_WITHIN_BUDGET_BIT | MemoryUsageVmaFlags(usage),
         .usage = MemoryUsageVma(usage),
         .requiredFlags = 0,
         .preferredFlags = MemoryUsagePreferedVmaFlags(usage),
-        .memoryTypeBits = 0,
+        .memoryTypeBits = usage == MemoryUsage::Stream ? 0u : valid_memory_types,
         .pool = VK_NULL_HANDLE,
         .pUserData = nullptr,
         .priority = 0.f,
diff --git a/src/video_core/vulkan_common/vulkan_memory_allocator.h b/src/video_core/vulkan_common/vulkan_memory_allocator.h
index f449bc8d0..38a182bcb 100644
--- a/src/video_core/vulkan_common/vulkan_memory_allocator.h
+++ b/src/video_core/vulkan_common/vulkan_memory_allocator.h
@@ -7,6 +7,7 @@
 #include <span>
 #include <vector>
 #include "common/common_types.h"
+#include "video_core/vulkan_common/vulkan_device.h"
 #include "video_core/vulkan_common/vulkan_wrapper.h"
 
 VK_DEFINE_HANDLE(VmaAllocator)
@@ -26,6 +27,18 @@ enum class MemoryUsage {
     Stream,      ///< Requests device local host visible buffer, falling back host memory.
 };
 
+template <typename F>
+void ForEachDeviceLocalHostVisibleHeap(const Device& device, F&& f) {
+    auto memory_props = device.GetPhysical().GetMemoryProperties().memoryProperties;
+    for (size_t i = 0; i < memory_props.memoryTypeCount; i++) {
+        auto& memory_type = memory_props.memoryTypes[i];
+        if ((memory_type.propertyFlags & VK_MEMORY_PROPERTY_DEVICE_LOCAL_BIT) &&
+            (memory_type.propertyFlags & VK_MEMORY_PROPERTY_HOST_VISIBLE_BIT)) {
+            f(memory_type.heapIndex, memory_props.memoryHeaps[memory_type.heapIndex]);
+        }
+    }
+}
+
 /// Ownership handle of a memory commitment.
 /// Points to a subregion of a memory allocation.
 class MemoryCommit {
@@ -124,6 +137,7 @@ private:
     std::vector<std::unique_ptr<MemoryAllocation>> allocations; ///< Current allocations.
     VkDeviceSize buffer_image_granularity; // The granularity for adjacent offsets between buffers
                                            // and optimal images
+    u32 valid_memory_types{~0u};
 };
 
 } // namespace Vulkan
diff --git a/src/video_core/vulkan_common/vulkan_wrapper.cpp b/src/video_core/vulkan_common/vulkan_wrapper.cpp
index 2f3254a97..70cf14afa 100644
--- a/src/video_core/vulkan_common/vulkan_wrapper.cpp
+++ b/src/video_core/vulkan_common/vulkan_wrapper.cpp
@@ -522,7 +522,7 @@ Instance Instance::Create(u32 version, Span<const char*> layers, Span<const char
         .applicationVersion = VK_MAKE_VERSION(0, 1, 0),
         .pEngineName = "yuzu Emulator",
         .engineVersion = VK_MAKE_VERSION(0, 1, 0),
-        .apiVersion = version,
+        .apiVersion = VK_API_VERSION_1_3,
     };
     const VkInstanceCreateInfo ci{
         .sType = VK_STRUCTURE_TYPE_INSTANCE_CREATE_INFO,
diff --git a/src/video_core/vulkan_common/vulkan_wrapper.h b/src/video_core/vulkan_common/vulkan_wrapper.h
index 1e3c0fa64..0487cd3b6 100644
--- a/src/video_core/vulkan_common/vulkan_wrapper.h
+++ b/src/video_core/vulkan_common/vulkan_wrapper.h
@@ -117,6 +117,9 @@ public:
     virtual ~Exception() = default;
 
     const char* what() const noexcept override;
+    VkResult GetResult() const noexcept {
+        return result;
+    }
 
 private:
     VkResult result;
diff --git a/src/yuzu/CMakeLists.txt b/src/yuzu/CMakeLists.txt
index d23c1e9d0..33e1fb663 100644
--- a/src/yuzu/CMakeLists.txt
+++ b/src/yuzu/CMakeLists.txt
@@ -195,6 +195,8 @@ add_executable(yuzu
     multiplayer/state.cpp
     multiplayer/state.h
     multiplayer/validation.h
+    play_time_manager.cpp
+    play_time_manager.h
     precompiled_headers.h
     qt_common.cpp
     qt_common.h
@@ -382,7 +384,7 @@ if (USE_DISCORD_PRESENCE)
         discord_impl.cpp
         discord_impl.h
     )
-    target_link_libraries(yuzu PRIVATE DiscordRPC::discord-rpc httplib::httplib)
+    target_link_libraries(yuzu PRIVATE DiscordRPC::discord-rpc httplib::httplib Qt${QT_MAJOR_VERSION}::Network)
     target_compile_definitions(yuzu PRIVATE -DUSE_DISCORD_PRESENCE)
 endif()
 
diff --git a/src/yuzu/applets/qt_controller.cpp b/src/yuzu/applets/qt_controller.cpp
index d15559518..515cb7ce6 100644
--- a/src/yuzu/applets/qt_controller.cpp
+++ b/src/yuzu/applets/qt_controller.cpp
@@ -23,6 +23,7 @@
 #include "yuzu/configuration/configure_vibration.h"
 #include "yuzu/configuration/input_profiles.h"
 #include "yuzu/main.h"
+#include "yuzu/util/controller_navigation.h"
 
 namespace {
 
@@ -132,6 +133,8 @@ QtControllerSelectorDialog::QtControllerSelectorDialog(
         ui->checkboxPlayer7Connected, ui->checkboxPlayer8Connected,
     };
 
+    ui->labelError->setVisible(false);
+
     // Setup/load everything prior to setting up connections.
     // This avoids unintentionally changing the states of elements while loading them in.
     SetSupportedControllers();
@@ -143,6 +146,8 @@ QtControllerSelectorDialog::QtControllerSelectorDialog(
 
     LoadConfiguration();
 
+    controller_navigation = new ControllerNavigation(system.HIDCore(), this);
+
     for (std::size_t i = 0; i < NUM_PLAYERS; ++i) {
         SetExplainText(i);
         UpdateControllerIcon(i);
@@ -150,16 +155,27 @@ QtControllerSelectorDialog::QtControllerSelectorDialog(
         UpdateBorderColor(i);
 
         connect(player_groupboxes[i], &QGroupBox::toggled, [this, i](bool checked) {
-            if (checked) {
-                for (std::size_t index = 0; index <= i; ++index) {
-                    connected_controller_checkboxes[index]->setChecked(checked);
-                }
-            } else {
-                for (std::size_t index = i; index < NUM_PLAYERS; ++index) {
-                    connected_controller_checkboxes[index]->setChecked(checked);
-                }
+            // Reconnect current controller if it was the last one checked
+            // (player number was reduced by more than one)
+            const bool reconnect_first = !checked && i < player_groupboxes.size() - 1 &&
+                                         player_groupboxes[i + 1]->isChecked();
+
+            // Ensures that connecting a controller changes the number of players
+            if (connected_controller_checkboxes[i]->isChecked() != checked) {
+                // Ensures that the players are always connected in sequential order
+                PropagatePlayerNumberChanged(i, checked, reconnect_first);
             }
         });
+        connect(connected_controller_checkboxes[i], &QCheckBox::clicked, [this, i](bool checked) {
+            // Reconnect current controller if it was the last one checked
+            // (player number was reduced by more than one)
+            const bool reconnect_first = !checked &&
+                                         i < connected_controller_checkboxes.size() - 1 &&
+                                         connected_controller_checkboxes[i + 1]->isChecked();
+
+            // Ensures that the players are always connected in sequential order
+            PropagatePlayerNumberChanged(i, checked, reconnect_first);
+        });
 
         connect(emulated_controllers[i], qOverload<int>(&QComboBox::currentIndexChanged),
                 [this, i](int) {
@@ -199,6 +215,12 @@ QtControllerSelectorDialog::QtControllerSelectorDialog(
     connect(ui->buttonBox, &QDialogButtonBox::accepted, this,
             &QtControllerSelectorDialog::ApplyConfiguration);
 
+    connect(controller_navigation, &ControllerNavigation::TriggerKeyboardEvent,
+            [this](Qt::Key key) {
+                QKeyEvent* event = new QKeyEvent(QEvent::KeyPress, key, Qt::NoModifier);
+                QCoreApplication::postEvent(this, event);
+            });
+
     // Enhancement: Check if the parameters have already been met before disconnecting controllers.
     // If all the parameters are met AND only allows a single player,
     // stop the constructor here as we do not need to continue.
@@ -217,6 +239,7 @@ QtControllerSelectorDialog::QtControllerSelectorDialog(
 }
 
 QtControllerSelectorDialog::~QtControllerSelectorDialog() {
+    controller_navigation->UnloadController();
     system.HIDCore().DisableAllControllerConfiguration();
 }
 
@@ -291,6 +314,31 @@ void QtControllerSelectorDialog::CallConfigureInputProfileDialog() {
     dialog.exec();
 }
 
+void QtControllerSelectorDialog::keyPressEvent(QKeyEvent* evt) {
+    const auto num_connected_players = static_cast<int>(
+        std::count_if(player_groupboxes.begin(), player_groupboxes.end(),
+                      [](const QGroupBox* player) { return player->isChecked(); }));
+
+    const auto min_supported_players = parameters.enable_single_mode ? 1 : parameters.min_players;
+    const auto max_supported_players = parameters.enable_single_mode ? 1 : parameters.max_players;
+
+    if ((evt->key() == Qt::Key_Enter || evt->key() == Qt::Key_Return) && !parameters_met) {
+        // Display error message when trying to validate using "Enter" and "OK" button is disabled
+        ui->labelError->setVisible(true);
+        return;
+    } else if (evt->key() == Qt::Key_Left && num_connected_players > min_supported_players) {
+        // Remove a player if possible
+        connected_controller_checkboxes[num_connected_players - 1]->setChecked(false);
+        return;
+    } else if (evt->key() == Qt::Key_Right && num_connected_players < max_supported_players) {
+        // Add a player, if possible
+        ui->labelError->setVisible(false);
+        connected_controller_checkboxes[num_connected_players]->setChecked(true);
+        return;
+    }
+    QDialog::keyPressEvent(evt);
+}
+
 bool QtControllerSelectorDialog::CheckIfParametersMet() {
     // Here, we check and validate the current configuration against all applicable parameters.
     const auto num_connected_players = static_cast<int>(
@@ -629,6 +677,29 @@ void QtControllerSelectorDialog::UpdateDockedState(bool is_handheld) {
     }
 }
 
+void QtControllerSelectorDialog::PropagatePlayerNumberChanged(size_t player_index, bool checked,
+                                                              bool reconnect_current) {
+    connected_controller_checkboxes[player_index]->setChecked(checked);
+    // Hide eventual error message about number of controllers
+    ui->labelError->setVisible(false);
+
+    if (checked) {
+        // Check all previous buttons when checked
+        if (player_index > 0) {
+            PropagatePlayerNumberChanged(player_index - 1, checked);
+        }
+    } else {
+        // Unchecked all following buttons when unchecked
+        if (player_index < connected_controller_checkboxes.size() - 1) {
+            PropagatePlayerNumberChanged(player_index + 1, checked);
+        }
+    }
+
+    if (reconnect_current) {
+        connected_controller_checkboxes[player_index]->setCheckState(Qt::Checked);
+    }
+}
+
 void QtControllerSelectorDialog::DisableUnsupportedPlayers() {
     const auto max_supported_players = parameters.enable_single_mode ? 1 : parameters.max_players;
 
diff --git a/src/yuzu/applets/qt_controller.h b/src/yuzu/applets/qt_controller.h
index 2fdc35857..e5372495d 100644
--- a/src/yuzu/applets/qt_controller.h
+++ b/src/yuzu/applets/qt_controller.h
@@ -34,6 +34,8 @@ class HIDCore;
 enum class NpadStyleIndex : u8;
 } // namespace Core::HID
 
+class ControllerNavigation;
+
 class QtControllerSelectorDialog final : public QDialog {
     Q_OBJECT
 
@@ -46,6 +48,8 @@ public:
 
     int exec() override;
 
+    void keyPressEvent(QKeyEvent* evt) override;
+
 private:
     // Applies the current configuration.
     void ApplyConfiguration();
@@ -96,6 +100,10 @@ private:
     // Updates the console mode.
     void UpdateDockedState(bool is_handheld);
 
+    // Enable preceding controllers or disable following ones
+    void PropagatePlayerNumberChanged(size_t player_index, bool checked,
+                                      bool reconnect_current = false);
+
     // Disables and disconnects unsupported players based on the given parameters.
     void DisableUnsupportedPlayers();
 
@@ -110,6 +118,8 @@ private:
 
     Core::System& system;
 
+    ControllerNavigation* controller_navigation = nullptr;
+
     // This is true if and only if all parameters are met. Otherwise, this is false.
     // This determines whether the "OK" button can be clicked to exit the applet.
     bool parameters_met{false};
diff --git a/src/yuzu/applets/qt_controller.ui b/src/yuzu/applets/qt_controller.ui
index 729e921ee..6f7cb3c13 100644
--- a/src/yuzu/applets/qt_controller.ui
+++ b/src/yuzu/applets/qt_controller.ui
@@ -2624,13 +2624,53 @@
           </spacer>
          </item>
          <item alignment="Qt::AlignBottom">
-          <widget class="QDialogButtonBox" name="buttonBox">
-           <property name="enabled">
-            <bool>true</bool>
-           </property>
-           <property name="standardButtons">
-            <set>QDialogButtonBox::Cancel|QDialogButtonBox::Ok</set>
-           </property>
+          <widget class="QWidget" name="closeButtons" native="true">
+           <layout class="QVBoxLayout" name="verticalLayout_46">
+            <property name="spacing">
+             <number>7</number>
+            </property>
+            <property name="leftMargin">
+             <number>0</number>
+            </property>
+            <property name="topMargin">
+             <number>0</number>
+            </property>
+            <property name="rightMargin">
+             <number>0</number>
+            </property>
+            <property name="bottomMargin">
+             <number>0</number>
+            </property>
+            <item>
+             <widget class="QLabel" name="labelError">
+              <property name="enabled">
+               <bool>true</bool>
+              </property>
+              <property name="styleSheet">
+               <string notr="true">QLabel { color : red; }</string>
+              </property>
+              <property name="text">
+               <string>Not enough controllers</string>
+              </property>
+              <property name="alignment">
+               <set>Qt::AlignCenter</set>
+              </property>
+              <property name="margin">
+               <number>0</number>
+              </property>
+             </widget>
+            </item>
+            <item>
+             <widget class="QDialogButtonBox" name="buttonBox">
+              <property name="enabled">
+               <bool>true</bool>
+              </property>
+              <property name="standardButtons">
+               <set>QDialogButtonBox::Cancel|QDialogButtonBox::Ok</set>
+              </property>
+             </widget>
+            </item>
+           </layout>
           </widget>
          </item>
         </layout>
diff --git a/src/yuzu/configuration/config.cpp b/src/yuzu/configuration/config.cpp
index 1de093447..baa3e55f3 100644
--- a/src/yuzu/configuration/config.cpp
+++ b/src/yuzu/configuration/config.cpp
@@ -114,7 +114,7 @@ const std::map<Settings::ShaderBackend, QString> Config::shader_backend_texts_ma
 // This must be in alphabetical order according to action name as it must have the same order as
 // UISetting::values.shortcuts, which is alphabetically ordered.
 // clang-format off
-const std::array<UISettings::Shortcut, 22> Config::default_hotkeys{{
+const std::array<UISettings::Shortcut, 23> Config::default_hotkeys{{
     {QStringLiteral(QT_TRANSLATE_NOOP("Hotkeys", "Audio Mute/Unmute")),        QStringLiteral(QT_TRANSLATE_NOOP("Hotkeys", "Main Window")), {QStringLiteral("Ctrl+M"),  QStringLiteral("Home+Dpad_Right"), Qt::WindowShortcut, false}},
     {QStringLiteral(QT_TRANSLATE_NOOP("Hotkeys", "Audio Volume Down")),        QStringLiteral(QT_TRANSLATE_NOOP("Hotkeys", "Main Window")), {QStringLiteral("-"),       QStringLiteral("Home+Dpad_Down"), Qt::ApplicationShortcut, true}},
     {QStringLiteral(QT_TRANSLATE_NOOP("Hotkeys", "Audio Volume Up")),          QStringLiteral(QT_TRANSLATE_NOOP("Hotkeys", "Main Window")), {QStringLiteral("="),       QStringLiteral("Home+Dpad_Up"), Qt::ApplicationShortcut, true}},
@@ -128,14 +128,15 @@ const std::array<UISettings::Shortcut, 22> Config::default_hotkeys{{
     {QStringLiteral(QT_TRANSLATE_NOOP("Hotkeys", "Fullscreen")),               QStringLiteral(QT_TRANSLATE_NOOP("Hotkeys", "Main Window")), {QStringLiteral("F11"),     QStringLiteral("Home+B"), Qt::WindowShortcut, false}},
     {QStringLiteral(QT_TRANSLATE_NOOP("Hotkeys", "Load File")),                QStringLiteral(QT_TRANSLATE_NOOP("Hotkeys", "Main Window")), {QStringLiteral("Ctrl+O"),  QStringLiteral(""), Qt::WidgetWithChildrenShortcut, false}},
     {QStringLiteral(QT_TRANSLATE_NOOP("Hotkeys", "Load/Remove Amiibo")),       QStringLiteral(QT_TRANSLATE_NOOP("Hotkeys", "Main Window")), {QStringLiteral("F2"),      QStringLiteral("Home+A"), Qt::WidgetWithChildrenShortcut, false}},
-    {QStringLiteral(QT_TRANSLATE_NOOP("Hotkeys", "Restart Emulation")),        QStringLiteral(QT_TRANSLATE_NOOP("Hotkeys", "Main Window")), {QStringLiteral("F6"),      QStringLiteral(""), Qt::WindowShortcut, false}},
-    {QStringLiteral(QT_TRANSLATE_NOOP("Hotkeys", "Stop Emulation")),           QStringLiteral(QT_TRANSLATE_NOOP("Hotkeys", "Main Window")), {QStringLiteral("F5"),      QStringLiteral(""), Qt::WindowShortcut, false}},
+    {QStringLiteral(QT_TRANSLATE_NOOP("Hotkeys", "Restart Emulation")),        QStringLiteral(QT_TRANSLATE_NOOP("Hotkeys", "Main Window")), {QStringLiteral("F6"),      QStringLiteral("R+Plus+Minus"), Qt::WindowShortcut, false}},
+    {QStringLiteral(QT_TRANSLATE_NOOP("Hotkeys", "Stop Emulation")),           QStringLiteral(QT_TRANSLATE_NOOP("Hotkeys", "Main Window")), {QStringLiteral("F5"),      QStringLiteral("L+Plus+Minus"), Qt::WindowShortcut, false}},
     {QStringLiteral(QT_TRANSLATE_NOOP("Hotkeys", "TAS Record")),               QStringLiteral(QT_TRANSLATE_NOOP("Hotkeys", "Main Window")), {QStringLiteral("Ctrl+F7"), QStringLiteral(""), Qt::ApplicationShortcut, false}},
     {QStringLiteral(QT_TRANSLATE_NOOP("Hotkeys", "TAS Reset")),                QStringLiteral(QT_TRANSLATE_NOOP("Hotkeys", "Main Window")), {QStringLiteral("Ctrl+F6"), QStringLiteral(""), Qt::ApplicationShortcut, false}},
     {QStringLiteral(QT_TRANSLATE_NOOP("Hotkeys", "TAS Start/Stop")),           QStringLiteral(QT_TRANSLATE_NOOP("Hotkeys", "Main Window")), {QStringLiteral("Ctrl+F5"), QStringLiteral(""), Qt::ApplicationShortcut, false}},
     {QStringLiteral(QT_TRANSLATE_NOOP("Hotkeys", "Toggle Filter Bar")),        QStringLiteral(QT_TRANSLATE_NOOP("Hotkeys", "Main Window")), {QStringLiteral("Ctrl+F"),  QStringLiteral(""), Qt::WindowShortcut, false}},
     {QStringLiteral(QT_TRANSLATE_NOOP("Hotkeys", "Toggle Framerate Limit")),   QStringLiteral(QT_TRANSLATE_NOOP("Hotkeys", "Main Window")), {QStringLiteral("Ctrl+U"),  QStringLiteral("Home+Y"), Qt::ApplicationShortcut, false}},
     {QStringLiteral(QT_TRANSLATE_NOOP("Hotkeys", "Toggle Mouse Panning")),     QStringLiteral(QT_TRANSLATE_NOOP("Hotkeys", "Main Window")), {QStringLiteral("Ctrl+F9"), QStringLiteral(""), Qt::ApplicationShortcut, false}},
+    {QStringLiteral(QT_TRANSLATE_NOOP("Hotkeys", "Toggle Renderdoc Capture")), QStringLiteral(QT_TRANSLATE_NOOP("Hotkeys", "Main Window")), {QStringLiteral(""),        QStringLiteral(""), Qt::ApplicationShortcut, false}},
     {QStringLiteral(QT_TRANSLATE_NOOP("Hotkeys", "Toggle Status Bar")),        QStringLiteral(QT_TRANSLATE_NOOP("Hotkeys", "Main Window")), {QStringLiteral("Ctrl+S"),  QStringLiteral(""), Qt::WindowShortcut, false}},
 }};
 // clang-format on
diff --git a/src/yuzu/configuration/config.h b/src/yuzu/configuration/config.h
index 727feebfb..74ec4f771 100644
--- a/src/yuzu/configuration/config.h
+++ b/src/yuzu/configuration/config.h
@@ -48,7 +48,7 @@ public:
         default_mouse_buttons;
     static const std::array<int, Settings::NativeKeyboard::NumKeyboardKeys> default_keyboard_keys;
     static const std::array<int, Settings::NativeKeyboard::NumKeyboardMods> default_keyboard_mods;
-    static const std::array<UISettings::Shortcut, 22> default_hotkeys;
+    static const std::array<UISettings::Shortcut, 23> default_hotkeys;
 
     static const std::map<Settings::AntiAliasing, QString> anti_aliasing_texts_map;
     static const std::map<Settings::ScalingFilter, QString> scaling_filter_texts_map;
diff --git a/src/yuzu/configuration/configure_audio.cpp b/src/yuzu/configuration/configure_audio.cpp
index 9ccfb2435..81dd51ad3 100644
--- a/src/yuzu/configuration/configure_audio.cpp
+++ b/src/yuzu/configuration/configure_audio.cpp
@@ -42,6 +42,9 @@ void ConfigureAudio::Setup(const ConfigurationShared::Builder& builder) {
         for (auto* setting : Settings::values.linkage.by_category[category]) {
             settings.push_back(setting);
         }
+        for (auto* setting : UISettings::values.linkage.by_category[category]) {
+            settings.push_back(setting);
+        }
     };
 
     push(Settings::Category::Audio);
diff --git a/src/yuzu/configuration/configure_hotkeys.cpp b/src/yuzu/configuration/configure_hotkeys.cpp
index 0b2a965f8..68e21cd84 100644
--- a/src/yuzu/configuration/configure_hotkeys.cpp
+++ b/src/yuzu/configuration/configure_hotkeys.cpp
@@ -319,6 +319,13 @@ void ConfigureHotkeys::ApplyConfiguration(HotkeyRegistry& registry) {
 void ConfigureHotkeys::RestoreDefaults() {
     for (int r = 0; r < model->rowCount(); ++r) {
         const QStandardItem* parent = model->item(r, 0);
+        const int hotkey_size = static_cast<int>(Config::default_hotkeys.size());
+
+        if (hotkey_size != parent->rowCount()) {
+            QMessageBox::warning(this, tr("Invalid hotkey settings"),
+                                 tr("An error occurred. Please report this issue on github."));
+            return;
+        }
 
         for (int r2 = 0; r2 < parent->rowCount(); ++r2) {
             model->item(r, 0)
diff --git a/src/yuzu/configuration/configure_input.cpp b/src/yuzu/configuration/configure_input.cpp
index e8f9ebfd8..3dcad2701 100644
--- a/src/yuzu/configuration/configure_input.cpp
+++ b/src/yuzu/configuration/configure_input.cpp
@@ -101,13 +101,13 @@ void ConfigureInput::Initialize(InputCommon::InputSubsystem* input_subsystem,
         ui->tabPlayer5, ui->tabPlayer6, ui->tabPlayer7, ui->tabPlayer8,
     };
 
-    player_connected = {
+    connected_controller_checkboxes = {
         ui->checkboxPlayer1Connected, ui->checkboxPlayer2Connected, ui->checkboxPlayer3Connected,
         ui->checkboxPlayer4Connected, ui->checkboxPlayer5Connected, ui->checkboxPlayer6Connected,
         ui->checkboxPlayer7Connected, ui->checkboxPlayer8Connected,
     };
 
-    std::array<QLabel*, 8> player_connected_labels = {
+    std::array<QLabel*, 8> connected_controller_labels = {
         ui->label,   ui->label_3, ui->label_4, ui->label_5,
         ui->label_6, ui->label_7, ui->label_8, ui->label_9,
     };
@@ -115,31 +115,37 @@ void ConfigureInput::Initialize(InputCommon::InputSubsystem* input_subsystem,
     for (std::size_t i = 0; i < player_tabs.size(); ++i) {
         player_tabs[i]->setLayout(new QHBoxLayout(player_tabs[i]));
         player_tabs[i]->layout()->addWidget(player_controllers[i]);
-        connect(player_controllers[i], &ConfigureInputPlayer::Connected, [&, i](bool is_connected) {
-            // Ensures that the controllers are always connected in sequential order
-            if (is_connected) {
-                for (std::size_t index = 0; index <= i; ++index) {
-                    player_connected[index]->setChecked(is_connected);
-                }
-            } else {
-                for (std::size_t index = i; index < player_tabs.size(); ++index) {
-                    player_connected[index]->setChecked(is_connected);
-                }
+        connect(player_controllers[i], &ConfigureInputPlayer::Connected, [this, i](bool checked) {
+            // Ensures that connecting a controller changes the number of players
+            if (connected_controller_checkboxes[i]->isChecked() != checked) {
+                // Ensures that the players are always connected in sequential order
+                PropagatePlayerNumberChanged(i, checked);
             }
         });
+        connect(connected_controller_checkboxes[i], &QCheckBox::clicked, [this, i](bool checked) {
+            // Reconnect current controller if it was the last one checked
+            // (player number was reduced by more than one)
+            const bool reconnect_first = !checked &&
+                                         i < connected_controller_checkboxes.size() - 1 &&
+                                         connected_controller_checkboxes[i + 1]->isChecked();
+
+            // Ensures that the players are always connected in sequential order
+            PropagatePlayerNumberChanged(i, checked, reconnect_first);
+        });
         connect(player_controllers[i], &ConfigureInputPlayer::RefreshInputDevices, this,
                 &ConfigureInput::UpdateAllInputDevices);
         connect(player_controllers[i], &ConfigureInputPlayer::RefreshInputProfiles, this,
                 &ConfigureInput::UpdateAllInputProfiles, Qt::QueuedConnection);
-        connect(player_connected[i], &QCheckBox::stateChanged, [this, i](int state) {
+        connect(connected_controller_checkboxes[i], &QCheckBox::stateChanged, [this, i](int state) {
+            // Keep activated controllers synced with the "Connected Controllers" checkboxes
             player_controllers[i]->ConnectPlayer(state == Qt::Checked);
         });
 
         // Remove/hide all the elements that exceed max_players, if applicable.
         if (i >= max_players) {
             ui->tabWidget->removeTab(static_cast<int>(max_players));
-            player_connected[i]->hide();
-            player_connected_labels[i]->hide();
+            connected_controller_checkboxes[i]->hide();
+            connected_controller_labels[i]->hide();
         }
     }
     // Only the first player can choose handheld mode so connect the signal just to player 1
@@ -183,6 +189,27 @@ void ConfigureInput::Initialize(InputCommon::InputSubsystem* input_subsystem,
     LoadConfiguration();
 }
 
+void ConfigureInput::PropagatePlayerNumberChanged(size_t player_index, bool checked,
+                                                  bool reconnect_current) {
+    connected_controller_checkboxes[player_index]->setChecked(checked);
+
+    if (checked) {
+        // Check all previous buttons when checked
+        if (player_index > 0) {
+            PropagatePlayerNumberChanged(player_index - 1, checked);
+        }
+    } else {
+        // Unchecked all following buttons when unchecked
+        if (player_index < connected_controller_checkboxes.size() - 1) {
+            PropagatePlayerNumberChanged(player_index + 1, checked);
+        }
+    }
+
+    if (reconnect_current) {
+        connected_controller_checkboxes[player_index]->setCheckState(Qt::Checked);
+    }
+}
+
 QList<QWidget*> ConfigureInput::GetSubTabs() const {
     return {
         ui->tabPlayer1, ui->tabPlayer2, ui->tabPlayer3, ui->tabPlayer4,  ui->tabPlayer5,
@@ -233,17 +260,17 @@ void ConfigureInput::LoadConfiguration() {
 }
 
 void ConfigureInput::LoadPlayerControllerIndices() {
-    for (std::size_t i = 0; i < player_connected.size(); ++i) {
+    for (std::size_t i = 0; i < connected_controller_checkboxes.size(); ++i) {
         if (i == 0) {
             auto* handheld =
                 system.HIDCore().GetEmulatedController(Core::HID::NpadIdType::Handheld);
             if (handheld->IsConnected()) {
-                player_connected[i]->setChecked(true);
+                connected_controller_checkboxes[i]->setChecked(true);
                 continue;
             }
         }
         const auto* controller = system.HIDCore().GetEmulatedControllerByIndex(i);
-        player_connected[i]->setChecked(controller->IsConnected());
+        connected_controller_checkboxes[i]->setChecked(controller->IsConnected());
     }
 }
 
diff --git a/src/yuzu/configuration/configure_input.h b/src/yuzu/configuration/configure_input.h
index c89189c36..136cd3a0a 100644
--- a/src/yuzu/configuration/configure_input.h
+++ b/src/yuzu/configuration/configure_input.h
@@ -56,6 +56,9 @@ private:
     void UpdateDockedState(bool is_handheld);
     void UpdateAllInputDevices();
     void UpdateAllInputProfiles(std::size_t player_index);
+    // Enable preceding controllers or disable following ones
+    void PropagatePlayerNumberChanged(size_t player_index, bool checked,
+                                      bool reconnect_current = false);
 
     /// Load configuration settings.
     void LoadConfiguration();
@@ -70,7 +73,8 @@ private:
 
     std::array<ConfigureInputPlayer*, 8> player_controllers;
     std::array<QWidget*, 8> player_tabs;
-    std::array<QCheckBox*, 8> player_connected;
+    // Checkboxes representing the "Connected Controllers".
+    std::array<QCheckBox*, 8> connected_controller_checkboxes;
     ConfigureInputAdvanced* advanced;
 
     Core::System& system;
diff --git a/src/yuzu/configuration/configure_input_player.h b/src/yuzu/configuration/configure_input_player.h
index d4df43d73..d3255d2b4 100644
--- a/src/yuzu/configuration/configure_input_player.h
+++ b/src/yuzu/configuration/configure_input_player.h
@@ -75,7 +75,7 @@ public:
     void ClearAll();
 
 signals:
-    /// Emitted when this controller is connected by the user.
+    /// Emitted when this controller is (dis)connected by the user.
     void Connected(bool connected);
     /// Emitted when the Handheld mode is selected (undocked with dual joycons attached).
     void HandheldStateChanged(bool is_handheld);
@@ -183,9 +183,6 @@ private:
     /// Stores a pair of "Connected Controllers" combobox index and Controller Type enum.
     std::vector<std::pair<int, Core::HID::NpadStyleIndex>> index_controller_type_pairs;
 
-    static constexpr int PLAYER_COUNT = 8;
-    std::array<QCheckBox*, PLAYER_COUNT> player_connected_checkbox;
-
     /// This will be the the setting function when an input is awaiting configuration.
     std::optional<std::function<void(const Common::ParamPackage&)>> input_setter;
 
diff --git a/src/yuzu/configuration/configure_ui.cpp b/src/yuzu/configuration/configure_ui.cpp
index a9fde9f4f..82f3b6e78 100644
--- a/src/yuzu/configuration/configure_ui.cpp
+++ b/src/yuzu/configuration/configure_ui.cpp
@@ -123,6 +123,8 @@ ConfigureUi::ConfigureUi(Core::System& system_, QWidget* parent)
     connect(ui->show_compat, &QCheckBox::stateChanged, this, &ConfigureUi::RequestGameListUpdate);
     connect(ui->show_size, &QCheckBox::stateChanged, this, &ConfigureUi::RequestGameListUpdate);
     connect(ui->show_types, &QCheckBox::stateChanged, this, &ConfigureUi::RequestGameListUpdate);
+    connect(ui->show_play_time, &QCheckBox::stateChanged, this,
+            &ConfigureUi::RequestGameListUpdate);
     connect(ui->game_icon_size_combobox, QOverload<int>::of(&QComboBox::currentIndexChanged), this,
             &ConfigureUi::RequestGameListUpdate);
     connect(ui->folder_icon_size_combobox, QOverload<int>::of(&QComboBox::currentIndexChanged),
@@ -167,6 +169,7 @@ void ConfigureUi::ApplyConfiguration() {
     UISettings::values.show_compat = ui->show_compat->isChecked();
     UISettings::values.show_size = ui->show_size->isChecked();
     UISettings::values.show_types = ui->show_types->isChecked();
+    UISettings::values.show_play_time = ui->show_play_time->isChecked();
     UISettings::values.game_icon_size = ui->game_icon_size_combobox->currentData().toUInt();
     UISettings::values.folder_icon_size = ui->folder_icon_size_combobox->currentData().toUInt();
     UISettings::values.row_1_text_id = ui->row_1_text_combobox->currentData().toUInt();
@@ -179,6 +182,7 @@ void ConfigureUi::ApplyConfiguration() {
     const u32 height = ScreenshotDimensionToInt(ui->screenshot_height->currentText());
     UISettings::values.screenshot_height.SetValue(height);
 
+    RequestGameListUpdate();
     system.ApplySettings();
 }
 
@@ -194,6 +198,7 @@ void ConfigureUi::SetConfiguration() {
     ui->show_compat->setChecked(UISettings::values.show_compat.GetValue());
     ui->show_size->setChecked(UISettings::values.show_size.GetValue());
     ui->show_types->setChecked(UISettings::values.show_types.GetValue());
+    ui->show_play_time->setChecked(UISettings::values.show_play_time.GetValue());
     ui->game_icon_size_combobox->setCurrentIndex(
         ui->game_icon_size_combobox->findData(UISettings::values.game_icon_size.GetValue()));
     ui->folder_icon_size_combobox->setCurrentIndex(
diff --git a/src/yuzu/configuration/configure_ui.ui b/src/yuzu/configuration/configure_ui.ui
index cb66ef104..b8e648381 100644
--- a/src/yuzu/configuration/configure_ui.ui
+++ b/src/yuzu/configuration/configure_ui.ui
@@ -105,6 +105,13 @@
          </widget>
         </item>
         <item>
+         <widget class="QCheckBox" name="show_play_time">
+          <property name="text">
+           <string>Show Play Time Column</string>
+          </property>
+         </widget>
+        </item>
+        <item>
          <layout class="QHBoxLayout" name="game_icon_size_qhbox_layout_2">
           <item>
            <widget class="QLabel" name="game_icon_size_label">
diff --git a/src/yuzu/configuration/configure_vibration.cpp b/src/yuzu/configuration/configure_vibration.cpp
index d765e808a..68c28b320 100644
--- a/src/yuzu/configuration/configure_vibration.cpp
+++ b/src/yuzu/configuration/configure_vibration.cpp
@@ -89,7 +89,7 @@ void ConfigureVibration::VibrateController(Core::HID::ControllerTriggerType type
 
     auto& player = Settings::values.players.GetValue()[player_index];
     auto controller = hid_core.GetEmulatedControllerByIndex(player_index);
-    const int vibration_strenght = vibration_spinboxes[player_index]->value();
+    const int vibration_strength = vibration_spinboxes[player_index]->value();
     const auto& buttons = controller->GetButtonsValues();
 
     bool button_is_pressed = false;
@@ -105,10 +105,10 @@ void ConfigureVibration::VibrateController(Core::HID::ControllerTriggerType type
         return;
     }
 
-    const int old_vibration_enabled = player.vibration_enabled;
-    const bool old_vibration_strenght = player.vibration_strength;
+    const bool old_vibration_enabled = player.vibration_enabled;
+    const int old_vibration_strength = player.vibration_strength;
     player.vibration_enabled = true;
-    player.vibration_strength = vibration_strenght;
+    player.vibration_strength = vibration_strength;
 
     const Core::HID::VibrationValue vibration{
         .low_amplitude = 1.0f,
@@ -121,7 +121,7 @@ void ConfigureVibration::VibrateController(Core::HID::ControllerTriggerType type
 
     // Restore previous values
     player.vibration_enabled = old_vibration_enabled;
-    player.vibration_strength = old_vibration_strenght;
+    player.vibration_strength = old_vibration_strength;
 }
 
 void ConfigureVibration::StopVibrations() {
diff --git a/src/yuzu/configuration/shared_translation.cpp b/src/yuzu/configuration/shared_translation.cpp
index 276bdbaba..3fe448f27 100644
--- a/src/yuzu/configuration/shared_translation.cpp
+++ b/src/yuzu/configuration/shared_translation.cpp
@@ -29,9 +29,10 @@ std::unique_ptr<TranslationMap> InitializeTranslations(QWidget* parent) {
     INSERT(Settings, sink_id, "Output Engine:", "");
     INSERT(Settings, audio_output_device_id, "Output Device:", "");
     INSERT(Settings, audio_input_device_id, "Input Device:", "");
-    INSERT(Settings, audio_muted, "Mute audio when in background", "");
+    INSERT(Settings, audio_muted, "Mute audio", "");
     INSERT(Settings, volume, "Volume:", "");
     INSERT(Settings, dump_audio_commands, "", "");
+    INSERT(UISettings, mute_when_in_background, "Mute audio when in background", "");
 
     // Core
     INSERT(Settings, use_multi_core, "Multicore CPU Emulation", "");
@@ -156,6 +157,7 @@ std::unique_ptr<TranslationMap> InitializeTranslations(QWidget* parent) {
     INSERT(UISettings, select_user_on_boot, "Prompt for user on game boot", "");
     INSERT(UISettings, pause_when_in_background, "Pause emulation when in background", "");
     INSERT(UISettings, confirm_before_closing, "Confirm exit while emulation is running", "");
+    INSERT(UISettings, confirm_before_stopping, "Confirm before stopping emulation", "");
     INSERT(UISettings, hide_mouse, "Hide mouse on inactivity", "");
     INSERT(UISettings, controller_applet_disabled, "Disable controller applet", "");
 
@@ -382,6 +384,13 @@ std::unique_ptr<ComboboxTranslationMap> ComboboxEnumeration(QWidget* parent) {
     translations->insert(
         {Settings::EnumMetadata<Settings::ConsoleMode>::Index(),
          {PAIR(ConsoleMode, Docked, "Docked"), PAIR(ConsoleMode, Handheld, "Handheld")}});
+    translations->insert(
+        {Settings::EnumMetadata<Settings::ConfirmStop>::Index(),
+         {
+             PAIR(ConfirmStop, Ask_Always, "Always ask (Default)"),
+             PAIR(ConfirmStop, Ask_Based_On_Game, "Only if game specifies not to stop"),
+             PAIR(ConfirmStop, Ask_Never, "Never ask"),
+         }});
 
 #undef PAIR
 #undef CTX_PAIR
diff --git a/src/yuzu/debugger/wait_tree.cpp b/src/yuzu/debugger/wait_tree.cpp
index 0783a2430..7049c57b6 100644
--- a/src/yuzu/debugger/wait_tree.cpp
+++ b/src/yuzu/debugger/wait_tree.cpp
@@ -127,7 +127,7 @@ std::vector<std::unique_ptr<WaitTreeItem>> WaitTreeCallstack::GetChildren() cons
         return list;
     }
 
-    if (thread.GetOwnerProcess() == nullptr || !thread.GetOwnerProcess()->Is64BitProcess()) {
+    if (thread.GetOwnerProcess() == nullptr || !thread.GetOwnerProcess()->Is64Bit()) {
         return list;
     }
 
diff --git a/src/yuzu/game_list.cpp b/src/yuzu/game_list.cpp
index f254c1e1c..7e7d8e252 100644
--- a/src/yuzu/game_list.cpp
+++ b/src/yuzu/game_list.cpp
@@ -312,8 +312,10 @@ void GameList::OnFilterCloseClicked() {
 }
 
 GameList::GameList(FileSys::VirtualFilesystem vfs_, FileSys::ManualContentProvider* provider_,
-                   Core::System& system_, GMainWindow* parent)
-    : QWidget{parent}, vfs{std::move(vfs_)}, provider{provider_}, system{system_} {
+                   PlayTime::PlayTimeManager& play_time_manager_, Core::System& system_,
+                   GMainWindow* parent)
+    : QWidget{parent}, vfs{std::move(vfs_)}, provider{provider_},
+      play_time_manager{play_time_manager_}, system{system_} {
     watcher = new QFileSystemWatcher(this);
     connect(watcher, &QFileSystemWatcher::directoryChanged, this, &GameList::RefreshGameDirectory);
 
@@ -340,6 +342,7 @@ GameList::GameList(FileSys::VirtualFilesystem vfs_, FileSys::ManualContentProvid
 
     tree_view->setColumnHidden(COLUMN_ADD_ONS, !UISettings::values.show_add_ons);
     tree_view->setColumnHidden(COLUMN_COMPATIBILITY, !UISettings::values.show_compat);
+    tree_view->setColumnHidden(COLUMN_PLAY_TIME, !UISettings::values.show_play_time);
     item_model->setSortRole(GameListItemPath::SortRole);
 
     connect(main_window, &GMainWindow::UpdateThemedIcons, this, &GameList::OnUpdateThemedIcons);
@@ -377,7 +380,6 @@ void GameList::UnloadController() {
 
 GameList::~GameList() {
     UnloadController();
-    emit ShouldCancelWorker();
 }
 
 void GameList::SetFilterFocus() {
@@ -394,6 +396,10 @@ void GameList::ClearFilter() {
     search_field->clear();
 }
 
+void GameList::WorkerEvent() {
+    current_worker->ProcessEvents(this);
+}
+
 void GameList::AddDirEntry(GameListDir* entry_items) {
     item_model->invisibleRootItem()->appendRow(entry_items);
     tree_view->setExpanded(
@@ -548,6 +554,7 @@ void GameList::AddGamePopup(QMenu& context_menu, u64 program_id, const std::stri
     QAction* remove_update = remove_menu->addAction(tr("Remove Installed Update"));
     QAction* remove_dlc = remove_menu->addAction(tr("Remove All Installed DLC"));
     QAction* remove_custom_config = remove_menu->addAction(tr("Remove Custom Configuration"));
+    QAction* remove_play_time_data = remove_menu->addAction(tr("Remove Play Time Data"));
     QAction* remove_cache_storage = remove_menu->addAction(tr("Remove Cache Storage"));
     QAction* remove_gl_shader_cache = remove_menu->addAction(tr("Remove OpenGL Pipeline Cache"));
     QAction* remove_vk_shader_cache = remove_menu->addAction(tr("Remove Vulkan Pipeline Cache"));
@@ -560,9 +567,9 @@ void GameList::AddGamePopup(QMenu& context_menu, u64 program_id, const std::stri
     QAction* verify_integrity = context_menu.addAction(tr("Verify Integrity"));
     QAction* copy_tid = context_menu.addAction(tr("Copy Title ID to Clipboard"));
     QAction* navigate_to_gamedb_entry = context_menu.addAction(tr("Navigate to GameDB entry"));
-#ifndef WIN32
     QMenu* shortcut_menu = context_menu.addMenu(tr("Create Shortcut"));
     QAction* create_desktop_shortcut = shortcut_menu->addAction(tr("Add to Desktop"));
+#ifndef WIN32
     QAction* create_applications_menu_shortcut =
         shortcut_menu->addAction(tr("Add to Applications Menu"));
 #endif
@@ -622,6 +629,8 @@ void GameList::AddGamePopup(QMenu& context_menu, u64 program_id, const std::stri
     connect(remove_custom_config, &QAction::triggered, [this, program_id, path]() {
         emit RemoveFileRequested(program_id, GameListRemoveTarget::CustomConfiguration, path);
     });
+    connect(remove_play_time_data, &QAction::triggered,
+            [this, program_id]() { emit RemovePlayTimeRequested(program_id); });
     connect(remove_cache_storage, &QAction::triggered, [this, program_id, path] {
         emit RemoveFileRequested(program_id, GameListRemoveTarget::CacheStorage, path);
     });
@@ -638,10 +647,10 @@ void GameList::AddGamePopup(QMenu& context_menu, u64 program_id, const std::stri
     connect(navigate_to_gamedb_entry, &QAction::triggered, [this, program_id]() {
         emit NavigateToGamedbEntryRequested(program_id, compatibility_list);
     });
-#ifndef WIN32
     connect(create_desktop_shortcut, &QAction::triggered, [this, program_id, path]() {
         emit CreateShortcut(program_id, path, GameListShortcutTarget::Desktop);
     });
+#ifndef WIN32
     connect(create_applications_menu_shortcut, &QAction::triggered, [this, program_id, path]() {
         emit CreateShortcut(program_id, path, GameListShortcutTarget::Applications);
     });
@@ -790,6 +799,7 @@ void GameList::RetranslateUI() {
     item_model->setHeaderData(COLUMN_ADD_ONS, Qt::Horizontal, tr("Add-ons"));
     item_model->setHeaderData(COLUMN_FILE_TYPE, Qt::Horizontal, tr("File type"));
     item_model->setHeaderData(COLUMN_SIZE, Qt::Horizontal, tr("Size"));
+    item_model->setHeaderData(COLUMN_PLAY_TIME, Qt::Horizontal, tr("Play time"));
 }
 
 void GameListSearchField::changeEvent(QEvent* event) {
@@ -817,28 +827,23 @@ void GameList::PopulateAsync(QVector<UISettings::GameDir>& game_dirs) {
     tree_view->setColumnHidden(COLUMN_COMPATIBILITY, !UISettings::values.show_compat);
     tree_view->setColumnHidden(COLUMN_FILE_TYPE, !UISettings::values.show_types);
     tree_view->setColumnHidden(COLUMN_SIZE, !UISettings::values.show_size);
+    tree_view->setColumnHidden(COLUMN_PLAY_TIME, !UISettings::values.show_play_time);
+
+    // Cancel any existing worker.
+    current_worker.reset();
 
     // Delete any rows that might already exist if we're repopulating
     item_model->removeRows(0, item_model->rowCount());
     search_field->clear();
 
-    emit ShouldCancelWorker();
+    current_worker = std::make_unique<GameListWorker>(vfs, provider, game_dirs, compatibility_list,
+                                                      play_time_manager, system);
 
-    GameListWorker* worker =
-        new GameListWorker(vfs, provider, game_dirs, compatibility_list, system);
-
-    connect(worker, &GameListWorker::EntryReady, this, &GameList::AddEntry, Qt::QueuedConnection);
-    connect(worker, &GameListWorker::DirEntryReady, this, &GameList::AddDirEntry,
-            Qt::QueuedConnection);
-    connect(worker, &GameListWorker::Finished, this, &GameList::DonePopulating,
+    // Get events from the worker as data becomes available
+    connect(current_worker.get(), &GameListWorker::DataAvailable, this, &GameList::WorkerEvent,
             Qt::QueuedConnection);
-    // Use DirectConnection here because worker->Cancel() is thread-safe and we want it to
-    // cancel without delay.
-    connect(this, &GameList::ShouldCancelWorker, worker, &GameListWorker::Cancel,
-            Qt::DirectConnection);
 
-    QThreadPool::globalInstance()->start(worker);
-    current_worker = std::move(worker);
+    QThreadPool::globalInstance()->start(current_worker.get());
 }
 
 void GameList::SaveInterfaceLayout() {
diff --git a/src/yuzu/game_list.h b/src/yuzu/game_list.h
index 1fcbbf0ba..563a3a35b 100644
--- a/src/yuzu/game_list.h
+++ b/src/yuzu/game_list.h
@@ -18,6 +18,7 @@
 #include "core/core.h"
 #include "uisettings.h"
 #include "yuzu/compatibility_list.h"
+#include "yuzu/play_time_manager.h"
 
 namespace Core {
 class System;
@@ -75,11 +76,13 @@ public:
         COLUMN_ADD_ONS,
         COLUMN_FILE_TYPE,
         COLUMN_SIZE,
+        COLUMN_PLAY_TIME,
         COLUMN_COUNT, // Number of columns
     };
 
     explicit GameList(std::shared_ptr<FileSys::VfsFilesystem> vfs_,
-                      FileSys::ManualContentProvider* provider_, Core::System& system_,
+                      FileSys::ManualContentProvider* provider_,
+                      PlayTime::PlayTimeManager& play_time_manager_, Core::System& system_,
                       GMainWindow* parent = nullptr);
     ~GameList() override;
 
@@ -106,13 +109,13 @@ signals:
     void BootGame(const QString& game_path, u64 program_id, std::size_t program_index,
                   StartGameType type, AmLaunchType launch_type);
     void GameChosen(const QString& game_path, const u64 title_id = 0);
-    void ShouldCancelWorker();
     void OpenFolderRequested(u64 program_id, GameListOpenTarget target,
                              const std::string& game_path);
     void OpenTransferableShaderCacheRequested(u64 program_id);
     void RemoveInstalledEntryRequested(u64 program_id, InstalledEntryType type);
     void RemoveFileRequested(u64 program_id, GameListRemoveTarget target,
                              const std::string& game_path);
+    void RemovePlayTimeRequested(u64 program_id);
     void DumpRomFSRequested(u64 program_id, const std::string& game_path, DumpRomFSTarget target);
     void VerifyIntegrityRequested(const std::string& game_path);
     void CopyTIDRequested(u64 program_id);
@@ -134,11 +137,16 @@ private slots:
     void OnUpdateThemedIcons();
 
 private:
+    friend class GameListWorker;
+    void WorkerEvent();
+
     void AddDirEntry(GameListDir* entry_items);
     void AddEntry(const QList<QStandardItem*>& entry_items, GameListDir* parent);
-    void ValidateEntry(const QModelIndex& item);
     void DonePopulating(const QStringList& watch_list);
 
+private:
+    void ValidateEntry(const QModelIndex& item);
+
     void RefreshGameDirectory();
 
     void ToggleFavorite(u64 program_id);
@@ -161,13 +169,14 @@ private:
     QVBoxLayout* layout = nullptr;
     QTreeView* tree_view = nullptr;
     QStandardItemModel* item_model = nullptr;
-    GameListWorker* current_worker = nullptr;
+    std::unique_ptr<GameListWorker> current_worker;
     QFileSystemWatcher* watcher = nullptr;
     ControllerNavigation* controller_navigation = nullptr;
     CompatibilityList compatibility_list;
 
     friend class GameListSearchField;
 
+    const PlayTime::PlayTimeManager& play_time_manager;
     Core::System& system;
 };
 
diff --git a/src/yuzu/game_list_p.h b/src/yuzu/game_list_p.h
index 1800f090f..86a0c41d9 100644
--- a/src/yuzu/game_list_p.h
+++ b/src/yuzu/game_list_p.h
@@ -18,6 +18,7 @@
 #include "common/common_types.h"
 #include "common/logging/log.h"
 #include "common/string_util.h"
+#include "yuzu/play_time_manager.h"
 #include "yuzu/uisettings.h"
 #include "yuzu/util/util.h"
 
@@ -221,6 +222,31 @@ public:
     }
 };
 
+/**
+ * GameListItem for Play Time values.
+ * This object stores the play time of a game in seconds, and its readable
+ * representation in minutes/hours
+ */
+class GameListItemPlayTime : public GameListItem {
+public:
+    static constexpr int PlayTimeRole = SortRole;
+
+    GameListItemPlayTime() = default;
+    explicit GameListItemPlayTime(const qulonglong time_seconds) {
+        setData(time_seconds, PlayTimeRole);
+    }
+
+    void setData(const QVariant& value, int role) override {
+        qulonglong time_seconds = value.toULongLong();
+        GameListItem::setData(PlayTime::ReadablePlayTime(time_seconds), Qt::DisplayRole);
+        GameListItem::setData(value, PlayTimeRole);
+    }
+
+    bool operator<(const QStandardItem& other) const override {
+        return data(PlayTimeRole).toULongLong() < other.data(PlayTimeRole).toULongLong();
+    }
+};
+
 class GameListDir : public GameListItem {
 public:
     static constexpr int GameDirRole = Qt::UserRole + 2;
diff --git a/src/yuzu/game_list_worker.cpp b/src/yuzu/game_list_worker.cpp
index e7fb8a282..69be21027 100644
--- a/src/yuzu/game_list_worker.cpp
+++ b/src/yuzu/game_list_worker.cpp
@@ -194,6 +194,7 @@ QList<QStandardItem*> MakeGameListEntry(const std::string& path, const std::stri
                                         const std::size_t size, const std::vector<u8>& icon,
                                         Loader::AppLoader& loader, u64 program_id,
                                         const CompatibilityList& compatibility_list,
+                                        const PlayTime::PlayTimeManager& play_time_manager,
                                         const FileSys::PatchManager& patch) {
     const auto it = FindMatchingCompatibilityEntry(compatibility_list, program_id);
 
@@ -212,6 +213,7 @@ QList<QStandardItem*> MakeGameListEntry(const std::string& path, const std::stri
         new GameListItemCompat(compatibility),
         new GameListItem(file_type_string),
         new GameListItemSize(size),
+        new GameListItemPlayTime(play_time_manager.GetPlayTime(program_id)),
     };
 
     const auto patch_versions = GetGameListCachedObject(
@@ -227,11 +229,57 @@ QList<QStandardItem*> MakeGameListEntry(const std::string& path, const std::stri
 GameListWorker::GameListWorker(FileSys::VirtualFilesystem vfs_,
                                FileSys::ManualContentProvider* provider_,
                                QVector<UISettings::GameDir>& game_dirs_,
-                               const CompatibilityList& compatibility_list_, Core::System& system_)
+                               const CompatibilityList& compatibility_list_,
+                               const PlayTime::PlayTimeManager& play_time_manager_,
+                               Core::System& system_)
     : vfs{std::move(vfs_)}, provider{provider_}, game_dirs{game_dirs_},
-      compatibility_list{compatibility_list_}, system{system_} {}
+      compatibility_list{compatibility_list_}, play_time_manager{play_time_manager_}, system{
+                                                                                          system_} {
+    // We want the game list to manage our lifetime.
+    setAutoDelete(false);
+}
+
+GameListWorker::~GameListWorker() {
+    this->disconnect();
+    stop_requested.store(true);
+    processing_completed.Wait();
+}
 
-GameListWorker::~GameListWorker() = default;
+void GameListWorker::ProcessEvents(GameList* game_list) {
+    while (true) {
+        std::function<void(GameList*)> func;
+        {
+            // Lock queue to protect concurrent modification.
+            std::scoped_lock lk(lock);
+
+            // If we can't pop a function, return.
+            if (queued_events.empty()) {
+                return;
+            }
+
+            // Pop a function.
+            func = std::move(queued_events.back());
+            queued_events.pop_back();
+        }
+
+        // Run the function.
+        func(game_list);
+    }
+}
+
+template <typename F>
+void GameListWorker::RecordEvent(F&& func) {
+    {
+        // Lock queue to protect concurrent modification.
+        std::scoped_lock lk(lock);
+
+        // Add the function into the front of the queue.
+        queued_events.emplace_front(std::move(func));
+    }
+
+    // Data now available.
+    emit DataAvailable();
+}
 
 void GameListWorker::AddTitlesToGameList(GameListDir* parent_dir) {
     using namespace FileSys;
@@ -279,16 +327,16 @@ void GameListWorker::AddTitlesToGameList(GameListDir* parent_dir) {
             GetMetadataFromControlNCA(patch, *control, icon, name);
         }
 
-        emit EntryReady(MakeGameListEntry(file->GetFullPath(), name, file->GetSize(), icon, *loader,
-                                          program_id, compatibility_list, patch),
-                        parent_dir);
+        auto entry = MakeGameListEntry(file->GetFullPath(), name, file->GetSize(), icon, *loader,
+                                       program_id, compatibility_list, play_time_manager, patch);
+        RecordEvent([=](GameList* game_list) { game_list->AddEntry(entry, parent_dir); });
     }
 }
 
 void GameListWorker::ScanFileSystem(ScanTarget target, const std::string& dir_path, bool deep_scan,
                                     GameListDir* parent_dir) {
     const auto callback = [this, target, parent_dir](const std::filesystem::path& path) -> bool {
-        if (stop_processing) {
+        if (stop_requested) {
             // Breaks the callback loop.
             return false;
         }
@@ -355,10 +403,12 @@ void GameListWorker::ScanFileSystem(ScanTarget target, const std::string& dir_pa
                         const FileSys::PatchManager patch{id, system.GetFileSystemController(),
                                                           system.GetContentProvider()};
 
-                        emit EntryReady(MakeGameListEntry(physical_name, name,
-                                                          Common::FS::GetSize(physical_name), icon,
-                                                          *loader, id, compatibility_list, patch),
-                                        parent_dir);
+                        auto entry = MakeGameListEntry(
+                            physical_name, name, Common::FS::GetSize(physical_name), icon, *loader,
+                            id, compatibility_list, play_time_manager, patch);
+
+                        RecordEvent(
+                            [=](GameList* game_list) { game_list->AddEntry(entry, parent_dir); });
                     }
                 } else {
                     std::vector<u8> icon;
@@ -370,10 +420,12 @@ void GameListWorker::ScanFileSystem(ScanTarget target, const std::string& dir_pa
                     const FileSys::PatchManager patch{program_id, system.GetFileSystemController(),
                                                       system.GetContentProvider()};
 
-                    emit EntryReady(
-                        MakeGameListEntry(physical_name, name, Common::FS::GetSize(physical_name),
-                                          icon, *loader, program_id, compatibility_list, patch),
-                        parent_dir);
+                    auto entry = MakeGameListEntry(
+                        physical_name, name, Common::FS::GetSize(physical_name), icon, *loader,
+                        program_id, compatibility_list, play_time_manager, patch);
+
+                    RecordEvent(
+                        [=](GameList* game_list) { game_list->AddEntry(entry, parent_dir); });
                 }
             }
         } else if (is_dir) {
@@ -392,26 +444,34 @@ void GameListWorker::ScanFileSystem(ScanTarget target, const std::string& dir_pa
 }
 
 void GameListWorker::run() {
-    stop_processing = false;
+    watch_list.clear();
     provider->ClearAllEntries();
 
+    const auto DirEntryReady = [&](GameListDir* game_list_dir) {
+        RecordEvent([=](GameList* game_list) { game_list->AddDirEntry(game_list_dir); });
+    };
+
     for (UISettings::GameDir& game_dir : game_dirs) {
+        if (stop_requested) {
+            break;
+        }
+
         if (game_dir.path == QStringLiteral("SDMC")) {
             auto* const game_list_dir = new GameListDir(game_dir, GameListItemType::SdmcDir);
-            emit DirEntryReady(game_list_dir);
+            DirEntryReady(game_list_dir);
             AddTitlesToGameList(game_list_dir);
         } else if (game_dir.path == QStringLiteral("UserNAND")) {
             auto* const game_list_dir = new GameListDir(game_dir, GameListItemType::UserNandDir);
-            emit DirEntryReady(game_list_dir);
+            DirEntryReady(game_list_dir);
             AddTitlesToGameList(game_list_dir);
         } else if (game_dir.path == QStringLiteral("SysNAND")) {
             auto* const game_list_dir = new GameListDir(game_dir, GameListItemType::SysNandDir);
-            emit DirEntryReady(game_list_dir);
+            DirEntryReady(game_list_dir);
             AddTitlesToGameList(game_list_dir);
         } else {
             watch_list.append(game_dir.path);
             auto* const game_list_dir = new GameListDir(game_dir);
-            emit DirEntryReady(game_list_dir);
+            DirEntryReady(game_list_dir);
             ScanFileSystem(ScanTarget::FillManualContentProvider, game_dir.path.toStdString(),
                            game_dir.deep_scan, game_list_dir);
             ScanFileSystem(ScanTarget::PopulateGameList, game_dir.path.toStdString(),
@@ -419,10 +479,6 @@ void GameListWorker::run() {
         }
     }
 
-    emit Finished(watch_list);
-}
-
-void GameListWorker::Cancel() {
-    this->disconnect();
-    stop_processing = true;
+    RecordEvent([=](GameList* game_list) { game_list->DonePopulating(watch_list); });
+    processing_completed.Set();
 }
diff --git a/src/yuzu/game_list_worker.h b/src/yuzu/game_list_worker.h
index 24a4e92c3..d5990fcde 100644
--- a/src/yuzu/game_list_worker.h
+++ b/src/yuzu/game_list_worker.h
@@ -4,6 +4,7 @@
 #pragma once
 
 #include <atomic>
+#include <deque>
 #include <memory>
 #include <string>
 
@@ -12,12 +13,15 @@
 #include <QRunnable>
 #include <QString>
 
+#include "common/thread.h"
 #include "yuzu/compatibility_list.h"
+#include "yuzu/play_time_manager.h"
 
 namespace Core {
 class System;
 }
 
+class GameList;
 class QStandardItem;
 
 namespace FileSys {
@@ -36,30 +40,30 @@ public:
     explicit GameListWorker(std::shared_ptr<FileSys::VfsFilesystem> vfs_,
                             FileSys::ManualContentProvider* provider_,
                             QVector<UISettings::GameDir>& game_dirs_,
-                            const CompatibilityList& compatibility_list_, Core::System& system_);
+                            const CompatibilityList& compatibility_list_,
+                            const PlayTime::PlayTimeManager& play_time_manager_,
+                            Core::System& system_);
     ~GameListWorker() override;
 
     /// Starts the processing of directory tree information.
     void run() override;
 
-    /// Tells the worker that it should no longer continue processing. Thread-safe.
-    void Cancel();
-
-signals:
+public:
     /**
-     * The `EntryReady` signal is emitted once an entry has been prepared and is ready
-     * to be added to the game list.
-     * @param entry_items a list with `QStandardItem`s that make up the columns of the new
-     * entry.
+     * Synchronously processes any events queued by the worker.
+     *
+     * AddDirEntry is called on the game list for every discovered directory.
+     * AddEntry is called on the game list for every discovered program.
+     * DonePopulating is called on the game list when processing completes.
      */
-    void DirEntryReady(GameListDir* entry_items);
-    void EntryReady(QList<QStandardItem*> entry_items, GameListDir* parent_dir);
+    void ProcessEvents(GameList* game_list);
 
-    /**
-     * After the worker has traversed the game directory looking for entries, this signal is
-     * emitted with a list of folders that should be watched for changes as well.
-     */
-    void Finished(QStringList watch_list);
+signals:
+    void DataAvailable();
+
+private:
+    template <typename F>
+    void RecordEvent(F&& func);
 
 private:
     void AddTitlesToGameList(GameListDir* parent_dir);
@@ -76,9 +80,15 @@ private:
     FileSys::ManualContentProvider* provider;
     QVector<UISettings::GameDir>& game_dirs;
     const CompatibilityList& compatibility_list;
+    const PlayTime::PlayTimeManager& play_time_manager;
 
     QStringList watch_list;
-    std::atomic_bool stop_processing;
+
+    std::mutex lock;
+    std::condition_variable cv;
+    std::deque<std::function<void(GameList*)>> queued_events;
+    std::atomic_bool stop_requested = false;
+    Common::Event processing_completed;
 
     Core::System& system;
 };
diff --git a/src/yuzu/main.cpp b/src/yuzu/main.cpp
index eb69da4ba..7cc11ae3b 100644
--- a/src/yuzu/main.cpp
+++ b/src/yuzu/main.cpp
@@ -67,6 +67,7 @@ static FileSys::VirtualFile VfsDirectoryCreateFileWrapper(const FileSys::Virtual
 #define QT_NO_OPENGL
 #include <QClipboard>
 #include <QDesktopServices>
+#include <QDir>
 #include <QFile>
 #include <QFileDialog>
 #include <QInputDialog>
@@ -76,6 +77,7 @@ static FileSys::VirtualFile VfsDirectoryCreateFileWrapper(const FileSys::Virtual
 #include <QPushButton>
 #include <QScreen>
 #include <QShortcut>
+#include <QStandardPaths>
 #include <QStatusBar>
 #include <QString>
 #include <QSysInfo>
@@ -98,6 +100,7 @@ static FileSys::VirtualFile VfsDirectoryCreateFileWrapper(const FileSys::Virtual
 #include "common/scm_rev.h"
 #include "common/scope_exit.h"
 #ifdef _WIN32
+#include <shlobj.h>
 #include "common/windows/timer_resolution.h"
 #endif
 #ifdef ARCHITECTURE_x86_64
@@ -150,6 +153,7 @@ static FileSys::VirtualFile VfsDirectoryCreateFileWrapper(const FileSys::Virtual
 #include "yuzu/install_dialog.h"
 #include "yuzu/loading_screen.h"
 #include "yuzu/main.h"
+#include "yuzu/play_time_manager.h"
 #include "yuzu/startup_checks.h"
 #include "yuzu/uisettings.h"
 #include "yuzu/util/clickable_label.h"
@@ -207,7 +211,7 @@ void GMainWindow::ShowTelemetryCallout() {
         tr("<a href='https://yuzu-emu.org/help/feature/telemetry/'>Anonymous "
            "data is collected</a> to help improve yuzu. "
            "<br/><br/>Would you like to share your usage data with us?");
-    if (QMessageBox::question(this, tr("Telemetry"), telemetry_message) != QMessageBox::Yes) {
+    if (!question(this, tr("Telemetry"), telemetry_message)) {
         Settings::values.enable_telemetry = false;
         system->ApplySettings();
     }
@@ -338,6 +342,8 @@ GMainWindow::GMainWindow(std::unique_ptr<Config> config_, bool has_broken_vulkan
     SetDiscordEnabled(UISettings::values.enable_discord_presence.GetValue());
     discord_rpc->Update();
 
+    play_time_manager = std::make_unique<PlayTime::PlayTimeManager>();
+
     system->GetRoomNetwork().Init();
 
     RegisterMetaTypes();
@@ -986,7 +992,7 @@ void GMainWindow::InitializeWidgets() {
     render_window = new GRenderWindow(this, emu_thread.get(), input_subsystem, *system);
     render_window->hide();
 
-    game_list = new GameList(vfs, provider.get(), *system, this);
+    game_list = new GameList(vfs, provider.get(), *play_time_manager, *system, this);
     ui->horizontalLayout->addWidget(game_list);
 
     game_list_placeholder = new GameListPlaceholder(this);
@@ -1447,6 +1453,7 @@ void GMainWindow::OnAppFocusStateChanged(Qt::ApplicationState state) {
             Settings::values.audio_muted = false;
             auto_muted = false;
         }
+        UpdateVolumeUI();
     }
 }
 
@@ -1460,6 +1467,8 @@ void GMainWindow::ConnectWidgetEvents() {
     connect(game_list, &GameList::RemoveInstalledEntryRequested, this,
             &GMainWindow::OnGameListRemoveInstalledEntry);
     connect(game_list, &GameList::RemoveFileRequested, this, &GMainWindow::OnGameListRemoveFile);
+    connect(game_list, &GameList::RemovePlayTimeRequested, this,
+            &GMainWindow::OnGameListRemovePlayTimeData);
     connect(game_list, &GameList::DumpRomFSRequested, this, &GMainWindow::OnGameListDumpRomFS);
     connect(game_list, &GameList::VerifyIntegrityRequested, this,
             &GMainWindow::OnGameListVerifyIntegrity);
@@ -1553,6 +1562,7 @@ void GMainWindow::ConnectMenuEvents() {
     // Tools
     connect_menu(ui->action_Rederive, std::bind(&GMainWindow::OnReinitializeKeys, this,
                                                 ReinitializeKeyBehavior::Warning));
+    connect_menu(ui->action_Load_Album, &GMainWindow::OnAlbum);
     connect_menu(ui->action_Load_Cabinet_Nickname_Owner,
                  [this]() { OnCabinet(Service::NFP::CabinetMode::StartNicknameAndOwnerSettings); });
     connect_menu(ui->action_Load_Cabinet_Eraser,
@@ -1590,6 +1600,7 @@ void GMainWindow::UpdateMenuState() {
     };
 
     const std::array applet_actions{
+        ui->action_Load_Album,
         ui->action_Load_Cabinet_Nickname_Owner,
         ui->action_Load_Cabinet_Eraser,
         ui->action_Load_Cabinet_Restorer,
@@ -2008,7 +2019,7 @@ void GMainWindow::BootGame(const QString& filename, u64 program_id, std::size_t
             std::filesystem::path{Common::U16StringFromBuffer(filename.utf16(), filename.size())}
                 .filename());
     }
-    const bool is_64bit = system->Kernel().ApplicationProcess()->Is64BitProcess();
+    const bool is_64bit = system->Kernel().ApplicationProcess()->Is64Bit();
     const auto instruction_set_suffix = is_64bit ? tr("(64-bit)") : tr("(32-bit)");
     title_name = tr("%1 %2", "%1 is the title name. %2 indicates if the title is 64-bit or 32-bit")
                      .arg(QString::fromStdString(title_name), instruction_set_suffix)
@@ -2409,9 +2420,8 @@ void GMainWindow::OnGameListRemoveInstalledEntry(u64 program_id, InstalledEntryT
         }
     }();
 
-    if (QMessageBox::question(this, tr("Remove Entry"), entry_question,
-                              QMessageBox::Yes | QMessageBox::No,
-                              QMessageBox::No) != QMessageBox::Yes) {
+    if (!question(this, tr("Remove Entry"), entry_question, QMessageBox::Yes | QMessageBox::No,
+                  QMessageBox::No)) {
         return;
     }
 
@@ -2510,8 +2520,8 @@ void GMainWindow::OnGameListRemoveFile(u64 program_id, GameListRemoveTarget targ
         }
     }();
 
-    if (QMessageBox::question(this, tr("Remove File"), question, QMessageBox::Yes | QMessageBox::No,
-                              QMessageBox::No) != QMessageBox::Yes) {
+    if (!GMainWindow::question(this, tr("Remove File"), question,
+                               QMessageBox::Yes | QMessageBox::No, QMessageBox::No)) {
         return;
     }
 
@@ -2534,6 +2544,17 @@ void GMainWindow::OnGameListRemoveFile(u64 program_id, GameListRemoveTarget targ
     }
 }
 
+void GMainWindow::OnGameListRemovePlayTimeData(u64 program_id) {
+    if (QMessageBox::question(this, tr("Remove Play Time Data"), tr("Reset play time?"),
+                              QMessageBox::Yes | QMessageBox::No,
+                              QMessageBox::No) != QMessageBox::Yes) {
+        return;
+    }
+
+    play_time_manager->ResetProgramPlayTime(program_id);
+    game_list->PopulateAsync(UISettings::values.game_dirs);
+}
+
 void GMainWindow::RemoveTransferableShaderCache(u64 program_id, GameListRemoveTarget target) {
     const auto target_file_name = [target] {
         switch (target) {
@@ -2825,7 +2846,6 @@ void GMainWindow::OnGameListCreateShortcut(u64 program_id, const std::string& ga
     const QStringList args = QApplication::arguments();
     std::filesystem::path yuzu_command = args[0].toStdString();
 
-#if defined(__linux__) || defined(__FreeBSD__)
     // If relative path, make it an absolute path
     if (yuzu_command.c_str()[0] == '.') {
         yuzu_command = Common::FS::GetCurrentDir() / yuzu_command;
@@ -2848,44 +2868,52 @@ void GMainWindow::OnGameListCreateShortcut(u64 program_id, const std::string& ga
         UISettings::values.shortcut_already_warned = true;
     }
 #endif // __linux__
-#endif // __linux__ || __FreeBSD__
 
     std::filesystem::path target_directory{};
-    // Determine target directory for shortcut
-#if defined(__linux__) || defined(__FreeBSD__)
-    const char* home = std::getenv("HOME");
-    const std::filesystem::path home_path = (home == nullptr ? "~" : home);
-    const char* xdg_data_home = std::getenv("XDG_DATA_HOME");
 
-    if (target == GameListShortcutTarget::Desktop) {
-        target_directory = home_path / "Desktop";
-        if (!Common::FS::IsDir(target_directory)) {
-            QMessageBox::critical(
-                this, tr("Create Shortcut"),
-                tr("Cannot create shortcut on desktop. Path \"%1\" does not exist.")
-                    .arg(QString::fromStdString(target_directory)),
-                QMessageBox::StandardButton::Ok);
-            return;
-        }
-    } else if (target == GameListShortcutTarget::Applications) {
-        target_directory = (xdg_data_home == nullptr ? home_path / ".local/share" : xdg_data_home) /
-                           "applications";
-        if (!Common::FS::CreateDirs(target_directory)) {
-            QMessageBox::critical(this, tr("Create Shortcut"),
-                                  tr("Cannot create shortcut in applications menu. Path \"%1\" "
-                                     "does not exist and cannot be created.")
-                                      .arg(QString::fromStdString(target_directory)),
-                                  QMessageBox::StandardButton::Ok);
-            return;
+    switch (target) {
+    case GameListShortcutTarget::Desktop: {
+        const QString desktop_path =
+            QStandardPaths::writableLocation(QStandardPaths::DesktopLocation);
+        target_directory = desktop_path.toUtf8().toStdString();
+        break;
+    }
+    case GameListShortcutTarget::Applications: {
+        const QString applications_path =
+            QStandardPaths::writableLocation(QStandardPaths::ApplicationsLocation);
+        if (applications_path.isEmpty()) {
+            const char* home = std::getenv("HOME");
+            if (home != nullptr) {
+                target_directory = std::filesystem::path(home) / ".local/share/applications";
+            }
+        } else {
+            target_directory = applications_path.toUtf8().toStdString();
         }
+        break;
+    }
+    default:
+        return;
+    }
+
+    const QDir dir(QString::fromStdString(target_directory.generic_string()));
+    if (!dir.exists()) {
+        QMessageBox::critical(this, tr("Create Shortcut"),
+                              tr("Cannot create shortcut. Path \"%1\" does not exist.")
+                                  .arg(QString::fromStdString(target_directory.generic_string())),
+                              QMessageBox::StandardButton::Ok);
+        return;
     }
-#endif
 
     const std::string game_file_name = std::filesystem::path(game_path).filename().string();
     // Determine full paths for icon and shortcut
 #if defined(__linux__) || defined(__FreeBSD__)
+    const char* home = std::getenv("HOME");
+    const std::filesystem::path home_path = (home == nullptr ? "~" : home);
+    const char* xdg_data_home = std::getenv("XDG_DATA_HOME");
+
     std::filesystem::path system_icons_path =
-        (xdg_data_home == nullptr ? home_path / ".local/share/" : xdg_data_home) /
+        (xdg_data_home == nullptr ? home_path / ".local/share/"
+                                  : std::filesystem::path(xdg_data_home)) /
         "icons/hicolor/256x256";
     if (!Common::FS::CreateDirs(system_icons_path)) {
         QMessageBox::critical(
@@ -2901,9 +2929,14 @@ void GMainWindow::OnGameListCreateShortcut(u64 program_id, const std::string& ga
     const std::filesystem::path shortcut_path =
         target_directory / (program_id == 0 ? fmt::format("yuzu-{}.desktop", game_file_name)
                                             : fmt::format("yuzu-{:016X}.desktop", program_id));
+#elif defined(WIN32)
+    std::filesystem::path icons_path =
+        Common::FS::GetYuzuPathString(Common::FS::YuzuPath::IconsDir);
+    std::filesystem::path icon_path =
+        icons_path / ((program_id == 0 ? fmt::format("yuzu-{}.ico", game_file_name)
+                                       : fmt::format("yuzu-{:016X}.ico", program_id)));
 #else
-    const std::filesystem::path icon_path{};
-    const std::filesystem::path shortcut_path{};
+    std::string icon_extension;
 #endif
 
     // Get title from game file
@@ -2928,29 +2961,37 @@ void GMainWindow::OnGameListCreateShortcut(u64 program_id, const std::string& ga
         LOG_WARNING(Frontend, "Could not read icon from {:s}", game_path);
     }
 
-    QImage icon_jpeg =
+    QImage icon_data =
         QImage::fromData(icon_image_file.data(), static_cast<int>(icon_image_file.size()));
 #if defined(__linux__) || defined(__FreeBSD__)
     // Convert and write the icon as a PNG
-    if (!icon_jpeg.save(QString::fromStdString(icon_path.string()))) {
+    if (!icon_data.save(QString::fromStdString(icon_path.string()))) {
         LOG_ERROR(Frontend, "Could not write icon as PNG to file");
     } else {
         LOG_INFO(Frontend, "Wrote an icon to {}", icon_path.string());
     }
+#elif defined(WIN32)
+    if (!SaveIconToFile(icon_path.string(), icon_data)) {
+        LOG_ERROR(Frontend, "Could not write icon to file");
+        return;
+    }
 #endif // __linux__
 
-#if defined(__linux__) || defined(__FreeBSD__)
+#ifdef _WIN32
+    // Replace characters that are illegal in Windows filenames by a dash
+    const std::string illegal_chars = "<>:\"/\\|?*";
+    for (char c : illegal_chars) {
+        std::replace(title.begin(), title.end(), c, '_');
+    }
+    const std::filesystem::path shortcut_path = target_directory / (title + ".lnk").c_str();
+#endif
+
     const std::string comment =
         tr("Start %1 with the yuzu Emulator").arg(QString::fromStdString(title)).toStdString();
     const std::string arguments = fmt::format("-g \"{:s}\"", game_path);
     const std::string categories = "Game;Emulator;Qt;";
     const std::string keywords = "Switch;Nintendo;";
-#else
-    const std::string comment{};
-    const std::string arguments{};
-    const std::string categories{};
-    const std::string keywords{};
-#endif
+
     if (!CreateShortcut(shortcut_path.string(), title, comment, icon_path.string(),
                         yuzu_command.string(), arguments, categories, keywords)) {
         QMessageBox::critical(this, tr("Create Shortcut"),
@@ -3357,6 +3398,9 @@ void GMainWindow::OnStartGame() {
     UpdateMenuState();
     OnTasStateChanged();
 
+    play_time_manager->SetProgramId(system->GetApplicationProcessProgramID());
+    play_time_manager->Start();
+
     discord_rpc->Update();
 }
 
@@ -3364,14 +3408,18 @@ void GMainWindow::OnRestartGame() {
     if (!system->IsPoweredOn()) {
         return;
     }
-    // Make a copy since ShutdownGame edits game_path
-    const auto current_game = QString(current_game_path);
-    ShutdownGame();
-    BootGame(current_game);
+
+    if (ConfirmShutdownGame()) {
+        // Make a copy since ShutdownGame edits game_path
+        const auto current_game = QString(current_game_path);
+        ShutdownGame();
+        BootGame(current_game);
+    }
 }
 
 void GMainWindow::OnPauseGame() {
     emu_thread->SetRunning(false);
+    play_time_manager->Stop();
     UpdateMenuState();
     AllowOSSleep();
 }
@@ -3388,15 +3436,39 @@ void GMainWindow::OnPauseContinueGame() {
 }
 
 void GMainWindow::OnStopGame() {
-    if (system->GetExitLocked() && !ConfirmForceLockedExit()) {
-        return;
+    if (ConfirmShutdownGame()) {
+        play_time_manager->Stop();
+        // Update game list to show new play time
+        game_list->PopulateAsync(UISettings::values.game_dirs);
+        if (OnShutdownBegin()) {
+            OnShutdownBeginDialog();
+        } else {
+            OnEmulationStopped();
+        }
     }
+}
 
-    if (OnShutdownBegin()) {
-        OnShutdownBeginDialog();
+bool GMainWindow::ConfirmShutdownGame() {
+    if (UISettings::values.confirm_before_stopping.GetValue() == ConfirmStop::Ask_Always) {
+        if (system->GetExitLocked()) {
+            if (!ConfirmForceLockedExit()) {
+                return false;
+            }
+        } else {
+            if (!ConfirmChangeGame()) {
+                return false;
+            }
+        }
     } else {
-        OnEmulationStopped();
+        if (UISettings::values.confirm_before_stopping.GetValue() ==
+                ConfirmStop::Ask_Based_On_Game &&
+            system->GetExitLocked()) {
+            if (!ConfirmForceLockedExit()) {
+                return false;
+            }
+        }
     }
+    return true;
 }
 
 void GMainWindow::OnLoadComplete() {
@@ -3776,22 +3848,11 @@ void GMainWindow::OnTasRecord() {
     const bool is_recording = input_subsystem->GetTas()->Record();
     if (!is_recording) {
         is_tas_recording_dialog_active = true;
-        ControllerNavigation* controller_navigation =
-            new ControllerNavigation(system->HIDCore(), this);
-        // Use QMessageBox instead of question so we can link controller navigation
-        QMessageBox* box_dialog = new QMessageBox();
-        box_dialog->setWindowTitle(tr("TAS Recording"));
-        box_dialog->setText(tr("Overwrite file of player 1?"));
-        box_dialog->setStandardButtons(QMessageBox::Yes | QMessageBox::No);
-        box_dialog->setDefaultButton(QMessageBox::Yes);
-        connect(controller_navigation, &ControllerNavigation::TriggerKeyboardEvent,
-                [box_dialog](Qt::Key key) {
-                    QKeyEvent* event = new QKeyEvent(QEvent::KeyPress, key, Qt::NoModifier);
-                    QCoreApplication::postEvent(box_dialog, event);
-                });
-        int res = box_dialog->exec();
-        controller_navigation->UnloadController();
-        input_subsystem->GetTas()->SaveRecording(res == QMessageBox::Yes);
+
+        bool answer = question(this, tr("TAS Recording"), tr("Overwrite file of player 1?"),
+                               QMessageBox::Yes | QMessageBox::No, QMessageBox::Yes);
+
+        input_subsystem->GetTas()->SaveRecording(answer);
         is_tas_recording_dialog_active = false;
     }
     OnTasStateChanged();
@@ -3965,6 +4026,34 @@ bool GMainWindow::CreateShortcut(const std::string& shortcut_path, const std::st
     shortcut_stream.close();
 
     return true;
+#elif defined(WIN32)
+    IShellLinkW* shell_link;
+    auto hres = CoCreateInstance(CLSID_ShellLink, NULL, CLSCTX_INPROC_SERVER, IID_IShellLinkW,
+                                 (void**)&shell_link);
+    if (FAILED(hres)) {
+        return false;
+    }
+    shell_link->SetPath(
+        Common::UTF8ToUTF16W(command).data()); // Path to the object we are referring to
+    shell_link->SetArguments(Common::UTF8ToUTF16W(arguments).data());
+    shell_link->SetDescription(Common::UTF8ToUTF16W(comment).data());
+    shell_link->SetIconLocation(Common::UTF8ToUTF16W(icon_path).data(), 0);
+
+    IPersistFile* persist_file;
+    hres = shell_link->QueryInterface(IID_IPersistFile, (void**)&persist_file);
+    if (FAILED(hres)) {
+        return false;
+    }
+
+    hres = persist_file->Save(Common::UTF8ToUTF16W(shortcut_path).data(), TRUE);
+    if (FAILED(hres)) {
+        return false;
+    }
+
+    persist_file->Release();
+    shell_link->Release();
+
+    return true;
 #endif
     return false;
 }
@@ -4004,6 +4093,29 @@ void GMainWindow::OnLoadAmiibo() {
     LoadAmiibo(filename);
 }
 
+bool GMainWindow::question(QWidget* parent, const QString& title, const QString& text,
+                           QMessageBox::StandardButtons buttons,
+                           QMessageBox::StandardButton defaultButton) {
+
+    QMessageBox* box_dialog = new QMessageBox(parent);
+    box_dialog->setWindowTitle(title);
+    box_dialog->setText(text);
+    box_dialog->setStandardButtons(buttons);
+    box_dialog->setDefaultButton(defaultButton);
+
+    ControllerNavigation* controller_navigation =
+        new ControllerNavigation(system->HIDCore(), box_dialog);
+    connect(controller_navigation, &ControllerNavigation::TriggerKeyboardEvent,
+            [box_dialog](Qt::Key key) {
+                QKeyEvent* event = new QKeyEvent(QEvent::KeyPress, key, Qt::NoModifier);
+                QCoreApplication::postEvent(box_dialog, event);
+            });
+    int res = box_dialog->exec();
+
+    controller_navigation->UnloadController();
+    return res == QMessageBox::Yes;
+}
+
 void GMainWindow::LoadAmiibo(const QString& filename) {
     auto* virtual_amiibo = input_subsystem->GetVirtualAmiibo();
     const QString title = tr("Error loading Amiibo data");
@@ -4157,6 +4269,29 @@ void GMainWindow::OnToggleStatusBar() {
     statusBar()->setVisible(ui->action_Show_Status_Bar->isChecked());
 }
 
+void GMainWindow::OnAlbum() {
+    constexpr u64 AlbumId = 0x010000000000100Dull;
+    auto bis_system = system->GetFileSystemController().GetSystemNANDContents();
+    if (!bis_system) {
+        QMessageBox::warning(this, tr("No firmware available"),
+                             tr("Please install the firmware to use the Album applet."));
+        return;
+    }
+
+    auto album_nca = bis_system->GetEntry(AlbumId, FileSys::ContentRecordType::Program);
+    if (!album_nca) {
+        QMessageBox::warning(this, tr("Album Applet"),
+                             tr("Album applet is not available. Please reinstall firmware."));
+        return;
+    }
+
+    system->GetAppletManager().SetCurrentAppletId(Service::AM::Applets::AppletId::PhotoViewer);
+
+    const auto filename = QString::fromStdString(album_nca->GetFullPath());
+    UISettings::values.roms_path = QFileInfo(filename).path();
+    BootGame(filename);
+}
+
 void GMainWindow::OnCabinet(Service::NFP::CabinetMode mode) {
     constexpr u64 CabinetId = 0x0100000000001002ull;
     auto bis_system = system->GetFileSystemController().GetSystemNANDContents();
@@ -4714,8 +4849,7 @@ bool GMainWindow::ConfirmClose() {
         return true;
     }
     const auto text = tr("Are you sure you want to close yuzu?");
-    const auto answer = QMessageBox::question(this, tr("yuzu"), text);
-    return answer != QMessageBox::No;
+    return question(this, tr("yuzu"), text);
 }
 
 void GMainWindow::closeEvent(QCloseEvent* event) {
@@ -4808,11 +4942,11 @@ bool GMainWindow::ConfirmChangeGame() {
     if (emu_thread == nullptr)
         return true;
 
-    const auto answer = QMessageBox::question(
+    // Use custom question to link controller navigation
+    return question(
         this, tr("yuzu"),
         tr("Are you sure you want to stop the emulation? Any unsaved progress will be lost."),
-        QMessageBox::Yes | QMessageBox::No, QMessageBox::No);
-    return answer != QMessageBox::No;
+        QMessageBox::Yes | QMessageBox::No, QMessageBox::Yes);
 }
 
 bool GMainWindow::ConfirmForceLockedExit() {
@@ -4822,8 +4956,7 @@ bool GMainWindow::ConfirmForceLockedExit() {
     const auto text = tr("The currently running application has requested yuzu to not exit.\n\n"
                          "Would you like to bypass this and exit anyway?");
 
-    const auto answer = QMessageBox::question(this, tr("yuzu"), text);
-    return answer != QMessageBox::No;
+    return question(this, tr("yuzu"), text);
 }
 
 void GMainWindow::RequestGameExit() {
diff --git a/src/yuzu/main.h b/src/yuzu/main.h
index 52028234c..270a40c5f 100644
--- a/src/yuzu/main.h
+++ b/src/yuzu/main.h
@@ -7,6 +7,7 @@
 #include <optional>
 
 #include <QMainWindow>
+#include <QMessageBox>
 #include <QTimer>
 #include <QTranslator>
 
@@ -15,6 +16,7 @@
 #include "input_common/drivers/tas_input.h"
 #include "yuzu/compatibility_list.h"
 #include "yuzu/hotkeys.h"
+#include "yuzu/util/controller_navigation.h"
 
 #ifdef __unix__
 #include <QVariant>
@@ -81,6 +83,10 @@ namespace DiscordRPC {
 class DiscordInterface;
 }
 
+namespace PlayTime {
+class PlayTimeManager;
+}
+
 namespace FileSys {
 class ContentProvider;
 class ManualContentProvider;
@@ -323,6 +329,7 @@ private slots:
     void OnGameListRemoveInstalledEntry(u64 program_id, InstalledEntryType type);
     void OnGameListRemoveFile(u64 program_id, GameListRemoveTarget target,
                               const std::string& game_path);
+    void OnGameListRemovePlayTimeData(u64 program_id);
     void OnGameListDumpRomFS(u64 program_id, const std::string& game_path, DumpRomFSTarget target);
     void OnGameListVerifyIntegrity(const std::string& game_path);
     void OnGameListCopyTID(u64 program_id);
@@ -369,6 +376,7 @@ private slots:
     void ResetWindowSize720();
     void ResetWindowSize900();
     void ResetWindowSize1080();
+    void OnAlbum();
     void OnCabinet(Service::NFP::CabinetMode mode);
     void OnMiiEdit();
     void OnCaptureScreenshot();
@@ -389,6 +397,7 @@ private:
     void RemoveVulkanDriverPipelineCache(u64 program_id);
     void RemoveAllTransferableShaderCaches(u64 program_id);
     void RemoveCustomConfiguration(u64 program_id, const std::string& game_path);
+    void RemovePlayTimeData(u64 program_id);
     void RemoveCacheStorage(u64 program_id);
     bool SelectRomFSDumpTarget(const FileSys::ContentProvider&, u64 program_id,
                                u64* selected_title_id, u8* selected_content_record_type);
@@ -417,6 +426,11 @@ private:
     bool CheckSystemArchiveDecryption();
     bool CheckFirmwarePresence();
     void ConfigureFilesystemProvider(const std::string& filepath);
+    /**
+     * Open (or not) the right confirm dialog based on current setting and game exit lock
+     * @returns true if the player confirmed or the settings do no require it
+     */
+    bool ConfirmShutdownGame();
 
     QString GetTasStateDescription() const;
     bool CreateShortcut(const std::string& shortcut_path, const std::string& title,
@@ -424,10 +438,22 @@ private:
                         const std::string& command, const std::string& arguments,
                         const std::string& categories, const std::string& keywords);
 
+    /**
+     * Mimic the behavior of QMessageBox::question but link controller navigation to the dialog
+     * The only difference is that it returns a boolean.
+     *
+     * @returns true if buttons contains QMessageBox::Yes and the user clicks on the "Yes" button.
+     */
+    bool question(QWidget* parent, const QString& title, const QString& text,
+                  QMessageBox::StandardButtons buttons =
+                      QMessageBox::StandardButtons(QMessageBox::Yes | QMessageBox::No),
+                  QMessageBox::StandardButton defaultButton = QMessageBox::NoButton);
+
     std::unique_ptr<Ui::MainWindow> ui;
 
     std::unique_ptr<Core::System> system;
     std::unique_ptr<DiscordRPC::DiscordInterface> discord_rpc;
+    std::unique_ptr<PlayTime::PlayTimeManager> play_time_manager;
     std::shared_ptr<InputCommon::InputSubsystem> input_subsystem;
 
     MultiplayerState* multiplayer_state = nullptr;
diff --git a/src/yuzu/main.ui b/src/yuzu/main.ui
index 31c3de9ef..88684ffb5 100644
--- a/src/yuzu/main.ui
+++ b/src/yuzu/main.ui
@@ -160,6 +160,7 @@
     <addaction name="action_Verify_installed_contents"/>
     <addaction name="separator"/>
     <addaction name="menu_cabinet_applet"/>
+    <addaction name="action_Load_Album"/>
     <addaction name="action_Load_Mii_Edit"/>
     <addaction name="separator"/>
     <addaction name="action_Capture_Screenshot"/>
@@ -380,6 +381,11 @@
     <string>&amp;Capture Screenshot</string>
    </property>
   </action>
+  <action name="action_Load_Album">
+    <property name="text">
+      <string>Open &amp;Album</string>
+    </property>
+  </action>
   <action name="action_Load_Cabinet_Nickname_Owner">
    <property name="text">
     <string>&amp;Set Nickname and Owner</string>
diff --git a/src/yuzu/play_time_manager.cpp b/src/yuzu/play_time_manager.cpp
new file mode 100644
index 000000000..155c36b7d
--- /dev/null
+++ b/src/yuzu/play_time_manager.cpp
@@ -0,0 +1,179 @@
+// SPDX-FileCopyrightText: 2023 yuzu Emulator Project
+// SPDX-License-Identifier: GPL-2.0-or-later
+
+#include "common/alignment.h"
+#include "common/fs/file.h"
+#include "common/fs/fs.h"
+#include "common/fs/path_util.h"
+#include "common/logging/log.h"
+#include "common/settings.h"
+#include "common/thread.h"
+#include "core/hle/service/acc/profile_manager.h"
+#include "yuzu/play_time_manager.h"
+
+namespace PlayTime {
+
+namespace {
+
+struct PlayTimeElement {
+    ProgramId program_id;
+    PlayTime play_time;
+};
+
+std::optional<std::filesystem::path> GetCurrentUserPlayTimePath() {
+    const Service::Account::ProfileManager manager;
+    const auto uuid = manager.GetUser(static_cast<s32>(Settings::values.current_user));
+    if (!uuid.has_value()) {
+        return std::nullopt;
+    }
+    return Common::FS::GetYuzuPath(Common::FS::YuzuPath::PlayTimeDir) /
+           uuid->RawString().append(".bin");
+}
+
+[[nodiscard]] bool ReadPlayTimeFile(PlayTimeDatabase& out_play_time_db) {
+    const auto filename = GetCurrentUserPlayTimePath();
+
+    if (!filename.has_value()) {
+        LOG_ERROR(Frontend, "Failed to get current user path");
+        return false;
+    }
+
+    out_play_time_db.clear();
+
+    if (Common::FS::Exists(filename.value())) {
+        Common::FS::IOFile file{filename.value(), Common::FS::FileAccessMode::Read,
+                                Common::FS::FileType::BinaryFile};
+        if (!file.IsOpen()) {
+            LOG_ERROR(Frontend, "Failed to open play time file: {}",
+                      Common::FS::PathToUTF8String(filename.value()));
+            return false;
+        }
+
+        const size_t num_elements = file.GetSize() / sizeof(PlayTimeElement);
+        std::vector<PlayTimeElement> elements(num_elements);
+
+        if (file.ReadSpan<PlayTimeElement>(elements) != num_elements) {
+            return false;
+        }
+
+        for (const auto& [program_id, play_time] : elements) {
+            if (program_id != 0) {
+                out_play_time_db[program_id] = play_time;
+            }
+        }
+    }
+
+    return true;
+}
+
+[[nodiscard]] bool WritePlayTimeFile(const PlayTimeDatabase& play_time_db) {
+    const auto filename = GetCurrentUserPlayTimePath();
+
+    if (!filename.has_value()) {
+        LOG_ERROR(Frontend, "Failed to get current user path");
+        return false;
+    }
+
+    Common::FS::IOFile file{filename.value(), Common::FS::FileAccessMode::Write,
+                            Common::FS::FileType::BinaryFile};
+    if (!file.IsOpen()) {
+        LOG_ERROR(Frontend, "Failed to open play time file: {}",
+                  Common::FS::PathToUTF8String(filename.value()));
+        return false;
+    }
+
+    std::vector<PlayTimeElement> elements;
+    elements.reserve(play_time_db.size());
+
+    for (auto& [program_id, play_time] : play_time_db) {
+        if (program_id != 0) {
+            elements.push_back(PlayTimeElement{program_id, play_time});
+        }
+    }
+
+    return file.WriteSpan<PlayTimeElement>(elements) == elements.size();
+}
+
+} // namespace
+
+PlayTimeManager::PlayTimeManager() {
+    if (!ReadPlayTimeFile(database)) {
+        LOG_ERROR(Frontend, "Failed to read play time database! Resetting to default.");
+    }
+}
+
+PlayTimeManager::~PlayTimeManager() {
+    Save();
+}
+
+void PlayTimeManager::SetProgramId(u64 program_id) {
+    running_program_id = program_id;
+}
+
+void PlayTimeManager::Start() {
+    play_time_thread = std::jthread([&](std::stop_token stop_token) { AutoTimestamp(stop_token); });
+}
+
+void PlayTimeManager::Stop() {
+    play_time_thread = {};
+}
+
+void PlayTimeManager::AutoTimestamp(std::stop_token stop_token) {
+    Common::SetCurrentThreadName("PlayTimeReport");
+
+    using namespace std::literals::chrono_literals;
+    using std::chrono::seconds;
+    using std::chrono::steady_clock;
+
+    auto timestamp = steady_clock::now();
+
+    const auto GetDuration = [&]() -> u64 {
+        const auto last_timestamp = std::exchange(timestamp, steady_clock::now());
+        const auto duration = std::chrono::duration_cast<seconds>(timestamp - last_timestamp);
+        return static_cast<u64>(duration.count());
+    };
+
+    while (!stop_token.stop_requested()) {
+        Common::StoppableTimedWait(stop_token, 30s);
+
+        database[running_program_id] += GetDuration();
+        Save();
+    }
+}
+
+void PlayTimeManager::Save() {
+    if (!WritePlayTimeFile(database)) {
+        LOG_ERROR(Frontend, "Failed to update play time database!");
+    }
+}
+
+u64 PlayTimeManager::GetPlayTime(u64 program_id) const {
+    auto it = database.find(program_id);
+    if (it != database.end()) {
+        return it->second;
+    } else {
+        return 0;
+    }
+}
+
+void PlayTimeManager::ResetProgramPlayTime(u64 program_id) {
+    database.erase(program_id);
+    Save();
+}
+
+QString ReadablePlayTime(qulonglong time_seconds) {
+    if (time_seconds == 0) {
+        return {};
+    }
+    const auto time_minutes = std::max(static_cast<double>(time_seconds) / 60, 1.0);
+    const auto time_hours = static_cast<double>(time_seconds) / 3600;
+    const bool is_minutes = time_minutes < 60;
+    const char* unit = is_minutes ? "m" : "h";
+    const auto value = is_minutes ? time_minutes : time_hours;
+
+    return QStringLiteral("%L1 %2")
+        .arg(value, 0, 'f', !is_minutes && time_seconds % 60 != 0)
+        .arg(QString::fromUtf8(unit));
+}
+
+} // namespace PlayTime
diff --git a/src/yuzu/play_time_manager.h b/src/yuzu/play_time_manager.h
new file mode 100644
index 000000000..5f96f3447
--- /dev/null
+++ b/src/yuzu/play_time_manager.h
@@ -0,0 +1,44 @@
+// SPDX-FileCopyrightText: 2023 yuzu Emulator Project
+// SPDX-License-Identifier: GPL-2.0-or-later
+
+#pragma once
+
+#include <QString>
+
+#include <map>
+
+#include "common/common_funcs.h"
+#include "common/common_types.h"
+#include "common/polyfill_thread.h"
+
+namespace PlayTime {
+
+using ProgramId = u64;
+using PlayTime = u64;
+using PlayTimeDatabase = std::map<ProgramId, PlayTime>;
+
+class PlayTimeManager {
+public:
+    explicit PlayTimeManager();
+    ~PlayTimeManager();
+
+    YUZU_NON_COPYABLE(PlayTimeManager);
+    YUZU_NON_MOVEABLE(PlayTimeManager);
+
+    u64 GetPlayTime(u64 program_id) const;
+    void ResetProgramPlayTime(u64 program_id);
+    void SetProgramId(u64 program_id);
+    void Start();
+    void Stop();
+
+private:
+    PlayTimeDatabase database;
+    u64 running_program_id;
+    std::jthread play_time_thread;
+    void AutoTimestamp(std::stop_token stop_token);
+    void Save();
+};
+
+QString ReadablePlayTime(qulonglong time_seconds);
+
+} // namespace PlayTime
diff --git a/src/yuzu/uisettings.h b/src/yuzu/uisettings.h
index 8efd63f31..b62ff620c 100644
--- a/src/yuzu/uisettings.h
+++ b/src/yuzu/uisettings.h
@@ -16,7 +16,9 @@
 #include "common/settings_enums.h"
 
 using Settings::Category;
+using Settings::ConfirmStop;
 using Settings::Setting;
+using Settings::SwitchableSetting;
 
 #ifndef CANNOT_EXPLICITLY_INSTANTIATE
 namespace Settings {
@@ -94,6 +96,15 @@ struct Values {
     Setting<bool> confirm_before_closing{
         linkage, true, "confirmClose", Category::UiGeneral, Settings::Specialization::Default,
         true,    true};
+
+    SwitchableSetting<ConfirmStop> confirm_before_stopping{linkage,
+                                                           ConfirmStop::Ask_Always,
+                                                           "confirmStop",
+                                                           Category::UiGeneral,
+                                                           Settings::Specialization::Default,
+                                                           true,
+                                                           true};
+
     Setting<bool> first_start{linkage, true, "firstStart", Category::Ui};
     Setting<bool> pause_when_in_background{linkage,
                                            false,
@@ -103,7 +114,7 @@ struct Values {
                                            true,
                                            true};
     Setting<bool> mute_when_in_background{
-        linkage, false, "muteWhenInBackground", Category::Ui, Settings::Specialization::Default,
+        linkage, false, "muteWhenInBackground", Category::Audio, Settings::Specialization::Default,
         true,    true};
     Setting<bool> hide_mouse{
         linkage, true, "hideInactiveMouse", Category::UiGeneral, Settings::Specialization::Default,
@@ -183,6 +194,9 @@ struct Values {
     Setting<bool> show_size{linkage, true, "show_size", Category::UiGameList};
     Setting<bool> show_types{linkage, true, "show_types", Category::UiGameList};
 
+    // Play time
+    Setting<bool> show_play_time{linkage, true, "show_play_time", Category::UiGameList};
+
     bool configuration_applied;
     bool reset_to_defaults;
     bool shortcut_already_warned{false};
diff --git a/src/yuzu/util/util.cpp b/src/yuzu/util/util.cpp
index 5c3e4589e..f2854c8ec 100644
--- a/src/yuzu/util/util.cpp
+++ b/src/yuzu/util/util.cpp
@@ -5,6 +5,10 @@
 #include <cmath>
 #include <QPainter>
 #include "yuzu/util/util.h"
+#ifdef _WIN32
+#include <windows.h>
+#include "common/fs/file.h"
+#endif
 
 QFont GetMonospaceFont() {
     QFont font(QStringLiteral("monospace"));
@@ -37,3 +41,101 @@ QPixmap CreateCirclePixmapFromColor(const QColor& color) {
     painter.drawEllipse({circle_pixmap.width() / 2.0, circle_pixmap.height() / 2.0}, 7.0, 7.0);
     return circle_pixmap;
 }
+
+bool SaveIconToFile(const std::string_view path, const QImage& image) {
+#if defined(WIN32)
+#pragma pack(push, 2)
+    struct IconDir {
+        WORD id_reserved;
+        WORD id_type;
+        WORD id_count;
+    };
+
+    struct IconDirEntry {
+        BYTE width;
+        BYTE height;
+        BYTE color_count;
+        BYTE reserved;
+        WORD planes;
+        WORD bit_count;
+        DWORD bytes_in_res;
+        DWORD image_offset;
+    };
+#pragma pack(pop)
+
+    const QImage source_image = image.convertToFormat(QImage::Format_RGB32);
+    constexpr std::array<int, 7> scale_sizes{256, 128, 64, 48, 32, 24, 16};
+    constexpr int bytes_per_pixel = 4;
+
+    const IconDir icon_dir{
+        .id_reserved = 0,
+        .id_type = 1,
+        .id_count = static_cast<WORD>(scale_sizes.size()),
+    };
+
+    Common::FS::IOFile icon_file(path, Common::FS::FileAccessMode::Write,
+                                 Common::FS::FileType::BinaryFile);
+    if (!icon_file.IsOpen()) {
+        return false;
+    }
+
+    if (!icon_file.Write(icon_dir)) {
+        return false;
+    }
+
+    std::size_t image_offset = sizeof(IconDir) + (sizeof(IconDirEntry) * scale_sizes.size());
+    for (std::size_t i = 0; i < scale_sizes.size(); i++) {
+        const int image_size = scale_sizes[i] * scale_sizes[i] * bytes_per_pixel;
+        const IconDirEntry icon_entry{
+            .width = static_cast<BYTE>(scale_sizes[i]),
+            .height = static_cast<BYTE>(scale_sizes[i]),
+            .color_count = 0,
+            .reserved = 0,
+            .planes = 1,
+            .bit_count = bytes_per_pixel * 8,
+            .bytes_in_res = static_cast<DWORD>(sizeof(BITMAPINFOHEADER) + image_size),
+            .image_offset = static_cast<DWORD>(image_offset),
+        };
+        image_offset += icon_entry.bytes_in_res;
+        if (!icon_file.Write(icon_entry)) {
+            return false;
+        }
+    }
+
+    for (std::size_t i = 0; i < scale_sizes.size(); i++) {
+        const QImage scaled_image = source_image.scaled(
+            scale_sizes[i], scale_sizes[i], Qt::IgnoreAspectRatio, Qt::SmoothTransformation);
+        const BITMAPINFOHEADER info_header{
+            .biSize = sizeof(BITMAPINFOHEADER),
+            .biWidth = scaled_image.width(),
+            .biHeight = scaled_image.height() * 2,
+            .biPlanes = 1,
+            .biBitCount = bytes_per_pixel * 8,
+            .biCompression = BI_RGB,
+            .biSizeImage{},
+            .biXPelsPerMeter{},
+            .biYPelsPerMeter{},
+            .biClrUsed{},
+            .biClrImportant{},
+        };
+
+        if (!icon_file.Write(info_header)) {
+            return false;
+        }
+
+        for (int y = 0; y < scaled_image.height(); y++) {
+            const auto* line = scaled_image.scanLine(scaled_image.height() - 1 - y);
+            std::vector<u8> line_data(scaled_image.width() * bytes_per_pixel);
+            std::memcpy(line_data.data(), line, line_data.size());
+            if (!icon_file.Write(line_data)) {
+                return false;
+            }
+        }
+    }
+    icon_file.Close();
+
+    return true;
+#else
+    return false;
+#endif
+}
diff --git a/src/yuzu/util/util.h b/src/yuzu/util/util.h
index 39dd2d895..09c14ce3f 100644
--- a/src/yuzu/util/util.h
+++ b/src/yuzu/util/util.h
@@ -7,14 +7,22 @@
 #include <QString>
 
 /// Returns a QFont object appropriate to use as a monospace font for debugging widgets, etc.
-QFont GetMonospaceFont();
+[[nodiscard]] QFont GetMonospaceFont();
 
 /// Convert a size in bytes into a readable format (KiB, MiB, etc.)
-QString ReadableByteSize(qulonglong size);
+[[nodiscard]] QString ReadableByteSize(qulonglong size);
 
 /**
  * Creates a circle pixmap from a specified color
  * @param color The color the pixmap shall have
  * @return QPixmap circle pixmap
  */
-QPixmap CreateCirclePixmapFromColor(const QColor& color);
+[[nodiscard]] QPixmap CreateCirclePixmapFromColor(const QColor& color);
+
+/**
+ * Saves a windows icon to a file
+ * @param path The icons path
+ * @param image The image to save
+ * @return bool If the operation succeeded
+ */
+[[nodiscard]] bool SaveIconToFile(const std::string_view path, const QImage& image);