102 files changed, 5463 insertions, 1609 deletions

diff --git a/.gitmodules b/.gitmodules
index 9d9356151..6931a641d 100644
--- a/.gitmodules
+++ b/.gitmodules
@@ -1,6 +1,6 @@
 [submodule "inih"]
     path = externals/inih/inih
-    url = https://github.com/svn2github/inih
+    url = https://github.com/benhoyt/inih.git
 [submodule "cubeb"]
     path = externals/cubeb
     url = https://github.com/kinetiknz/cubeb.git
diff --git a/CMakeLists.txt b/CMakeLists.txt
index 0479efaab..1bd9d2aa7 100644
--- a/CMakeLists.txt
+++ b/CMakeLists.txt
@@ -263,6 +263,7 @@ if (CONAN_REQUIRED_LIBS)
         libzip:with_openssl=False
         libzip:enable_windows_crypto=False
     )
+
     conan_check(VERSION 1.24.0 REQUIRED)
     # Add the bincrafters remote
     conan_add_remote(NAME bincrafters
@@ -354,6 +355,19 @@ if (NOT LIBUSB_FOUND)
     set(LIBUSB_LIBRARIES usb)
 endif()
 
+# Use system installed ffmpeg.
+if (NOT MSVC)
+    find_package(FFmpeg REQUIRED)
+else()
+    set(FFMPEG_EXT_NAME "ffmpeg-4.2.1")
+    set(FFMPEG_PATH "${CMAKE_BINARY_DIR}/externals/${FFMPEG_EXT_NAME}")
+    download_bundled_external("ffmpeg/" ${FFMPEG_EXT_NAME} "")
+    set(FFMPEG_FOUND YES)
+    set(FFMPEG_INCLUDE_DIR "${FFMPEG_PATH}/include" CACHE PATH "Path to FFmpeg headers" FORCE)
+    set(FFMPEG_LIBRARY_DIR "${FFMPEG_PATH}/bin" CACHE PATH "Path to FFmpeg library" FORCE)
+    set(FFMPEG_DLL_DIR "${FFMPEG_PATH}/bin" CACHE PATH "Path to FFmpeg dll's" FORCE)
+endif()
+
 # Prefer the -pthread flag on Linux.
 set(THREADS_PREFER_PTHREAD_FLAG ON)
 find_package(Threads REQUIRED)
diff --git a/CMakeModules/CopyYuzuFFmpegDeps.cmake b/CMakeModules/CopyYuzuFFmpegDeps.cmake
new file mode 100644
index 000000000..cca1eeeab
--- /dev/null
+++ b/CMakeModules/CopyYuzuFFmpegDeps.cmake
@@ -0,0 +1,10 @@
+function(copy_yuzu_FFmpeg_deps target_dir)
+    include(WindowsCopyFiles)
+    set(DLL_DEST "${CMAKE_BINARY_DIR}/bin/$<CONFIG>/")
+    windows_copy_files(${target_dir} ${FFMPEG_DLL_DIR} ${DLL_DEST}
+        avcodec-58.dll
+        avutil-56.dll
+        swresample-3.dll
+        swscale-5.dll
+    )
+endfunction(copy_yuzu_FFmpeg_deps)
diff --git a/externals/CMakeLists.txt b/externals/CMakeLists.txt
index e01ff6930..421b35890 100644
--- a/externals/CMakeLists.txt
+++ b/externals/CMakeLists.txt
@@ -73,17 +73,20 @@ if (NOT LIBZIP_FOUND)
 endif()
 
 if (ENABLE_WEB_SERVICE)
-    # LibreSSL
-    set(LIBRESSL_SKIP_INSTALL ON CACHE BOOL "")
-    add_subdirectory(libressl EXCLUDE_FROM_ALL)
-    target_include_directories(ssl INTERFACE ./libressl/include)
-    target_compile_definitions(ssl PRIVATE -DHAVE_INET_NTOP)
-    get_directory_property(OPENSSL_LIBRARIES
-        DIRECTORY libressl
-        DEFINITION OPENSSL_LIBS)
-
-    # lurlparser
-    add_subdirectory(lurlparser EXCLUDE_FROM_ALL)
+    find_package(OpenSSL 1.1)
+    if (OPENSSL_FOUND)
+        set(OPENSSL_LIBRARIES OpenSSL::SSL OpenSSL::Crypto)
+    else()
+        # LibreSSL
+        set(LIBRESSL_SKIP_INSTALL ON CACHE BOOL "")
+        set(OPENSSLDIR "/etc/ssl/")
+        add_subdirectory(libressl EXCLUDE_FROM_ALL)
+        target_include_directories(ssl INTERFACE ./libressl/include)
+        target_compile_definitions(ssl PRIVATE -DHAVE_INET_NTOP)
+        get_directory_property(OPENSSL_LIBRARIES
+            DIRECTORY libressl
+            DEFINITION OPENSSL_LIBS)
+    endif()
 
     # httplib
     add_library(httplib INTERFACE)
diff --git a/externals/find-modules/FindFFmpeg.cmake b/externals/find-modules/FindFFmpeg.cmake
new file mode 100644
index 000000000..77b331e00
--- /dev/null
+++ b/externals/find-modules/FindFFmpeg.cmake
@@ -0,0 +1,100 @@
+# - Try to find ffmpeg libraries (libavcodec, libavformat and libavutil)
+# Once done this will define
+#
+# FFMPEG_FOUND - system has ffmpeg or libav
+# FFMPEG_INCLUDE_DIR - the ffmpeg include directory
+# FFMPEG_LIBRARIES - Link these to use ffmpeg
+# FFMPEG_LIBAVCODEC
+# FFMPEG_LIBAVFORMAT
+# FFMPEG_LIBAVUTIL
+#
+# Copyright (c) 2008 Andreas Schneider <mail@cynapses.org>
+# Modified for other libraries by Lasse Kärkkäinen <tronic>
+# Modified for Hedgewars by Stepik777
+# Modified for FFmpeg-example Tuukka Pasanen 2018
+# Modified for yuzu toastUnlimted 2020
+#
+# Redistribution and use is allowed according to the terms of the New
+# BSD license.
+#
+
+include(FindPackageHandleStandardArgs)
+
+find_package_handle_standard_args(FFMPEG
+  FOUND_VAR FFMPEG_FOUND
+  REQUIRED_VARS
+      FFMPEG_LIBRARY
+      FFMPEG_INCLUDE_DIR
+  VERSION_VAR FFMPEG_VERSION
+)
+
+if(FFMPEG_LIBRARIES AND FFMPEG_INCLUDE_DIR)
+  # in cache already
+  set(FFMPEG_FOUND TRUE)
+else()
+  # use pkg-config to get the directories and then use these values
+  # in the FIND_PATH() and FIND_LIBRARY() calls
+  find_package(PkgConfig)
+  if(PKG_CONFIG_FOUND)
+    pkg_check_modules(_FFMPEG_AVCODEC libavcodec)
+    pkg_check_modules(_FFMPEG_AVUTIL libavutil)
+    pkg_check_modules(_FFMPEG_SWSCALE libswscale)
+  endif()
+
+  find_path(FFMPEG_AVCODEC_INCLUDE_DIR
+    NAMES libavcodec/avcodec.h
+    PATHS ${_FFMPEG_AVCODEC_INCLUDE_DIRS}
+      /usr/include
+      /usr/local/include
+      /opt/local/include
+      /sw/include
+    PATH_SUFFIXES ffmpeg libav)
+
+  find_library(FFMPEG_LIBAVCODEC
+    NAMES avcodec
+    PATHS ${_FFMPEG_AVCODEC_LIBRARY_DIRS}
+      /usr/lib
+      /usr/local/lib
+      /opt/local/lib
+      /sw/lib)
+
+  find_library(FFMPEG_LIBAVUTIL
+    NAMES avutil
+    PATHS ${_FFMPEG_AVUTIL_LIBRARY_DIRS}
+      /usr/lib
+      /usr/local/lib
+      /opt/local/lib
+      /sw/lib)
+
+  find_library(FFMPEG_LIBSWSCALE
+    NAMES swscale
+    PATHS ${_FFMPEG_SWSCALE_LIBRARY_DIRS}
+      /usr/lib
+      /usr/local/lib
+      /opt/local/lib
+      /sw/lib)
+
+  if(FFMPEG_LIBAVCODEC AND FFMPEG_LIBAVUTIL AND FFMPEG_LIBSWSCALE)
+    set(FFMPEG_FOUND TRUE)
+  endif()
+
+  if(FFMPEG_FOUND)
+    set(FFMPEG_INCLUDE_DIR ${FFMPEG_AVCODEC_INCLUDE_DIR})
+    set(FFMPEG_LIBRARIES
+      ${FFMPEG_LIBAVCODEC}
+      ${FFMPEG_LIBAVUTIL}
+      ${FFMPEG_LIBSWSCALE})
+  endif()
+
+  if(FFMPEG_FOUND)
+    if(NOT FFMPEG_FIND_QUIETLY)
+      message(STATUS
+      "Found FFMPEG or Libav: ${FFMPEG_LIBRARIES}, ${FFMPEG_INCLUDE_DIR}")
+    endif()
+  else()
+    if(FFMPEG_FIND_REQUIRED)
+      message(FATAL_ERROR
+      "Could not find libavcodec or libavutil or libswscale")
+    endif()
+  endif()
+endif()
diff --git a/externals/httplib/README.md b/externals/httplib/README.md
index 73037d297..1940e446c 100644
--- a/externals/httplib/README.md
+++ b/externals/httplib/README.md
@@ -1,4 +1,4 @@
-From https://github.com/yhirose/cpp-httplib/tree/fce8e6fefdab4ad48bc5b25c98e5ebfda4f3cf53
+From https://github.com/yhirose/cpp-httplib/tree/ff5677ad197947177c158fe857caff4f0e242045 with https://github.com/yhirose/cpp-httplib/pull/701
 
 MIT License
 
diff --git a/externals/httplib/httplib.h b/externals/httplib/httplib.h
index 5139b7f05..8982054e2 100644
--- a/externals/httplib/httplib.h
+++ b/externals/httplib/httplib.h
@@ -173,6 +173,7 @@ using socket_t = int;
 #define INVALID_SOCKET (-1)
 #endif //_WIN32
 
+#include <algorithm>
 #include <array>
 #include <atomic>
 #include <cassert>
@@ -237,6 +238,27 @@ namespace httplib {
 
 namespace detail {
 
+/*
+ * Backport std::make_unique from C++14.
+ *
+ * NOTE: This code came up with the following stackoverflow post:
+ * https://stackoverflow.com/questions/10149840/c-arrays-and-make-unique
+ *
+ */
+
+template <class T, class... Args>
+typename std::enable_if<!std::is_array<T>::value, std::unique_ptr<T>>::type
+make_unique(Args &&... args) {
+  return std::unique_ptr<T>(new T(std::forward<Args>(args)...));
+}
+
+template <class T>
+typename std::enable_if<std::is_array<T>::value, std::unique_ptr<T>>::type
+make_unique(std::size_t n) {
+  typedef typename std::remove_extent<T>::type RT;
+  return std::unique_ptr<T>(new RT[n]);
+}
+
 struct ci {
   bool operator()(const std::string &s1, const std::string &s2) const {
     return std::lexicographical_compare(
@@ -304,6 +326,10 @@ using ContentProvider =
 using ContentProviderWithoutLength =
     std::function<bool(size_t offset, DataSink &sink)>;
 
+using ContentReceiverWithProgress =
+    std::function<bool(const char *data, size_t data_length, uint64_t offset,
+                       uint64_t total_length)>;
+
 using ContentReceiver =
     std::function<bool(const char *data, size_t data_length)>;
 
@@ -317,14 +343,17 @@ public:
                                              ContentReceiver receiver)>;
 
   ContentReader(Reader reader, MultipartReader multipart_reader)
-      : reader_(reader), multipart_reader_(multipart_reader) {}
+      : reader_(std::move(reader)),
+        multipart_reader_(std::move(multipart_reader)) {}
 
   bool operator()(MultipartContentHeader header,
                   ContentReceiver receiver) const {
-    return multipart_reader_(header, receiver);
+    return multipart_reader_(std::move(header), std::move(receiver));
   }
 
-  bool operator()(ContentReceiver receiver) const { return reader_(receiver); }
+  bool operator()(ContentReceiver receiver) const {
+    return reader_(std::move(receiver));
+  }
 
   Reader reader_;
   MultipartReader multipart_reader_;
@@ -353,7 +382,7 @@ struct Request {
   // for client
   size_t redirect_count = CPPHTTPLIB_REDIRECT_MAX_COUNT;
   ResponseHandler response_handler;
-  ContentReceiver content_receiver;
+  ContentReceiverWithProgress content_receiver;
   size_t content_length = 0;
   ContentProvider content_provider;
   Progress progress;
@@ -475,7 +504,7 @@ public:
 
   void enqueue(std::function<void()> fn) override {
     std::unique_lock<std::mutex> lock(mutex_);
-    jobs_.push_back(fn);
+    jobs_.push_back(std::move(fn));
     cond_.notify_one();
   }
 
@@ -664,8 +693,8 @@ private:
                          ContentReceiver multipart_receiver);
 
   virtual bool process_and_close_socket(socket_t sock);
-  
-  struct MountPointEntry { 
+
+  struct MountPointEntry {
     std::string mount_point;
     std::string base_dir;
     Headers headers;
@@ -704,23 +733,27 @@ enum Error {
   Canceled,
   SSLConnection,
   SSLLoadingCerts,
-  SSLServerVerification
+  SSLServerVerification,
+  UnsupportedMultipartBoundaryChars
 };
 
 class Result {
 public:
-  Result(const std::shared_ptr<Response> &res, Error err)
-      : res_(res), err_(err) {}
+  Result(std::unique_ptr<Response> res, Error err)
+      : res_(std::move(res)), err_(err) {}
   operator bool() const { return res_ != nullptr; }
   bool operator==(std::nullptr_t) const { return res_ == nullptr; }
   bool operator!=(std::nullptr_t) const { return res_ != nullptr; }
   const Response &value() const { return *res_; }
+  Response &value() { return *res_; }
   const Response &operator*() const { return *res_; }
+  Response &operator*() { return *res_; }
   const Response *operator->() const { return res_.get(); }
+  Response *operator->() { return res_.get(); }
   Error error() const { return err_; }
 
 private:
-  std::shared_ptr<Response> res_;
+  std::unique_ptr<Response> res_;
   Error err_;
 };
 
@@ -777,6 +810,8 @@ public:
   Result Post(const char *path, const MultipartFormDataItems &items);
   Result Post(const char *path, const Headers &headers,
               const MultipartFormDataItems &items);
+  Result Post(const char *path, const Headers &headers,
+              const MultipartFormDataItems &items, const std::string &boundary);
 
   Result Put(const char *path);
   Result Put(const char *path, const std::string &body,
@@ -863,7 +898,21 @@ protected:
   };
 
   virtual bool create_and_connect_socket(Socket &socket);
-  virtual void close_socket(Socket &socket, bool process_socket_ret);
+
+  // All of:
+  //   shutdown_ssl
+  //   shutdown_socket
+  //   close_socket
+  // should ONLY be called when socket_mutex_ is locked.
+  // Also, shutdown_ssl and close_socket should also NOT be called concurrently
+  // with a DIFFERENT thread sending requests using that socket.
+  virtual void shutdown_ssl(Socket &socket, bool shutdown_gracefully);
+  void shutdown_socket(Socket &socket);
+  void close_socket(Socket &socket);
+
+  // Similar to shutdown_ssl and close_socket, this should NOT be called
+  // concurrently with a DIFFERENT thread sending requests from the socket
+  void lock_socket_and_shutdown_and_close();
 
   bool process_request(Stream &strm, const Request &req, Response &res,
                        bool close_connection);
@@ -873,7 +922,7 @@ protected:
   void copy_settings(const ClientImpl &rhs);
 
   // Error state
-  mutable Error error_ = Error::Success;
+  mutable std::atomic<Error> error_;
 
   // Socket endoint information
   const std::string host_;
@@ -885,6 +934,11 @@ protected:
   mutable std::mutex socket_mutex_;
   std::recursive_mutex request_mutex_;
 
+  // These are all protected under socket_mutex
+  int socket_requests_in_flight_ = 0;
+  std::thread::id socket_requests_are_from_thread_ = std::thread::id();
+  bool socket_should_be_closed_when_request_is_done_ = false;
+
   // Default headers
   Headers default_headers_;
 
@@ -942,13 +996,13 @@ private:
   bool redirect(const Request &req, Response &res);
   bool handle_request(Stream &strm, const Request &req, Response &res,
                       bool close_connection);
-  void stop_core();
-  std::shared_ptr<Response> send_with_content_provider(
+  std::unique_ptr<Response> send_with_content_provider(
       const char *method, const char *path, const Headers &headers,
       const std::string &body, size_t content_length,
       ContentProvider content_provider, const char *content_type);
 
-  virtual bool process_socket(Socket &socket,
+  // socket is const because this function is called when socket_mutex_ is not locked
+  virtual bool process_socket(const Socket &socket,
                               std::function<bool(Stream &strm)> callback);
   virtual bool is_ssl() const;
 };
@@ -1012,6 +1066,8 @@ public:
   Result Post(const char *path, const MultipartFormDataItems &items);
   Result Post(const char *path, const Headers &headers,
               const MultipartFormDataItems &items);
+  Result Post(const char *path, const Headers &headers,
+              const MultipartFormDataItems &items, const std::string &boundary);
   Result Put(const char *path);
   Result Put(const char *path, const std::string &body,
              const char *content_type);
@@ -1098,7 +1154,7 @@ public:
 #endif
 
 private:
-  std::shared_ptr<ClientImpl> cli_;
+  std::unique_ptr<ClientImpl> cli_;
 
 #ifdef CPPHTTPLIB_OPENSSL_SUPPORT
   bool is_ssl_ = false;
@@ -1154,9 +1210,9 @@ public:
 
 private:
   bool create_and_connect_socket(Socket &socket) override;
-  void close_socket(Socket &socket, bool process_socket_ret) override;
+  void shutdown_ssl(Socket &socket, bool shutdown_gracefully) override;
 
-  bool process_socket(Socket &socket,
+  bool process_socket(const Socket &socket,
                       std::function<bool(Stream &strm)> callback) override;
   bool is_ssl() const override;
 
@@ -1942,9 +1998,9 @@ inline socket_t create_client_socket(const char *host, int port,
                                      bool tcp_nodelay,
                                      SocketOptions socket_options,
                                      time_t timeout_sec, time_t timeout_usec,
-                                     const std::string &intf, Error &error) {
+                                     const std::string &intf, std::atomic<Error> &error) {
   auto sock = create_socket(
-      host, port, 0, tcp_nodelay, socket_options,
+      host, port, 0, tcp_nodelay, std::move(socket_options),
       [&](socket_t sock, struct addrinfo &ai) -> bool {
         if (!intf.empty()) {
 #ifdef USE_IF2IP
@@ -2478,7 +2534,8 @@ inline bool read_headers(Stream &strm, Headers &headers) {
 }
 
 inline bool read_content_with_length(Stream &strm, uint64_t len,
-                                     Progress progress, ContentReceiver out) {
+                                     Progress progress,
+                                     ContentReceiverWithProgress out) {
   char buf[CPPHTTPLIB_RECV_BUFSIZ];
 
   uint64_t r = 0;
@@ -2487,8 +2544,7 @@ inline bool read_content_with_length(Stream &strm, uint64_t len,
     auto n = strm.read(buf, (std::min)(read_len, CPPHTTPLIB_RECV_BUFSIZ));
     if (n <= 0) { return false; }
 
-    if (!out(buf, static_cast<size_t>(n))) { return false; }
-
+    if (!out(buf, static_cast<size_t>(n), r, len)) { return false; }
     r += static_cast<uint64_t>(n);
 
     if (progress) {
@@ -2510,8 +2566,10 @@ inline void skip_content_with_length(Stream &strm, uint64_t len) {
   }
 }
 
-inline bool read_content_without_length(Stream &strm, ContentReceiver out) {
+inline bool read_content_without_length(Stream &strm,
+                                        ContentReceiverWithProgress out) {
   char buf[CPPHTTPLIB_RECV_BUFSIZ];
+  uint64_t r = 0;
   for (;;) {
     auto n = strm.read(buf, CPPHTTPLIB_RECV_BUFSIZ);
     if (n < 0) {
@@ -2519,13 +2577,16 @@ inline bool read_content_without_length(Stream &strm, ContentReceiver out) {
     } else if (n == 0) {
       return true;
     }
-    if (!out(buf, static_cast<size_t>(n))) { return false; }
+
+    if (!out(buf, static_cast<size_t>(n), r, 0)) { return false; }
+    r += static_cast<uint64_t>(n);
   }
 
   return true;
 }
 
-inline bool read_content_chunked(Stream &strm, ContentReceiver out) {
+inline bool read_content_chunked(Stream &strm,
+                                 ContentReceiverWithProgress out) {
   const auto bufsiz = 16;
   char buf[bufsiz];
 
@@ -2570,23 +2631,24 @@ inline bool is_chunked_transfer_encoding(const Headers &headers) {
 }
 
 template <typename T, typename U>
-bool prepare_content_receiver(T &x, int &status, ContentReceiver receiver,
+bool prepare_content_receiver(T &x, int &status,
+                              ContentReceiverWithProgress receiver,
                               bool decompress, U callback) {
   if (decompress) {
     std::string encoding = x.get_header_value("Content-Encoding");
-    std::shared_ptr<decompressor> decompressor;
+    std::unique_ptr<decompressor> decompressor;
 
     if (encoding.find("gzip") != std::string::npos ||
         encoding.find("deflate") != std::string::npos) {
 #ifdef CPPHTTPLIB_ZLIB_SUPPORT
-      decompressor = std::make_shared<gzip_decompressor>();
+      decompressor = detail::make_unique<gzip_decompressor>();
 #else
       status = 415;
       return false;
 #endif
     } else if (encoding.find("br") != std::string::npos) {
 #ifdef CPPHTTPLIB_BROTLI_SUPPORT
-      decompressor = std::make_shared<brotli_decompressor>();
+      decompressor = detail::make_unique<brotli_decompressor>();
 #else
       status = 415;
       return false;
@@ -2595,12 +2657,14 @@ bool prepare_content_receiver(T &x, int &status, ContentReceiver receiver,
 
     if (decompressor) {
       if (decompressor->is_valid()) {
-        ContentReceiver out = [&](const char *buf, size_t n) {
-          return decompressor->decompress(
-              buf, n,
-              [&](const char *buf, size_t n) { return receiver(buf, n); });
+        ContentReceiverWithProgress out = [&](const char *buf, size_t n,
+                                              uint64_t off, uint64_t len) {
+          return decompressor->decompress(buf, n,
+                                          [&](const char *buf, size_t n) {
+                                            return receiver(buf, n, off, len);
+                                          });
         };
-        return callback(out);
+        return callback(std::move(out));
       } else {
         status = 500;
         return false;
@@ -2608,18 +2672,20 @@ bool prepare_content_receiver(T &x, int &status, ContentReceiver receiver,
     }
   }
 
-  ContentReceiver out = [&](const char *buf, size_t n) {
-    return receiver(buf, n);
+  ContentReceiverWithProgress out = [&](const char *buf, size_t n, uint64_t off,
+                                        uint64_t len) {
+    return receiver(buf, n, off, len);
   };
-  return callback(out);
+  return callback(std::move(out));
 }
 
 template <typename T>
 bool read_content(Stream &strm, T &x, size_t payload_max_length, int &status,
-                  Progress progress, ContentReceiver receiver,
+                  Progress progress, ContentReceiverWithProgress receiver,
                   bool decompress) {
   return prepare_content_receiver(
-      x, status, receiver, decompress, [&](const ContentReceiver &out) {
+      x, status, std::move(receiver), decompress,
+      [&](const ContentReceiverWithProgress &out) {
         auto ret = true;
         auto exceed_payload_max_length = false;
 
@@ -2634,7 +2700,7 @@ bool read_content(Stream &strm, T &x, size_t payload_max_length, int &status,
             skip_content_with_length(strm, len);
             ret = false;
           } else if (len > 0) {
-            ret = read_content_with_length(strm, len, progress, out);
+            ret = read_content_with_length(strm, len, std::move(progress), out);
           }
         }
 
@@ -2875,7 +2941,7 @@ inline bool parse_multipart_boundary(const std::string &content_type,
   return !boundary.empty();
 }
 
-inline bool parse_range_header(const std::string &s, Ranges &ranges) {
+inline bool parse_range_header(const std::string &s, Ranges &ranges) try {
   static auto re_first_range = std::regex(R"(bytes=(\d*-\d*(?:,\s*\d*-\d*)*))");
   std::smatch m;
   if (std::regex_match(s, m, re_first_range)) {
@@ -2907,7 +2973,7 @@ inline bool parse_range_header(const std::string &s, Ranges &ranges) {
     return all_valid_ranges;
   }
   return false;
-}
+} catch (...) { return false; }
 
 class MultipartFormDataParser {
 public:
@@ -2918,7 +2984,8 @@ public:
   bool is_valid() const { return is_valid_; }
 
   template <typename T, typename U>
-  bool parse(const char *buf, size_t n, T content_callback, U header_callback) {
+  bool parse(const char *buf, size_t n, const T &content_callback,
+             const U &header_callback) {
 
     static const std::regex re_content_disposition(
         "^Content-Disposition:\\s*form-data;\\s*name=\"(.*?)\"(?:;\\s*filename="
@@ -3090,8 +3157,13 @@ inline std::string make_multipart_data_boundary() {
   static const char data[] =
       "0123456789ABCDEFGHIJKLMNOPQRSTUVWXYZabcdefghijklmnopqrstuvwxyz";
 
+  // std::random_device might actually be deterministic on some
+  // platforms, but due to lack of support in the c++ standard library,
+  // doing better requires either some ugly hacks or breaking portability.
   std::random_device seed_gen;
-  std::mt19937 engine(seed_gen());
+  // Request 128 bits of entropy for initialization
+  std::seed_seq seed_sequence{seed_gen(), seed_gen(), seed_gen(), seed_gen()};
+  std::mt19937 engine(seed_sequence);
 
   std::string result = "--cpp-httplib-multipart-data-";
 
@@ -3114,7 +3186,7 @@ get_range_offset_and_length(const Request &req, size_t content_length,
   auto slen = static_cast<ssize_t>(content_length);
 
   if (r.first == -1) {
-    r.first = slen - r.second;
+    r.first = (std::max)(static_cast<ssize_t>(0), slen - r.second);
     r.second = slen - 1;
   }
 
@@ -3451,7 +3523,7 @@ inline std::pair<std::string, std::string> make_range_header(Ranges ranges) {
     if (r.second != -1) { field += std::to_string(r.second); }
     i++;
   }
-  return std::make_pair("Range", field);
+  return std::make_pair("Range", std::move(field));
 }
 
 inline std::pair<std::string, std::string>
@@ -3460,7 +3532,7 @@ make_basic_authentication_header(const std::string &username,
                                  bool is_proxy = false) {
   auto field = "Basic " + detail::base64_encode(username + ":" + password);
   auto key = is_proxy ? "Proxy-Authorization" : "Authorization";
-  return std::make_pair(key, field);
+  return std::make_pair(key, std::move(field));
 }
 
 inline std::pair<std::string, std::string>
@@ -3468,7 +3540,7 @@ make_bearer_token_authentication_header(const std::string &token,
                                         bool is_proxy = false) {
   auto field = "Bearer " + token;
   auto key = is_proxy ? "Proxy-Authorization" : "Authorization";
-  return std::make_pair(key, field);
+  return std::make_pair(key, std::move(field));
 }
 
 // Request implementation
@@ -3761,60 +3833,66 @@ inline Server::Server()
 inline Server::~Server() {}
 
 inline Server &Server::Get(const char *pattern, Handler handler) {
-  get_handlers_.push_back(std::make_pair(std::regex(pattern), handler));
+  get_handlers_.push_back(
+      std::make_pair(std::regex(pattern), std::move(handler)));
   return *this;
 }
 
 inline Server &Server::Post(const char *pattern, Handler handler) {
-  post_handlers_.push_back(std::make_pair(std::regex(pattern), handler));
+  post_handlers_.push_back(
+      std::make_pair(std::regex(pattern), std::move(handler)));
   return *this;
 }
 
 inline Server &Server::Post(const char *pattern,
                             HandlerWithContentReader handler) {
   post_handlers_for_content_reader_.push_back(
-      std::make_pair(std::regex(pattern), handler));
+      std::make_pair(std::regex(pattern), std::move(handler)));
   return *this;
 }
 
 inline Server &Server::Put(const char *pattern, Handler handler) {
-  put_handlers_.push_back(std::make_pair(std::regex(pattern), handler));
+  put_handlers_.push_back(
+      std::make_pair(std::regex(pattern), std::move(handler)));
   return *this;
 }
 
 inline Server &Server::Put(const char *pattern,
                            HandlerWithContentReader handler) {
   put_handlers_for_content_reader_.push_back(
-      std::make_pair(std::regex(pattern), handler));
+      std::make_pair(std::regex(pattern), std::move(handler)));
   return *this;
 }
 
 inline Server &Server::Patch(const char *pattern, Handler handler) {
-  patch_handlers_.push_back(std::make_pair(std::regex(pattern), handler));
+  patch_handlers_.push_back(
+      std::make_pair(std::regex(pattern), std::move(handler)));
   return *this;
 }
 
 inline Server &Server::Patch(const char *pattern,
                              HandlerWithContentReader handler) {
   patch_handlers_for_content_reader_.push_back(
-      std::make_pair(std::regex(pattern), handler));
+      std::make_pair(std::regex(pattern), std::move(handler)));
   return *this;
 }
 
 inline Server &Server::Delete(const char *pattern, Handler handler) {
-  delete_handlers_.push_back(std::make_pair(std::regex(pattern), handler));
+  delete_handlers_.push_back(
+      std::make_pair(std::regex(pattern), std::move(handler)));
   return *this;
 }
 
 inline Server &Server::Delete(const char *pattern,
                               HandlerWithContentReader handler) {
   delete_handlers_for_content_reader_.push_back(
-      std::make_pair(std::regex(pattern), handler));
+      std::make_pair(std::regex(pattern), std::move(handler)));
   return *this;
 }
 
 inline Server &Server::Options(const char *pattern, Handler handler) {
-  options_handlers_.push_back(std::make_pair(std::regex(pattern), handler));
+  options_handlers_.push_back(
+      std::make_pair(std::regex(pattern), std::move(handler)));
   return *this;
 }
 
@@ -3860,7 +3938,7 @@ inline void Server::set_error_handler(Handler handler) {
 inline void Server::set_tcp_nodelay(bool on) { tcp_nodelay_ = on; }
 
 inline void Server::set_socket_options(SocketOptions socket_options) {
-  socket_options_ = socket_options;
+  socket_options_ = std::move(socket_options);
 }
 
 inline void Server::set_logger(Logger logger) { logger_ = std::move(logger); }
@@ -4045,16 +4123,16 @@ inline bool Server::write_response(Stream &strm, bool close_connection,
     }
 
     if (type != detail::EncodingType::None) {
-      std::shared_ptr<detail::compressor> compressor;
+      std::unique_ptr<detail::compressor> compressor;
 
       if (type == detail::EncodingType::Gzip) {
 #ifdef CPPHTTPLIB_ZLIB_SUPPORT
-        compressor = std::make_shared<detail::gzip_compressor>();
+        compressor = detail::make_unique<detail::gzip_compressor>();
         res.set_header("Content-Encoding", "gzip");
 #endif
       } else if (type == detail::EncodingType::Brotli) {
 #ifdef CPPHTTPLIB_BROTLI_SUPPORT
-        compressor = std::make_shared<detail::brotli_compressor>();
+        compressor = detail::make_unique<detail::brotli_compressor>();
         res.set_header("Content-Encoding", "brotli");
 #endif
       }
@@ -4136,17 +4214,17 @@ Server::write_content_with_provider(Stream &strm, const Request &req,
     if (res.is_chunked_content_provider) {
       auto type = detail::encoding_type(req, res);
 
-      std::shared_ptr<detail::compressor> compressor;
+      std::unique_ptr<detail::compressor> compressor;
       if (type == detail::EncodingType::Gzip) {
 #ifdef CPPHTTPLIB_ZLIB_SUPPORT
-        compressor = std::make_shared<detail::gzip_compressor>();
+        compressor = detail::make_unique<detail::gzip_compressor>();
 #endif
       } else if (type == detail::EncodingType::Brotli) {
 #ifdef CPPHTTPLIB_BROTLI_SUPPORT
-        compressor = std::make_shared<detail::brotli_compressor>();
+        compressor = detail::make_unique<detail::brotli_compressor>();
 #endif
       } else {
-        compressor = std::make_shared<detail::nocompressor>();
+        compressor = detail::make_unique<detail::nocompressor>();
       }
       assert(compressor != nullptr);
 
@@ -4198,8 +4276,9 @@ inline bool Server::read_content_with_content_receiver(
     Stream &strm, Request &req, Response &res, ContentReceiver receiver,
     MultipartContentHeader multipart_header,
     ContentReceiver multipart_receiver) {
-  return read_content_core(strm, req, res, receiver, multipart_header,
-                           multipart_receiver);
+  return read_content_core(strm, req, res, std::move(receiver),
+                           std::move(multipart_header),
+                           std::move(multipart_receiver));
 }
 
 inline bool Server::read_content_core(Stream &strm, Request &req, Response &res,
@@ -4207,7 +4286,7 @@ inline bool Server::read_content_core(Stream &strm, Request &req, Response &res,
                                       MultipartContentHeader mulitpart_header,
                                       ContentReceiver multipart_receiver) {
   detail::MultipartFormDataParser multipart_form_data_parser;
-  ContentReceiver out;
+  ContentReceiverWithProgress out;
 
   if (req.is_multipart_form_data()) {
     const auto &content_type = req.get_header_value("Content-Type");
@@ -4218,7 +4297,7 @@ inline bool Server::read_content_core(Stream &strm, Request &req, Response &res,
     }
 
     multipart_form_data_parser.set_boundary(std::move(boundary));
-    out = [&](const char *buf, size_t n) {
+    out = [&](const char *buf, size_t n, uint64_t /*off*/, uint64_t /*len*/) {
       /* For debug
       size_t pos = 0;
       while (pos < n) {
@@ -4234,7 +4313,8 @@ inline bool Server::read_content_core(Stream &strm, Request &req, Response &res,
                                               mulitpart_header);
     };
   } else {
-    out = receiver;
+    out = [receiver](const char *buf, size_t n, uint64_t /*off*/,
+                     uint64_t /*len*/) { return receiver(buf, n); };
   }
 
   if (req.method == "DELETE" && !req.has_header("Content-Length")) {
@@ -4271,7 +4351,7 @@ inline bool Server::handle_file_request(Request &req, Response &res,
           auto type =
               detail::find_content_type(path, file_extension_and_mimetype_map_);
           if (type) { res.set_header("Content-Type", type); }
-          for (const auto& kv : entry.headers) {
+          for (const auto &kv : entry.headers) {
             res.set_header(kv.first.c_str(), kv.second);
           }
           res.status = 200;
@@ -4290,7 +4370,7 @@ inline socket_t
 Server::create_server_socket(const char *host, int port, int socket_flags,
                              SocketOptions socket_options) const {
   return detail::create_socket(
-      host, port, socket_flags, tcp_nodelay_, socket_options,
+      host, port, socket_flags, tcp_nodelay_, std::move(socket_options),
       [](socket_t sock, struct addrinfo &ai) -> bool {
         if (::bind(sock, ai.ai_addr, static_cast<socklen_t>(ai.ai_addrlen))) {
           return false;
@@ -4392,32 +4472,37 @@ inline bool Server::routing(Request &req, Response &res, Stream &strm) {
     {
       ContentReader reader(
           [&](ContentReceiver receiver) {
-            return read_content_with_content_receiver(strm, req, res, receiver,
-                                                      nullptr, nullptr);
+            return read_content_with_content_receiver(
+                strm, req, res, std::move(receiver), nullptr, nullptr);
           },
           [&](MultipartContentHeader header, ContentReceiver receiver) {
             return read_content_with_content_receiver(strm, req, res, nullptr,
-                                                      header, receiver);
+                                                      std::move(header),
+                                                      std::move(receiver));
           });
 
       if (req.method == "POST") {
         if (dispatch_request_for_content_reader(
-                req, res, reader, post_handlers_for_content_reader_)) {
+                req, res, std::move(reader),
+                post_handlers_for_content_reader_)) {
           return true;
         }
       } else if (req.method == "PUT") {
         if (dispatch_request_for_content_reader(
-                req, res, reader, put_handlers_for_content_reader_)) {
+                req, res, std::move(reader),
+                put_handlers_for_content_reader_)) {
           return true;
         }
       } else if (req.method == "PATCH") {
         if (dispatch_request_for_content_reader(
-                req, res, reader, patch_handlers_for_content_reader_)) {
+                req, res, std::move(reader),
+                patch_handlers_for_content_reader_)) {
           return true;
         }
       } else if (req.method == "DELETE") {
         if (dispatch_request_for_content_reader(
-                req, res, reader, delete_handlers_for_content_reader_)) {
+                req, res, std::move(reader),
+                delete_handlers_for_content_reader_)) {
           return true;
         }
       }
@@ -4448,7 +4533,6 @@ inline bool Server::routing(Request &req, Response &res, Stream &strm) {
 
 inline bool Server::dispatch_request(Request &req, Response &res,
                                      const Handlers &handlers) {
-
   try {
     for (const auto &x : handlers) {
       const auto &pattern = x.first;
@@ -4531,7 +4615,8 @@ Server::process_request(Stream &strm, bool close_connection,
   if (req.has_header("Range")) {
     const auto &range_header_value = req.get_header_value("Range");
     if (!detail::parse_range_header(range_header_value, req.ranges)) {
-      // TODO: error
+      res.status = 416;
+      return write_response(strm, close_connection, req, res);
     }
   }
 
@@ -4588,11 +4673,11 @@ inline ClientImpl::ClientImpl(const std::string &host, int port)
 inline ClientImpl::ClientImpl(const std::string &host, int port,
                               const std::string &client_cert_path,
                               const std::string &client_key_path)
-    : host_(host), port_(port),
+    : error_(Error::Success), host_(host), port_(port),
       host_and_port_(host_ + ":" + std::to_string(port_)),
       client_cert_path_(client_cert_path), client_key_path_(client_key_path) {}
 
-inline ClientImpl::~ClientImpl() { stop_core(); }
+inline ClientImpl::~ClientImpl() { lock_socket_and_shutdown_and_close(); }
 
 inline bool ClientImpl::is_valid() const { return true; }
 
@@ -4653,15 +4738,47 @@ inline bool ClientImpl::create_and_connect_socket(Socket &socket) {
   return true;
 }
 
-inline void ClientImpl::close_socket(Socket &socket,
-                                     bool /*process_socket_ret*/) {
-  detail::close_socket(socket.sock);
-  socket_.sock = INVALID_SOCKET;
+inline void ClientImpl::shutdown_ssl(Socket &socket, bool shutdown_gracefully) {
+  (void)socket;
+  (void)shutdown_gracefully;
+  //If there are any requests in flight from threads other than us, then it's
+  //a thread-unsafe race because individual ssl* objects are not thread-safe. 
+  assert(socket_requests_in_flight_ == 0 ||
+         socket_requests_are_from_thread_ == std::this_thread::get_id());
+}
+
+inline void ClientImpl::shutdown_socket(Socket &socket) {
+  if (socket.sock == INVALID_SOCKET)
+    return;  
+  detail::shutdown_socket(socket.sock);
+}
+ 
+inline void ClientImpl::close_socket(Socket &socket) {
+  // If there are requests in flight in another thread, usually closing
+  // the socket will be fine and they will simply receive an error when
+  // using the closed socket, but it is still a bug since rarely the OS
+  // may reassign the socket id to be used for a new socket, and then
+  // suddenly they will be operating on a live socket that is different
+  // than the one they intended!
+  assert(socket_requests_in_flight_ == 0 ||
+         socket_requests_are_from_thread_ == std::this_thread::get_id());
+  // It is also a bug if this happens while SSL is still active
 #ifdef CPPHTTPLIB_OPENSSL_SUPPORT
-  socket_.ssl = nullptr;
+  assert(socket.ssl == nullptr);
 #endif
+  if (socket.sock == INVALID_SOCKET)
+    return;
+  detail::close_socket(socket.sock);
+  socket.sock = INVALID_SOCKET;
 }
 
+inline void ClientImpl::lock_socket_and_shutdown_and_close() {
+  std::lock_guard<std::mutex> guard(socket_mutex_);
+  shutdown_ssl(socket_, true);
+  shutdown_socket(socket_);
+  close_socket(socket_);
+}
+ 
 inline bool ClientImpl::read_response_line(Stream &strm, Response &res) {
   std::array<char, 2048> buf;
 
@@ -4696,11 +4813,23 @@ inline bool ClientImpl::send(const Request &req, Response &res) {
 
   {
     std::lock_guard<std::mutex> guard(socket_mutex_);
+    // Set this to false immediately - if it ever gets set to true by the end of the
+    // request, we know another thread instructed us to close the socket.
+    socket_should_be_closed_when_request_is_done_ = false;
 
     auto is_alive = false;
     if (socket_.is_open()) {
       is_alive = detail::select_write(socket_.sock, 0, 0) > 0;
-      if (!is_alive) { close_socket(socket_, false); }
+      if (!is_alive) {
+        // Attempt to avoid sigpipe by shutting down nongracefully if it seems like
+        // the other side has already closed the connection
+        // Also, there cannot be any requests in flight from other threads since we locked
+        // request_mutex_, so safe to close everything immediately
+        const bool shutdown_gracefully = false;
+        shutdown_ssl(socket_, shutdown_gracefully);
+        shutdown_socket(socket_);
+        close_socket(socket_);
+      }
     }
 
     if (!is_alive) {
@@ -4721,15 +4850,38 @@ inline bool ClientImpl::send(const Request &req, Response &res) {
       }
 #endif
     }
+
+    // Mark the current socket as being in use so that it cannot be closed by anyone
+    // else while this request is ongoing, even though we will be releasing the mutex.
+    if (socket_requests_in_flight_ > 1) {
+      assert(socket_requests_are_from_thread_ == std::this_thread::get_id());
+    }
+    socket_requests_in_flight_ += 1;
+    socket_requests_are_from_thread_ = std::this_thread::get_id();
   }
 
   auto close_connection = !keep_alive_;
-
   auto ret = process_socket(socket_, [&](Stream &strm) {
     return handle_request(strm, req, res, close_connection);
   });
 
-  if (close_connection || !ret) { stop_core(); }
+  //Briefly lock mutex in order to mark that a request is no longer ongoing
+  {
+    std::lock_guard<std::mutex> guard(socket_mutex_);
+    socket_requests_in_flight_ -= 1;
+    if (socket_requests_in_flight_ <= 0) {
+      assert(socket_requests_in_flight_ == 0);
+      socket_requests_are_from_thread_ = std::thread::id();
+    }
+
+    if (socket_should_be_closed_when_request_is_done_ ||
+        close_connection ||
+        !ret ) {
+      shutdown_ssl(socket_, true);
+      shutdown_socket(socket_);
+      close_socket(socket_);
+    }
+  }
 
   if (!ret) {
     if (error_ == Error::Success) { error_ = Error::Unknown; }
@@ -4973,7 +5125,7 @@ inline bool ClientImpl::write_request(Stream &strm, const Request &req,
   return true;
 }
 
-inline std::shared_ptr<Response> ClientImpl::send_with_content_provider(
+inline std::unique_ptr<Response> ClientImpl::send_with_content_provider(
     const char *method, const char *path, const Headers &headers,
     const std::string &body, size_t content_length,
     ContentProvider content_provider, const char *content_type) {
@@ -5036,15 +5188,15 @@ inline std::shared_ptr<Response> ClientImpl::send_with_content_provider(
   {
     if (content_provider) {
       req.content_length = content_length;
-      req.content_provider = content_provider;
+      req.content_provider = std::move(content_provider);
     } else {
       req.body = body;
     }
   }
 
-  auto res = std::make_shared<Response>();
+  auto res = detail::make_unique<Response>();
 
-  return send(req, *res) ? res : nullptr;
+  return send(req, *res) ? std::move(res) : nullptr;
 }
 
 inline bool ClientImpl::process_request(Stream &strm, const Request &req,
@@ -5070,16 +5222,21 @@ inline bool ClientImpl::process_request(Stream &strm, const Request &req,
   if (req.method != "HEAD" && req.method != "CONNECT") {
     auto out =
         req.content_receiver
-            ? static_cast<ContentReceiver>([&](const char *buf, size_t n) {
-                auto ret = req.content_receiver(buf, n);
-                if (!ret) { error_ = Error::Canceled; }
-                return ret;
-              })
-            : static_cast<ContentReceiver>([&](const char *buf, size_t n) {
-                if (res.body.size() + n > res.body.max_size()) { return false; }
-                res.body.append(buf, n);
-                return true;
-              });
+            ? static_cast<ContentReceiverWithProgress>(
+                  [&](const char *buf, size_t n, uint64_t off, uint64_t len) {
+                    auto ret = req.content_receiver(buf, n, off, len);
+                    if (!ret) { error_ = Error::Canceled; }
+                    return ret;
+                  })
+            : static_cast<ContentReceiverWithProgress>(
+                  [&](const char *buf, size_t n, uint64_t /*off*/,
+                      uint64_t /*len*/) {
+                    if (res.body.size() + n > res.body.max_size()) {
+                      return false;
+                    }
+                    res.body.append(buf, n);
+                    return true;
+                  });
 
     auto progress = [&](uint64_t current, uint64_t total) {
       if (!req.progress) { return true; }
@@ -5090,7 +5247,8 @@ inline bool ClientImpl::process_request(Stream &strm, const Request &req,
 
     int dummy_status;
     if (!detail::read_content(strm, res, (std::numeric_limits<size_t>::max)(),
-                              dummy_status, progress, out, decompress_)) {
+                              dummy_status, std::move(progress), std::move(out),
+                              decompress_)) {
       if (error_ != Error::Canceled) { error_ = Error::Read; }
       return false;
     }
@@ -5098,7 +5256,16 @@ inline bool ClientImpl::process_request(Stream &strm, const Request &req,
 
   if (res.get_header_value("Connection") == "close" ||
       (res.version == "HTTP/1.0" && res.reason != "Connection established")) {
-    stop_core();
+    // TODO this requires a not-entirely-obvious chain of calls to be correct
+    // for this to be safe. Maybe a code refactor (such as moving this out to
+    // the send function and getting rid of the recursiveness of the mutex)
+    // could make this more obvious.
+    
+    // This is safe to call because process_request is only called by handle_request
+    // which is only called by send, which locks the request mutex during the process.
+    // It would be a bug to call it from a different thread since it's a thread-safety
+    // issue to do these things to the socket if another thread is using the socket.
+    lock_socket_and_shutdown_and_close();
   }
 
   // Log
@@ -5108,11 +5275,11 @@ inline bool ClientImpl::process_request(Stream &strm, const Request &req,
 }
 
 inline bool
-ClientImpl::process_socket(Socket &socket,
+ClientImpl::process_socket(const Socket &socket,
                            std::function<bool(Stream &strm)> callback) {
-  return detail::process_client_socket(socket.sock, read_timeout_sec_,
-                                       read_timeout_usec_, write_timeout_sec_,
-                                       write_timeout_usec_, callback);
+  return detail::process_client_socket(
+      socket.sock, read_timeout_sec_, read_timeout_usec_, write_timeout_sec_,
+      write_timeout_usec_, std::move(callback));
 }
 
 inline bool ClientImpl::is_ssl() const { return false; }
@@ -5138,9 +5305,9 @@ inline Result ClientImpl::Get(const char *path, const Headers &headers,
   req.headers.insert(headers.begin(), headers.end());
   req.progress = std::move(progress);
 
-  auto res = std::make_shared<Response>();
+  auto res = detail::make_unique<Response>();
   auto ret = send(req, *res);
-  return Result{ret ? res : nullptr, get_last_error()};
+  return Result{ret ? std::move(res) : nullptr, get_last_error()};
 }
 
 inline Result ClientImpl::Get(const char *path,
@@ -5170,23 +5337,23 @@ inline Result ClientImpl::Get(const char *path, const Headers &headers,
 inline Result ClientImpl::Get(const char *path,
                               ResponseHandler response_handler,
                               ContentReceiver content_receiver) {
-  return Get(path, Headers(), std::move(response_handler), content_receiver,
-             nullptr);
+  return Get(path, Headers(), std::move(response_handler),
+             std::move(content_receiver), nullptr);
 }
 
 inline Result ClientImpl::Get(const char *path, const Headers &headers,
                               ResponseHandler response_handler,
                               ContentReceiver content_receiver) {
-  return Get(path, headers, std::move(response_handler), content_receiver,
-             nullptr);
+  return Get(path, headers, std::move(response_handler),
+             std::move(content_receiver), nullptr);
 }
 
 inline Result ClientImpl::Get(const char *path,
                               ResponseHandler response_handler,
                               ContentReceiver content_receiver,
                               Progress progress) {
-  return Get(path, Headers(), std::move(response_handler), content_receiver,
-             progress);
+  return Get(path, Headers(), std::move(response_handler),
+             std::move(content_receiver), std::move(progress));
 }
 
 inline Result ClientImpl::Get(const char *path, const Headers &headers,
@@ -5199,12 +5366,16 @@ inline Result ClientImpl::Get(const char *path, const Headers &headers,
   req.headers = default_headers_;
   req.headers.insert(headers.begin(), headers.end());
   req.response_handler = std::move(response_handler);
-  req.content_receiver = std::move(content_receiver);
+  req.content_receiver =
+      [content_receiver](const char *data, size_t data_length,
+                         uint64_t /*offset*/, uint64_t /*total_length*/) {
+        return content_receiver(data, data_length);
+      };
   req.progress = std::move(progress);
 
-  auto res = std::make_shared<Response>();
+  auto res = detail::make_unique<Response>();
   auto ret = send(req, *res);
-  return Result{ret ? res : nullptr, get_last_error()};
+  return Result{ret ? std::move(res) : nullptr, get_last_error()};
 }
 
 inline Result ClientImpl::Head(const char *path) {
@@ -5218,9 +5389,9 @@ inline Result ClientImpl::Head(const char *path, const Headers &headers) {
   req.headers.insert(headers.begin(), headers.end());
   req.path = path;
 
-  auto res = std::make_shared<Response>();
+  auto res = detail::make_unique<Response>();
   auto ret = send(req, *res);
-  return Result{ret ? res : nullptr, get_last_error()};
+  return Result{ret ? std::move(res) : nullptr, get_last_error()};
 }
 
 inline Result ClientImpl::Post(const char *path) {
@@ -5237,7 +5408,7 @@ inline Result ClientImpl::Post(const char *path, const Headers &headers,
                                const char *content_type) {
   auto ret = send_with_content_provider("POST", path, headers, body, 0, nullptr,
                                         content_type);
-  return Result{ret, get_last_error()};
+  return Result{std::move(ret), get_last_error()};
 }
 
 inline Result ClientImpl::Post(const char *path, const Params &params) {
@@ -5247,17 +5418,18 @@ inline Result ClientImpl::Post(const char *path, const Params &params) {
 inline Result ClientImpl::Post(const char *path, size_t content_length,
                                ContentProvider content_provider,
                                const char *content_type) {
-  return Post(path, Headers(), content_length, content_provider, content_type);
+  return Post(path, Headers(), content_length, std::move(content_provider),
+              content_type);
 }
 
 inline Result ClientImpl::Post(const char *path, const Headers &headers,
                                size_t content_length,
                                ContentProvider content_provider,
                                const char *content_type) {
-  auto ret = send_with_content_provider("POST", path, headers, std::string(),
-                                        content_length, content_provider,
-                                        content_type);
-  return Result{ret, get_last_error()};
+  auto ret = send_with_content_provider(
+      "POST", path, headers, std::string(), content_length,
+      std::move(content_provider), content_type);
+  return Result{std::move(ret), get_last_error()};
 }
 
 inline Result ClientImpl::Post(const char *path, const Headers &headers,
@@ -5273,7 +5445,18 @@ inline Result ClientImpl::Post(const char *path,
 
 inline Result ClientImpl::Post(const char *path, const Headers &headers,
                                const MultipartFormDataItems &items) {
-  auto boundary = detail::make_multipart_data_boundary();
+  return Post(path, headers, items, detail::make_multipart_data_boundary());
+}
+inline Result ClientImpl::Post(const char *path, const Headers &headers,
+                               const MultipartFormDataItems &items,
+                               const std::string &boundary) {
+  for (size_t i = 0; i < boundary.size(); i++) {
+    char c = boundary[i];
+    if (!std::isalnum(c) && c != '-' && c != '_') {
+      error_ = Error::UnsupportedMultipartBoundaryChars;
+      return Result{nullptr, error_};
+    }
+  }
 
   std::string body;
 
@@ -5311,23 +5494,24 @@ inline Result ClientImpl::Put(const char *path, const Headers &headers,
                               const char *content_type) {
   auto ret = send_with_content_provider("PUT", path, headers, body, 0, nullptr,
                                         content_type);
-  return Result{ret, get_last_error()};
+  return Result{std::move(ret), get_last_error()};
 }
 
 inline Result ClientImpl::Put(const char *path, size_t content_length,
                               ContentProvider content_provider,
                               const char *content_type) {
-  return Put(path, Headers(), content_length, content_provider, content_type);
+  return Put(path, Headers(), content_length, std::move(content_provider),
+             content_type);
 }
 
 inline Result ClientImpl::Put(const char *path, const Headers &headers,
                               size_t content_length,
                               ContentProvider content_provider,
                               const char *content_type) {
-  auto ret = send_with_content_provider("PUT", path, headers, std::string(),
-                                        content_length, content_provider,
-                                        content_type);
-  return Result{ret, get_last_error()};
+  auto ret = send_with_content_provider(
+      "PUT", path, headers, std::string(), content_length,
+      std::move(content_provider), content_type);
+  return Result{std::move(ret), get_last_error()};
 }
 
 inline Result ClientImpl::Put(const char *path, const Params &params) {
@@ -5350,23 +5534,24 @@ inline Result ClientImpl::Patch(const char *path, const Headers &headers,
                                 const char *content_type) {
   auto ret = send_with_content_provider("PATCH", path, headers, body, 0,
                                         nullptr, content_type);
-  return Result{ret, get_last_error()};
+  return Result{std::move(ret), get_last_error()};
 }
 
 inline Result ClientImpl::Patch(const char *path, size_t content_length,
                                 ContentProvider content_provider,
                                 const char *content_type) {
-  return Patch(path, Headers(), content_length, content_provider, content_type);
+  return Patch(path, Headers(), content_length, std::move(content_provider),
+               content_type);
 }
 
 inline Result ClientImpl::Patch(const char *path, const Headers &headers,
                                 size_t content_length,
                                 ContentProvider content_provider,
                                 const char *content_type) {
-  auto ret = send_with_content_provider("PATCH", path, headers, std::string(),
-                                        content_length, content_provider,
-                                        content_type);
-  return Result{ret, get_last_error()};
+  auto ret = send_with_content_provider(
+      "PATCH", path, headers, std::string(), content_length,
+      std::move(content_provider), content_type);
+  return Result{std::move(ret), get_last_error()};
 }
 
 inline Result ClientImpl::Delete(const char *path) {
@@ -5394,9 +5579,9 @@ inline Result ClientImpl::Delete(const char *path, const Headers &headers,
   if (content_type) { req.headers.emplace("Content-Type", content_type); }
   req.body = body;
 
-  auto res = std::make_shared<Response>();
+  auto res = detail::make_unique<Response>();
   auto ret = send(req, *res);
-  return Result{ret ? res : nullptr, get_last_error()};
+  return Result{ret ? std::move(res) : nullptr, get_last_error()};
 }
 
 inline Result ClientImpl::Options(const char *path) {
@@ -5410,9 +5595,9 @@ inline Result ClientImpl::Options(const char *path, const Headers &headers) {
   req.headers.insert(headers.begin(), headers.end());
   req.path = path;
 
-  auto res = std::make_shared<Response>();
+  auto res = detail::make_unique<Response>();
   auto ret = send(req, *res);
-  return Result{ret ? res : nullptr, get_last_error()};
+  return Result{ret ? std::move(res) : nullptr, get_last_error()};
 }
 
 inline size_t ClientImpl::is_socket_open() const {
@@ -5421,18 +5606,27 @@ inline size_t ClientImpl::is_socket_open() const {
 }
 
 inline void ClientImpl::stop() {
-  stop_core();
-  error_ = Error::Canceled;
-}
-
-inline void ClientImpl::stop_core() {
   std::lock_guard<std::mutex> guard(socket_mutex_);
-  if (socket_.is_open()) {
-    detail::shutdown_socket(socket_.sock);
-    std::this_thread::sleep_for(std::chrono::milliseconds(1));
-    close_socket(socket_, true);
-    std::this_thread::sleep_for(std::chrono::milliseconds(1));
+  // There is no guarantee that this doesn't get overwritten later, but set it so that
+  // there is a good chance that any threads stopping as a result pick up this error.
+  error_ = Error::Canceled;
+  
+  // If there is anything ongoing right now, the ONLY thread-safe thing we can do
+  // is to shutdown_socket, so that threads using this socket suddenly discover
+  // they can't read/write any more and error out.
+  // Everything else (closing the socket, shutting ssl down) is unsafe because these
+  // actions are not thread-safe.
+  if (socket_requests_in_flight_ > 0) {
+    shutdown_socket(socket_);
+    // Aside from that, we set a flag for the socket to be closed when we're done.
+    socket_should_be_closed_when_request_is_done_ = true;
+    return;
   }
+
+  //Otherwise, sitll holding the mutex, we can shut everything down ourselves
+  shutdown_ssl(socket_, true);
+  shutdown_socket(socket_);
+  close_socket(socket_);
 }
 
 inline void ClientImpl::set_connection_timeout(time_t sec, time_t usec) {
@@ -5479,7 +5673,7 @@ inline void ClientImpl::set_default_headers(Headers headers) {
 inline void ClientImpl::set_tcp_nodelay(bool on) { tcp_nodelay_ = on; }
 
 inline void ClientImpl::set_socket_options(SocketOptions socket_options) {
-  socket_options_ = socket_options;
+  socket_options_ = std::move(socket_options);
 }
 
 inline void ClientImpl::set_compress(bool on) { compress_ = on; }
@@ -5554,9 +5748,12 @@ inline SSL *ssl_new(socket_t sock, SSL_CTX *ctx, std::mutex &ctx_mutex,
 }
 
 inline void ssl_delete(std::mutex &ctx_mutex, SSL *ssl,
-                       bool process_socket_ret) {
-  if (process_socket_ret) {
-    SSL_shutdown(ssl); // shutdown only if not already closed by remote
+                       bool shutdown_gracefully) {
+  // sometimes we may want to skip this to try to avoid SIGPIPE if we know
+  // the remote has closed the network connection
+  // Note that it is not always possible to avoid SIGPIPE, this is merely a best-efforts.
+  if (shutdown_gracefully) {
+    SSL_shutdown(ssl); 
   }
 
   std::lock_guard<std::mutex> guard(ctx_mutex);
@@ -5650,22 +5847,7 @@ inline SSLSocketStream::SSLSocketStream(socket_t sock, SSL *ssl,
       read_timeout_usec_(read_timeout_usec),
       write_timeout_sec_(write_timeout_sec),
       write_timeout_usec_(write_timeout_usec) {
-  {
-    timeval tv;
-    tv.tv_sec = static_cast<long>(read_timeout_sec);
-    tv.tv_usec = static_cast<decltype(tv.tv_usec)>(read_timeout_usec);
-
-    setsockopt(sock, SOL_SOCKET, SO_RCVTIMEO, reinterpret_cast<char *>(&tv),
-               sizeof(tv));
-  }
-  {
-    timeval tv;
-    tv.tv_sec = static_cast<long>(write_timeout_sec);
-    tv.tv_usec = static_cast<decltype(tv.tv_usec)>(write_timeout_usec);
-
-    setsockopt(sock, SOL_SOCKET, SO_SNDTIMEO, reinterpret_cast<char *>(&tv),
-               sizeof(tv));
-  }
+  SSL_clear_mode(ssl, SSL_MODE_AUTO_RETRY);
 }
 
 inline SSLSocketStream::~SSLSocketStream() {}
@@ -5680,8 +5862,27 @@ inline bool SSLSocketStream::is_writable() const {
 }
 
 inline ssize_t SSLSocketStream::read(char *ptr, size_t size) {
-  if (SSL_pending(ssl_) > 0 || is_readable()) {
+  if (SSL_pending(ssl_) > 0) {
     return SSL_read(ssl_, ptr, static_cast<int>(size));
+  } else if (is_readable()) {
+    auto ret = SSL_read(ssl_, ptr, static_cast<int>(size));
+    if (ret < 0) {
+      auto err = SSL_get_error(ssl_, ret);
+      while (err == SSL_ERROR_WANT_READ) {
+        if (SSL_pending(ssl_) > 0) {
+          return SSL_read(ssl_, ptr, static_cast<int>(size));
+        } else if (is_readable()) {
+          ret = SSL_read(ssl_, ptr, static_cast<int>(size));
+          if (ret >= 0) {
+            return ret;
+          }
+          err = SSL_get_error(ssl_, ret);
+        } else {
+          return -1;
+        }
+      }
+    }
+    return ret;
   }
   return -1;
 }
@@ -5788,9 +5989,12 @@ inline bool SSLServer::process_and_close_socket(socket_t sock) {
         });
 
     detail::ssl_delete(ctx_mutex_, ssl, ret);
+    detail::shutdown_socket(sock);
+    detail::close_socket(sock);
     return ret;
   }
 
+  detail::shutdown_socket(sock);
   detail::close_socket(sock);
   return false;
 }
@@ -5843,6 +6047,10 @@ inline SSLClient::SSLClient(const std::string &host, int port,
 
 inline SSLClient::~SSLClient() {
   if (ctx_) { SSL_CTX_free(ctx_); }
+  // Make sure to shut down SSL since shutdown_ssl will resolve to the
+  // base function rather than the derived function once we get to the
+  // base class destructor, and won't free the SSL (causing a leak).
+  SSLClient::shutdown_ssl(socket_, true);
 }
 
 inline bool SSLClient::is_valid() const { return ctx_; }
@@ -5876,11 +6084,11 @@ inline bool SSLClient::create_and_connect_socket(Socket &socket) {
   return is_valid() && ClientImpl::create_and_connect_socket(socket);
 }
 
+// Assumes that socket_mutex_ is locked and that there are no requests in flight
 inline bool SSLClient::connect_with_proxy(Socket &socket, Response &res,
                                           bool &success) {
   success = true;
   Response res2;
-
   if (!detail::process_client_socket(
           socket.sock, read_timeout_sec_, read_timeout_usec_,
           write_timeout_sec_, write_timeout_usec_, [&](Stream &strm) {
@@ -5889,7 +6097,10 @@ inline bool SSLClient::connect_with_proxy(Socket &socket, Response &res,
             req2.path = host_and_port_;
             return process_request(strm, req2, res2, false);
           })) {
-    close_socket(socket, true);
+    // Thread-safe to close everything because we are assuming there are no requests in flight
+    shutdown_ssl(socket, true);
+    shutdown_socket(socket);
+    close_socket(socket);
     success = false;
     return false;
   }
@@ -5912,7 +6123,10 @@ inline bool SSLClient::connect_with_proxy(Socket &socket, Response &res,
                       true));
                   return process_request(strm, req3, res3, false);
                 })) {
-          close_socket(socket, true);
+          // Thread-safe to close everything because we are assuming there are no requests in flight
+          shutdown_ssl(socket, true);
+          shutdown_socket(socket);
+          close_socket(socket);
           success = false;
           return false;
         }
@@ -6005,26 +6219,30 @@ inline bool SSLClient::initialize_ssl(Socket &socket) {
     return true;
   }
 
-  close_socket(socket, false);
+  shutdown_socket(socket);
+  close_socket(socket);
   return false;
 }
 
-inline void SSLClient::close_socket(Socket &socket, bool process_socket_ret) {
-  detail::close_socket(socket.sock);
-  socket_.sock = INVALID_SOCKET;
+inline void SSLClient::shutdown_ssl(Socket &socket, bool shutdown_gracefully) {
+  if (socket.sock == INVALID_SOCKET) {
+    assert(socket.ssl == nullptr);
+    return;
+  }
   if (socket.ssl) {
-    detail::ssl_delete(ctx_mutex_, socket.ssl, process_socket_ret);
-    socket_.ssl = nullptr;
+    detail::ssl_delete(ctx_mutex_, socket.ssl, shutdown_gracefully);
+    socket.ssl = nullptr;
   }
+  assert(socket.ssl == nullptr);
 }
 
 inline bool
-SSLClient::process_socket(Socket &socket,
+SSLClient::process_socket(const Socket &socket,
                           std::function<bool(Stream &strm)> callback) {
   assert(socket.ssl);
   return detail::process_client_socket_ssl(
       socket.ssl, socket.sock, read_timeout_sec_, read_timeout_usec_,
-      write_timeout_sec_, write_timeout_usec_, callback);
+      write_timeout_sec_, write_timeout_usec_, std::move(callback));
 }
 
 inline bool SSLClient::is_ssl() const { return true; }
@@ -6175,6 +6393,8 @@ inline Client::Client(const char *scheme_host_port,
 #else
     if (!scheme.empty() && scheme != "http") {
 #endif
+      std::string msg = "'" + scheme + "' scheme is not supported.";
+      throw std::invalid_argument(msg);
       return;
     }
 
@@ -6187,28 +6407,28 @@ inline Client::Client(const char *scheme_host_port,
 
     if (is_ssl) {
 #ifdef CPPHTTPLIB_OPENSSL_SUPPORT
-      cli_ = std::make_shared<SSLClient>(host.c_str(), port, client_cert_path,
-                                         client_key_path);
+      cli_ = detail::make_unique<SSLClient>(host.c_str(), port,
+                                            client_cert_path, client_key_path);
       is_ssl_ = is_ssl;
 #endif
     } else {
-      cli_ = std::make_shared<ClientImpl>(host.c_str(), port, client_cert_path,
-                                          client_key_path);
+      cli_ = detail::make_unique<ClientImpl>(host.c_str(), port,
+                                             client_cert_path, client_key_path);
     }
   } else {
-    cli_ = std::make_shared<ClientImpl>(scheme_host_port, 80, client_cert_path,
-                                        client_key_path);
+    cli_ = detail::make_unique<ClientImpl>(scheme_host_port, 80,
+                                           client_cert_path, client_key_path);
   }
 }
 
 inline Client::Client(const std::string &host, int port)
-    : cli_(std::make_shared<ClientImpl>(host, port)) {}
+    : cli_(detail::make_unique<ClientImpl>(host, port)) {}
 
 inline Client::Client(const std::string &host, int port,
                       const std::string &client_cert_path,
                       const std::string &client_key_path)
-    : cli_(std::make_shared<ClientImpl>(host, port, client_cert_path,
-                                        client_key_path)) {}
+    : cli_(detail::make_unique<ClientImpl>(host, port, client_cert_path,
+                                           client_key_path)) {}
 
 inline Client::~Client() {}
 
@@ -6221,11 +6441,11 @@ inline Result Client::Get(const char *path, const Headers &headers) {
   return cli_->Get(path, headers);
 }
 inline Result Client::Get(const char *path, Progress progress) {
-  return cli_->Get(path, progress);
+  return cli_->Get(path, std::move(progress));
 }
 inline Result Client::Get(const char *path, const Headers &headers,
                           Progress progress) {
-  return cli_->Get(path, headers, progress);
+  return cli_->Get(path, headers, std::move(progress));
 }
 inline Result Client::Get(const char *path, ContentReceiver content_receiver) {
   return cli_->Get(path, std::move(content_receiver));
@@ -6262,7 +6482,8 @@ inline Result Client::Get(const char *path, ResponseHandler response_handler,
 inline Result Client::Get(const char *path, const Headers &headers,
                           ResponseHandler response_handler,
                           ContentReceiver content_receiver, Progress progress) {
-  return cli_->Get(path, headers, response_handler, content_receiver, progress);
+  return cli_->Get(path, headers, std::move(response_handler),
+                   std::move(content_receiver), std::move(progress));
 }
 
 inline Result Client::Head(const char *path) { return cli_->Head(path); }
@@ -6282,13 +6503,14 @@ inline Result Client::Post(const char *path, const Headers &headers,
 inline Result Client::Post(const char *path, size_t content_length,
                            ContentProvider content_provider,
                            const char *content_type) {
-  return cli_->Post(path, content_length, content_provider, content_type);
+  return cli_->Post(path, content_length, std::move(content_provider),
+                    content_type);
 }
 inline Result Client::Post(const char *path, const Headers &headers,
                            size_t content_length,
                            ContentProvider content_provider,
                            const char *content_type) {
-  return cli_->Post(path, headers, content_length, content_provider,
+  return cli_->Post(path, headers, content_length, std::move(content_provider),
                     content_type);
 }
 inline Result Client::Post(const char *path, const Params &params) {
@@ -6306,6 +6528,11 @@ inline Result Client::Post(const char *path, const Headers &headers,
                            const MultipartFormDataItems &items) {
   return cli_->Post(path, headers, items);
 }
+inline Result Client::Post(const char *path, const Headers &headers,
+                           const MultipartFormDataItems &items,
+                           const std::string &boundary) {
+  return cli_->Post(path, headers, items, boundary);
+}
 inline Result Client::Put(const char *path) { return cli_->Put(path); }
 inline Result Client::Put(const char *path, const std::string &body,
                           const char *content_type) {
@@ -6318,13 +6545,14 @@ inline Result Client::Put(const char *path, const Headers &headers,
 inline Result Client::Put(const char *path, size_t content_length,
                           ContentProvider content_provider,
                           const char *content_type) {
-  return cli_->Put(path, content_length, content_provider, content_type);
+  return cli_->Put(path, content_length, std::move(content_provider),
+                   content_type);
 }
 inline Result Client::Put(const char *path, const Headers &headers,
                           size_t content_length,
                           ContentProvider content_provider,
                           const char *content_type) {
-  return cli_->Put(path, headers, content_length, content_provider,
+  return cli_->Put(path, headers, content_length, std::move(content_provider),
                    content_type);
 }
 inline Result Client::Put(const char *path, const Params &params) {
@@ -6345,13 +6573,14 @@ inline Result Client::Patch(const char *path, const Headers &headers,
 inline Result Client::Patch(const char *path, size_t content_length,
                             ContentProvider content_provider,
                             const char *content_type) {
-  return cli_->Patch(path, content_length, content_provider, content_type);
+  return cli_->Patch(path, content_length, std::move(content_provider),
+                     content_type);
 }
 inline Result Client::Patch(const char *path, const Headers &headers,
                             size_t content_length,
                             ContentProvider content_provider,
                             const char *content_type) {
-  return cli_->Patch(path, headers, content_length, content_provider,
+  return cli_->Patch(path, headers, content_length, std::move(content_provider),
                      content_type);
 }
 inline Result Client::Delete(const char *path) { return cli_->Delete(path); }
@@ -6386,7 +6615,7 @@ inline void Client::set_default_headers(Headers headers) {
 
 inline void Client::set_tcp_nodelay(bool on) { cli_->set_tcp_nodelay(on); }
 inline void Client::set_socket_options(SocketOptions socket_options) {
-  cli_->set_socket_options(socket_options);
+  cli_->set_socket_options(std::move(socket_options));
 }
 
 inline void Client::set_connection_timeout(time_t sec, time_t usec) {
diff --git a/externals/inih/inih b/externals/inih/inih
-Subproject 603729dec89aaca42d7bd08f08bc333165b7d5d
+Subproject 1e80a47dffbda813604f0913e2ad68c7054c14e
diff --git a/externals/libressl b/externals/libressl
-Subproject 7d01cb01cb1a926ecb4c9c98b107ef3c26f59df
+Subproject 8289d0d07de6553bf4b900bf60e808ea3f7f59d
diff --git a/externals/lurlparser/CMakeLists.txt b/externals/lurlparser/CMakeLists.txt
deleted file mode 100644
index 45046ffd3..000000000
--- a/externals/lurlparser/CMakeLists.txt
+++ /dev/null
@@ -1,8 +0,0 @@
-add_library(lurlparser
-        LUrlParser.cpp
-        LUrlParser.h
-)
-
-create_target_directory_groups(lurlparser)
-
-target_include_directories(lurlparser INTERFACE ${CMAKE_CURRENT_SOURCE_DIR})
diff --git a/externals/lurlparser/LUrlParser.cpp b/externals/lurlparser/LUrlParser.cpp
deleted file mode 100644
index 9c134e330..000000000
--- a/externals/lurlparser/LUrlParser.cpp
+++ /dev/null
@@ -1,265 +0,0 @@
-/*
- * Lightweight URL & URI parser (RFC 1738, RFC 3986)
- * https://github.com/corporateshark/LUrlParser
- *
- * The MIT License (MIT)
- *
- * Copyright (C) 2015 Sergey Kosarevsky (sk@linderdaum.com)
- *
- * 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.
- */
-
-#include "LUrlParser.h"
-
-#include <algorithm>
-#include <cstring>
-#include <stdlib.h>
-
-// check if the scheme name is valid
-static bool IsSchemeValid( const std::string& SchemeName )
-{
-    for ( auto c : SchemeName  )
-    {
-        if ( !isalpha( c ) && c != '+' && c != '-' && c != '.' ) return false;
-    }
-
-    return true;
-}
-
-bool LUrlParser::clParseURL::GetPort( int* OutPort ) const
-{
-    if ( !IsValid() ) { return false; }
-
-    int Port = atoi( m_Port.c_str() );
-
-    if ( Port <= 0 || Port > 65535 ) { return false; }
-
-    if ( OutPort ) { *OutPort = Port; }
-
-    return true;
-}
-
-// based on RFC 1738 and RFC 3986
-LUrlParser::clParseURL LUrlParser::clParseURL::ParseURL( const std::string& URL )
-{
-    LUrlParser::clParseURL Result;
-
-    const char* CurrentString = URL.c_str();
-
-    /*
-     *  <scheme>:<scheme-specific-part>
-     *  <scheme> := [a-z\+\-\.]+
-     *  For resiliency, programs interpreting URLs should treat upper case letters as equivalent to lower case in scheme names
-     */
-
-    // try to read scheme
-    {
-        const char* LocalString = strchr( CurrentString, ':' );
-
-        if ( !LocalString )
-        {
-            return clParseURL( LUrlParserError_NoUrlCharacter );
-        }
-
-        // save the scheme name
-        Result.m_Scheme = std::string( CurrentString, LocalString - CurrentString );
-
-        if ( !IsSchemeValid( Result.m_Scheme ) )
-        {
-            return clParseURL( LUrlParserError_InvalidSchemeName );
-        }
-
-        // scheme should be lowercase
-        std::transform( Result.m_Scheme.begin(), Result.m_Scheme.end(), Result.m_Scheme.begin(), ::tolower );
-
-        // skip ':'
-        CurrentString = LocalString+1;
-    }
-
-    /*
-     *  //<user>:<password>@<host>:<port>/<url-path>
-     *  any ":", "@" and "/" must be normalized
-     */
-
-    // skip "//"
-    if ( *CurrentString++ != '/' ) return clParseURL( LUrlParserError_NoDoubleSlash );
-    if ( *CurrentString++ != '/' ) return clParseURL( LUrlParserError_NoDoubleSlash );
-
-    // check if the user name and password are specified
-    bool bHasUserName = false;
-
-    const char* LocalString = CurrentString;
-
-    while ( *LocalString )
-    {
-        if ( *LocalString == '@' )
-        {
-            // user name and password are specified
-            bHasUserName = true;
-            break;
-        }
-        else if ( *LocalString == '/' )
-        {
-            // end of <host>:<port> specification
-            bHasUserName = false;
-            break;
-        }
-
-        LocalString++;
-    }
-
-    // user name and password
-    LocalString = CurrentString;
-
-    if ( bHasUserName )
-    {
-        // read user name
-        while ( *LocalString && *LocalString != ':' && *LocalString != '@' ) LocalString++;
-
-        Result.m_UserName = std::string( CurrentString, LocalString - CurrentString );
-
-        // proceed with the current pointer
-        CurrentString = LocalString;
-
-        if ( *CurrentString == ':' )
-        {
-            // skip ':'
-            CurrentString++;
-
-            // read password
-            LocalString = CurrentString;
-
-            while ( *LocalString && *LocalString != '@' ) LocalString++;
-
-            Result.m_Password = std::string( CurrentString, LocalString - CurrentString );
-
-            CurrentString = LocalString;
-        }
-
-        // skip '@'
-        if ( *CurrentString != '@' )
-        {
-            return clParseURL( LUrlParserError_NoAtSign );
-        }
-
-        CurrentString++;
-    }
-
-    bool bHasBracket = ( *CurrentString == '[' );
-
-    // go ahead, read the host name
-    LocalString = CurrentString;
-
-    while ( *LocalString )
-    {
-        if ( bHasBracket && *LocalString == ']' )
-        {
-            // end of IPv6 address
-            LocalString++;
-            break;
-        }
-        else if ( !bHasBracket && ( *LocalString == ':' || *LocalString == '/' ) )
-        {
-            // port number is specified
-            break;
-        }
-
-        LocalString++;
-    }
-
-    Result.m_Host = std::string( CurrentString, LocalString - CurrentString );
-
-    CurrentString = LocalString;
-
-    // is port number specified?
-    if ( *CurrentString == ':' )
-    {
-        CurrentString++;
-
-        // read port number
-        LocalString = CurrentString;
-
-        while ( *LocalString && *LocalString != '/' ) LocalString++;
-
-        Result.m_Port = std::string( CurrentString, LocalString - CurrentString );
-
-        CurrentString = LocalString;
-    }
-
-    // end of string
-    if ( !*CurrentString )
-    {
-        Result.m_ErrorCode = LUrlParserError_Ok;
-
-        return Result;
-    }
-
-    // skip '/'
-    if ( *CurrentString != '/' )
-    {
-        return clParseURL( LUrlParserError_NoSlash );
-    }
-
-    CurrentString++;
-
-    // parse the path
-    LocalString = CurrentString;
-
-    while ( *LocalString && *LocalString != '#' && *LocalString != '?' ) LocalString++;
-
-    Result.m_Path = std::string( CurrentString, LocalString - CurrentString );
-
-    CurrentString = LocalString;
-
-    // check for query
-    if ( *CurrentString == '?' )
-    {
-        // skip '?'
-        CurrentString++;
-
-        // read query
-        LocalString = CurrentString;
-
-        while ( *LocalString && *LocalString != '#' ) LocalString++;
-
-        Result.m_Query = std::string( CurrentString, LocalString - CurrentString );
-
-        CurrentString = LocalString;
-    }
-
-    // check for fragment
-    if ( *CurrentString == '#' )
-    {
-        // skip '#'
-        CurrentString++;
-
-        // read fragment
-        LocalString = CurrentString;
-
-        while ( *LocalString ) LocalString++;
-
-        Result.m_Fragment = std::string( CurrentString, LocalString - CurrentString );
-
-        CurrentString = LocalString;
-    }
-
-    Result.m_ErrorCode = LUrlParserError_Ok;
-
-    return Result;
-}
diff --git a/externals/lurlparser/LUrlParser.h b/externals/lurlparser/LUrlParser.h
deleted file mode 100644
index 25d210981..000000000
--- a/externals/lurlparser/LUrlParser.h
+++ /dev/null
@@ -1,78 +0,0 @@
-/*
- * Lightweight URL & URI parser (RFC 1738, RFC 3986)
- * https://github.com/corporateshark/LUrlParser
- *
- * The MIT License (MIT)
- *
- * Copyright (C) 2015 Sergey Kosarevsky (sk@linderdaum.com)
- *
- * 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.
- */
-
-#pragma once
-
-#include <string>
-
-namespace LUrlParser
-{
-enum LUrlParserError
-{
-    LUrlParserError_Ok = 0,
-    LUrlParserError_Uninitialized = 1,
-    LUrlParserError_NoUrlCharacter = 2,
-    LUrlParserError_InvalidSchemeName = 3,
-    LUrlParserError_NoDoubleSlash = 4,
-    LUrlParserError_NoAtSign = 5,
-    LUrlParserError_UnexpectedEndOfLine = 6,
-    LUrlParserError_NoSlash = 7,
-};
-
-class clParseURL
-{
-public:
-    LUrlParserError m_ErrorCode;
-    std::string m_Scheme;
-    std::string m_Host;
-    std::string m_Port;
-    std::string m_Path;
-    std::string m_Query;
-    std::string m_Fragment;
-    std::string m_UserName;
-    std::string m_Password;
-
-    clParseURL()
-            : m_ErrorCode( LUrlParserError_Uninitialized )
-    {}
-
-    /// return 'true' if the parsing was successful
-    bool IsValid() const { return m_ErrorCode == LUrlParserError_Ok; }
-
-    /// helper to convert the port number to int, return 'true' if the port is valid (within the 0..65535 range)
-    bool GetPort( int* OutPort ) const;
-
-    /// parse the URL
-    static clParseURL ParseURL( const std::string& URL );
-
-private:
-    explicit clParseURL( LUrlParserError ErrorCode )
-            : m_ErrorCode( ErrorCode )
-    {}
-};
-
-} // namespace LUrlParser
diff --git a/externals/lurlparser/README.md b/externals/lurlparser/README.md
deleted file mode 100644
index be7f0135a..000000000
--- a/externals/lurlparser/README.md
+++ /dev/null
@@ -1,19 +0,0 @@
-From https://github.com/corporateshark/LUrlParser/commit/455d5e2d27e3946f11ad0328fee9ee2628e6a8e2
-
-MIT License
-
-===
-
-Lightweight URL & URI parser (RFC 1738, RFC 3986)
-
-(C) Sergey Kosarevsky, 2015
-
-@corporateshark sk@linderdaum.com
-
-http://www.linderdaum.com
-
-http://blog.linderdaum.com
-
-=============================
-
-A tiny and lightweight URL & URI parser (RFC 1738, RFC 3986) written in C++.
diff --git a/src/common/CMakeLists.txt b/src/common/CMakeLists.txt
index 0fb5d9708..e50ab2922 100644
--- a/src/common/CMakeLists.txt
+++ b/src/common/CMakeLists.txt
@@ -150,6 +150,8 @@ add_library(common STATIC
     scope_exit.h
     spin_lock.cpp
     spin_lock.h
+    stream.cpp
+    stream.h
     string_util.cpp
     string_util.h
     swap.h
diff --git a/src/common/fiber.cpp b/src/common/fiber.cpp
index 1c1d09ccb..e186ed880 100644
--- a/src/common/fiber.cpp
+++ b/src/common/fiber.cpp
@@ -91,7 +91,7 @@ void Fiber::Rewind() {
     SwitchToFiber(impl->rewind_handle);
 }
 
-void Fiber::YieldTo(std::shared_ptr<Fiber>& from, std::shared_ptr<Fiber>& to) {
+void Fiber::YieldTo(std::shared_ptr<Fiber> from, std::shared_ptr<Fiber> to) {
     ASSERT_MSG(from != nullptr, "Yielding fiber is null!");
     ASSERT_MSG(to != nullptr, "Next fiber is null!");
     to->guard.lock();
@@ -199,7 +199,7 @@ void Fiber::Rewind() {
     boost::context::detail::jump_fcontext(impl->rewind_context, this);
 }
 
-void Fiber::YieldTo(std::shared_ptr<Fiber>& from, std::shared_ptr<Fiber>& to) {
+void Fiber::YieldTo(std::shared_ptr<Fiber> from, std::shared_ptr<Fiber> to) {
     ASSERT_MSG(from != nullptr, "Yielding fiber is null!");
     ASSERT_MSG(to != nullptr, "Next fiber is null!");
     to->guard.lock();
diff --git a/src/common/fiber.h b/src/common/fiber.h
index 89dde5e36..cefd61df9 100644
--- a/src/common/fiber.h
+++ b/src/common/fiber.h
@@ -46,7 +46,7 @@ public:
 
     /// Yields control from Fiber 'from' to Fiber 'to'
     /// Fiber 'from' must be the currently running fiber.
-    static void YieldTo(std::shared_ptr<Fiber>& from, std::shared_ptr<Fiber>& to);
+    static void YieldTo(std::shared_ptr<Fiber> from, std::shared_ptr<Fiber> to);
     [[nodiscard]] static std::shared_ptr<Fiber> ThreadToFiber();
 
     void SetRewindPoint(std::function<void(void*)>&& rewind_func, void* start_parameter);
diff --git a/src/common/stream.cpp b/src/common/stream.cpp
new file mode 100644
index 000000000..bf0496c26
--- /dev/null
+++ b/src/common/stream.cpp
@@ -0,0 +1,47 @@
+// Copyright 2020 yuzu Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#include <stdexcept>
+#include "common/common_types.h"
+#include "common/stream.h"
+
+namespace Common {
+
+Stream::Stream() = default;
+Stream::~Stream() = default;
+
+void Stream::Seek(s32 offset, SeekOrigin origin) {
+    if (origin == SeekOrigin::SetOrigin) {
+        if (offset < 0) {
+            position = 0;
+        } else if (position >= buffer.size()) {
+            position = buffer.size();
+        } else {
+            position = offset;
+        }
+    } else if (origin == SeekOrigin::FromCurrentPos) {
+        Seek(static_cast<s32>(position) + offset, SeekOrigin::SetOrigin);
+    } else if (origin == SeekOrigin::FromEnd) {
+        Seek(static_cast<s32>(buffer.size()) - offset, SeekOrigin::SetOrigin);
+    }
+}
+
+u8 Stream::ReadByte() {
+    if (position < buffer.size()) {
+        return buffer[position++];
+    } else {
+        throw std::out_of_range("Attempting to read a byte not within the buffer range");
+    }
+}
+
+void Stream::WriteByte(u8 byte) {
+    if (position == buffer.size()) {
+        buffer.push_back(byte);
+        position++;
+    } else {
+        buffer.insert(buffer.begin() + position, byte);
+    }
+}
+
+} // namespace Common
diff --git a/src/common/stream.h b/src/common/stream.h
new file mode 100644
index 000000000..2585c16af
--- /dev/null
+++ b/src/common/stream.h
@@ -0,0 +1,50 @@
+// Copyright 2020 yuzu Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#pragma once
+
+#include <vector>
+#include "common/common_types.h"
+
+namespace Common {
+
+enum class SeekOrigin {
+    SetOrigin,
+    FromCurrentPos,
+    FromEnd,
+};
+
+class Stream {
+public:
+    /// Stream creates a bitstream and provides common functionality on the stream.
+    explicit Stream();
+    ~Stream();
+
+    /// Reposition bitstream "cursor" to the specified offset from origin
+    void Seek(s32 offset, SeekOrigin origin);
+
+    /// Reads next byte in the stream buffer and increments position
+    u8 ReadByte();
+
+    /// Writes byte at current position
+    void WriteByte(u8 byte);
+
+    std::size_t GetPosition() const {
+        return position;
+    }
+
+    std::vector<u8>& GetBuffer() {
+        return buffer;
+    }
+
+    const std::vector<u8>& GetBuffer() const {
+        return buffer;
+    }
+
+private:
+    std::vector<u8> buffer;
+    std::size_t position{0};
+};
+
+} // namespace Common
diff --git a/src/core/CMakeLists.txt b/src/core/CMakeLists.txt
index db1c9fdef..e0f207f3e 100644
--- a/src/core/CMakeLists.txt
+++ b/src/core/CMakeLists.txt
@@ -439,6 +439,8 @@ add_library(core STATIC
     hle/service/nvdrv/devices/nvhost_gpu.h
     hle/service/nvdrv/devices/nvhost_nvdec.cpp
     hle/service/nvdrv/devices/nvhost_nvdec.h
+    hle/service/nvdrv/devices/nvhost_nvdec_common.cpp
+    hle/service/nvdrv/devices/nvhost_nvdec_common.h
     hle/service/nvdrv/devices/nvhost_nvjpg.cpp
     hle/service/nvdrv/devices/nvhost_nvjpg.h
     hle/service/nvdrv/devices/nvhost_vic.cpp
diff --git a/src/core/core.cpp b/src/core/core.cpp
index 81e8cc338..fde2ccc09 100644
--- a/src/core/core.cpp
+++ b/src/core/core.cpp
@@ -40,6 +40,7 @@
 #include "core/hle/service/lm/manager.h"
 #include "core/hle/service/service.h"
 #include "core/hle/service/sm/sm.h"
+#include "core/hle/service/time/time_manager.h"
 #include "core/loader/loader.h"
 #include "core/memory.h"
 #include "core/memory/cheat_engine.h"
@@ -121,7 +122,7 @@ FileSys::VirtualFile GetGameFileFromPath(const FileSys::VirtualFilesystem& vfs,
 struct System::Impl {
     explicit Impl(System& system)
         : kernel{system}, fs_controller{system}, memory{system},
-          cpu_manager{system}, reporter{system}, applet_manager{system} {}
+          cpu_manager{system}, reporter{system}, applet_manager{system}, time_manager{system} {}
 
     ResultStatus Run() {
         status = ResultStatus::Success;
@@ -189,6 +190,9 @@ struct System::Impl {
             return ResultStatus::ErrorVideoCore;
         }
 
+        // Initialize time manager, which must happen after kernel is created
+        time_manager.Initialize();
+
         is_powered_on = true;
         exit_lock = false;
 
@@ -387,6 +391,7 @@ struct System::Impl {
     /// Service State
     Service::Glue::ARPManager arp_manager;
     Service::LM::Manager lm_manager{reporter};
+    Service::Time::TimeManager time_manager;
 
     /// Service manager
     std::shared_ptr<Service::SM::ServiceManager> service_manager;
@@ -717,6 +722,14 @@ const Service::LM::Manager& System::GetLogManager() const {
     return impl->lm_manager;
 }
 
+Service::Time::TimeManager& System::GetTimeManager() {
+    return impl->time_manager;
+}
+
+const Service::Time::TimeManager& System::GetTimeManager() const {
+    return impl->time_manager;
+}
+
 void System::SetExitLock(bool locked) {
     impl->exit_lock = locked;
 }
diff --git a/src/core/core.h b/src/core/core.h
index 27efe30bb..6db896bae 100644
--- a/src/core/core.h
+++ b/src/core/core.h
@@ -69,6 +69,10 @@ namespace SM {
 class ServiceManager;
 } // namespace SM
 
+namespace Time {
+class TimeManager;
+} // namespace Time
+
 } // namespace Service
 
 namespace Tegra {
@@ -361,6 +365,10 @@ public:
 
     const Service::LM::Manager& GetLogManager() const;
 
+    Service::Time::TimeManager& GetTimeManager();
+
+    const Service::Time::TimeManager& GetTimeManager() const;
+
     void SetExitLock(bool locked);
 
     bool GetExitLock() const;
diff --git a/src/core/cpu_manager.cpp b/src/core/cpu_manager.cpp
index 688b99eba..983210197 100644
--- a/src/core/cpu_manager.cpp
+++ b/src/core/cpu_manager.cpp
@@ -365,6 +365,8 @@ void CpuManager::RunThread(std::size_t core) {
     data.enter_barrier.reset();
     data.exit_barrier.reset();
     data.initialized = false;
+
+    MicroProfileOnThreadExit();
 }
 
 } // namespace Core
diff --git a/src/core/frontend/applets/controller.cpp b/src/core/frontend/applets/controller.cpp
index c5d65f2d0..5582091f4 100644
--- a/src/core/frontend/applets/controller.cpp
+++ b/src/core/frontend/applets/controller.cpp
@@ -19,7 +19,7 @@ DefaultControllerApplet::DefaultControllerApplet(Service::SM::ServiceManager& se
 DefaultControllerApplet::~DefaultControllerApplet() = default;
 
 void DefaultControllerApplet::ReconfigureControllers(std::function<void()> callback,
-                                                     ControllerParameters parameters) const {
+                                                     const ControllerParameters& parameters) const {
     LOG_INFO(Service_HID, "called, deducing the best configuration based on the given parameters!");
 
     auto& npad =
diff --git a/src/core/frontend/applets/controller.h b/src/core/frontend/applets/controller.h
index 3e49cdbb9..dff71d8d9 100644
--- a/src/core/frontend/applets/controller.h
+++ b/src/core/frontend/applets/controller.h
@@ -38,7 +38,7 @@ public:
     virtual ~ControllerApplet();
 
     virtual void ReconfigureControllers(std::function<void()> callback,
-                                        ControllerParameters parameters) const = 0;
+                                        const ControllerParameters& parameters) const = 0;
 };
 
 class DefaultControllerApplet final : public ControllerApplet {
@@ -47,7 +47,7 @@ public:
     ~DefaultControllerApplet() override;
 
     void ReconfigureControllers(std::function<void()> callback,
-                                ControllerParameters parameters) const override;
+                                const ControllerParameters& parameters) const override;
 
 private:
     Service::SM::ServiceManager& service_manager;
diff --git a/src/core/hle/kernel/kernel.cpp b/src/core/hle/kernel/kernel.cpp
index b2b5b8adf..bb3e312a7 100644
--- a/src/core/hle/kernel/kernel.cpp
+++ b/src/core/hle/kernel/kernel.cpp
@@ -86,8 +86,6 @@ struct KernelCore::Impl {
         }
         cores.clear();
 
-        registered_core_threads.reset();
-
         process_list.clear();
         current_process = nullptr;
 
@@ -199,9 +197,7 @@ struct KernelCore::Impl {
         const auto it = std::find(register_host_thread_keys.begin(), end, this_id);
         ASSERT(core_id < Core::Hardware::NUM_CPU_CORES);
         ASSERT(it == end);
-        ASSERT(!registered_core_threads[core_id]);
         InsertHostThread(static_cast<u32>(core_id));
-        registered_core_threads.set(core_id);
     }
 
     void RegisterHostThread() {
@@ -332,7 +328,6 @@ struct KernelCore::Impl {
 
     // 0-3 IDs represent core threads, >3 represent others
     std::atomic<u32> registered_thread_ids{Core::Hardware::NUM_CPU_CORES};
-    std::bitset<Core::Hardware::NUM_CPU_CORES> registered_core_threads;
 
     // Number of host threads is a relatively high number to avoid overflowing
     static constexpr size_t NUM_REGISTRABLE_HOST_THREADS = 64;
diff --git a/src/core/hle/kernel/process.cpp b/src/core/hle/kernel/process.cpp
index ff9d9248b..b17529dee 100644
--- a/src/core/hle/kernel/process.cpp
+++ b/src/core/hle/kernel/process.cpp
@@ -4,6 +4,7 @@
 
 #include <algorithm>
 #include <bitset>
+#include <ctime>
 #include <memory>
 #include <random>
 #include "common/alignment.h"
@@ -123,7 +124,7 @@ std::shared_ptr<Process> Process::Create(Core::System& system, std::string name,
                                                               : kernel.CreateNewUserProcessID();
     process->capabilities.InitializeForMetadatalessProcess();
 
-    std::mt19937 rng(Settings::values.rng_seed.GetValue().value_or(0));
+    std::mt19937 rng(Settings::values.rng_seed.GetValue().value_or(std::time(nullptr)));
     std::uniform_int_distribution<u64> distribution;
     std::generate(process->random_entropy.begin(), process->random_entropy.end(),
                   [&] { return distribution(rng); });
diff --git a/src/core/hle/service/am/am.cpp b/src/core/hle/service/am/am.cpp
index d7a81f64a..2ce742e35 100644
--- a/src/core/hle/service/am/am.cpp
+++ b/src/core/hle/service/am/am.cpp
@@ -1201,6 +1201,8 @@ IApplicationFunctions::IApplicationFunctions(Core::System& system_)
         {151, nullptr, "TryPopFromNotificationStorageChannel"},
         {160, nullptr, "GetHealthWarningDisappearedSystemEvent"},
         {170, nullptr, "SetHdcpAuthenticationActivated"},
+        {180, nullptr, "GetLaunchRequiredVersion"},
+        {181, nullptr, "UpgradeLaunchRequiredVersion"},
         {500, nullptr, "StartContinuousRecordingFlushForDebug"},
         {1000, nullptr, "CreateMovieMaker"},
         {1001, nullptr, "PrepareForJit"},
diff --git a/src/core/hle/service/hid/hid.cpp b/src/core/hle/service/hid/hid.cpp
index 8918946a1..50f709b25 100644
--- a/src/core/hle/service/hid/hid.cpp
+++ b/src/core/hle/service/hid/hid.cpp
@@ -260,7 +260,7 @@ Hid::Hid(Core::System& system) : ServiceFramework("hid"), system(system) {
         {404, nullptr, "HasLeftRightBattery"},
         {405, nullptr, "GetNpadInterfaceType"},
         {406, nullptr, "GetNpadLeftRightInterfaceType"},
-        {407, nullptr, "GetNpadOfHighestBatteryLevelForJoyLeft"},
+        {407, nullptr, "GetNpadOfHighestBatteryLevel"},
         {408, nullptr, "GetNpadOfHighestBatteryLevelForJoyRight"},
         {500, nullptr, "GetPalmaConnectionHandle"},
         {501, nullptr, "InitializePalma"},
diff --git a/src/core/hle/service/mii/mii.cpp b/src/core/hle/service/mii/mii.cpp
index b81bf6277..d7080b715 100644
--- a/src/core/hle/service/mii/mii.cpp
+++ b/src/core/hle/service/mii/mii.cpp
@@ -47,6 +47,7 @@ public:
             {23, nullptr, "Convert"},
             {24, nullptr, "ConvertCoreDataToCharInfo"},
             {25, nullptr, "ConvertCharInfoToCoreData"},
+            {26, nullptr, "Append"},
         };
         // clang-format on
 
diff --git a/src/core/hle/service/nvdrv/devices/nvhost_nvdec.cpp b/src/core/hle/service/nvdrv/devices/nvhost_nvdec.cpp
index fcb612864..b6df48360 100644
--- a/src/core/hle/service/nvdrv/devices/nvhost_nvdec.cpp
+++ b/src/core/hle/service/nvdrv/devices/nvhost_nvdec.cpp
@@ -2,15 +2,17 @@
 // Licensed under GPLv2 or any later version
 // Refer to the license.txt file included.
 
-#include <cstring>
-
 #include "common/assert.h"
 #include "common/logging/log.h"
+#include "core/core.h"
 #include "core/hle/service/nvdrv/devices/nvhost_nvdec.h"
+#include "video_core/memory_manager.h"
+#include "video_core/renderer_base.h"
 
 namespace Service::Nvidia::Devices {
 
-nvhost_nvdec::nvhost_nvdec(Core::System& system) : nvdevice(system) {}
+nvhost_nvdec::nvhost_nvdec(Core::System& system, std::shared_ptr<nvmap> nvmap_dev)
+    : nvhost_nvdec_common(system, std::move(nvmap_dev)) {}
 nvhost_nvdec::~nvhost_nvdec() = default;
 
 u32 nvhost_nvdec::ioctl(Ioctl command, const std::vector<u8>& input, const std::vector<u8>& input2,
@@ -21,7 +23,7 @@ u32 nvhost_nvdec::ioctl(Ioctl command, const std::vector<u8>& input, const std::
 
     switch (static_cast<IoctlCommand>(command.raw)) {
     case IoctlCommand::IocSetNVMAPfdCommand:
-        return SetNVMAPfd(input, output);
+        return SetNVMAPfd(input);
     case IoctlCommand::IocSubmit:
         return Submit(input, output);
     case IoctlCommand::IocGetSyncpoint:
@@ -29,79 +31,29 @@ u32 nvhost_nvdec::ioctl(Ioctl command, const std::vector<u8>& input, const std::
     case IoctlCommand::IocGetWaitbase:
         return GetWaitbase(input, output);
     case IoctlCommand::IocMapBuffer:
-        return MapBuffer(input, output);
+    case IoctlCommand::IocMapBuffer2:
+    case IoctlCommand::IocMapBuffer3:
     case IoctlCommand::IocMapBufferEx:
-        return MapBufferEx(input, output);
-    case IoctlCommand::IocUnmapBufferEx:
-        return UnmapBufferEx(input, output);
+        return MapBuffer(input, output);
+    case IoctlCommand::IocUnmapBufferEx: {
+        // This command is sent when the video stream has ended, flush all video contexts
+        // This is usually sent in the folowing order: vic, nvdec, vic.
+        // Inform the GPU to clear any remaining nvdec buffers when this is detected.
+        LOG_INFO(Service_NVDRV, "NVDEC video stream ended");
+        Tegra::ChCommandHeaderList cmdlist(1);
+        cmdlist[0] = Tegra::ChCommandHeader{0xDEADB33F};
+        system.GPU().PushCommandBuffer(cmdlist);
+        [[fallthrough]]; // fallthrough to unmap buffers
+    };
+    case IoctlCommand::IocUnmapBuffer:
+    case IoctlCommand::IocUnmapBuffer2:
+    case IoctlCommand::IocUnmapBuffer3:
+        return UnmapBuffer(input, output);
+    case IoctlCommand::IocSetSubmitTimeout:
+        return SetSubmitTimeout(input, output);
     }
 
-    UNIMPLEMENTED_MSG("Unimplemented ioctl");
-    return 0;
-}
-
-u32 nvhost_nvdec::SetNVMAPfd(const std::vector<u8>& input, std::vector<u8>& output) {
-    IoctlSetNvmapFD params{};
-    std::memcpy(&params, input.data(), sizeof(IoctlSetNvmapFD));
-    LOG_DEBUG(Service_NVDRV, "called, fd={}", params.nvmap_fd);
-
-    nvmap_fd = params.nvmap_fd;
-    return 0;
-}
-
-u32 nvhost_nvdec::Submit(const std::vector<u8>& input, std::vector<u8>& output) {
-    IoctlSubmit params{};
-    std::memcpy(&params, input.data(), sizeof(IoctlSubmit));
-    LOG_WARNING(Service_NVDRV, "(STUBBED) called");
-    std::memcpy(output.data(), &params, sizeof(IoctlSubmit));
-    return 0;
-}
-
-u32 nvhost_nvdec::GetSyncpoint(const std::vector<u8>& input, std::vector<u8>& output) {
-    IoctlGetSyncpoint params{};
-    std::memcpy(&params, input.data(), sizeof(IoctlGetSyncpoint));
-    LOG_INFO(Service_NVDRV, "called, unknown=0x{:X}", params.unknown);
-    params.value = 0; // Seems to be hard coded at 0
-    std::memcpy(output.data(), &params, sizeof(IoctlGetSyncpoint));
-    return 0;
-}
-
-u32 nvhost_nvdec::GetWaitbase(const std::vector<u8>& input, std::vector<u8>& output) {
-    IoctlGetWaitbase params{};
-    std::memcpy(&params, input.data(), sizeof(IoctlGetWaitbase));
-    LOG_INFO(Service_NVDRV, "called, unknown=0x{:X}", params.unknown);
-    params.value = 0; // Seems to be hard coded at 0
-    std::memcpy(output.data(), &params, sizeof(IoctlGetWaitbase));
-    return 0;
-}
-
-u32 nvhost_nvdec::MapBuffer(const std::vector<u8>& input, std::vector<u8>& output) {
-    IoctlMapBuffer params{};
-    std::memcpy(&params, input.data(), sizeof(IoctlMapBuffer));
-    LOG_WARNING(Service_NVDRV, "(STUBBED) called with address={:08X}{:08X}", params.address_2,
-                params.address_1);
-    params.address_1 = 0;
-    params.address_2 = 0;
-    std::memcpy(output.data(), &params, sizeof(IoctlMapBuffer));
-    return 0;
-}
-
-u32 nvhost_nvdec::MapBufferEx(const std::vector<u8>& input, std::vector<u8>& output) {
-    IoctlMapBufferEx params{};
-    std::memcpy(&params, input.data(), sizeof(IoctlMapBufferEx));
-    LOG_WARNING(Service_NVDRV, "(STUBBED) called with address={:08X}{:08X}", params.address_2,
-                params.address_1);
-    params.address_1 = 0;
-    params.address_2 = 0;
-    std::memcpy(output.data(), &params, sizeof(IoctlMapBufferEx));
-    return 0;
-}
-
-u32 nvhost_nvdec::UnmapBufferEx(const std::vector<u8>& input, std::vector<u8>& output) {
-    IoctlUnmapBufferEx params{};
-    std::memcpy(&params, input.data(), sizeof(IoctlUnmapBufferEx));
-    LOG_WARNING(Service_NVDRV, "(STUBBED) called");
-    std::memcpy(output.data(), &params, sizeof(IoctlUnmapBufferEx));
+    UNIMPLEMENTED_MSG("Unimplemented ioctl 0x{:X}", command.raw);
     return 0;
 }
 
diff --git a/src/core/hle/service/nvdrv/devices/nvhost_nvdec.h b/src/core/hle/service/nvdrv/devices/nvhost_nvdec.h
index 4332db118..102777ddd 100644
--- a/src/core/hle/service/nvdrv/devices/nvhost_nvdec.h
+++ b/src/core/hle/service/nvdrv/devices/nvhost_nvdec.h
@@ -4,16 +4,14 @@
 
 #pragma once
 
-#include <vector>
-#include "common/common_types.h"
-#include "common/swap.h"
-#include "core/hle/service/nvdrv/devices/nvdevice.h"
+#include <memory>
+#include "core/hle/service/nvdrv/devices/nvhost_nvdec_common.h"
 
 namespace Service::Nvidia::Devices {
 
-class nvhost_nvdec final : public nvdevice {
+class nvhost_nvdec final : public nvhost_nvdec_common {
 public:
-    explicit nvhost_nvdec(Core::System& system);
+    explicit nvhost_nvdec(Core::System& system, std::shared_ptr<nvmap> nvmap_dev);
     ~nvhost_nvdec() override;
 
     u32 ioctl(Ioctl command, const std::vector<u8>& input, const std::vector<u8>& input2,
@@ -27,62 +25,15 @@ private:
         IocGetSyncpoint = 0xC0080002,
         IocGetWaitbase = 0xC0080003,
         IocMapBuffer = 0xC01C0009,
+        IocMapBuffer2 = 0xC16C0009,
+        IocMapBuffer3 = 0xC15C0009,
         IocMapBufferEx = 0xC0A40009,
-        IocUnmapBufferEx = 0xC0A4000A,
+        IocUnmapBuffer = 0xC0A4000A,
+        IocUnmapBuffer2 = 0xC16C000A,
+        IocUnmapBufferEx = 0xC01C000A,
+        IocUnmapBuffer3 = 0xC15C000A,
+        IocSetSubmitTimeout = 0x40040007,
     };
-
-    struct IoctlSetNvmapFD {
-        u32_le nvmap_fd;
-    };
-    static_assert(sizeof(IoctlSetNvmapFD) == 0x4, "IoctlSetNvmapFD is incorrect size");
-
-    struct IoctlSubmit {
-        INSERT_PADDING_BYTES(0x40); // TODO(DarkLordZach): RE this structure
-    };
-    static_assert(sizeof(IoctlSubmit) == 0x40, "IoctlSubmit has incorrect size");
-
-    struct IoctlGetSyncpoint {
-        u32 unknown; // seems to be ignored? Nintendo added this
-        u32 value;
-    };
-    static_assert(sizeof(IoctlGetSyncpoint) == 0x08, "IoctlGetSyncpoint has incorrect size");
-
-    struct IoctlGetWaitbase {
-        u32 unknown; // seems to be ignored? Nintendo added this
-        u32 value;
-    };
-    static_assert(sizeof(IoctlGetWaitbase) == 0x08, "IoctlGetWaitbase has incorrect size");
-
-    struct IoctlMapBuffer {
-        u32 unknown;
-        u32 address_1;
-        u32 address_2;
-        INSERT_PADDING_BYTES(0x10); // TODO(DarkLordZach): RE this structure
-    };
-    static_assert(sizeof(IoctlMapBuffer) == 0x1C, "IoctlMapBuffer is incorrect size");
-
-    struct IoctlMapBufferEx {
-        u32 unknown;
-        u32 address_1;
-        u32 address_2;
-        INSERT_PADDING_BYTES(0x98); // TODO(DarkLordZach): RE this structure
-    };
-    static_assert(sizeof(IoctlMapBufferEx) == 0xA4, "IoctlMapBufferEx has incorrect size");
-
-    struct IoctlUnmapBufferEx {
-        INSERT_PADDING_BYTES(0xA4); // TODO(DarkLordZach): RE this structure
-    };
-    static_assert(sizeof(IoctlUnmapBufferEx) == 0xA4, "IoctlUnmapBufferEx has incorrect size");
-
-    u32_le nvmap_fd{};
-
-    u32 SetNVMAPfd(const std::vector<u8>& input, std::vector<u8>& output);
-    u32 Submit(const std::vector<u8>& input, std::vector<u8>& output);
-    u32 GetSyncpoint(const std::vector<u8>& input, std::vector<u8>& output);
-    u32 GetWaitbase(const std::vector<u8>& input, std::vector<u8>& output);
-    u32 MapBuffer(const std::vector<u8>& input, std::vector<u8>& output);
-    u32 MapBufferEx(const std::vector<u8>& input, std::vector<u8>& output);
-    u32 UnmapBufferEx(const std::vector<u8>& input, std::vector<u8>& output);
 };
 
 } // namespace Service::Nvidia::Devices
diff --git a/src/core/hle/service/nvdrv/devices/nvhost_nvdec_common.cpp b/src/core/hle/service/nvdrv/devices/nvhost_nvdec_common.cpp
new file mode 100644
index 000000000..85792495f
--- /dev/null
+++ b/src/core/hle/service/nvdrv/devices/nvhost_nvdec_common.cpp
@@ -0,0 +1,234 @@
+// Copyright 2020 yuzu emulator team
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#include <algorithm>
+#include <cstring>
+
+#include "common/assert.h"
+#include "common/common_types.h"
+#include "common/logging/log.h"
+#include "core/core.h"
+#include "core/hle/service/nvdrv/devices/nvhost_nvdec_common.h"
+#include "core/hle/service/nvdrv/devices/nvmap.h"
+#include "core/memory.h"
+#include "video_core/memory_manager.h"
+#include "video_core/renderer_base.h"
+
+namespace Service::Nvidia::Devices {
+
+namespace {
+// Splice vectors will copy count amount of type T from the input vector into the dst vector.
+template <typename T>
+std::size_t SpliceVectors(const std::vector<u8>& input, std::vector<T>& dst, std::size_t count,
+                          std::size_t offset) {
+    std::memcpy(dst.data(), input.data() + offset, count * sizeof(T));
+    offset += count * sizeof(T);
+    return offset;
+}
+
+// Write vectors will write data to the output buffer
+template <typename T>
+std::size_t WriteVectors(std::vector<u8>& dst, const std::vector<T>& src, std::size_t offset) {
+    std::memcpy(dst.data() + offset, src.data(), src.size() * sizeof(T));
+    offset += src.size() * sizeof(T);
+    return offset;
+}
+} // Anonymous namespace
+
+namespace NvErrCodes {
+constexpr u32 Success{};
+constexpr u32 OutOfMemory{static_cast<u32>(-12)};
+constexpr u32 InvalidInput{static_cast<u32>(-22)};
+} // namespace NvErrCodes
+
+nvhost_nvdec_common::nvhost_nvdec_common(Core::System& system, std::shared_ptr<nvmap> nvmap_dev)
+    : nvdevice(system), nvmap_dev(std::move(nvmap_dev)) {}
+nvhost_nvdec_common::~nvhost_nvdec_common() = default;
+
+u32 nvhost_nvdec_common::SetNVMAPfd(const std::vector<u8>& input) {
+    IoctlSetNvmapFD params{};
+    std::memcpy(&params, input.data(), sizeof(IoctlSetNvmapFD));
+    LOG_DEBUG(Service_NVDRV, "called, fd={}", params.nvmap_fd);
+
+    nvmap_fd = params.nvmap_fd;
+    return 0;
+}
+
+u32 nvhost_nvdec_common::Submit(const std::vector<u8>& input, std::vector<u8>& output) {
+    IoctlSubmit params{};
+    std::memcpy(&params, input.data(), sizeof(IoctlSubmit));
+    LOG_DEBUG(Service_NVDRV, "called NVDEC Submit, cmd_buffer_count={}", params.cmd_buffer_count);
+
+    // Instantiate param buffers
+    std::size_t offset = sizeof(IoctlSubmit);
+    std::vector<CommandBuffer> command_buffers(params.cmd_buffer_count);
+    std::vector<Reloc> relocs(params.relocation_count);
+    std::vector<u32> reloc_shifts(params.relocation_count);
+    std::vector<SyncptIncr> syncpt_increments(params.syncpoint_count);
+    std::vector<SyncptIncr> wait_checks(params.syncpoint_count);
+    std::vector<Fence> fences(params.fence_count);
+
+    // Splice input into their respective buffers
+    offset = SpliceVectors(input, command_buffers, params.cmd_buffer_count, offset);
+    offset = SpliceVectors(input, relocs, params.relocation_count, offset);
+    offset = SpliceVectors(input, reloc_shifts, params.relocation_count, offset);
+    offset = SpliceVectors(input, syncpt_increments, params.syncpoint_count, offset);
+    offset = SpliceVectors(input, wait_checks, params.syncpoint_count, offset);
+    offset = SpliceVectors(input, fences, params.fence_count, offset);
+
+    // TODO(ameerj): For async gpu, utilize fences for syncpoint 'max' increment
+
+    auto& gpu = system.GPU();
+
+    for (const auto& cmd_buffer : command_buffers) {
+        auto object = nvmap_dev->GetObject(cmd_buffer.memory_id);
+        ASSERT_OR_EXECUTE(object, return NvErrCodes::InvalidInput;);
+        const auto map = FindBufferMap(object->dma_map_addr);
+        if (!map) {
+            LOG_ERROR(Service_NVDRV, "Tried to submit an invalid offset 0x{:X} dma 0x{:X}",
+                      object->addr, object->dma_map_addr);
+            return 0;
+        }
+        Tegra::ChCommandHeaderList cmdlist(cmd_buffer.word_count);
+        gpu.MemoryManager().ReadBlock(map->StartAddr() + cmd_buffer.offset, cmdlist.data(),
+                                      cmdlist.size() * sizeof(u32));
+        gpu.PushCommandBuffer(cmdlist);
+    }
+
+    std::memcpy(output.data(), &params, sizeof(IoctlSubmit));
+    // Some games expect command_buffers to be written back
+    offset = sizeof(IoctlSubmit);
+    offset = WriteVectors(output, command_buffers, offset);
+    offset = WriteVectors(output, relocs, offset);
+    offset = WriteVectors(output, reloc_shifts, offset);
+    offset = WriteVectors(output, syncpt_increments, offset);
+    offset = WriteVectors(output, wait_checks, offset);
+
+    return NvErrCodes::Success;
+}
+
+u32 nvhost_nvdec_common::GetSyncpoint(const std::vector<u8>& input, std::vector<u8>& output) {
+    IoctlGetSyncpoint params{};
+    std::memcpy(&params, input.data(), sizeof(IoctlGetSyncpoint));
+    LOG_DEBUG(Service_NVDRV, "called GetSyncpoint, id={}", params.param);
+
+    // We found that implementing this causes deadlocks with async gpu, along with degraded
+    // performance. TODO: RE the nvdec async implementation
+    params.value = 0;
+    std::memcpy(output.data(), &params, sizeof(IoctlGetSyncpoint));
+
+    return NvErrCodes::Success;
+}
+
+u32 nvhost_nvdec_common::GetWaitbase(const std::vector<u8>& input, std::vector<u8>& output) {
+    IoctlGetWaitbase params{};
+    std::memcpy(&params, input.data(), sizeof(IoctlGetWaitbase));
+    params.value = 0; // Seems to be hard coded at 0
+    std::memcpy(output.data(), &params, sizeof(IoctlGetWaitbase));
+    return 0;
+}
+
+u32 nvhost_nvdec_common::MapBuffer(const std::vector<u8>& input, std::vector<u8>& output) {
+    IoctlMapBuffer params{};
+    std::memcpy(&params, input.data(), sizeof(IoctlMapBuffer));
+    std::vector<MapBufferEntry> cmd_buffer_handles(params.num_entries);
+
+    SpliceVectors(input, cmd_buffer_handles, params.num_entries, sizeof(IoctlMapBuffer));
+
+    auto& gpu = system.GPU();
+
+    for (auto& cmf_buff : cmd_buffer_handles) {
+        auto object{nvmap_dev->GetObject(cmf_buff.map_handle)};
+        if (!object) {
+            LOG_ERROR(Service_NVDRV, "invalid cmd_buffer nvmap_handle={:X}", cmf_buff.map_handle);
+            std::memcpy(output.data(), &params, output.size());
+            return NvErrCodes::InvalidInput;
+        }
+        if (object->dma_map_addr == 0) {
+            // NVDEC and VIC memory is in the 32-bit address space
+            // MapAllocate32 will attempt to map a lower 32-bit value in the shared gpu memory space
+            const GPUVAddr low_addr = gpu.MemoryManager().MapAllocate32(object->addr, object->size);
+            object->dma_map_addr = static_cast<u32>(low_addr);
+            // Ensure that the dma_map_addr is indeed in the lower 32-bit address space.
+            ASSERT(object->dma_map_addr == low_addr);
+        }
+        if (!object->dma_map_addr) {
+            LOG_ERROR(Service_NVDRV, "failed to map size={}", object->size);
+        } else {
+            cmf_buff.map_address = object->dma_map_addr;
+            AddBufferMap(object->dma_map_addr, object->size, object->addr,
+                         object->status == nvmap::Object::Status::Allocated);
+        }
+    }
+    std::memcpy(output.data(), &params, sizeof(IoctlMapBuffer));
+    std::memcpy(output.data() + sizeof(IoctlMapBuffer), cmd_buffer_handles.data(),
+                cmd_buffer_handles.size() * sizeof(MapBufferEntry));
+
+    return NvErrCodes::Success;
+}
+
+u32 nvhost_nvdec_common::UnmapBuffer(const std::vector<u8>& input, std::vector<u8>& output) {
+    IoctlMapBuffer params{};
+    std::memcpy(&params, input.data(), sizeof(IoctlMapBuffer));
+    std::vector<MapBufferEntry> cmd_buffer_handles(params.num_entries);
+    SpliceVectors(input, cmd_buffer_handles, params.num_entries, sizeof(IoctlMapBuffer));
+
+    auto& gpu = system.GPU();
+
+    for (auto& cmf_buff : cmd_buffer_handles) {
+        const auto object{nvmap_dev->GetObject(cmf_buff.map_handle)};
+        if (!object) {
+            LOG_ERROR(Service_NVDRV, "invalid cmd_buffer nvmap_handle={:X}", cmf_buff.map_handle);
+            std::memcpy(output.data(), &params, output.size());
+            return NvErrCodes::InvalidInput;
+        }
+        if (const auto size{RemoveBufferMap(object->dma_map_addr)}; size) {
+            gpu.MemoryManager().Unmap(object->dma_map_addr, *size);
+        } else {
+            // This occurs quite frequently, however does not seem to impact functionality
+            LOG_DEBUG(Service_NVDRV, "invalid offset=0x{:X} dma=0x{:X}", object->addr,
+                      object->dma_map_addr);
+        }
+        object->dma_map_addr = 0;
+    }
+    std::memset(output.data(), 0, output.size());
+    return NvErrCodes::Success;
+}
+
+u32 nvhost_nvdec_common::SetSubmitTimeout(const std::vector<u8>& input, std::vector<u8>& output) {
+    std::memcpy(&submit_timeout, input.data(), input.size());
+    LOG_WARNING(Service_NVDRV, "(STUBBED) called");
+    return NvErrCodes::Success;
+}
+
+std::optional<nvhost_nvdec_common::BufferMap> nvhost_nvdec_common::FindBufferMap(
+    GPUVAddr gpu_addr) const {
+    const auto it = std::find_if(
+        buffer_mappings.begin(), buffer_mappings.upper_bound(gpu_addr), [&](const auto& entry) {
+            return (gpu_addr >= entry.second.StartAddr() && gpu_addr < entry.second.EndAddr());
+        });
+
+    ASSERT(it != buffer_mappings.end());
+    return it->second;
+}
+
+void nvhost_nvdec_common::AddBufferMap(GPUVAddr gpu_addr, std::size_t size, VAddr cpu_addr,
+                                       bool is_allocated) {
+    buffer_mappings.insert_or_assign(gpu_addr, BufferMap{gpu_addr, size, cpu_addr, is_allocated});
+}
+
+std::optional<std::size_t> nvhost_nvdec_common::RemoveBufferMap(GPUVAddr gpu_addr) {
+    const auto iter{buffer_mappings.find(gpu_addr)};
+    if (iter == buffer_mappings.end()) {
+        return std::nullopt;
+    }
+    std::size_t size = 0;
+    if (iter->second.IsAllocated()) {
+        size = iter->second.Size();
+    }
+    buffer_mappings.erase(iter);
+    return size;
+}
+
+} // namespace Service::Nvidia::Devices
diff --git a/src/core/hle/service/nvdrv/devices/nvhost_nvdec_common.h b/src/core/hle/service/nvdrv/devices/nvhost_nvdec_common.h
new file mode 100644
index 000000000..c249c5349
--- /dev/null
+++ b/src/core/hle/service/nvdrv/devices/nvhost_nvdec_common.h
@@ -0,0 +1,168 @@
+// Copyright 2020 yuzu emulator team
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#pragma once
+
+#include <map>
+#include <vector>
+#include "common/common_types.h"
+#include "common/swap.h"
+#include "core/hle/service/nvdrv/devices/nvdevice.h"
+
+namespace Service::Nvidia::Devices {
+class nvmap;
+
+class nvhost_nvdec_common : public nvdevice {
+public:
+    explicit nvhost_nvdec_common(Core::System& system, std::shared_ptr<nvmap> nvmap_dev);
+    ~nvhost_nvdec_common() override;
+
+    virtual u32 ioctl(Ioctl command, const std::vector<u8>& input, const std::vector<u8>& input2,
+                      std::vector<u8>& output, std::vector<u8>& output2, IoctlCtrl& ctrl,
+                      IoctlVersion version) = 0;
+
+protected:
+    class BufferMap final {
+    public:
+        constexpr BufferMap() = default;
+
+        constexpr BufferMap(GPUVAddr start_addr, std::size_t size)
+            : start_addr{start_addr}, end_addr{start_addr + size} {}
+
+        constexpr BufferMap(GPUVAddr start_addr, std::size_t size, VAddr cpu_addr,
+                            bool is_allocated)
+            : start_addr{start_addr}, end_addr{start_addr + size}, cpu_addr{cpu_addr},
+              is_allocated{is_allocated} {}
+
+        constexpr VAddr StartAddr() const {
+            return start_addr;
+        }
+
+        constexpr VAddr EndAddr() const {
+            return end_addr;
+        }
+
+        constexpr std::size_t Size() const {
+            return end_addr - start_addr;
+        }
+
+        constexpr VAddr CpuAddr() const {
+            return cpu_addr;
+        }
+
+        constexpr bool IsAllocated() const {
+            return is_allocated;
+        }
+
+    private:
+        GPUVAddr start_addr{};
+        GPUVAddr end_addr{};
+        VAddr cpu_addr{};
+        bool is_allocated{};
+    };
+
+    struct IoctlSetNvmapFD {
+        u32_le nvmap_fd;
+    };
+    static_assert(sizeof(IoctlSetNvmapFD) == 4, "IoctlSetNvmapFD is incorrect size");
+
+    struct IoctlSubmitCommandBuffer {
+        u32_le id;
+        u32_le offset;
+        u32_le count;
+    };
+    static_assert(sizeof(IoctlSubmitCommandBuffer) == 0xC,
+                  "IoctlSubmitCommandBuffer is incorrect size");
+    struct IoctlSubmit {
+        u32_le cmd_buffer_count;
+        u32_le relocation_count;
+        u32_le syncpoint_count;
+        u32_le fence_count;
+    };
+    static_assert(sizeof(IoctlSubmit) == 0x10, "IoctlSubmit has incorrect size");
+
+    struct CommandBuffer {
+        s32 memory_id;
+        u32 offset;
+        s32 word_count;
+    };
+    static_assert(sizeof(CommandBuffer) == 0xC, "CommandBuffer has incorrect size");
+
+    struct Reloc {
+        s32 cmdbuffer_memory;
+        s32 cmdbuffer_offset;
+        s32 target;
+        s32 target_offset;
+    };
+    static_assert(sizeof(Reloc) == 0x10, "CommandBuffer has incorrect size");
+
+    struct SyncptIncr {
+        u32 id;
+        u32 increments;
+    };
+    static_assert(sizeof(SyncptIncr) == 0x8, "CommandBuffer has incorrect size");
+
+    struct Fence {
+        u32 id;
+        u32 value;
+    };
+    static_assert(sizeof(Fence) == 0x8, "CommandBuffer has incorrect size");
+
+    struct IoctlGetSyncpoint {
+        // Input
+        u32_le param;
+        // Output
+        u32_le value;
+    };
+    static_assert(sizeof(IoctlGetSyncpoint) == 8, "IocGetIdParams has wrong size");
+
+    struct IoctlGetWaitbase {
+        u32_le unknown; // seems to be ignored? Nintendo added this
+        u32_le value;
+    };
+    static_assert(sizeof(IoctlGetWaitbase) == 0x8, "IoctlGetWaitbase is incorrect size");
+
+    struct IoctlMapBuffer {
+        u32_le num_entries;
+        u32_le data_address; // Ignored by the driver.
+        u32_le attach_host_ch_das;
+    };
+    static_assert(sizeof(IoctlMapBuffer) == 0x0C, "IoctlMapBuffer is incorrect size");
+
+    struct IocGetIdParams {
+        // Input
+        u32_le param;
+        // Output
+        u32_le value;
+    };
+    static_assert(sizeof(IocGetIdParams) == 8, "IocGetIdParams has wrong size");
+
+    // Used for mapping and unmapping command buffers
+    struct MapBufferEntry {
+        u32_le map_handle;
+        u32_le map_address;
+    };
+    static_assert(sizeof(IoctlMapBuffer) == 0x0C, "IoctlMapBuffer is incorrect size");
+
+    /// Ioctl command implementations
+    u32 SetNVMAPfd(const std::vector<u8>& input);
+    u32 Submit(const std::vector<u8>& input, std::vector<u8>& output);
+    u32 GetSyncpoint(const std::vector<u8>& input, std::vector<u8>& output);
+    u32 GetWaitbase(const std::vector<u8>& input, std::vector<u8>& output);
+    u32 MapBuffer(const std::vector<u8>& input, std::vector<u8>& output);
+    u32 UnmapBuffer(const std::vector<u8>& input, std::vector<u8>& output);
+    u32 SetSubmitTimeout(const std::vector<u8>& input, std::vector<u8>& output);
+
+    std::optional<BufferMap> FindBufferMap(GPUVAddr gpu_addr) const;
+    void AddBufferMap(GPUVAddr gpu_addr, std::size_t size, VAddr cpu_addr, bool is_allocated);
+    std::optional<std::size_t> RemoveBufferMap(GPUVAddr gpu_addr);
+
+    u32_le nvmap_fd{};
+    u32_le submit_timeout{};
+    std::shared_ptr<nvmap> nvmap_dev;
+
+    // This is expected to be ordered, therefore we must use a map, not unordered_map
+    std::map<GPUVAddr, BufferMap> buffer_mappings;
+};
+}; // namespace Service::Nvidia::Devices
diff --git a/src/core/hle/service/nvdrv/devices/nvhost_vic.cpp b/src/core/hle/service/nvdrv/devices/nvhost_vic.cpp
index 9da19ad56..60db54d00 100644
--- a/src/core/hle/service/nvdrv/devices/nvhost_vic.cpp
+++ b/src/core/hle/service/nvdrv/devices/nvhost_vic.cpp
@@ -2,15 +2,17 @@
 // Licensed under GPLv2 or any later version
 // Refer to the license.txt file included.
 
-#include <cstring>
-
 #include "common/assert.h"
 #include "common/logging/log.h"
+#include "core/core.h"
 #include "core/hle/service/nvdrv/devices/nvhost_vic.h"
+#include "video_core/memory_manager.h"
+#include "video_core/renderer_base.h"
 
 namespace Service::Nvidia::Devices {
+nvhost_vic::nvhost_vic(Core::System& system, std::shared_ptr<nvmap> nvmap_dev)
+    : nvhost_nvdec_common(system, std::move(nvmap_dev)) {}
 
-nvhost_vic::nvhost_vic(Core::System& system) : nvdevice(system) {}
 nvhost_vic::~nvhost_vic() = default;
 
 u32 nvhost_vic::ioctl(Ioctl command, const std::vector<u8>& input, const std::vector<u8>& input2,
@@ -21,7 +23,7 @@ u32 nvhost_vic::ioctl(Ioctl command, const std::vector<u8>& input, const std::ve
 
     switch (static_cast<IoctlCommand>(command.raw)) {
     case IoctlCommand::IocSetNVMAPfdCommand:
-        return SetNVMAPfd(input, output);
+        return SetNVMAPfd(input);
     case IoctlCommand::IocSubmit:
         return Submit(input, output);
     case IoctlCommand::IocGetSyncpoint:
@@ -29,83 +31,19 @@ u32 nvhost_vic::ioctl(Ioctl command, const std::vector<u8>& input, const std::ve
     case IoctlCommand::IocGetWaitbase:
         return GetWaitbase(input, output);
     case IoctlCommand::IocMapBuffer:
-        return MapBuffer(input, output);
+    case IoctlCommand::IocMapBuffer2:
+    case IoctlCommand::IocMapBuffer3:
+    case IoctlCommand::IocMapBuffer4:
     case IoctlCommand::IocMapBufferEx:
         return MapBuffer(input, output);
+    case IoctlCommand::IocUnmapBuffer:
+    case IoctlCommand::IocUnmapBuffer2:
+    case IoctlCommand::IocUnmapBuffer3:
     case IoctlCommand::IocUnmapBufferEx:
-        return UnmapBufferEx(input, output);
+        return UnmapBuffer(input, output);
     }
 
-    UNIMPLEMENTED_MSG("Unimplemented ioctl");
-    return 0;
-}
-
-u32 nvhost_vic::SetNVMAPfd(const std::vector<u8>& input, std::vector<u8>& output) {
-    IoctlSetNvmapFD params{};
-    std::memcpy(&params, input.data(), sizeof(IoctlSetNvmapFD));
-    LOG_DEBUG(Service_NVDRV, "called, fd={}", params.nvmap_fd);
-
-    nvmap_fd = params.nvmap_fd;
-    return 0;
-}
-
-u32 nvhost_vic::Submit(const std::vector<u8>& input, std::vector<u8>& output) {
-    IoctlSubmit params{};
-    std::memcpy(&params, input.data(), sizeof(IoctlSubmit));
-    LOG_WARNING(Service_NVDRV, "(STUBBED) called");
-
-    // Workaround for Luigi's Mansion 3, as nvhost_vic is not implemented for asynch GPU
-    params.command_buffer = {};
-
-    std::memcpy(output.data(), &params, sizeof(IoctlSubmit));
-    return 0;
-}
-
-u32 nvhost_vic::GetSyncpoint(const std::vector<u8>& input, std::vector<u8>& output) {
-    IoctlGetSyncpoint params{};
-    std::memcpy(&params, input.data(), sizeof(IoctlGetSyncpoint));
-    LOG_INFO(Service_NVDRV, "called, unknown=0x{:X}", params.unknown);
-    params.value = 0; // Seems to be hard coded at 0
-    std::memcpy(output.data(), &params, sizeof(IoctlGetSyncpoint));
-    return 0;
-}
-
-u32 nvhost_vic::GetWaitbase(const std::vector<u8>& input, std::vector<u8>& output) {
-    IoctlGetWaitbase params{};
-    std::memcpy(&params, input.data(), sizeof(IoctlGetWaitbase));
-    LOG_INFO(Service_NVDRV, "called, unknown=0x{:X}", params.unknown);
-    params.value = 0; // Seems to be hard coded at 0
-    std::memcpy(output.data(), &params, sizeof(IoctlGetWaitbase));
-    return 0;
-}
-
-u32 nvhost_vic::MapBuffer(const std::vector<u8>& input, std::vector<u8>& output) {
-    IoctlMapBuffer params{};
-    std::memcpy(&params, input.data(), sizeof(IoctlMapBuffer));
-    LOG_WARNING(Service_NVDRV, "(STUBBED) called with address={:08X}{:08X}", params.address_2,
-                params.address_1);
-    params.address_1 = 0;
-    params.address_2 = 0;
-    std::memcpy(output.data(), &params, sizeof(IoctlMapBuffer));
-    return 0;
-}
-
-u32 nvhost_vic::MapBufferEx(const std::vector<u8>& input, std::vector<u8>& output) {
-    IoctlMapBufferEx params{};
-    std::memcpy(&params, input.data(), sizeof(IoctlMapBufferEx));
-    LOG_WARNING(Service_NVDRV, "(STUBBED) called with address={:08X}{:08X}", params.address_2,
-                params.address_1);
-    params.address_1 = 0;
-    params.address_2 = 0;
-    std::memcpy(output.data(), &params, sizeof(IoctlMapBufferEx));
-    return 0;
-}
-
-u32 nvhost_vic::UnmapBufferEx(const std::vector<u8>& input, std::vector<u8>& output) {
-    IoctlUnmapBufferEx params{};
-    std::memcpy(&params, input.data(), sizeof(IoctlUnmapBufferEx));
-    LOG_WARNING(Service_NVDRV, "(STUBBED) called");
-    std::memcpy(output.data(), &params, sizeof(IoctlUnmapBufferEx));
+    UNIMPLEMENTED_MSG("Unimplemented ioctl 0x{:X}", command.raw);
     return 0;
 }
 
diff --git a/src/core/hle/service/nvdrv/devices/nvhost_vic.h b/src/core/hle/service/nvdrv/devices/nvhost_vic.h
index a7bb7bbd5..f975b190c 100644
--- a/src/core/hle/service/nvdrv/devices/nvhost_vic.h
+++ b/src/core/hle/service/nvdrv/devices/nvhost_vic.h
@@ -4,19 +4,15 @@
 
 #pragma once
 
-#include <array>
-#include <vector>
-#include "common/common_types.h"
-#include "common/swap.h"
-#include "core/hle/service/nvdrv/devices/nvdevice.h"
+#include "core/hle/service/nvdrv/devices/nvhost_nvdec_common.h"
 
 namespace Service::Nvidia::Devices {
+class nvmap;
 
-class nvhost_vic final : public nvdevice {
+class nvhost_vic final : public nvhost_nvdec_common {
 public:
-    explicit nvhost_vic(Core::System& system);
-    ~nvhost_vic() override;
-
+    explicit nvhost_vic(Core::System& system, std::shared_ptr<nvmap> nvmap_dev);
+    ~nvhost_vic();
     u32 ioctl(Ioctl command, const std::vector<u8>& input, const std::vector<u8>& input2,
               std::vector<u8>& output, std::vector<u8>& output2, IoctlCtrl& ctrl,
               IoctlVersion version) override;
@@ -28,74 +24,14 @@ private:
         IocGetSyncpoint = 0xC0080002,
         IocGetWaitbase = 0xC0080003,
         IocMapBuffer = 0xC01C0009,
+        IocMapBuffer2 = 0xC0340009,
+        IocMapBuffer3 = 0xC0140009,
+        IocMapBuffer4 = 0xC00C0009,
         IocMapBufferEx = 0xC03C0009,
-        IocUnmapBufferEx = 0xC03C000A,
-    };
-
-    struct IoctlSetNvmapFD {
-        u32_le nvmap_fd;
-    };
-    static_assert(sizeof(IoctlSetNvmapFD) == 4, "IoctlSetNvmapFD is incorrect size");
-
-    struct IoctlSubmitCommandBuffer {
-        u32 id;
-        u32 offset;
-        u32 count;
-    };
-    static_assert(sizeof(IoctlSubmitCommandBuffer) == 0xC,
-                  "IoctlSubmitCommandBuffer is incorrect size");
-
-    struct IoctlSubmit {
-        u32 command_buffer_count;
-        u32 relocations_count;
-        u32 syncpt_count;
-        u32 wait_count;
-        std::array<IoctlSubmitCommandBuffer, 4> command_buffer;
-    };
-    static_assert(sizeof(IoctlSubmit) == 0x40, "IoctlSubmit is incorrect size");
-
-    struct IoctlGetSyncpoint {
-        u32 unknown; // seems to be ignored? Nintendo added this
-        u32 value;
-    };
-    static_assert(sizeof(IoctlGetSyncpoint) == 0x8, "IoctlGetSyncpoint is incorrect size");
-
-    struct IoctlGetWaitbase {
-        u32 unknown; // seems to be ignored? Nintendo added this
-        u32 value;
-    };
-    static_assert(sizeof(IoctlGetWaitbase) == 0x8, "IoctlGetWaitbase is incorrect size");
-
-    struct IoctlMapBuffer {
-        u32 unknown;
-        u32 address_1;
-        u32 address_2;
-        INSERT_PADDING_BYTES(0x10); // TODO(DarkLordZach): RE this structure
-    };
-    static_assert(sizeof(IoctlMapBuffer) == 0x1C, "IoctlMapBuffer is incorrect size");
-
-    struct IoctlMapBufferEx {
-        u32 unknown;
-        u32 address_1;
-        u32 address_2;
-        INSERT_PADDING_BYTES(0x30); // TODO(DarkLordZach): RE this structure
+        IocUnmapBuffer = 0xC03C000A,
+        IocUnmapBuffer2 = 0xC034000A,
+        IocUnmapBuffer3 = 0xC00C000A,
+        IocUnmapBufferEx = 0xC01C000A,
     };
-    static_assert(sizeof(IoctlMapBufferEx) == 0x3C, "IoctlMapBufferEx is incorrect size");
-
-    struct IoctlUnmapBufferEx {
-        INSERT_PADDING_BYTES(0x3C); // TODO(DarkLordZach): RE this structure
-    };
-    static_assert(sizeof(IoctlUnmapBufferEx) == 0x3C, "IoctlUnmapBufferEx is incorrect size");
-
-    u32_le nvmap_fd{};
-
-    u32 SetNVMAPfd(const std::vector<u8>& input, std::vector<u8>& output);
-    u32 Submit(const std::vector<u8>& input, std::vector<u8>& output);
-    u32 GetSyncpoint(const std::vector<u8>& input, std::vector<u8>& output);
-    u32 GetWaitbase(const std::vector<u8>& input, std::vector<u8>& output);
-    u32 MapBuffer(const std::vector<u8>& input, std::vector<u8>& output);
-    u32 MapBufferEx(const std::vector<u8>& input, std::vector<u8>& output);
-    u32 UnmapBufferEx(const std::vector<u8>& input, std::vector<u8>& output);
 };
-
 } // namespace Service::Nvidia::Devices
diff --git a/src/core/hle/service/nvdrv/devices/nvmap.h b/src/core/hle/service/nvdrv/devices/nvmap.h
index 84624be00..04b9ef540 100644
--- a/src/core/hle/service/nvdrv/devices/nvmap.h
+++ b/src/core/hle/service/nvdrv/devices/nvmap.h
@@ -37,6 +37,7 @@ public:
         VAddr addr;
         Status status;
         u32 refcount;
+        u32 dma_map_addr;
     };
 
     std::shared_ptr<Object> GetObject(u32 handle) const {
diff --git a/src/core/hle/service/nvdrv/nvdrv.cpp b/src/core/hle/service/nvdrv/nvdrv.cpp
index 197c77db0..803c1a984 100644
--- a/src/core/hle/service/nvdrv/nvdrv.cpp
+++ b/src/core/hle/service/nvdrv/nvdrv.cpp
@@ -51,9 +51,9 @@ Module::Module(Core::System& system) {
     devices["/dev/nvmap"] = nvmap_dev;
     devices["/dev/nvdisp_disp0"] = std::make_shared<Devices::nvdisp_disp0>(system, nvmap_dev);
     devices["/dev/nvhost-ctrl"] = std::make_shared<Devices::nvhost_ctrl>(system, events_interface);
-    devices["/dev/nvhost-nvdec"] = std::make_shared<Devices::nvhost_nvdec>(system);
+    devices["/dev/nvhost-nvdec"] = std::make_shared<Devices::nvhost_nvdec>(system, nvmap_dev);
     devices["/dev/nvhost-nvjpg"] = std::make_shared<Devices::nvhost_nvjpg>(system);
-    devices["/dev/nvhost-vic"] = std::make_shared<Devices::nvhost_vic>(system);
+    devices["/dev/nvhost-vic"] = std::make_shared<Devices::nvhost_vic>(system, nvmap_dev);
 }
 
 Module::~Module() = default;
diff --git a/src/core/hle/service/set/set.cpp b/src/core/hle/service/set/set.cpp
index e64777668..ffbf90b00 100644
--- a/src/core/hle/service/set/set.cpp
+++ b/src/core/hle/service/set/set.cpp
@@ -202,6 +202,7 @@ SET::SET() : ServiceFramework("set") {
         {8, &SET::GetQuestFlag, "GetQuestFlag"},
         {9, &SET::GetKeyCodeMap2, "GetKeyCodeMap2"},
         {10, nullptr, "GetFirmwareVersionForDebug"},
+        {11, nullptr, "GetDeviceNickName"},
     };
     // clang-format on
 
diff --git a/src/core/hle/service/set/set_sys.cpp b/src/core/hle/service/set/set_sys.cpp
index 8bd4c7e79..080b5743e 100644
--- a/src/core/hle/service/set/set_sys.cpp
+++ b/src/core/hle/service/set/set_sys.cpp
@@ -300,6 +300,8 @@ SET_SYS::SET_SYS() : ServiceFramework("set:sys") {
         {198, nullptr, "SetButtonConfigRegisteredSettingsEmbedded"},
         {199, nullptr, "GetButtonConfigRegisteredSettings"},
         {200, nullptr, "SetButtonConfigRegisteredSettings"},
+        {201, nullptr, "GetFieldTestingFlag"},
+        {202, nullptr, "SetFieldTestingFlag"},
     };
     // clang-format on
 
diff --git a/src/core/hle/service/time/time.cpp b/src/core/hle/service/time/time.cpp
index ee4fa4b48..7d0474e0b 100644
--- a/src/core/hle/service/time/time.cpp
+++ b/src/core/hle/service/time/time.cpp
@@ -10,6 +10,7 @@
 #include "core/hle/ipc_helpers.h"
 #include "core/hle/kernel/client_port.h"
 #include "core/hle/kernel/client_session.h"
+#include "core/hle/kernel/kernel.h"
 #include "core/hle/kernel/scheduler.h"
 #include "core/hle/service/time/interface.h"
 #include "core/hle/service/time/time.h"
@@ -125,7 +126,7 @@ ResultCode Module::Interface::GetClockSnapshotFromSystemClockContextInternal(
     Kernel::Thread* thread, Clock::SystemClockContext user_context,
     Clock::SystemClockContext network_context, u8 type, Clock::ClockSnapshot& clock_snapshot) {
 
-    auto& time_manager{module->GetTimeManager()};
+    auto& time_manager{system.GetTimeManager()};
 
     clock_snapshot.is_automatic_correction_enabled =
         time_manager.GetStandardUserSystemClockCore().IsAutomaticCorrectionEnabled();
@@ -182,7 +183,7 @@ void Module::Interface::GetStandardUserSystemClock(Kernel::HLERequestContext& ct
     LOG_DEBUG(Service_Time, "called");
     IPC::ResponseBuilder rb{ctx, 2, 0, 1};
     rb.Push(RESULT_SUCCESS);
-    rb.PushIpcInterface<ISystemClock>(module->GetTimeManager().GetStandardUserSystemClockCore(),
+    rb.PushIpcInterface<ISystemClock>(system.GetTimeManager().GetStandardUserSystemClockCore(),
                                       system);
 }
 
@@ -190,7 +191,7 @@ void Module::Interface::GetStandardNetworkSystemClock(Kernel::HLERequestContext&
     LOG_DEBUG(Service_Time, "called");
     IPC::ResponseBuilder rb{ctx, 2, 0, 1};
     rb.Push(RESULT_SUCCESS);
-    rb.PushIpcInterface<ISystemClock>(module->GetTimeManager().GetStandardNetworkSystemClockCore(),
+    rb.PushIpcInterface<ISystemClock>(system.GetTimeManager().GetStandardNetworkSystemClockCore(),
                                       system);
 }
 
@@ -198,29 +199,28 @@ void Module::Interface::GetStandardSteadyClock(Kernel::HLERequestContext& ctx) {
     LOG_DEBUG(Service_Time, "called");
     IPC::ResponseBuilder rb{ctx, 2, 0, 1};
     rb.Push(RESULT_SUCCESS);
-    rb.PushIpcInterface<ISteadyClock>(module->GetTimeManager().GetStandardSteadyClockCore(),
-                                      system);
+    rb.PushIpcInterface<ISteadyClock>(system.GetTimeManager().GetStandardSteadyClockCore(), system);
 }
 
 void Module::Interface::GetTimeZoneService(Kernel::HLERequestContext& ctx) {
     LOG_DEBUG(Service_Time, "called");
     IPC::ResponseBuilder rb{ctx, 2, 0, 1};
     rb.Push(RESULT_SUCCESS);
-    rb.PushIpcInterface<ITimeZoneService>(module->GetTimeManager().GetTimeZoneContentManager());
+    rb.PushIpcInterface<ITimeZoneService>(system.GetTimeManager().GetTimeZoneContentManager());
 }
 
 void Module::Interface::GetStandardLocalSystemClock(Kernel::HLERequestContext& ctx) {
     LOG_DEBUG(Service_Time, "called");
     IPC::ResponseBuilder rb{ctx, 2, 0, 1};
     rb.Push(RESULT_SUCCESS);
-    rb.PushIpcInterface<ISystemClock>(module->GetTimeManager().GetStandardLocalSystemClockCore(),
+    rb.PushIpcInterface<ISystemClock>(system.GetTimeManager().GetStandardLocalSystemClockCore(),
                                       system);
 }
 
 void Module::Interface::IsStandardNetworkSystemClockAccuracySufficient(
     Kernel::HLERequestContext& ctx) {
     LOG_DEBUG(Service_Time, "called");
-    auto& clock_core{module->GetTimeManager().GetStandardNetworkSystemClockCore()};
+    auto& clock_core{system.GetTimeManager().GetStandardNetworkSystemClockCore()};
     IPC::ResponseBuilder rb{ctx, 3};
     rb.Push(RESULT_SUCCESS);
     rb.Push<u32>(clock_core.IsStandardNetworkSystemClockAccuracySufficient(system));
@@ -229,7 +229,7 @@ void Module::Interface::IsStandardNetworkSystemClockAccuracySufficient(
 void Module::Interface::CalculateMonotonicSystemClockBaseTimePoint(Kernel::HLERequestContext& ctx) {
     LOG_DEBUG(Service_Time, "called");
 
-    auto& steady_clock_core{module->GetTimeManager().GetStandardSteadyClockCore()};
+    auto& steady_clock_core{system.GetTimeManager().GetStandardSteadyClockCore()};
     if (!steady_clock_core.IsInitialized()) {
         IPC::ResponseBuilder rb{ctx, 2};
         rb.Push(ERROR_UNINITIALIZED_CLOCK);
@@ -262,8 +262,8 @@ void Module::Interface::GetClockSnapshot(Kernel::HLERequestContext& ctx) {
 
     Clock::SystemClockContext user_context{};
     if (const ResultCode result{
-            module->GetTimeManager().GetStandardUserSystemClockCore().GetClockContext(
-                system, user_context)};
+            system.GetTimeManager().GetStandardUserSystemClockCore().GetClockContext(system,
+                                                                                     user_context)};
         result.IsError()) {
         IPC::ResponseBuilder rb{ctx, 2};
         rb.Push(result);
@@ -271,7 +271,7 @@ void Module::Interface::GetClockSnapshot(Kernel::HLERequestContext& ctx) {
     }
     Clock::SystemClockContext network_context{};
     if (const ResultCode result{
-            module->GetTimeManager().GetStandardNetworkSystemClockCore().GetClockContext(
+            system.GetTimeManager().GetStandardNetworkSystemClockCore().GetClockContext(
                 system, network_context)};
         result.IsError()) {
         IPC::ResponseBuilder rb{ctx, 2};
@@ -372,7 +372,7 @@ void Module::Interface::GetSharedMemoryNativeHandle(Kernel::HLERequestContext& c
     LOG_DEBUG(Service_Time, "called");
     IPC::ResponseBuilder rb{ctx, 2, 1};
     rb.Push(RESULT_SUCCESS);
-    rb.PushCopyObjects(module->GetTimeManager().GetSharedMemory().GetSharedMemoryHolder());
+    rb.PushCopyObjects(SharedFrom(&system.Kernel().GetTimeSharedMem()));
 }
 
 Module::Interface::Interface(std::shared_ptr<Module> module, Core::System& system, const char* name)
@@ -381,7 +381,7 @@ Module::Interface::Interface(std::shared_ptr<Module> module, Core::System& syste
 Module::Interface::~Interface() = default;
 
 void InstallInterfaces(Core::System& system) {
-    auto module{std::make_shared<Module>(system)};
+    auto module{std::make_shared<Module>()};
     std::make_shared<Time>(module, system, "time:a")->InstallAsService(system.ServiceManager());
     std::make_shared<Time>(module, system, "time:s")->InstallAsService(system.ServiceManager());
     std::make_shared<Time>(module, system, "time:u")->InstallAsService(system.ServiceManager());
diff --git a/src/core/hle/service/time/time.h b/src/core/hle/service/time/time.h
index 41f3002e9..49f4aac0a 100644
--- a/src/core/hle/service/time/time.h
+++ b/src/core/hle/service/time/time.h
@@ -16,7 +16,7 @@ namespace Service::Time {
 
 class Module final {
 public:
-    Module(Core::System& system) : time_manager{system} {}
+    Module() = default;
 
     class Interface : public ServiceFramework<Interface> {
     public:
@@ -46,13 +46,6 @@ public:
         std::shared_ptr<Module> module;
         Core::System& system;
     };
-
-    TimeManager& GetTimeManager() {
-        return time_manager;
-    }
-
-private:
-    TimeManager time_manager;
 };
 
 /// Registers all Time services with the specified service manager.
diff --git a/src/core/hle/service/time/time_manager.cpp b/src/core/hle/service/time/time_manager.cpp
index b4dfe45e5..858623e2b 100644
--- a/src/core/hle/service/time/time_manager.cpp
+++ b/src/core/hle/service/time/time_manager.cpp
@@ -22,125 +22,282 @@ static std::chrono::seconds GetSecondsSinceEpoch() {
            Settings::values.custom_rtc_differential;
 }
 
-static s64 GetExternalTimeZoneOffset() {
-    // With "auto" timezone setting, we use the external system's timezone offset
-    if (Settings::GetTimeZoneString() == "auto") {
-        return Common::TimeZone::GetCurrentOffsetSeconds().count();
-    }
-    return 0;
-}
-
 static s64 GetExternalRtcValue() {
-    return GetSecondsSinceEpoch().count() + GetExternalTimeZoneOffset();
-}
-
-TimeManager::TimeManager(Core::System& system)
-    : shared_memory{system}, standard_local_system_clock_core{standard_steady_clock_core},
-      standard_network_system_clock_core{standard_steady_clock_core},
-      standard_user_system_clock_core{standard_local_system_clock_core,
-                                      standard_network_system_clock_core, system},
-      ephemeral_network_system_clock_core{tick_based_steady_clock_core},
-      local_system_clock_context_writer{
-          std::make_shared<Clock::LocalSystemClockContextWriter>(shared_memory)},
-      network_system_clock_context_writer{
-          std::make_shared<Clock::NetworkSystemClockContextWriter>(shared_memory)},
-      ephemeral_network_system_clock_context_writer{
-          std::make_shared<Clock::EphemeralNetworkSystemClockContextWriter>()},
-      time_zone_content_manager{*this, system} {
-
-    const auto system_time{Clock::TimeSpanType::FromSeconds(GetExternalRtcValue())};
-    SetupStandardSteadyClock(system, Common::UUID::Generate(), system_time, {}, {});
-    SetupStandardLocalSystemClock(system, {}, system_time.ToSeconds());
-    SetupStandardNetworkSystemClock({}, standard_network_clock_accuracy);
-    SetupStandardUserSystemClock(system, {}, Clock::SteadyClockTimePoint::GetRandom());
-    SetupEphemeralNetworkSystemClock();
+    return GetSecondsSinceEpoch().count() + TimeManager::GetExternalTimeZoneOffset();
 }
 
-TimeManager::~TimeManager() = default;
+struct TimeManager::Impl final {
+    explicit Impl(Core::System& system)
+        : shared_memory{system}, standard_local_system_clock_core{standard_steady_clock_core},
+          standard_network_system_clock_core{standard_steady_clock_core},
+          standard_user_system_clock_core{standard_local_system_clock_core,
+                                          standard_network_system_clock_core, system},
+          ephemeral_network_system_clock_core{tick_based_steady_clock_core},
+          local_system_clock_context_writer{
+              std::make_shared<Clock::LocalSystemClockContextWriter>(shared_memory)},
+          network_system_clock_context_writer{
+              std::make_shared<Clock::NetworkSystemClockContextWriter>(shared_memory)},
+          ephemeral_network_system_clock_context_writer{
+              std::make_shared<Clock::EphemeralNetworkSystemClockContextWriter>()},
+          time_zone_content_manager{system} {
 
-void TimeManager::SetupTimeZoneManager(std::string location_name,
-                                       Clock::SteadyClockTimePoint time_zone_updated_time_point,
-                                       std::size_t total_location_name_count,
-                                       u128 time_zone_rule_version,
-                                       FileSys::VirtualFile& vfs_file) {
-    if (time_zone_content_manager.GetTimeZoneManager().SetDeviceLocationNameWithTimeZoneRule(
-            location_name, vfs_file) != RESULT_SUCCESS) {
-        UNREACHABLE();
-        return;
-    }
-
-    time_zone_content_manager.GetTimeZoneManager().SetUpdatedTime(time_zone_updated_time_point);
-    time_zone_content_manager.GetTimeZoneManager().SetTotalLocationNameCount(
-        total_location_name_count);
-    time_zone_content_manager.GetTimeZoneManager().SetTimeZoneRuleVersion(time_zone_rule_version);
-    time_zone_content_manager.GetTimeZoneManager().MarkAsInitialized();
-}
-
-void TimeManager::SetupStandardSteadyClock(Core::System& system, Common::UUID clock_source_id,
-                                           Clock::TimeSpanType setup_value,
-                                           Clock::TimeSpanType internal_offset,
-                                           bool is_rtc_reset_detected) {
-    standard_steady_clock_core.SetClockSourceId(clock_source_id);
-    standard_steady_clock_core.SetSetupValue(setup_value);
-    standard_steady_clock_core.SetInternalOffset(internal_offset);
-    standard_steady_clock_core.MarkAsInitialized();
-
-    const auto current_time_point{standard_steady_clock_core.GetCurrentRawTimePoint(system)};
-    shared_memory.SetupStandardSteadyClock(system, clock_source_id, current_time_point);
-}
-
-void TimeManager::SetupStandardLocalSystemClock(Core::System& system,
-                                                Clock::SystemClockContext clock_context,
-                                                s64 posix_time) {
-    standard_local_system_clock_core.SetUpdateCallbackInstance(local_system_clock_context_writer);
-
-    const auto current_time_point{
-        standard_local_system_clock_core.GetSteadyClockCore().GetCurrentTimePoint(system)};
-    if (current_time_point.clock_source_id == clock_context.steady_time_point.clock_source_id) {
-        standard_local_system_clock_core.SetSystemClockContext(clock_context);
-    } else {
-        if (standard_local_system_clock_core.SetCurrentTime(system, posix_time) != RESULT_SUCCESS) {
+        const auto system_time{Clock::TimeSpanType::FromSeconds(GetExternalRtcValue())};
+        SetupStandardSteadyClock(system, Common::UUID::Generate(), system_time, {}, {});
+        SetupStandardLocalSystemClock(system, {}, system_time.ToSeconds());
+        SetupStandardNetworkSystemClock({}, standard_network_clock_accuracy);
+        SetupStandardUserSystemClock(system, {}, Clock::SteadyClockTimePoint::GetRandom());
+        SetupEphemeralNetworkSystemClock();
+    }
+
+    ~Impl() = default;
+
+    Clock::StandardSteadyClockCore& GetStandardSteadyClockCore() {
+        return standard_steady_clock_core;
+    }
+
+    const Clock::StandardSteadyClockCore& GetStandardSteadyClockCore() const {
+        return standard_steady_clock_core;
+    }
+
+    Clock::StandardLocalSystemClockCore& GetStandardLocalSystemClockCore() {
+        return standard_local_system_clock_core;
+    }
+
+    const Clock::StandardLocalSystemClockCore& GetStandardLocalSystemClockCore() const {
+        return standard_local_system_clock_core;
+    }
+
+    Clock::StandardNetworkSystemClockCore& GetStandardNetworkSystemClockCore() {
+        return standard_network_system_clock_core;
+    }
+
+    const Clock::StandardNetworkSystemClockCore& GetStandardNetworkSystemClockCore() const {
+        return standard_network_system_clock_core;
+    }
+
+    Clock::StandardUserSystemClockCore& GetStandardUserSystemClockCore() {
+        return standard_user_system_clock_core;
+    }
+
+    const Clock::StandardUserSystemClockCore& GetStandardUserSystemClockCore() const {
+        return standard_user_system_clock_core;
+    }
+
+    TimeZone::TimeZoneContentManager& GetTimeZoneContentManager() {
+        return time_zone_content_manager;
+    }
+
+    const TimeZone::TimeZoneContentManager& GetTimeZoneContentManager() const {
+        return time_zone_content_manager;
+    }
+
+    SharedMemory& GetSharedMemory() {
+        return shared_memory;
+    }
+
+    const SharedMemory& GetSharedMemory() const {
+        return shared_memory;
+    }
+
+    void SetupTimeZoneManager(std::string location_name,
+                              Clock::SteadyClockTimePoint time_zone_updated_time_point,
+                              std::size_t total_location_name_count, u128 time_zone_rule_version,
+                              FileSys::VirtualFile& vfs_file) {
+        if (time_zone_content_manager.GetTimeZoneManager().SetDeviceLocationNameWithTimeZoneRule(
+                location_name, vfs_file) != RESULT_SUCCESS) {
             UNREACHABLE();
             return;
         }
+
+        time_zone_content_manager.GetTimeZoneManager().SetUpdatedTime(time_zone_updated_time_point);
+        time_zone_content_manager.GetTimeZoneManager().SetTotalLocationNameCount(
+            total_location_name_count);
+        time_zone_content_manager.GetTimeZoneManager().SetTimeZoneRuleVersion(
+            time_zone_rule_version);
+        time_zone_content_manager.GetTimeZoneManager().MarkAsInitialized();
     }
 
-    standard_local_system_clock_core.MarkAsInitialized();
-}
+    static s64 GetExternalTimeZoneOffset() {
+        // With "auto" timezone setting, we use the external system's timezone offset
+        if (Settings::GetTimeZoneString() == "auto") {
+            return Common::TimeZone::GetCurrentOffsetSeconds().count();
+        }
+        return 0;
+    }
 
-void TimeManager::SetupStandardNetworkSystemClock(Clock::SystemClockContext clock_context,
-                                                  Clock::TimeSpanType sufficient_accuracy) {
-    standard_network_system_clock_core.SetUpdateCallbackInstance(
-        network_system_clock_context_writer);
+    void SetupStandardSteadyClock(Core::System& system, Common::UUID clock_source_id,
+                                  Clock::TimeSpanType setup_value,
+                                  Clock::TimeSpanType internal_offset, bool is_rtc_reset_detected) {
+        standard_steady_clock_core.SetClockSourceId(clock_source_id);
+        standard_steady_clock_core.SetSetupValue(setup_value);
+        standard_steady_clock_core.SetInternalOffset(internal_offset);
+        standard_steady_clock_core.MarkAsInitialized();
 
-    if (standard_network_system_clock_core.SetSystemClockContext(clock_context) != RESULT_SUCCESS) {
-        UNREACHABLE();
-        return;
+        const auto current_time_point{standard_steady_clock_core.GetCurrentRawTimePoint(system)};
+        shared_memory.SetupStandardSteadyClock(system, clock_source_id, current_time_point);
     }
 
-    standard_network_system_clock_core.SetStandardNetworkClockSufficientAccuracy(
-        sufficient_accuracy);
-    standard_network_system_clock_core.MarkAsInitialized();
-}
+    void SetupStandardLocalSystemClock(Core::System& system,
+                                       Clock::SystemClockContext clock_context, s64 posix_time) {
+        standard_local_system_clock_core.SetUpdateCallbackInstance(
+            local_system_clock_context_writer);
+
+        const auto current_time_point{
+            standard_local_system_clock_core.GetSteadyClockCore().GetCurrentTimePoint(system)};
+        if (current_time_point.clock_source_id == clock_context.steady_time_point.clock_source_id) {
+            standard_local_system_clock_core.SetSystemClockContext(clock_context);
+        } else {
+            if (standard_local_system_clock_core.SetCurrentTime(system, posix_time) !=
+                RESULT_SUCCESS) {
+                UNREACHABLE();
+                return;
+            }
+        }
+
+        standard_local_system_clock_core.MarkAsInitialized();
+    }
+
+    void SetupStandardNetworkSystemClock(Clock::SystemClockContext clock_context,
+                                         Clock::TimeSpanType sufficient_accuracy) {
+        standard_network_system_clock_core.SetUpdateCallbackInstance(
+            network_system_clock_context_writer);
 
-void TimeManager::SetupStandardUserSystemClock(
-    Core::System& system, bool is_automatic_correction_enabled,
-    Clock::SteadyClockTimePoint steady_clock_time_point) {
-    if (standard_user_system_clock_core.SetAutomaticCorrectionEnabled(
-            system, is_automatic_correction_enabled) != RESULT_SUCCESS) {
-        UNREACHABLE();
-        return;
+        if (standard_network_system_clock_core.SetSystemClockContext(clock_context) !=
+            RESULT_SUCCESS) {
+            UNREACHABLE();
+            return;
+        }
+
+        standard_network_system_clock_core.SetStandardNetworkClockSufficientAccuracy(
+            sufficient_accuracy);
+        standard_network_system_clock_core.MarkAsInitialized();
     }
 
-    standard_user_system_clock_core.SetAutomaticCorrectionUpdatedTime(steady_clock_time_point);
-    standard_user_system_clock_core.MarkAsInitialized();
-    shared_memory.SetAutomaticCorrectionEnabled(is_automatic_correction_enabled);
+    void SetupStandardUserSystemClock(Core::System& system, bool is_automatic_correction_enabled,
+                                      Clock::SteadyClockTimePoint steady_clock_time_point) {
+        if (standard_user_system_clock_core.SetAutomaticCorrectionEnabled(
+                system, is_automatic_correction_enabled) != RESULT_SUCCESS) {
+            UNREACHABLE();
+            return;
+        }
+
+        standard_user_system_clock_core.SetAutomaticCorrectionUpdatedTime(steady_clock_time_point);
+        standard_user_system_clock_core.MarkAsInitialized();
+        shared_memory.SetAutomaticCorrectionEnabled(is_automatic_correction_enabled);
+    }
+
+    void SetupEphemeralNetworkSystemClock() {
+        ephemeral_network_system_clock_core.SetUpdateCallbackInstance(
+            ephemeral_network_system_clock_context_writer);
+        ephemeral_network_system_clock_core.MarkAsInitialized();
+    }
+
+    void UpdateLocalSystemClockTime(Core::System& system, s64 posix_time) {
+        const auto timespan{Service::Time::Clock::TimeSpanType::FromSeconds(posix_time)};
+        if (GetStandardLocalSystemClockCore()
+                .SetCurrentTime(system, timespan.ToSeconds())
+                .IsError()) {
+            UNREACHABLE();
+            return;
+        }
+    }
+
+    SharedMemory shared_memory;
+
+    Clock::StandardSteadyClockCore standard_steady_clock_core;
+    Clock::TickBasedSteadyClockCore tick_based_steady_clock_core;
+    Clock::StandardLocalSystemClockCore standard_local_system_clock_core;
+    Clock::StandardNetworkSystemClockCore standard_network_system_clock_core;
+    Clock::StandardUserSystemClockCore standard_user_system_clock_core;
+    Clock::EphemeralNetworkSystemClockCore ephemeral_network_system_clock_core;
+
+    std::shared_ptr<Clock::LocalSystemClockContextWriter> local_system_clock_context_writer;
+    std::shared_ptr<Clock::NetworkSystemClockContextWriter> network_system_clock_context_writer;
+    std::shared_ptr<Clock::EphemeralNetworkSystemClockContextWriter>
+        ephemeral_network_system_clock_context_writer;
+
+    TimeZone::TimeZoneContentManager time_zone_content_manager;
+};
+
+TimeManager::TimeManager(Core::System& system) : system{system} {}
+
+TimeManager::~TimeManager() = default;
+
+void TimeManager::Initialize() {
+    impl = std::make_unique<Impl>(system);
+
+    // Time zones can only be initialized after impl is valid
+    impl->time_zone_content_manager.Initialize(*this);
+}
+
+Clock::StandardSteadyClockCore& TimeManager::GetStandardSteadyClockCore() {
+    return impl->standard_steady_clock_core;
+}
+
+const Clock::StandardSteadyClockCore& TimeManager::GetStandardSteadyClockCore() const {
+    return impl->standard_steady_clock_core;
+}
+
+Clock::StandardLocalSystemClockCore& TimeManager::GetStandardLocalSystemClockCore() {
+    return impl->standard_local_system_clock_core;
+}
+
+const Clock::StandardLocalSystemClockCore& TimeManager::GetStandardLocalSystemClockCore() const {
+    return impl->standard_local_system_clock_core;
+}
+
+Clock::StandardNetworkSystemClockCore& TimeManager::GetStandardNetworkSystemClockCore() {
+    return impl->standard_network_system_clock_core;
 }
 
-void TimeManager::SetupEphemeralNetworkSystemClock() {
-    ephemeral_network_system_clock_core.SetUpdateCallbackInstance(
-        ephemeral_network_system_clock_context_writer);
-    ephemeral_network_system_clock_core.MarkAsInitialized();
+const Clock::StandardNetworkSystemClockCore& TimeManager::GetStandardNetworkSystemClockCore()
+    const {
+    return impl->standard_network_system_clock_core;
+}
+
+Clock::StandardUserSystemClockCore& TimeManager::GetStandardUserSystemClockCore() {
+    return impl->standard_user_system_clock_core;
+}
+
+const Clock::StandardUserSystemClockCore& TimeManager::GetStandardUserSystemClockCore() const {
+    return impl->standard_user_system_clock_core;
+}
+
+TimeZone::TimeZoneContentManager& TimeManager::GetTimeZoneContentManager() {
+    return impl->time_zone_content_manager;
+}
+
+const TimeZone::TimeZoneContentManager& TimeManager::GetTimeZoneContentManager() const {
+    return impl->time_zone_content_manager;
+}
+
+SharedMemory& TimeManager::GetSharedMemory() {
+    return impl->shared_memory;
+}
+
+const SharedMemory& TimeManager::GetSharedMemory() const {
+    return impl->shared_memory;
+}
+
+void TimeManager::UpdateLocalSystemClockTime(s64 posix_time) {
+    impl->UpdateLocalSystemClockTime(system, posix_time);
+}
+
+void TimeManager::SetupTimeZoneManager(std::string location_name,
+                                       Clock::SteadyClockTimePoint time_zone_updated_time_point,
+                                       std::size_t total_location_name_count,
+                                       u128 time_zone_rule_version,
+                                       FileSys::VirtualFile& vfs_file) {
+    impl->SetupTimeZoneManager(location_name, time_zone_updated_time_point,
+                               total_location_name_count, time_zone_rule_version, vfs_file);
+}
+
+/*static*/ s64 TimeManager::GetExternalTimeZoneOffset() {
+    // With "auto" timezone setting, we use the external system's timezone offset
+    if (Settings::GetTimeZoneString() == "auto") {
+        return Common::TimeZone::GetCurrentOffsetSeconds().count();
+    }
+    return 0;
 }
 
 } // namespace Service::Time
diff --git a/src/core/hle/service/time/time_manager.h b/src/core/hle/service/time/time_manager.h
index 8e65f0d22..993c7c288 100644
--- a/src/core/hle/service/time/time_manager.h
+++ b/src/core/hle/service/time/time_manager.h
@@ -5,6 +5,7 @@
 #pragma once
 
 #include "common/common_types.h"
+#include "common/time_zone.h"
 #include "core/file_sys/vfs_types.h"
 #include "core/hle/service/time/clock_types.h"
 #include "core/hle/service/time/ephemeral_network_system_clock_core.h"
@@ -32,86 +33,46 @@ public:
     explicit TimeManager(Core::System& system);
     ~TimeManager();
 
-    Clock::StandardSteadyClockCore& GetStandardSteadyClockCore() {
-        return standard_steady_clock_core;
-    }
+    void Initialize();
 
-    const Clock::StandardSteadyClockCore& GetStandardSteadyClockCore() const {
-        return standard_steady_clock_core;
-    }
+    Clock::StandardSteadyClockCore& GetStandardSteadyClockCore();
 
-    Clock::StandardLocalSystemClockCore& GetStandardLocalSystemClockCore() {
-        return standard_local_system_clock_core;
-    }
+    const Clock::StandardSteadyClockCore& GetStandardSteadyClockCore() const;
 
-    const Clock::StandardLocalSystemClockCore& GetStandardLocalSystemClockCore() const {
-        return standard_local_system_clock_core;
-    }
+    Clock::StandardLocalSystemClockCore& GetStandardLocalSystemClockCore();
 
-    Clock::StandardNetworkSystemClockCore& GetStandardNetworkSystemClockCore() {
-        return standard_network_system_clock_core;
-    }
+    const Clock::StandardLocalSystemClockCore& GetStandardLocalSystemClockCore() const;
 
-    const Clock::StandardNetworkSystemClockCore& GetStandardNetworkSystemClockCore() const {
-        return standard_network_system_clock_core;
-    }
+    Clock::StandardNetworkSystemClockCore& GetStandardNetworkSystemClockCore();
 
-    Clock::StandardUserSystemClockCore& GetStandardUserSystemClockCore() {
-        return standard_user_system_clock_core;
-    }
+    const Clock::StandardNetworkSystemClockCore& GetStandardNetworkSystemClockCore() const;
 
-    const Clock::StandardUserSystemClockCore& GetStandardUserSystemClockCore() const {
-        return standard_user_system_clock_core;
-    }
+    Clock::StandardUserSystemClockCore& GetStandardUserSystemClockCore();
 
-    TimeZone::TimeZoneContentManager& GetTimeZoneContentManager() {
-        return time_zone_content_manager;
-    }
+    const Clock::StandardUserSystemClockCore& GetStandardUserSystemClockCore() const;
 
-    const TimeZone::TimeZoneContentManager& GetTimeZoneContentManager() const {
-        return time_zone_content_manager;
-    }
+    TimeZone::TimeZoneContentManager& GetTimeZoneContentManager();
 
-    SharedMemory& GetSharedMemory() {
-        return shared_memory;
-    }
+    const TimeZone::TimeZoneContentManager& GetTimeZoneContentManager() const;
 
-    const SharedMemory& GetSharedMemory() const {
-        return shared_memory;
-    }
+    void UpdateLocalSystemClockTime(s64 posix_time);
+
+    SharedMemory& GetSharedMemory();
+
+    const SharedMemory& GetSharedMemory() const;
 
     void SetupTimeZoneManager(std::string location_name,
                               Clock::SteadyClockTimePoint time_zone_updated_time_point,
                               std::size_t total_location_name_count, u128 time_zone_rule_version,
                               FileSys::VirtualFile& vfs_file);
 
+    static s64 GetExternalTimeZoneOffset();
+
 private:
-    void SetupStandardSteadyClock(Core::System& system, Common::UUID clock_source_id,
-                                  Clock::TimeSpanType setup_value,
-                                  Clock::TimeSpanType internal_offset, bool is_rtc_reset_detected);
-    void SetupStandardLocalSystemClock(Core::System& system,
-                                       Clock::SystemClockContext clock_context, s64 posix_time);
-    void SetupStandardNetworkSystemClock(Clock::SystemClockContext clock_context,
-                                         Clock::TimeSpanType sufficient_accuracy);
-    void SetupStandardUserSystemClock(Core::System& system, bool is_automatic_correction_enabled,
-                                      Clock::SteadyClockTimePoint steady_clock_time_point);
-    void SetupEphemeralNetworkSystemClock();
-
-    SharedMemory shared_memory;
-
-    Clock::StandardSteadyClockCore standard_steady_clock_core;
-    Clock::TickBasedSteadyClockCore tick_based_steady_clock_core;
-    Clock::StandardLocalSystemClockCore standard_local_system_clock_core;
-    Clock::StandardNetworkSystemClockCore standard_network_system_clock_core;
-    Clock::StandardUserSystemClockCore standard_user_system_clock_core;
-    Clock::EphemeralNetworkSystemClockCore ephemeral_network_system_clock_core;
-
-    std::shared_ptr<Clock::LocalSystemClockContextWriter> local_system_clock_context_writer;
-    std::shared_ptr<Clock::NetworkSystemClockContextWriter> network_system_clock_context_writer;
-    std::shared_ptr<Clock::EphemeralNetworkSystemClockContextWriter>
-        ephemeral_network_system_clock_context_writer;
-
-    TimeZone::TimeZoneContentManager time_zone_content_manager;
+    Core::System& system;
+
+    struct Impl;
+    std::unique_ptr<Impl> impl;
 };
 
 } // namespace Service::Time
diff --git a/src/core/hle/service/time/time_zone_content_manager.cpp b/src/core/hle/service/time/time_zone_content_manager.cpp
index 320672add..4177d0a41 100644
--- a/src/core/hle/service/time/time_zone_content_manager.cpp
+++ b/src/core/hle/service/time/time_zone_content_manager.cpp
@@ -68,9 +68,10 @@ static std::vector<std::string> BuildLocationNameCache(Core::System& system) {
     return location_name_cache;
 }
 
-TimeZoneContentManager::TimeZoneContentManager(TimeManager& time_manager, Core::System& system)
-    : system{system}, location_name_cache{BuildLocationNameCache(system)} {
+TimeZoneContentManager::TimeZoneContentManager(Core::System& system)
+    : system{system}, location_name_cache{BuildLocationNameCache(system)} {}
 
+void TimeZoneContentManager::Initialize(TimeManager& time_manager) {
     std::string location_name;
     const auto timezone_setting = Settings::GetTimeZoneString();
     if (timezone_setting == "auto" || timezone_setting == "default") {
diff --git a/src/core/hle/service/time/time_zone_content_manager.h b/src/core/hle/service/time/time_zone_content_manager.h
index 4f302c3b9..52dd1a020 100644
--- a/src/core/hle/service/time/time_zone_content_manager.h
+++ b/src/core/hle/service/time/time_zone_content_manager.h
@@ -21,7 +21,9 @@ namespace Service::Time::TimeZone {
 
 class TimeZoneContentManager final {
 public:
-    TimeZoneContentManager(TimeManager& time_manager, Core::System& system);
+    explicit TimeZoneContentManager(Core::System& system);
+
+    void Initialize(TimeManager& time_manager);
 
     TimeZoneManager& GetTimeZoneManager() {
         return time_zone_manager;
diff --git a/src/core/settings.cpp b/src/core/settings.cpp
index 28d3f9099..e14c02045 100644
--- a/src/core/settings.cpp
+++ b/src/core/settings.cpp
@@ -63,6 +63,7 @@ void LogSettings() {
     log_setting("Renderer_GPUAccuracyLevel", values.gpu_accuracy.GetValue());
     log_setting("Renderer_UseAsynchronousGpuEmulation",
                 values.use_asynchronous_gpu_emulation.GetValue());
+    log_setting("Renderer_UseNvdecEmulation", values.use_nvdec_emulation.GetValue());
     log_setting("Renderer_UseVsync", values.use_vsync.GetValue());
     log_setting("Renderer_UseAssemblyShaders", values.use_assembly_shaders.GetValue());
     log_setting("Renderer_UseAsynchronousShaders", values.use_asynchronous_shaders.GetValue());
@@ -119,6 +120,7 @@ void RestoreGlobalState() {
     values.use_disk_shader_cache.SetGlobal(true);
     values.gpu_accuracy.SetGlobal(true);
     values.use_asynchronous_gpu_emulation.SetGlobal(true);
+    values.use_nvdec_emulation.SetGlobal(true);
     values.use_vsync.SetGlobal(true);
     values.use_assembly_shaders.SetGlobal(true);
     values.use_asynchronous_shaders.SetGlobal(true);
diff --git a/src/core/settings.h b/src/core/settings.h
index 9834f44bb..604805615 100644
--- a/src/core/settings.h
+++ b/src/core/settings.h
@@ -111,6 +111,7 @@ struct Values {
     Setting<bool> use_disk_shader_cache;
     Setting<GPUAccuracy> gpu_accuracy;
     Setting<bool> use_asynchronous_gpu_emulation;
+    Setting<bool> use_nvdec_emulation;
     Setting<bool> use_vsync;
     Setting<bool> use_assembly_shaders;
     Setting<bool> use_asynchronous_shaders;
diff --git a/src/core/telemetry_session.cpp b/src/core/telemetry_session.cpp
index da09c0dbc..ebc19e18a 100644
--- a/src/core/telemetry_session.cpp
+++ b/src/core/telemetry_session.cpp
@@ -206,6 +206,8 @@ void TelemetrySession::AddInitialInfo(Loader::AppLoader& app_loader) {
              TranslateGPUAccuracyLevel(Settings::values.gpu_accuracy.GetValue()));
     AddField(field_type, "Renderer_UseAsynchronousGpuEmulation",
              Settings::values.use_asynchronous_gpu_emulation.GetValue());
+    AddField(field_type, "Renderer_UseNvdecEmulation",
+             Settings::values.use_nvdec_emulation.GetValue());
     AddField(field_type, "Renderer_UseVsync", Settings::values.use_vsync.GetValue());
     AddField(field_type, "Renderer_UseAssemblyShaders",
              Settings::values.use_assembly_shaders.GetValue());
diff --git a/src/input_common/gcadapter/gc_adapter.cpp b/src/input_common/gcadapter/gc_adapter.cpp
index c95feb0d7..b912188b6 100644
--- a/src/input_common/gcadapter/gc_adapter.cpp
+++ b/src/input_common/gcadapter/gc_adapter.cpp
@@ -21,26 +21,6 @@
 
 namespace GCAdapter {
 
-// Used to loop through and assign button in poller
-constexpr std::array<PadButton, 12> PadButtonArray{
-    PadButton::PAD_BUTTON_LEFT, PadButton::PAD_BUTTON_RIGHT, PadButton::PAD_BUTTON_DOWN,
-    PadButton::PAD_BUTTON_UP,   PadButton::PAD_TRIGGER_Z,    PadButton::PAD_TRIGGER_R,
-    PadButton::PAD_TRIGGER_L,   PadButton::PAD_BUTTON_A,     PadButton::PAD_BUTTON_B,
-    PadButton::PAD_BUTTON_X,    PadButton::PAD_BUTTON_Y,     PadButton::PAD_BUTTON_START,
-};
-
-static void PadToState(const GCPadStatus& pad, GCState& out_state) {
-    for (const auto& button : PadButtonArray) {
-        const auto button_key = static_cast<u16>(button);
-        const auto button_value = (pad.button & button_key) != 0;
-        out_state.buttons.insert_or_assign(static_cast<s32>(button_key), button_value);
-    }
-
-    for (std::size_t i = 0; i < pad.axis_values.size(); ++i) {
-        out_state.axes.insert_or_assign(static_cast<u32>(i), pad.axis_values[i]);
-    }
-}
-
 Adapter::Adapter() {
     if (usb_adapter_handle != nullptr) {
         return;
@@ -49,168 +29,263 @@ Adapter::Adapter() {
 
     const int init_res = libusb_init(&libusb_ctx);
     if (init_res == LIBUSB_SUCCESS) {
-        Setup();
+        adapter_scan_thread = std::thread(&Adapter::AdapterScanThread, this);
     } else {
         LOG_ERROR(Input, "libusb could not be initialized. failed with error = {}", init_res);
     }
 }
 
-GCPadStatus Adapter::GetPadStatus(std::size_t port, const std::array<u8, 37>& adapter_payload) {
-    GCPadStatus pad = {};
-    const std::size_t offset = 1 + (9 * port);
+Adapter::~Adapter() {
+    Reset();
+}
+
+void Adapter::AdapterInputThread() {
+    LOG_DEBUG(Input, "GC Adapter input thread started");
+    s32 payload_size{};
+    AdapterPayload adapter_payload{};
+
+    if (adapter_scan_thread.joinable()) {
+        adapter_scan_thread.join();
+    }
+
+    while (adapter_input_thread_running) {
+        libusb_interrupt_transfer(usb_adapter_handle, input_endpoint, adapter_payload.data(),
+                                  static_cast<s32>(adapter_payload.size()), &payload_size, 16);
+        if (IsPayloadCorrect(adapter_payload, payload_size)) {
+            UpdateControllers(adapter_payload);
+            UpdateVibrations();
+        }
+        std::this_thread::yield();
+    }
 
-    adapter_controllers_status[port] = static_cast<ControllerTypes>(adapter_payload[offset] >> 4);
+    if (restart_scan_thread) {
+        adapter_scan_thread = std::thread(&Adapter::AdapterScanThread, this);
+        restart_scan_thread = false;
+    }
+}
+
+bool Adapter::IsPayloadCorrect(const AdapterPayload& adapter_payload, s32 payload_size) {
+    if (payload_size != static_cast<s32>(adapter_payload.size()) ||
+        adapter_payload[0] != LIBUSB_DT_HID) {
+        LOG_DEBUG(Input, "Error reading payload (size: {}, type: {:02x})", payload_size,
+                  adapter_payload[0]);
+        if (input_error_counter++ > 20) {
+            LOG_ERROR(Input, "GC adapter timeout, Is the adapter connected?");
+            adapter_input_thread_running = false;
+            restart_scan_thread = true;
+        }
+        return false;
+    }
+
+    input_error_counter = 0;
+    return true;
+}
+
+void Adapter::UpdateControllers(const AdapterPayload& adapter_payload) {
+    for (std::size_t port = 0; port < pads.size(); ++port) {
+        const std::size_t offset = 1 + (9 * port);
+        const auto type = static_cast<ControllerTypes>(adapter_payload[offset] >> 4);
+        UpdatePadType(port, type);
+        if (DeviceConnected(port)) {
+            const u8 b1 = adapter_payload[offset + 1];
+            const u8 b2 = adapter_payload[offset + 2];
+            UpdateStateButtons(port, b1, b2);
+            UpdateStateAxes(port, adapter_payload);
+            if (configuring) {
+                UpdateYuzuSettings(port);
+            }
+        }
+    }
+}
+
+void Adapter::UpdatePadType(std::size_t port, ControllerTypes pad_type) {
+    if (pads[port].type == pad_type) {
+        return;
+    }
+    // Device changed reset device and set new type
+    ResetDevice(port);
+    pads[port].type = pad_type;
+}
+
+void Adapter::UpdateStateButtons(std::size_t port, u8 b1, u8 b2) {
+    if (port >= pads.size()) {
+        return;
+    }
 
     static constexpr std::array<PadButton, 8> b1_buttons{
-        PadButton::PAD_BUTTON_A,    PadButton::PAD_BUTTON_B,    PadButton::PAD_BUTTON_X,
-        PadButton::PAD_BUTTON_Y,    PadButton::PAD_BUTTON_LEFT, PadButton::PAD_BUTTON_RIGHT,
-        PadButton::PAD_BUTTON_DOWN, PadButton::PAD_BUTTON_UP,
+        PadButton::ButtonA,    PadButton::ButtonB,     PadButton::ButtonX,    PadButton::ButtonY,
+        PadButton::ButtonLeft, PadButton::ButtonRight, PadButton::ButtonDown, PadButton::ButtonUp,
     };
 
     static constexpr std::array<PadButton, 4> b2_buttons{
-        PadButton::PAD_BUTTON_START,
-        PadButton::PAD_TRIGGER_Z,
-        PadButton::PAD_TRIGGER_R,
-        PadButton::PAD_TRIGGER_L,
+        PadButton::ButtonStart,
+        PadButton::TriggerZ,
+        PadButton::TriggerR,
+        PadButton::TriggerL,
     };
+    pads[port].buttons = 0;
+    for (std::size_t i = 0; i < b1_buttons.size(); ++i) {
+        if ((b1 & (1U << i)) != 0) {
+            pads[port].buttons =
+                static_cast<u16>(pads[port].buttons | static_cast<u16>(b1_buttons[i]));
+            pads[port].last_button = b1_buttons[i];
+        }
+    }
 
+    for (std::size_t j = 0; j < b2_buttons.size(); ++j) {
+        if ((b2 & (1U << j)) != 0) {
+            pads[port].buttons =
+                static_cast<u16>(pads[port].buttons | static_cast<u16>(b2_buttons[j]));
+            pads[port].last_button = b2_buttons[j];
+        }
+    }
+}
+
+void Adapter::UpdateStateAxes(std::size_t port, const AdapterPayload& adapter_payload) {
+    if (port >= pads.size()) {
+        return;
+    }
+
+    const std::size_t offset = 1 + (9 * port);
     static constexpr std::array<PadAxes, 6> axes{
         PadAxes::StickX,    PadAxes::StickY,      PadAxes::SubstickX,
         PadAxes::SubstickY, PadAxes::TriggerLeft, PadAxes::TriggerRight,
     };
 
-    if (adapter_controllers_status[port] == ControllerTypes::None && !get_origin[port]) {
-        // Controller may have been disconnected, recalibrate if reconnected.
-        get_origin[port] = true;
+    for (const PadAxes axis : axes) {
+        const auto index = static_cast<std::size_t>(axis);
+        const u8 axis_value = adapter_payload[offset + 3 + index];
+        if (pads[port].axis_origin[index] == 255) {
+            pads[port].axis_origin[index] = axis_value;
+        }
+        pads[port].axis_values[index] =
+            static_cast<s16>(axis_value - pads[port].axis_origin[index]);
     }
+}
 
-    if (adapter_controllers_status[port] != ControllerTypes::None) {
-        const u8 b1 = adapter_payload[offset + 1];
-        const u8 b2 = adapter_payload[offset + 2];
+void Adapter::UpdateYuzuSettings(std::size_t port) {
+    if (port >= pads.size()) {
+        return;
+    }
 
-        for (std::size_t i = 0; i < b1_buttons.size(); ++i) {
-            if ((b1 & (1U << i)) != 0) {
-                pad.button = static_cast<u16>(pad.button | static_cast<u16>(b1_buttons[i]));
-            }
-        }
+    constexpr u8 axis_threshold = 50;
+    GCPadStatus pad_status = {.port = port};
 
-        for (std::size_t j = 0; j < b2_buttons.size(); ++j) {
-            if ((b2 & (1U << j)) != 0) {
-                pad.button = static_cast<u16>(pad.button | static_cast<u16>(b2_buttons[j]));
-            }
-        }
-        for (PadAxes axis : axes) {
-            const auto index = static_cast<std::size_t>(axis);
-            pad.axis_values[index] = adapter_payload[offset + 3 + index];
-        }
+    if (pads[port].buttons != 0) {
+        pad_status.button = pads[port].last_button;
+        pad_queue.Push(pad_status);
+    }
+
+    // Accounting for a threshold here to ensure an intentional press
+    for (std::size_t i = 0; i < pads[port].axis_values.size(); ++i) {
+        const s16 value = pads[port].axis_values[i];
 
-        if (get_origin[port]) {
-            origin_status[port].axis_values = pad.axis_values;
-            get_origin[port] = false;
+        if (value > axis_threshold || value < -axis_threshold) {
+            pad_status.axis = static_cast<PadAxes>(i);
+            pad_status.axis_value = value;
+            pad_status.axis_threshold = axis_threshold;
+            pad_queue.Push(pad_status);
         }
     }
-    return pad;
 }
 
-void Adapter::Read() {
-    LOG_DEBUG(Input, "GC Adapter Read() thread started");
+void Adapter::UpdateVibrations() {
+    // Use 8 states to keep the switching between on/off fast enough for
+    // a human to not notice the difference between switching from on/off
+    // More states = more rumble strengths = slower update time
+    constexpr u8 vibration_states = 8;
 
-    int payload_size;
-    std::array<u8, 37> adapter_payload;
-    std::array<GCPadStatus, 4> pads;
-
-    while (adapter_thread_running) {
-        libusb_interrupt_transfer(usb_adapter_handle, input_endpoint, adapter_payload.data(),
-                                  sizeof(adapter_payload), &payload_size, 16);
-
-        if (payload_size != sizeof(adapter_payload) || adapter_payload[0] != LIBUSB_DT_HID) {
-            LOG_ERROR(Input,
-                      "Error reading payload (size: {}, type: {:02x}) Is the adapter connected?",
-                      payload_size, adapter_payload[0]);
-            adapter_thread_running = false; // error reading from adapter, stop reading.
-            break;
-        }
-        for (std::size_t port = 0; port < pads.size(); ++port) {
-            pads[port] = GetPadStatus(port, adapter_payload);
-            if (DeviceConnected(port) && configuring) {
-                if (pads[port].button != 0) {
-                    pad_queue[port].Push(pads[port]);
-                }
+    vibration_counter = (vibration_counter + 1) % vibration_states;
 
-                // Accounting for a threshold here to ensure an intentional press
-                for (size_t i = 0; i < pads[port].axis_values.size(); ++i) {
-                    const u8 value = pads[port].axis_values[i];
-                    const u8 origin = origin_status[port].axis_values[i];
-
-                    if (value > origin + pads[port].THRESHOLD ||
-                        value < origin - pads[port].THRESHOLD) {
-                        pads[port].axis = static_cast<PadAxes>(i);
-                        pads[port].axis_value = pads[port].axis_values[i];
-                        pad_queue[port].Push(pads[port]);
-                    }
-                }
-            }
-            PadToState(pads[port], state[port]);
-        }
-        std::this_thread::yield();
+    for (GCController& pad : pads) {
+        const bool vibrate = pad.rumble_amplitude > vibration_counter;
+        vibration_changed |= vibrate != pad.enable_vibration;
+        pad.enable_vibration = vibrate;
     }
+    SendVibrations();
 }
 
-void Adapter::Setup() {
-    // Initialize all controllers as unplugged
-    adapter_controllers_status.fill(ControllerTypes::None);
-    // Initialize all ports to store axis origin values
-    get_origin.fill(true);
-
-    // pointer to list of connected usb devices
-    libusb_device** devices{};
-
-    // populate the list of devices, get the count
-    const ssize_t device_count = libusb_get_device_list(libusb_ctx, &devices);
-    if (device_count < 0) {
-        LOG_ERROR(Input, "libusb_get_device_list failed with error: {}", device_count);
+void Adapter::SendVibrations() {
+    if (!rumble_enabled || !vibration_changed) {
         return;
     }
-
-    if (devices != nullptr) {
-        for (std::size_t index = 0; index < static_cast<std::size_t>(device_count); ++index) {
-            if (CheckDeviceAccess(devices[index])) {
-                // GC Adapter found and accessible, registering it
-                GetGCEndpoint(devices[index]);
-                break;
-            }
+    s32 size{};
+    constexpr u8 rumble_command = 0x11;
+    const u8 p1 = pads[0].enable_vibration;
+    const u8 p2 = pads[1].enable_vibration;
+    const u8 p3 = pads[2].enable_vibration;
+    const u8 p4 = pads[3].enable_vibration;
+    std::array<u8, 5> payload = {rumble_command, p1, p2, p3, p4};
+    const int err = libusb_interrupt_transfer(usb_adapter_handle, output_endpoint, payload.data(),
+                                              static_cast<s32>(payload.size()), &size, 16);
+    if (err) {
+        LOG_DEBUG(Input, "Adapter libusb write failed: {}", libusb_error_name(err));
+        if (output_error_counter++ > 5) {
+            LOG_ERROR(Input, "GC adapter output timeout, Rumble disabled");
+            rumble_enabled = false;
         }
-        libusb_free_device_list(devices, 1);
+        return;
     }
+    output_error_counter = 0;
+    vibration_changed = false;
 }
 
-bool Adapter::CheckDeviceAccess(libusb_device* device) {
-    libusb_device_descriptor desc;
-    const int get_descriptor_error = libusb_get_device_descriptor(device, &desc);
-    if (get_descriptor_error) {
-        // could not acquire the descriptor, no point in trying to use it.
-        LOG_ERROR(Input, "libusb_get_device_descriptor failed with error: {}",
-                  get_descriptor_error);
-        return false;
+bool Adapter::RumblePlay(std::size_t port, f32 amplitude) {
+    amplitude = std::clamp(amplitude, 0.0f, 1.0f);
+    const auto raw_amp = static_cast<u8>(amplitude * 0x8);
+    pads[port].rumble_amplitude = raw_amp;
+
+    return rumble_enabled;
+}
+
+void Adapter::AdapterScanThread() {
+    adapter_scan_thread_running = true;
+    adapter_input_thread_running = false;
+    if (adapter_input_thread.joinable()) {
+        adapter_input_thread.join();
+    }
+    ClearLibusbHandle();
+    ResetDevices();
+    while (adapter_scan_thread_running && !adapter_input_thread_running) {
+        Setup();
+        std::this_thread::sleep_for(std::chrono::seconds(1));
     }
+}
 
-    if (desc.idVendor != 0x057e || desc.idProduct != 0x0337) {
-        // This isn't the device we are looking for.
-        return false;
+void Adapter::Setup() {
+    usb_adapter_handle = libusb_open_device_with_vid_pid(libusb_ctx, 0x057e, 0x0337);
+
+    if (usb_adapter_handle == NULL) {
+        return;
+    }
+    if (!CheckDeviceAccess()) {
+        ClearLibusbHandle();
+        return;
     }
-    const int open_error = libusb_open(device, &usb_adapter_handle);
 
-    if (open_error == LIBUSB_ERROR_ACCESS) {
-        LOG_ERROR(Input, "Yuzu can not gain access to this device: ID {:04X}:{:04X}.",
-                  desc.idVendor, desc.idProduct);
-        return false;
+    libusb_device* device = libusb_get_device(usb_adapter_handle);
+
+    LOG_INFO(Input, "GC adapter is now connected");
+    // GC Adapter found and accessible, registering it
+    if (GetGCEndpoint(device)) {
+        adapter_scan_thread_running = false;
+        adapter_input_thread_running = true;
+        rumble_enabled = true;
+        input_error_counter = 0;
+        output_error_counter = 0;
+        adapter_input_thread = std::thread(&Adapter::AdapterInputThread, this);
     }
-    if (open_error) {
-        LOG_ERROR(Input, "libusb_open failed to open device with error = {}", open_error);
-        return false;
+}
+
+bool Adapter::CheckDeviceAccess() {
+    // This fixes payload problems from offbrand GCAdapters
+    const s32 control_transfer_error =
+        libusb_control_transfer(usb_adapter_handle, 0x21, 11, 0x0001, 0, nullptr, 0, 1000);
+    if (control_transfer_error < 0) {
+        LOG_ERROR(Input, "libusb_control_transfer failed with error= {}", control_transfer_error);
     }
 
-    int kernel_driver_error = libusb_kernel_driver_active(usb_adapter_handle, 0);
+    s32 kernel_driver_error = libusb_kernel_driver_active(usb_adapter_handle, 0);
     if (kernel_driver_error == 1) {
         kernel_driver_error = libusb_detach_kernel_driver(usb_adapter_handle, 0);
         if (kernel_driver_error != 0 && kernel_driver_error != LIBUSB_ERROR_NOT_SUPPORTED) {
@@ -236,13 +311,13 @@ bool Adapter::CheckDeviceAccess(libusb_device* device) {
     return true;
 }
 
-void Adapter::GetGCEndpoint(libusb_device* device) {
+bool Adapter::GetGCEndpoint(libusb_device* device) {
     libusb_config_descriptor* config = nullptr;
     const int config_descriptor_return = libusb_get_config_descriptor(device, 0, &config);
     if (config_descriptor_return != LIBUSB_SUCCESS) {
         LOG_ERROR(Input, "libusb_get_config_descriptor failed with error = {}",
                   config_descriptor_return);
-        return;
+        return false;
     }
 
     for (u8 ic = 0; ic < config->bNumInterfaces; ic++) {
@@ -264,31 +339,51 @@ void Adapter::GetGCEndpoint(libusb_device* device) {
     unsigned char clear_payload = 0x13;
     libusb_interrupt_transfer(usb_adapter_handle, output_endpoint, &clear_payload,
                               sizeof(clear_payload), nullptr, 16);
-
-    adapter_thread_running = true;
-    adapter_input_thread = std::thread(&Adapter::Read, this);
+    return true;
 }
 
-Adapter::~Adapter() {
-    Reset();
-}
+void Adapter::JoinThreads() {
+    restart_scan_thread = false;
+    adapter_input_thread_running = false;
+    adapter_scan_thread_running = false;
 
-void Adapter::Reset() {
-    if (adapter_thread_running) {
-        adapter_thread_running = false;
+    if (adapter_scan_thread.joinable()) {
+        adapter_scan_thread.join();
     }
+
     if (adapter_input_thread.joinable()) {
         adapter_input_thread.join();
     }
+}
 
-    adapter_controllers_status.fill(ControllerTypes::None);
-    get_origin.fill(true);
-
+void Adapter::ClearLibusbHandle() {
     if (usb_adapter_handle) {
         libusb_release_interface(usb_adapter_handle, 1);
         libusb_close(usb_adapter_handle);
         usb_adapter_handle = nullptr;
     }
+}
+
+void Adapter::ResetDevices() {
+    for (std::size_t i = 0; i < pads.size(); ++i) {
+        ResetDevice(i);
+    }
+}
+
+void Adapter::ResetDevice(std::size_t port) {
+    pads[port].type = ControllerTypes::None;
+    pads[port].enable_vibration = false;
+    pads[port].rumble_amplitude = 0;
+    pads[port].buttons = 0;
+    pads[port].last_button = PadButton::Undefined;
+    pads[port].axis_values.fill(0);
+    pads[port].axis_origin.fill(255);
+}
+
+void Adapter::Reset() {
+    JoinThreads();
+    ClearLibusbHandle();
+    ResetDevices();
 
     if (libusb_ctx) {
         libusb_exit(libusb_ctx);
@@ -297,11 +392,11 @@ void Adapter::Reset() {
 
 std::vector<Common::ParamPackage> Adapter::GetInputDevices() const {
     std::vector<Common::ParamPackage> devices;
-    for (std::size_t port = 0; port < state.size(); ++port) {
+    for (std::size_t port = 0; port < pads.size(); ++port) {
         if (!DeviceConnected(port)) {
             continue;
         }
-        std::string name = fmt::format("Gamecube Controller {}", port);
+        std::string name = fmt::format("Gamecube Controller {}", port + 1);
         devices.emplace_back(Common::ParamPackage{
             {"class", "gcpad"},
             {"display", std::move(name)},
@@ -318,18 +413,18 @@ InputCommon::ButtonMapping Adapter::GetButtonMappingForDevice(
     // This list also excludes any button that can't be really mapped
     static constexpr std::array<std::pair<Settings::NativeButton::Values, PadButton>, 12>
         switch_to_gcadapter_button = {
-            std::pair{Settings::NativeButton::A, PadButton::PAD_BUTTON_A},
-            {Settings::NativeButton::B, PadButton::PAD_BUTTON_B},
-            {Settings::NativeButton::X, PadButton::PAD_BUTTON_X},
-            {Settings::NativeButton::Y, PadButton::PAD_BUTTON_Y},
-            {Settings::NativeButton::Plus, PadButton::PAD_BUTTON_START},
-            {Settings::NativeButton::DLeft, PadButton::PAD_BUTTON_LEFT},
-            {Settings::NativeButton::DUp, PadButton::PAD_BUTTON_UP},
-            {Settings::NativeButton::DRight, PadButton::PAD_BUTTON_RIGHT},
-            {Settings::NativeButton::DDown, PadButton::PAD_BUTTON_DOWN},
-            {Settings::NativeButton::SL, PadButton::PAD_TRIGGER_L},
-            {Settings::NativeButton::SR, PadButton::PAD_TRIGGER_R},
-            {Settings::NativeButton::R, PadButton::PAD_TRIGGER_Z},
+            std::pair{Settings::NativeButton::A, PadButton::ButtonA},
+            {Settings::NativeButton::B, PadButton::ButtonB},
+            {Settings::NativeButton::X, PadButton::ButtonX},
+            {Settings::NativeButton::Y, PadButton::ButtonY},
+            {Settings::NativeButton::Plus, PadButton::ButtonStart},
+            {Settings::NativeButton::DLeft, PadButton::ButtonLeft},
+            {Settings::NativeButton::DUp, PadButton::ButtonUp},
+            {Settings::NativeButton::DRight, PadButton::ButtonRight},
+            {Settings::NativeButton::DDown, PadButton::ButtonDown},
+            {Settings::NativeButton::SL, PadButton::TriggerL},
+            {Settings::NativeButton::SR, PadButton::TriggerR},
+            {Settings::NativeButton::R, PadButton::TriggerZ},
         };
     if (!params.Has("port")) {
         return {};
@@ -352,8 +447,10 @@ InputCommon::ButtonMapping Adapter::GetButtonMappingForDevice(
     for (const auto& [switch_button, gcadapter_axis] : switch_to_gcadapter_axis) {
         Common::ParamPackage button_params({{"engine", "gcpad"}});
         button_params.Set("port", params.Get("port", 0));
-        button_params.Set("button", static_cast<int>(PadButton::PAD_STICK));
-        button_params.Set("axis", static_cast<int>(gcadapter_axis));
+        button_params.Set("button", static_cast<s32>(PadButton::Stick));
+        button_params.Set("axis", static_cast<s32>(gcadapter_axis));
+        button_params.Set("threshold", 0.5f);
+        button_params.Set("direction", "+");
         mapping.insert_or_assign(switch_button, std::move(button_params));
     }
     return mapping;
@@ -382,46 +479,33 @@ InputCommon::AnalogMapping Adapter::GetAnalogMappingForDevice(
 }
 
 bool Adapter::DeviceConnected(std::size_t port) const {
-    return adapter_controllers_status[port] != ControllerTypes::None;
-}
-
-void Adapter::ResetDeviceType(std::size_t port) {
-    adapter_controllers_status[port] = ControllerTypes::None;
+    return pads[port].type != ControllerTypes::None;
 }
 
 void Adapter::BeginConfiguration() {
-    get_origin.fill(true);
-    for (auto& pq : pad_queue) {
-        pq.Clear();
-    }
+    pad_queue.Clear();
     configuring = true;
 }
 
 void Adapter::EndConfiguration() {
-    for (auto& pq : pad_queue) {
-        pq.Clear();
-    }
+    pad_queue.Clear();
     configuring = false;
 }
 
-std::array<Common::SPSCQueue<GCPadStatus>, 4>& Adapter::GetPadQueue() {
+Common::SPSCQueue<GCPadStatus>& Adapter::GetPadQueue() {
     return pad_queue;
 }
 
-const std::array<Common::SPSCQueue<GCPadStatus>, 4>& Adapter::GetPadQueue() const {
+const Common::SPSCQueue<GCPadStatus>& Adapter::GetPadQueue() const {
     return pad_queue;
 }
 
-std::array<GCState, 4>& Adapter::GetPadState() {
-    return state;
-}
-
-const std::array<GCState, 4>& Adapter::GetPadState() const {
-    return state;
+GCController& Adapter::GetPadState(std::size_t port) {
+    return pads.at(port);
 }
 
-int Adapter::GetOriginValue(u32 port, u32 axis) const {
-    return origin_status[port].axis_values[axis];
+const GCController& Adapter::GetPadState(std::size_t port) const {
+    return pads.at(port);
 }
 
 } // namespace GCAdapter
diff --git a/src/input_common/gcadapter/gc_adapter.h b/src/input_common/gcadapter/gc_adapter.h
index 4f5f3de8e..d28dcfad3 100644
--- a/src/input_common/gcadapter/gc_adapter.h
+++ b/src/input_common/gcadapter/gc_adapter.h
@@ -19,24 +19,23 @@ struct libusb_device_handle;
 namespace GCAdapter {
 
 enum class PadButton {
-    PAD_BUTTON_LEFT = 0x0001,
-    PAD_BUTTON_RIGHT = 0x0002,
-    PAD_BUTTON_DOWN = 0x0004,
-    PAD_BUTTON_UP = 0x0008,
-    PAD_TRIGGER_Z = 0x0010,
-    PAD_TRIGGER_R = 0x0020,
-    PAD_TRIGGER_L = 0x0040,
-    PAD_BUTTON_A = 0x0100,
-    PAD_BUTTON_B = 0x0200,
-    PAD_BUTTON_X = 0x0400,
-    PAD_BUTTON_Y = 0x0800,
-    PAD_BUTTON_START = 0x1000,
+    Undefined = 0x0000,
+    ButtonLeft = 0x0001,
+    ButtonRight = 0x0002,
+    ButtonDown = 0x0004,
+    ButtonUp = 0x0008,
+    TriggerZ = 0x0010,
+    TriggerR = 0x0020,
+    TriggerL = 0x0040,
+    ButtonA = 0x0100,
+    ButtonB = 0x0200,
+    ButtonX = 0x0400,
+    ButtonY = 0x0800,
+    ButtonStart = 0x1000,
     // Below is for compatibility with "AxisButton" type
-    PAD_STICK = 0x2000,
+    Stick = 0x2000,
 };
 
-extern const std::array<PadButton, 12> PadButtonArray;
-
 enum class PadAxes : u8 {
     StickX,
     StickY,
@@ -47,87 +46,122 @@ enum class PadAxes : u8 {
     Undefined,
 };
 
+enum class ControllerTypes {
+    None,
+    Wired,
+    Wireless,
+};
+
 struct GCPadStatus {
-    u16 button{}; // Or-ed PAD_BUTTON_* and PAD_TRIGGER_* bits
+    std::size_t port{};
 
-    std::array<u8, 6> axis_values{};    // Triggers and sticks, following indices defined in PadAxes
-    static constexpr u8 THRESHOLD = 50; // Threshold for axis press for polling
+    PadButton button{PadButton::Undefined}; // Or-ed PAD_BUTTON_* and PAD_TRIGGER_* bits
 
-    u8 port{};
     PadAxes axis{PadAxes::Undefined};
-    u8 axis_value{255};
+    s16 axis_value{};
+    u8 axis_threshold{50};
 };
 
-struct GCState {
-    std::unordered_map<int, bool> buttons;
-    std::unordered_map<u32, u16> axes;
+struct GCController {
+    ControllerTypes type{};
+    bool enable_vibration{};
+    u8 rumble_amplitude{};
+    u16 buttons{};
+    PadButton last_button{};
+    std::array<s16, 6> axis_values{};
+    std::array<u8, 6> axis_origin{};
 };
 
-enum class ControllerTypes { None, Wired, Wireless };
-
 class Adapter {
 public:
-    /// Initialize the GC Adapter capture and read sequence
     Adapter();
-
-    /// Close the adapter read thread and release the adapter
     ~Adapter();
+
+    /// Request a vibration for a controlelr
+    bool RumblePlay(std::size_t port, f32 amplitude);
+
     /// Used for polling
     void BeginConfiguration();
     void EndConfiguration();
 
+    Common::SPSCQueue<GCPadStatus>& GetPadQueue();
+    const Common::SPSCQueue<GCPadStatus>& GetPadQueue() const;
+
+    GCController& GetPadState(std::size_t port);
+    const GCController& GetPadState(std::size_t port) const;
+
+    /// Returns true if there is a device connected to port
+    bool DeviceConnected(std::size_t port) const;
+
+    /// Used for automapping features
     std::vector<Common::ParamPackage> GetInputDevices() const;
     InputCommon::ButtonMapping GetButtonMappingForDevice(const Common::ParamPackage& params) const;
     InputCommon::AnalogMapping GetAnalogMappingForDevice(const Common::ParamPackage& params) const;
 
-    /// Returns true if there is a device connected to port
-    bool DeviceConnected(std::size_t port) const;
+private:
+    using AdapterPayload = std::array<u8, 37>;
 
-    std::array<Common::SPSCQueue<GCPadStatus>, 4>& GetPadQueue();
-    const std::array<Common::SPSCQueue<GCPadStatus>, 4>& GetPadQueue() const;
+    void UpdatePadType(std::size_t port, ControllerTypes pad_type);
+    void UpdateControllers(const AdapterPayload& adapter_payload);
+    void UpdateYuzuSettings(std::size_t port);
+    void UpdateStateButtons(std::size_t port, u8 b1, u8 b2);
+    void UpdateStateAxes(std::size_t port, const AdapterPayload& adapter_payload);
+    void UpdateVibrations();
 
-    std::array<GCState, 4>& GetPadState();
-    const std::array<GCState, 4>& GetPadState() const;
+    void AdapterInputThread();
 
-    int GetOriginValue(u32 port, u32 axis) const;
+    void AdapterScanThread();
 
-private:
-    GCPadStatus GetPadStatus(std::size_t port, const std::array<u8, 37>& adapter_payload);
+    bool IsPayloadCorrect(const AdapterPayload& adapter_payload, s32 payload_size);
+
+    // Updates vibration state of all controllers
+    void SendVibrations();
+
+    /// For use in initialization, querying devices to find the adapter
+    void Setup();
 
-    void Read();
+    /// Resets status of all GC controller devices to a disconected state
+    void ResetDevices();
 
-    /// Resets status of device connected to port
-    void ResetDeviceType(std::size_t port);
+    /// Resets status of device connected to a disconected state
+    void ResetDevice(std::size_t port);
 
     /// Returns true if we successfully gain access to GC Adapter
-    bool CheckDeviceAccess(libusb_device* device);
+    bool CheckDeviceAccess();
 
-    /// Captures GC Adapter endpoint address,
-    void GetGCEndpoint(libusb_device* device);
+    /// Captures GC Adapter endpoint address
+    /// Returns true if the endpoind was set correctly
+    bool GetGCEndpoint(libusb_device* device);
 
     /// For shutting down, clear all data, join all threads, release usb
     void Reset();
 
-    /// For use in initialization, querying devices to find the adapter
-    void Setup();
+    // Join all threads
+    void JoinThreads();
+
+    // Release usb handles
+    void ClearLibusbHandle();
 
     libusb_device_handle* usb_adapter_handle = nullptr;
+    std::array<GCController, 4> pads;
+    Common::SPSCQueue<GCPadStatus> pad_queue;
 
     std::thread adapter_input_thread;
-    bool adapter_thread_running;
+    std::thread adapter_scan_thread;
+    bool adapter_input_thread_running;
+    bool adapter_scan_thread_running;
+    bool restart_scan_thread;
 
     libusb_context* libusb_ctx;
 
-    u8 input_endpoint = 0;
-    u8 output_endpoint = 0;
-
-    bool configuring = false;
+    u8 input_endpoint{0};
+    u8 output_endpoint{0};
+    u8 input_error_counter{0};
+    u8 output_error_counter{0};
+    int vibration_counter{0};
 
-    std::array<GCState, 4> state;
-    std::array<bool, 4> get_origin;
-    std::array<GCPadStatus, 4> origin_status;
-    std::array<Common::SPSCQueue<GCPadStatus>, 4> pad_queue;
-    std::array<ControllerTypes, 4> adapter_controllers_status{};
+    bool configuring{false};
+    bool rumble_enabled{true};
+    bool vibration_changed{true};
 };
-
 } // namespace GCAdapter
diff --git a/src/input_common/gcadapter/gc_poller.cpp b/src/input_common/gcadapter/gc_poller.cpp
index 893556916..6bd6f57fc 100644
--- a/src/input_common/gcadapter/gc_poller.cpp
+++ b/src/input_common/gcadapter/gc_poller.cpp
@@ -15,22 +15,30 @@ namespace InputCommon {
 
 class GCButton final : public Input::ButtonDevice {
 public:
-    explicit GCButton(u32 port_, int button_, const GCAdapter::Adapter* adapter)
+    explicit GCButton(u32 port_, s32 button_, GCAdapter::Adapter* adapter)
         : port(port_), button(button_), gcadapter(adapter) {}
 
     ~GCButton() override;
 
     bool GetStatus() const override {
         if (gcadapter->DeviceConnected(port)) {
-            return gcadapter->GetPadState()[port].buttons.at(button);
+            return (gcadapter->GetPadState(port).buttons & button) != 0;
         }
         return false;
     }
 
+    bool SetRumblePlay(f32 amp_high, f32 amp_low, f32 freq_high, f32 freq_low) const override {
+        const float amplitude = amp_high + amp_low > 2.0f ? 1.0f : (amp_high + amp_low) * 0.5f;
+        const auto new_amp =
+            static_cast<f32>(pow(amplitude, 0.5f) * (3.0f - 2.0f * pow(amplitude, 0.15f)));
+
+        return gcadapter->RumblePlay(port, new_amp);
+    }
+
 private:
     const u32 port;
-    const int button;
-    const GCAdapter::Adapter* gcadapter;
+    const s32 button;
+    GCAdapter::Adapter* gcadapter;
 };
 
 class GCAxisButton final : public Input::ButtonDevice {
@@ -38,13 +46,12 @@ public:
     explicit GCAxisButton(u32 port_, u32 axis_, float threshold_, bool trigger_if_greater_,
                           const GCAdapter::Adapter* adapter)
         : port(port_), axis(axis_), threshold(threshold_), trigger_if_greater(trigger_if_greater_),
-          gcadapter(adapter),
-          origin_value(static_cast<float>(adapter->GetOriginValue(port_, axis_))) {}
+          gcadapter(adapter) {}
 
     bool GetStatus() const override {
         if (gcadapter->DeviceConnected(port)) {
-            const float current_axis_value = gcadapter->GetPadState()[port].axes.at(axis);
-            const float axis_value = (current_axis_value - origin_value) / 128.0f;
+            const float current_axis_value = gcadapter->GetPadState(port).axis_values.at(axis);
+            const float axis_value = current_axis_value / 128.0f;
             if (trigger_if_greater) {
                 // TODO: Might be worthwile to set a slider for the trigger threshold. It is
                 // currently always set to 0.5 in configure_input_player.cpp ZL/ZR HandleClick
@@ -61,7 +68,6 @@ private:
     float threshold;
     bool trigger_if_greater;
     const GCAdapter::Adapter* gcadapter;
-    const float origin_value;
 };
 
 GCButtonFactory::GCButtonFactory(std::shared_ptr<GCAdapter::Adapter> adapter_)
@@ -73,7 +79,7 @@ std::unique_ptr<Input::ButtonDevice> GCButtonFactory::Create(const Common::Param
     const auto button_id = params.Get("button", 0);
     const auto port = static_cast<u32>(params.Get("port", 0));
 
-    constexpr int PAD_STICK_ID = static_cast<u16>(GCAdapter::PadButton::PAD_STICK);
+    constexpr s32 PAD_STICK_ID = static_cast<s32>(GCAdapter::PadButton::Stick);
 
     // button is not an axis/stick button
     if (button_id != PAD_STICK_ID) {
@@ -106,32 +112,25 @@ Common::ParamPackage GCButtonFactory::GetNextInput() const {
     Common::ParamPackage params;
     GCAdapter::GCPadStatus pad;
     auto& queue = adapter->GetPadQueue();
-    for (std::size_t port = 0; port < queue.size(); ++port) {
-        while (queue[port].Pop(pad)) {
-            // This while loop will break on the earliest detected button
-            params.Set("engine", "gcpad");
-            params.Set("port", static_cast<int>(port));
-            for (const auto& button : GCAdapter::PadButtonArray) {
-                const u16 button_value = static_cast<u16>(button);
-                if (pad.button & button_value) {
-                    params.Set("button", button_value);
-                    break;
-                }
-            }
+    while (queue.Pop(pad)) {
+        // This while loop will break on the earliest detected button
+        params.Set("engine", "gcpad");
+        params.Set("port", static_cast<s32>(pad.port));
+        if (pad.button != GCAdapter::PadButton::Undefined) {
+            params.Set("button", static_cast<u16>(pad.button));
+        }
 
-            // For Axis button implementation
-            if (pad.axis != GCAdapter::PadAxes::Undefined) {
-                params.Set("axis", static_cast<u8>(pad.axis));
-                params.Set("button", static_cast<u16>(GCAdapter::PadButton::PAD_STICK));
-                if (pad.axis_value > 128) {
-                    params.Set("direction", "+");
-                    params.Set("threshold", "0.25");
-                } else {
-                    params.Set("direction", "-");
-                    params.Set("threshold", "-0.25");
-                }
-                break;
+        // For Axis button implementation
+        if (pad.axis != GCAdapter::PadAxes::Undefined) {
+            params.Set("axis", static_cast<u8>(pad.axis));
+            params.Set("button", static_cast<u16>(GCAdapter::PadButton::Stick));
+            params.Set("threshold", "0.25");
+            if (pad.axis_value > 0) {
+                params.Set("direction", "+");
+            } else {
+                params.Set("direction", "-");
             }
+            break;
         }
     }
     return params;
@@ -152,17 +151,14 @@ public:
     explicit GCAnalog(u32 port_, u32 axis_x_, u32 axis_y_, float deadzone_,
                       const GCAdapter::Adapter* adapter, float range_)
         : port(port_), axis_x(axis_x_), axis_y(axis_y_), deadzone(deadzone_), gcadapter(adapter),
-          origin_value_x(static_cast<float>(adapter->GetOriginValue(port_, axis_x_))),
-          origin_value_y(static_cast<float>(adapter->GetOriginValue(port_, axis_y_))),
           range(range_) {}
 
     float GetAxis(u32 axis) const {
         if (gcadapter->DeviceConnected(port)) {
             std::lock_guard lock{mutex};
-            const auto origin_value = axis % 2 == 0 ? origin_value_x : origin_value_y;
             const auto axis_value =
-                static_cast<float>(gcadapter->GetPadState()[port].axes.at(axis));
-            return (axis_value - origin_value) / (100.0f * range);
+                static_cast<float>(gcadapter->GetPadState(port).axis_values.at(axis));
+            return (axis_value) / (100.0f * range);
         }
         return 0.0f;
     }
@@ -215,8 +211,6 @@ private:
     const u32 axis_y;
     const float deadzone;
     const GCAdapter::Adapter* gcadapter;
-    const float origin_value_x;
-    const float origin_value_y;
     const float range;
     mutable std::mutex mutex;
 };
@@ -254,26 +248,44 @@ void GCAnalogFactory::EndConfiguration() {
 
 Common::ParamPackage GCAnalogFactory::GetNextInput() {
     GCAdapter::GCPadStatus pad;
+    Common::ParamPackage params;
     auto& queue = adapter->GetPadQueue();
-    for (std::size_t port = 0; port < queue.size(); ++port) {
-        while (queue[port].Pop(pad)) {
-            if (pad.axis == GCAdapter::PadAxes::Undefined ||
-                std::abs((static_cast<float>(pad.axis_value) - 128.0f) / 128.0f) < 0.1f) {
-                continue;
-            }
-            // An analog device needs two axes, so we need to store the axis for later and wait for
-            // a second input event. The axes also must be from the same joystick.
-            const u8 axis = static_cast<u8>(pad.axis);
-            if (analog_x_axis == -1) {
-                analog_x_axis = axis;
-                controller_number = static_cast<int>(port);
-            } else if (analog_y_axis == -1 && analog_x_axis != axis &&
-                       controller_number == static_cast<int>(port)) {
-                analog_y_axis = axis;
-            }
+    while (queue.Pop(pad)) {
+        if (pad.button != GCAdapter::PadButton::Undefined) {
+            params.Set("engine", "gcpad");
+            params.Set("port", static_cast<s32>(pad.port));
+            params.Set("button", static_cast<u16>(pad.button));
+            return params;
+        }
+        if (pad.axis == GCAdapter::PadAxes::Undefined ||
+            std::abs(static_cast<float>(pad.axis_value) / 128.0f) < 0.1f) {
+            continue;
+        }
+        // An analog device needs two axes, so we need to store the axis for later and wait for
+        // a second input event. The axes also must be from the same joystick.
+        const u8 axis = static_cast<u8>(pad.axis);
+        if (axis == 0 || axis == 1) {
+            analog_x_axis = 0;
+            analog_y_axis = 1;
+            controller_number = static_cast<s32>(pad.port);
+            break;
+        }
+        if (axis == 2 || axis == 3) {
+            analog_x_axis = 2;
+            analog_y_axis = 3;
+            controller_number = static_cast<s32>(pad.port);
+            break;
+        }
+
+        if (analog_x_axis == -1) {
+            analog_x_axis = axis;
+            controller_number = static_cast<s32>(pad.port);
+        } else if (analog_y_axis == -1 && analog_x_axis != axis &&
+                   controller_number == static_cast<s32>(pad.port)) {
+            analog_y_axis = axis;
+            break;
         }
     }
-    Common::ParamPackage params;
     if (analog_x_axis != -1 && analog_y_axis != -1) {
         params.Set("engine", "gcpad");
         params.Set("port", controller_number);
diff --git a/src/input_common/sdl/sdl_impl.cpp b/src/input_common/sdl/sdl_impl.cpp
index 9c3035920..10883e2d9 100644
--- a/src/input_common/sdl/sdl_impl.cpp
+++ b/src/input_common/sdl/sdl_impl.cpp
@@ -155,15 +155,15 @@ public:
         return sdl_joystick.get();
     }
 
-    void SetSDLJoystick(SDL_Joystick* joystick, SDL_GameController* controller) {
-        sdl_controller.reset(controller);
-        sdl_joystick.reset(joystick);
-    }
-
     SDL_GameController* GetSDLGameController() const {
         return sdl_controller.get();
     }
 
+    void SetSDLJoystick(SDL_Joystick* joystick, SDL_GameController* controller) {
+        sdl_joystick.reset(joystick);
+        sdl_controller.reset(controller);
+    }
+
 private:
     struct State {
         std::unordered_map<int, bool> buttons;
@@ -186,69 +186,58 @@ private:
 std::shared_ptr<SDLJoystick> SDLState::GetSDLJoystickByGUID(const std::string& guid, int port) {
     std::lock_guard lock{joystick_map_mutex};
     const auto it = joystick_map.find(guid);
+
     if (it != joystick_map.end()) {
         while (it->second.size() <= static_cast<std::size_t>(port)) {
             auto joystick = std::make_shared<SDLJoystick>(guid, static_cast<int>(it->second.size()),
                                                           nullptr, nullptr);
             it->second.emplace_back(std::move(joystick));
         }
+
         return it->second[static_cast<std::size_t>(port)];
     }
+
     auto joystick = std::make_shared<SDLJoystick>(guid, 0, nullptr, nullptr);
+
     return joystick_map[guid].emplace_back(std::move(joystick));
 }
 
 std::shared_ptr<SDLJoystick> SDLState::GetSDLJoystickBySDLID(SDL_JoystickID sdl_id) {
     auto sdl_joystick = SDL_JoystickFromInstanceID(sdl_id);
-    auto sdl_controller = SDL_GameControllerFromInstanceID(sdl_id);
     const std::string guid = GetGUID(sdl_joystick);
 
     std::lock_guard lock{joystick_map_mutex};
     const auto map_it = joystick_map.find(guid);
-    if (map_it != joystick_map.end()) {
-        const auto vec_it =
-            std::find_if(map_it->second.begin(), map_it->second.end(),
-                         [&sdl_joystick](const std::shared_ptr<SDLJoystick>& joystick) {
-                             return sdl_joystick == joystick->GetSDLJoystick();
-                         });
-        if (vec_it != map_it->second.end()) {
-            // This is the common case: There is already an existing SDL_Joystick mapped to a
-            // SDLJoystick. return the SDLJoystick
-            return *vec_it;
-        }
 
-        // Search for a SDLJoystick without a mapped SDL_Joystick...
-        const auto nullptr_it = std::find_if(map_it->second.begin(), map_it->second.end(),
-                                             [](const std::shared_ptr<SDLJoystick>& joystick) {
-                                                 return joystick->GetSDLJoystick() == nullptr;
-                                             });
-        if (nullptr_it != map_it->second.end()) {
-            // ... and map it
-            (*nullptr_it)->SetSDLJoystick(sdl_joystick, sdl_controller);
-            return *nullptr_it;
-        }
+    if (map_it == joystick_map.end()) {
+        return nullptr;
+    }
 
-        // There is no SDLJoystick without a mapped SDL_Joystick
-        // Create a new SDLJoystick
-        const int port = static_cast<int>(map_it->second.size());
-        auto joystick = std::make_shared<SDLJoystick>(guid, port, sdl_joystick, sdl_controller);
-        return map_it->second.emplace_back(std::move(joystick));
+    const auto vec_it = std::find_if(map_it->second.begin(), map_it->second.end(),
+                                     [&sdl_joystick](const auto& joystick) {
+                                         return joystick->GetSDLJoystick() == sdl_joystick;
+                                     });
+
+    if (vec_it == map_it->second.end()) {
+        return nullptr;
     }
 
-    auto joystick = std::make_shared<SDLJoystick>(guid, 0, sdl_joystick, sdl_controller);
-    return joystick_map[guid].emplace_back(std::move(joystick));
+    return *vec_it;
 }
 
 void SDLState::InitJoystick(int joystick_index) {
     SDL_Joystick* sdl_joystick = SDL_JoystickOpen(joystick_index);
     SDL_GameController* sdl_gamecontroller = nullptr;
+
     if (SDL_IsGameController(joystick_index)) {
         sdl_gamecontroller = SDL_GameControllerOpen(joystick_index);
     }
+
     if (!sdl_joystick) {
         LOG_ERROR(Input, "Failed to open joystick {}", joystick_index);
         return;
     }
+
     const std::string guid = GetGUID(sdl_joystick);
 
     std::lock_guard lock{joystick_map_mutex};
@@ -257,14 +246,17 @@ void SDLState::InitJoystick(int joystick_index) {
         joystick_map[guid].emplace_back(std::move(joystick));
         return;
     }
+
     auto& joystick_guid_list = joystick_map[guid];
-    const auto it = std::find_if(
-        joystick_guid_list.begin(), joystick_guid_list.end(),
-        [](const std::shared_ptr<SDLJoystick>& joystick) { return !joystick->GetSDLJoystick(); });
-    if (it != joystick_guid_list.end()) {
-        (*it)->SetSDLJoystick(sdl_joystick, sdl_gamecontroller);
+    const auto joystick_it =
+        std::find_if(joystick_guid_list.begin(), joystick_guid_list.end(),
+                     [](const auto& joystick) { return !joystick->GetSDLJoystick(); });
+
+    if (joystick_it != joystick_guid_list.end()) {
+        (*joystick_it)->SetSDLJoystick(sdl_joystick, sdl_gamecontroller);
         return;
     }
+
     const int port = static_cast<int>(joystick_guid_list.size());
     auto joystick = std::make_shared<SDLJoystick>(guid, port, sdl_joystick, sdl_gamecontroller);
     joystick_guid_list.emplace_back(std::move(joystick));
@@ -274,18 +266,14 @@ void SDLState::CloseJoystick(SDL_Joystick* sdl_joystick) {
     const std::string guid = GetGUID(sdl_joystick);
 
     std::lock_guard lock{joystick_map_mutex};
-    auto& joystick_guid_list = joystick_map[guid];
-    auto joystick_it = std::find_if(
-        joystick_guid_list.begin(), joystick_guid_list.end(),
-        [&sdl_joystick](auto& joystick) { return joystick->GetSDLJoystick() == sdl_joystick; });
-
-    if (joystick_it != joystick_guid_list.end()) {
-        (*joystick_it)->SetSDLJoystick(nullptr, nullptr);
-        joystick_guid_list.erase(joystick_it);
-        if (joystick_guid_list.empty()) {
-            joystick_map.erase(guid);
-        }
-    }
+    // This call to guid is safe since the joystick is guaranteed to be in the map
+    const auto& joystick_guid_list = joystick_map[guid];
+    const auto joystick_it = std::find_if(joystick_guid_list.begin(), joystick_guid_list.end(),
+                                          [&sdl_joystick](const auto& joystick) {
+                                              return joystick->GetSDLJoystick() == sdl_joystick;
+                                          });
+
+    (*joystick_it)->SetSDLJoystick(nullptr, nullptr);
 }
 
 void SDLState::HandleGameControllerEvent(const SDL_Event& event) {
@@ -720,8 +708,7 @@ std::vector<Common::ParamPackage> SDLState::GetInputDevices() {
     std::vector<Common::ParamPackage> devices;
     for (const auto& [key, value] : joystick_map) {
         for (const auto& joystick : value) {
-            auto* joy = joystick->GetSDLJoystick();
-            if (auto* controller = joystick->GetSDLGameController()) {
+            if (auto* const controller = joystick->GetSDLGameController()) {
                 std::string name =
                     fmt::format("{} {}", SDL_GameControllerName(controller), joystick->GetPort());
                 devices.emplace_back(Common::ParamPackage{
@@ -730,7 +717,7 @@ std::vector<Common::ParamPackage> SDLState::GetInputDevices() {
                     {"guid", joystick->GetGUID()},
                     {"port", std::to_string(joystick->GetPort())},
                 });
-            } else if (joy) {
+            } else if (auto* const joy = joystick->GetSDLJoystick()) {
                 std::string name = fmt::format("{} {}", SDL_JoystickName(joy), joystick->GetPort());
                 devices.emplace_back(Common::ParamPackage{
                     {"class", "sdl"},
@@ -797,21 +784,27 @@ Common::ParamPackage BuildHatParamPackageForButton(int port, std::string guid, s
 Common::ParamPackage SDLEventToButtonParamPackage(SDLState& state, const SDL_Event& event) {
     switch (event.type) {
     case SDL_JOYAXISMOTION: {
-        const auto joystick = state.GetSDLJoystickBySDLID(event.jaxis.which);
-        return BuildAnalogParamPackageForButton(joystick->GetPort(), joystick->GetGUID(),
-                                                static_cast<s32>(event.jaxis.axis),
-                                                event.jaxis.value);
+        if (const auto joystick = state.GetSDLJoystickBySDLID(event.jaxis.which)) {
+            return BuildAnalogParamPackageForButton(joystick->GetPort(), joystick->GetGUID(),
+                                                    static_cast<s32>(event.jaxis.axis),
+                                                    event.jaxis.value);
+        }
+        break;
     }
     case SDL_JOYBUTTONUP: {
-        const auto joystick = state.GetSDLJoystickBySDLID(event.jbutton.which);
-        return BuildButtonParamPackageForButton(joystick->GetPort(), joystick->GetGUID(),
-                                                static_cast<s32>(event.jbutton.button));
+        if (const auto joystick = state.GetSDLJoystickBySDLID(event.jbutton.which)) {
+            return BuildButtonParamPackageForButton(joystick->GetPort(), joystick->GetGUID(),
+                                                    static_cast<s32>(event.jbutton.button));
+        }
+        break;
     }
     case SDL_JOYHATMOTION: {
-        const auto joystick = state.GetSDLJoystickBySDLID(event.jhat.which);
-        return BuildHatParamPackageForButton(joystick->GetPort(), joystick->GetGUID(),
-                                             static_cast<s32>(event.jhat.hat),
-                                             static_cast<s32>(event.jhat.value));
+        if (const auto joystick = state.GetSDLJoystickBySDLID(event.jhat.which)) {
+            return BuildHatParamPackageForButton(joystick->GetPort(), joystick->GetGUID(),
+                                                 static_cast<s32>(event.jhat.hat),
+                                                 static_cast<s32>(event.jhat.value));
+        }
+        break;
     }
     }
     return {};
@@ -820,21 +813,27 @@ Common::ParamPackage SDLEventToButtonParamPackage(SDLState& state, const SDL_Eve
 Common::ParamPackage SDLEventToMotionParamPackage(SDLState& state, const SDL_Event& event) {
     switch (event.type) {
     case SDL_JOYAXISMOTION: {
-        const auto joystick = state.GetSDLJoystickBySDLID(event.jaxis.which);
-        return BuildAnalogParamPackageForButton(joystick->GetPort(), joystick->GetGUID(),
-                                                static_cast<s32>(event.jaxis.axis),
-                                                event.jaxis.value);
+        if (const auto joystick = state.GetSDLJoystickBySDLID(event.jaxis.which)) {
+            return BuildAnalogParamPackageForButton(joystick->GetPort(), joystick->GetGUID(),
+                                                    static_cast<s32>(event.jaxis.axis),
+                                                    event.jaxis.value);
+        }
+        break;
     }
     case SDL_JOYBUTTONUP: {
-        const auto joystick = state.GetSDLJoystickBySDLID(event.jbutton.which);
-        return BuildButtonParamPackageForButton(joystick->GetPort(), joystick->GetGUID(),
-                                                static_cast<s32>(event.jbutton.button));
+        if (const auto joystick = state.GetSDLJoystickBySDLID(event.jbutton.which)) {
+            return BuildButtonParamPackageForButton(joystick->GetPort(), joystick->GetGUID(),
+                                                    static_cast<s32>(event.jbutton.button));
+        }
+        break;
     }
     case SDL_JOYHATMOTION: {
-        const auto joystick = state.GetSDLJoystickBySDLID(event.jhat.which);
-        return BuildHatParamPackageForButton(joystick->GetPort(), joystick->GetGUID(),
-                                             static_cast<s32>(event.jhat.hat),
-                                             static_cast<s32>(event.jhat.value));
+        if (const auto joystick = state.GetSDLJoystickBySDLID(event.jhat.which)) {
+            return BuildHatParamPackageForButton(joystick->GetPort(), joystick->GetGUID(),
+                                                 static_cast<s32>(event.jhat.hat),
+                                                 static_cast<s32>(event.jhat.value));
+        }
+        break;
     }
     }
     return {};
@@ -1062,9 +1061,8 @@ public:
             // Simplify controller config by testing if game controller support is enabled.
             if (event.type == SDL_JOYAXISMOTION) {
                 const auto axis = event.jaxis.axis;
-                const auto joystick = state.GetSDLJoystickBySDLID(event.jaxis.which);
-                auto* const controller = joystick->GetSDLGameController();
-                if (controller) {
+                if (const auto joystick = state.GetSDLJoystickBySDLID(event.jaxis.which);
+                    auto* const controller = joystick->GetSDLGameController()) {
                     const auto axis_left_x =
                         SDL_GameControllerGetBindForAxis(controller, SDL_CONTROLLER_AXIS_LEFTX)
                             .value.axis;
@@ -1098,12 +1096,13 @@ public:
         }
 
         if (analog_x_axis != -1 && analog_y_axis != -1) {
-            const auto joystick = state.GetSDLJoystickBySDLID(event.jaxis.which);
-            auto params = BuildParamPackageForAnalog(joystick->GetPort(), joystick->GetGUID(),
-                                                     analog_x_axis, analog_y_axis);
-            analog_x_axis = -1;
-            analog_y_axis = -1;
-            return params;
+            if (const auto joystick = state.GetSDLJoystickBySDLID(event.jaxis.which)) {
+                auto params = BuildParamPackageForAnalog(joystick->GetPort(), joystick->GetGUID(),
+                                                         analog_x_axis, analog_y_axis);
+                analog_x_axis = -1;
+                analog_y_axis = -1;
+                return params;
+            }
         }
         return {};
     }
diff --git a/src/tests/common/fibers.cpp b/src/tests/common/fibers.cpp
index 4fd92428f..4757dd2b4 100644
--- a/src/tests/common/fibers.cpp
+++ b/src/tests/common/fibers.cpp
@@ -6,18 +6,40 @@
 #include <cstdlib>
 #include <functional>
 #include <memory>
+#include <mutex>
+#include <stdexcept>
 #include <thread>
 #include <unordered_map>
 #include <vector>
 
 #include <catch2/catch.hpp>
-#include <math.h>
+
 #include "common/common_types.h"
 #include "common/fiber.h"
-#include "common/spin_lock.h"
 
 namespace Common {
 
+class ThreadIds {
+public:
+    void Register(u32 id) {
+        const auto thread_id = std::this_thread::get_id();
+        std::scoped_lock lock{mutex};
+        if (ids.contains(thread_id)) {
+            throw std::logic_error{"Registering the same thread twice"};
+        }
+        ids.emplace(thread_id, id);
+    }
+
+    [[nodiscard]] u32 Get() const {
+        std::scoped_lock lock{mutex};
+        return ids.at(std::this_thread::get_id());
+    }
+
+private:
+    mutable std::mutex mutex;
+    std::unordered_map<std::thread::id, u32> ids;
+};
+
 class TestControl1 {
 public:
     TestControl1() = default;
@@ -26,7 +48,7 @@ public:
 
     void ExecuteThread(u32 id);
 
-    std::unordered_map<std::thread::id, u32> ids;
+    ThreadIds thread_ids;
     std::vector<std::shared_ptr<Common::Fiber>> thread_fibers;
     std::vector<std::shared_ptr<Common::Fiber>> work_fibers;
     std::vector<u32> items;
@@ -39,8 +61,7 @@ static void WorkControl1(void* control) {
 }
 
 void TestControl1::DoWork() {
-    std::thread::id this_id = std::this_thread::get_id();
-    u32 id = ids[this_id];
+    const u32 id = thread_ids.Get();
     u32 value = items[id];
     for (u32 i = 0; i < id; i++) {
         value++;
@@ -50,8 +71,7 @@ void TestControl1::DoWork() {
 }
 
 void TestControl1::ExecuteThread(u32 id) {
-    std::thread::id this_id = std::this_thread::get_id();
-    ids[this_id] = id;
+    thread_ids.Register(id);
     auto thread_fiber = Fiber::ThreadToFiber();
     thread_fibers[id] = thread_fiber;
     work_fibers[id] = std::make_shared<Fiber>(std::function<void(void*)>{WorkControl1}, this);
@@ -98,8 +118,7 @@ public:
             value1 += i;
         }
         Fiber::YieldTo(fiber1, fiber3);
-        std::thread::id this_id = std::this_thread::get_id();
-        u32 id = ids[this_id];
+        const u32 id = thread_ids.Get();
         assert1 = id == 1;
         value2 += 5000;
         Fiber::YieldTo(fiber1, thread_fibers[id]);
@@ -115,8 +134,7 @@ public:
     }
 
     void DoWork3() {
-        std::thread::id this_id = std::this_thread::get_id();
-        u32 id = ids[this_id];
+        const u32 id = thread_ids.Get();
         assert2 = id == 0;
         value1 += 1000;
         Fiber::YieldTo(fiber3, thread_fibers[id]);
@@ -125,14 +143,12 @@ public:
     void ExecuteThread(u32 id);
 
     void CallFiber1() {
-        std::thread::id this_id = std::this_thread::get_id();
-        u32 id = ids[this_id];
+        const u32 id = thread_ids.Get();
         Fiber::YieldTo(thread_fibers[id], fiber1);
     }
 
     void CallFiber2() {
-        std::thread::id this_id = std::this_thread::get_id();
-        u32 id = ids[this_id];
+        const u32 id = thread_ids.Get();
         Fiber::YieldTo(thread_fibers[id], fiber2);
     }
 
@@ -145,7 +161,7 @@ public:
     u32 value2{};
     std::atomic<bool> trap{true};
     std::atomic<bool> trap2{true};
-    std::unordered_map<std::thread::id, u32> ids;
+    ThreadIds thread_ids;
     std::vector<std::shared_ptr<Common::Fiber>> thread_fibers;
     std::shared_ptr<Common::Fiber> fiber1;
     std::shared_ptr<Common::Fiber> fiber2;
@@ -168,15 +184,13 @@ static void WorkControl2_3(void* control) {
 }
 
 void TestControl2::ExecuteThread(u32 id) {
-    std::thread::id this_id = std::this_thread::get_id();
-    ids[this_id] = id;
+    thread_ids.Register(id);
     auto thread_fiber = Fiber::ThreadToFiber();
     thread_fibers[id] = thread_fiber;
 }
 
 void TestControl2::Exit() {
-    std::thread::id this_id = std::this_thread::get_id();
-    u32 id = ids[this_id];
+    const u32 id = thread_ids.Get();
     thread_fibers[id]->Exit();
 }
 
@@ -228,24 +242,21 @@ public:
     void DoWork1() {
         value1 += 1;
         Fiber::YieldTo(fiber1, fiber2);
-        std::thread::id this_id = std::this_thread::get_id();
-        u32 id = ids[this_id];
+        const u32 id = thread_ids.Get();
         value3 += 1;
         Fiber::YieldTo(fiber1, thread_fibers[id]);
     }
 
     void DoWork2() {
         value2 += 1;
-        std::thread::id this_id = std::this_thread::get_id();
-        u32 id = ids[this_id];
+        const u32 id = thread_ids.Get();
         Fiber::YieldTo(fiber2, thread_fibers[id]);
     }
 
     void ExecuteThread(u32 id);
 
     void CallFiber1() {
-        std::thread::id this_id = std::this_thread::get_id();
-        u32 id = ids[this_id];
+        const u32 id = thread_ids.Get();
         Fiber::YieldTo(thread_fibers[id], fiber1);
     }
 
@@ -254,7 +265,7 @@ public:
     u32 value1{};
     u32 value2{};
     u32 value3{};
-    std::unordered_map<std::thread::id, u32> ids;
+    ThreadIds thread_ids;
     std::vector<std::shared_ptr<Common::Fiber>> thread_fibers;
     std::shared_ptr<Common::Fiber> fiber1;
     std::shared_ptr<Common::Fiber> fiber2;
@@ -271,15 +282,13 @@ static void WorkControl3_2(void* control) {
 }
 
 void TestControl3::ExecuteThread(u32 id) {
-    std::thread::id this_id = std::this_thread::get_id();
-    ids[this_id] = id;
+    thread_ids.Register(id);
     auto thread_fiber = Fiber::ThreadToFiber();
     thread_fibers[id] = thread_fiber;
 }
 
 void TestControl3::Exit() {
-    std::thread::id this_id = std::this_thread::get_id();
-    u32 id = ids[this_id];
+    const u32 id = thread_ids.Get();
     thread_fibers[id]->Exit();
 }
 
diff --git a/src/video_core/CMakeLists.txt b/src/video_core/CMakeLists.txt
index 77ebac19f..abcee2a1c 100644
--- a/src/video_core/CMakeLists.txt
+++ b/src/video_core/CMakeLists.txt
@@ -5,6 +5,24 @@ add_library(video_core STATIC
     buffer_cache/buffer_cache.h
     buffer_cache/map_interval.cpp
     buffer_cache/map_interval.h
+    cdma_pusher.cpp
+    cdma_pusher.h
+    command_classes/codecs/codec.cpp
+    command_classes/codecs/codec.h
+    command_classes/codecs/h264.cpp
+    command_classes/codecs/h264.h
+    command_classes/codecs/vp9.cpp
+    command_classes/codecs/vp9.h
+    command_classes/codecs/vp9_types.h
+    command_classes/host1x.cpp
+    command_classes/host1x.h
+    command_classes/nvdec.cpp
+    command_classes/nvdec.h
+    command_classes/nvdec_common.h
+    command_classes/sync_manager.cpp
+    command_classes/sync_manager.h
+    command_classes/vic.cpp
+    command_classes/vic.h
     compatible_formats.cpp
     compatible_formats.h
     dirty_flags.cpp
@@ -250,6 +268,14 @@ create_target_directory_groups(video_core)
 target_link_libraries(video_core PUBLIC common core)
 target_link_libraries(video_core PRIVATE glad xbyak)
 
+if (MSVC)
+    target_include_directories(video_core PRIVATE ${FFMPEG_INCLUDE_DIR})
+    target_link_libraries(video_core PUBLIC ${FFMPEG_LIBRARY_DIR}/swscale.lib ${FFMPEG_LIBRARY_DIR}/avcodec.lib ${FFMPEG_LIBRARY_DIR}/avutil.lib)
+else()
+    target_include_directories(video_core PRIVATE ${FFMPEG_INCLUDE_DIR})
+    target_link_libraries(video_core PRIVATE ${FFMPEG_LIBRARIES})
+endif()
+
 add_dependencies(video_core host_shaders)
 target_include_directories(video_core PRIVATE ${HOST_SHADERS_INCLUDE})
 
@@ -276,7 +302,10 @@ else()
     target_compile_options(video_core PRIVATE
         -Werror=conversion
         -Wno-error=sign-conversion
+        -Werror=pessimizing-move
+        -Werror=redundant-move
         -Werror=switch
+        -Werror=type-limits
         -Werror=unused-variable
 
         $<$<CXX_COMPILER_ID:GNU>:-Werror=class-memaccess>
diff --git a/src/video_core/cdma_pusher.cpp b/src/video_core/cdma_pusher.cpp
new file mode 100644
index 000000000..b60f86260
--- /dev/null
+++ b/src/video_core/cdma_pusher.cpp
@@ -0,0 +1,171 @@
+// MIT License
+//
+// Copyright (c) Ryujinx Team and Contributors
+//
+// 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.
+//
+
+#include "command_classes/host1x.h"
+#include "command_classes/nvdec.h"
+#include "command_classes/vic.h"
+#include "common/bit_util.h"
+#include "video_core/cdma_pusher.h"
+#include "video_core/command_classes/nvdec_common.h"
+#include "video_core/engines/maxwell_3d.h"
+#include "video_core/gpu.h"
+#include "video_core/memory_manager.h"
+
+namespace Tegra {
+CDmaPusher::CDmaPusher(GPU& gpu)
+    : gpu(gpu), nvdec_processor(std::make_shared<Nvdec>(gpu)),
+      vic_processor(std::make_unique<Vic>(gpu, nvdec_processor)),
+      host1x_processor(std::make_unique<Host1x>(gpu)),
+      nvdec_sync(std::make_unique<SyncptIncrManager>(gpu)),
+      vic_sync(std::make_unique<SyncptIncrManager>(gpu)) {}
+
+CDmaPusher::~CDmaPusher() = default;
+
+void CDmaPusher::Push(ChCommandHeaderList&& entries) {
+    cdma_queue.push(std::move(entries));
+}
+
+void CDmaPusher::DispatchCalls() {
+    while (!cdma_queue.empty()) {
+        Step();
+    }
+}
+
+void CDmaPusher::Step() {
+    const auto entries{cdma_queue.front()};
+    cdma_queue.pop();
+
+    std::vector<u32> values(entries.size());
+    std::memcpy(values.data(), entries.data(), entries.size() * sizeof(u32));
+
+    for (const u32 value : values) {
+        if (mask != 0) {
+            const u32 lbs = Common::CountTrailingZeroes32(mask);
+            mask &= ~(1U << lbs);
+            ExecuteCommand(static_cast<u32>(offset + lbs), value);
+            continue;
+        } else if (count != 0) {
+            --count;
+            ExecuteCommand(static_cast<u32>(offset), value);
+            if (incrementing) {
+                ++offset;
+            }
+            continue;
+        }
+        const auto mode = static_cast<ChSubmissionMode>((value >> 28) & 0xf);
+        switch (mode) {
+        case ChSubmissionMode::SetClass: {
+            mask = value & 0x3f;
+            offset = (value >> 16) & 0xfff;
+            current_class = static_cast<ChClassId>((value >> 6) & 0x3ff);
+            break;
+        }
+        case ChSubmissionMode::Incrementing:
+        case ChSubmissionMode::NonIncrementing:
+            count = value & 0xffff;
+            offset = (value >> 16) & 0xfff;
+            incrementing = mode == ChSubmissionMode::Incrementing;
+            break;
+        case ChSubmissionMode::Mask:
+            mask = value & 0xffff;
+            offset = (value >> 16) & 0xfff;
+            break;
+        case ChSubmissionMode::Immediate: {
+            const u32 data = value & 0xfff;
+            offset = (value >> 16) & 0xfff;
+            ExecuteCommand(static_cast<u32>(offset), data);
+            break;
+        }
+        default:
+            UNIMPLEMENTED_MSG("ChSubmission mode {} is not implemented!", static_cast<u32>(mode));
+            break;
+        }
+    }
+}
+
+void CDmaPusher::ExecuteCommand(u32 offset, u32 data) {
+    switch (current_class) {
+    case ChClassId::NvDec:
+        ThiStateWrite(nvdec_thi_state, offset, {data});
+        switch (static_cast<ThiMethod>(offset)) {
+        case ThiMethod::IncSyncpt: {
+            LOG_DEBUG(Service_NVDRV, "NVDEC Class IncSyncpt Method");
+            const auto syncpoint_id = static_cast<u32>(data & 0xFF);
+            const auto cond = static_cast<u32>((data >> 8) & 0xFF);
+            if (cond == 0) {
+                nvdec_sync->Increment(syncpoint_id);
+            } else {
+                nvdec_sync->IncrementWhenDone(static_cast<u32>(current_class), syncpoint_id);
+                nvdec_sync->SignalDone(syncpoint_id);
+            }
+            break;
+        }
+        case ThiMethod::SetMethod1:
+            LOG_DEBUG(Service_NVDRV, "NVDEC method 0x{:X}",
+                      static_cast<u32>(nvdec_thi_state.method_0));
+            nvdec_processor->ProcessMethod(
+                static_cast<Tegra::Nvdec::Method>(nvdec_thi_state.method_0), {data});
+            break;
+        default:
+            break;
+        }
+        break;
+    case ChClassId::GraphicsVic:
+        ThiStateWrite(vic_thi_state, static_cast<u32>(offset), {data});
+        switch (static_cast<ThiMethod>(offset)) {
+        case ThiMethod::IncSyncpt: {
+            LOG_DEBUG(Service_NVDRV, "VIC Class IncSyncpt Method");
+            const auto syncpoint_id = static_cast<u32>(data & 0xFF);
+            const auto cond = static_cast<u32>((data >> 8) & 0xFF);
+            if (cond == 0) {
+                vic_sync->Increment(syncpoint_id);
+            } else {
+                vic_sync->IncrementWhenDone(static_cast<u32>(current_class), syncpoint_id);
+                vic_sync->SignalDone(syncpoint_id);
+            }
+            break;
+        }
+        case ThiMethod::SetMethod1:
+            LOG_DEBUG(Service_NVDRV, "VIC method 0x{:X}, Args=({})",
+                      static_cast<u32>(vic_thi_state.method_0), data);
+            vic_processor->ProcessMethod(static_cast<Tegra::Vic::Method>(vic_thi_state.method_0),
+                                         {data});
+            break;
+        default:
+            break;
+        }
+        break;
+    case ChClassId::Host1x:
+        // This device is mainly for syncpoint synchronization
+        LOG_DEBUG(Service_NVDRV, "Host1X Class Method");
+        host1x_processor->ProcessMethod(static_cast<Tegra::Host1x::Method>(offset), {data});
+        break;
+    default:
+        UNIMPLEMENTED_MSG("Current class not implemented {:X}", static_cast<u32>(current_class));
+        break;
+    }
+}
+
+void CDmaPusher::ThiStateWrite(ThiRegisters& state, u32 offset, const std::vector<u32>& arguments) {
+    u8* const state_offset = reinterpret_cast<u8*>(&state) + sizeof(u32) * offset;
+    std::memcpy(state_offset, arguments.data(), sizeof(u32) * arguments.size());
+}
+
+} // namespace Tegra
diff --git a/src/video_core/cdma_pusher.h b/src/video_core/cdma_pusher.h
new file mode 100644
index 000000000..982f309c5
--- /dev/null
+++ b/src/video_core/cdma_pusher.h
@@ -0,0 +1,138 @@
+// Copyright 2020 yuzu Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#pragma once
+
+#include <memory>
+#include <unordered_map>
+#include <vector>
+#include <queue>
+
+#include "common/bit_field.h"
+#include "common/common_types.h"
+#include "video_core/command_classes/sync_manager.h"
+
+namespace Tegra {
+
+class GPU;
+class Nvdec;
+class Vic;
+class Host1x;
+
+enum class ChSubmissionMode : u32 {
+    SetClass = 0,
+    Incrementing = 1,
+    NonIncrementing = 2,
+    Mask = 3,
+    Immediate = 4,
+    Restart = 5,
+    Gather = 6,
+};
+
+enum class ChClassId : u32 {
+    NoClass = 0x0,
+    Host1x = 0x1,
+    VideoEncodeMpeg = 0x20,
+    VideoEncodeNvEnc = 0x21,
+    VideoStreamingVi = 0x30,
+    VideoStreamingIsp = 0x32,
+    VideoStreamingIspB = 0x34,
+    VideoStreamingViI2c = 0x36,
+    GraphicsVic = 0x5d,
+    Graphics3D = 0x60,
+    GraphicsGpu = 0x61,
+    Tsec = 0xe0,
+    TsecB = 0xe1,
+    NvJpg = 0xc0,
+    NvDec = 0xf0
+};
+
+enum class ChMethod : u32 {
+    Empty = 0,
+    SetMethod = 0x10,
+    SetData = 0x11,
+};
+
+union ChCommandHeader {
+    u32 raw;
+    BitField<0, 16, u32> value;
+    BitField<16, 12, ChMethod> method_offset;
+    BitField<28, 4, ChSubmissionMode> submission_mode;
+};
+static_assert(sizeof(ChCommandHeader) == sizeof(u32), "ChCommand header is an invalid size");
+
+struct ChCommand {
+    ChClassId class_id{};
+    int method_offset{};
+    std::vector<u32> arguments;
+};
+
+using ChCommandHeaderList = std::vector<Tegra::ChCommandHeader>;
+using ChCommandList = std::vector<Tegra::ChCommand>;
+
+struct ThiRegisters {
+    u32_le increment_syncpt{};
+    INSERT_PADDING_WORDS(1);
+    u32_le increment_syncpt_error{};
+    u32_le ctx_switch_incremement_syncpt{};
+    INSERT_PADDING_WORDS(4);
+    u32_le ctx_switch{};
+    INSERT_PADDING_WORDS(1);
+    u32_le ctx_syncpt_eof{};
+    INSERT_PADDING_WORDS(5);
+    u32_le method_0{};
+    u32_le method_1{};
+    INSERT_PADDING_WORDS(12);
+    u32_le int_status{};
+    u32_le int_mask{};
+};
+
+enum class ThiMethod : u32 {
+    IncSyncpt = offsetof(ThiRegisters, increment_syncpt) / sizeof(u32),
+    SetMethod0 = offsetof(ThiRegisters, method_0) / sizeof(u32),
+    SetMethod1 = offsetof(ThiRegisters, method_1) / sizeof(u32),
+};
+
+class CDmaPusher {
+public:
+    explicit CDmaPusher(GPU& gpu);
+    ~CDmaPusher();
+
+    /// Push NVDEC command buffer entries into queue
+    void Push(ChCommandHeaderList&& entries);
+
+    /// Process queued command buffer entries
+    void DispatchCalls();
+
+    /// Process one queue element
+    void Step();
+
+    /// Invoke command class devices to execute the command based on the current state
+    void ExecuteCommand(u32 offset, u32 data);
+
+private:
+    /// Write arguments value to the ThiRegisters member at the specified offset
+    void ThiStateWrite(ThiRegisters& state, u32 offset, const std::vector<u32>& arguments);
+
+    GPU& gpu;
+
+    std::shared_ptr<Tegra::Nvdec> nvdec_processor;
+    std::unique_ptr<Tegra::Vic> vic_processor;
+    std::unique_ptr<Tegra::Host1x> host1x_processor;
+    std::unique_ptr<SyncptIncrManager> nvdec_sync;
+    std::unique_ptr<SyncptIncrManager> vic_sync;
+    ChClassId current_class{};
+    ThiRegisters vic_thi_state{};
+    ThiRegisters nvdec_thi_state{};
+
+    s32 count{};
+    s32 offset{};
+    s32 mask{};
+    bool incrementing{};
+
+    // Queue of command lists to be processed
+    std::queue<ChCommandHeaderList> cdma_queue;
+};
+
+} // namespace Tegra
diff --git a/src/video_core/command_classes/codecs/codec.cpp b/src/video_core/command_classes/codecs/codec.cpp
new file mode 100644
index 000000000..1adf3cd13
--- /dev/null
+++ b/src/video_core/command_classes/codecs/codec.cpp
@@ -0,0 +1,115 @@
+// Copyright 2020 yuzu Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#include <cstring>
+#include <fstream>
+#include <vector>
+#include "common/assert.h"
+#include "video_core/command_classes/codecs/codec.h"
+#include "video_core/command_classes/codecs/h264.h"
+#include "video_core/command_classes/codecs/vp9.h"
+#include "video_core/gpu.h"
+#include "video_core/memory_manager.h"
+
+extern "C" {
+#include <libavutil/opt.h>
+}
+
+namespace Tegra {
+
+Codec::Codec(GPU& gpu_)
+    : gpu(gpu_), h264_decoder(std::make_unique<Decoder::H264>(gpu)),
+      vp9_decoder(std::make_unique<Decoder::VP9>(gpu)) {}
+
+Codec::~Codec() {
+    if (!initialized) {
+        return;
+    }
+    // Free libav memory
+    avcodec_send_packet(av_codec_ctx, nullptr);
+    avcodec_receive_frame(av_codec_ctx, av_frame);
+    avcodec_flush_buffers(av_codec_ctx);
+
+    av_frame_unref(av_frame);
+    av_free(av_frame);
+    avcodec_close(av_codec_ctx);
+}
+
+void Codec::SetTargetCodec(NvdecCommon::VideoCodec codec) {
+    LOG_INFO(Service_NVDRV, "NVDEC video codec initialized to {}", static_cast<u32>(codec));
+    current_codec = codec;
+}
+
+void Codec::StateWrite(u32 offset, u64 arguments) {
+    u8* const state_offset = reinterpret_cast<u8*>(&state) + offset * sizeof(u64);
+    std::memcpy(state_offset, &arguments, sizeof(u64));
+}
+
+void Codec::Decode() {
+    bool is_first_frame = false;
+
+    if (!initialized) {
+        if (current_codec == NvdecCommon::VideoCodec::H264) {
+            av_codec = avcodec_find_decoder(AV_CODEC_ID_H264);
+        } else if (current_codec == NvdecCommon::VideoCodec::Vp9) {
+            av_codec = avcodec_find_decoder(AV_CODEC_ID_VP9);
+        } else {
+            LOG_ERROR(Service_NVDRV, "Unknown video codec {}", static_cast<u32>(current_codec));
+            return;
+        }
+
+        av_codec_ctx = avcodec_alloc_context3(av_codec);
+        av_frame = av_frame_alloc();
+        av_opt_set(av_codec_ctx->priv_data, "tune", "zerolatency", 0);
+
+        // TODO(ameerj): libavcodec gpu hw acceleration
+
+        const auto av_error = avcodec_open2(av_codec_ctx, av_codec, nullptr);
+        if (av_error < 0) {
+            LOG_ERROR(Service_NVDRV, "avcodec_open2() Failed.");
+            av_frame_unref(av_frame);
+            av_free(av_frame);
+            avcodec_close(av_codec_ctx);
+            return;
+        }
+        initialized = true;
+        is_first_frame = true;
+    }
+    bool vp9_hidden_frame = false;
+
+    AVPacket packet{};
+    av_init_packet(&packet);
+    std::vector<u8> frame_data;
+
+    if (current_codec == NvdecCommon::VideoCodec::H264) {
+        frame_data = h264_decoder->ComposeFrameHeader(state, is_first_frame);
+    } else if (current_codec == NvdecCommon::VideoCodec::Vp9) {
+        frame_data = vp9_decoder->ComposeFrameHeader(state);
+        vp9_hidden_frame = vp9_decoder->WasFrameHidden();
+    }
+
+    packet.data = frame_data.data();
+    packet.size = static_cast<int>(frame_data.size());
+
+    avcodec_send_packet(av_codec_ctx, &packet);
+
+    if (!vp9_hidden_frame) {
+        // Only receive/store visible frames
+        avcodec_receive_frame(av_codec_ctx, av_frame);
+    }
+}
+
+AVFrame* Codec::GetCurrentFrame() {
+    return av_frame;
+}
+
+const AVFrame* Codec::GetCurrentFrame() const {
+    return av_frame;
+}
+
+NvdecCommon::VideoCodec Codec::GetCurrentCodec() const {
+    return current_codec;
+}
+
+} // namespace Tegra
diff --git a/src/video_core/command_classes/codecs/codec.h b/src/video_core/command_classes/codecs/codec.h
new file mode 100644
index 000000000..cb67094f6
--- /dev/null
+++ b/src/video_core/command_classes/codecs/codec.h
@@ -0,0 +1,66 @@
+// Copyright 2020 yuzu Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#pragma once
+
+#include <memory>
+#include "common/common_types.h"
+#include "video_core/command_classes/nvdec_common.h"
+
+extern "C" {
+#if defined(__GNUC__) || defined(__clang__)
+#pragma GCC diagnostic ignored "-Wconversion"
+#endif
+#include <libavcodec/avcodec.h>
+#if defined(__GNUC__) || defined(__clang__)
+#pragma GCC diagnostic pop
+#endif
+}
+
+namespace Tegra {
+class GPU;
+struct VicRegisters;
+
+namespace Decoder {
+class H264;
+class VP9;
+} // namespace Decoder
+
+class Codec {
+public:
+    explicit Codec(GPU& gpu);
+    ~Codec();
+
+    /// Sets NVDEC video stream codec
+    void SetTargetCodec(NvdecCommon::VideoCodec codec);
+
+    /// Populate NvdecRegisters state with argument value at the provided offset
+    void StateWrite(u32 offset, u64 arguments);
+
+    /// Call decoders to construct headers, decode AVFrame with ffmpeg
+    void Decode();
+
+    /// Returns most recently decoded frame
+    AVFrame* GetCurrentFrame();
+    const AVFrame* GetCurrentFrame() const;
+
+    /// Returns the value of current_codec
+    NvdecCommon::VideoCodec GetCurrentCodec() const;
+
+private:
+    bool initialized{};
+    NvdecCommon::VideoCodec current_codec{NvdecCommon::VideoCodec::None};
+
+    AVCodec* av_codec{nullptr};
+    AVCodecContext* av_codec_ctx{nullptr};
+    AVFrame* av_frame{nullptr};
+
+    GPU& gpu;
+    std::unique_ptr<Decoder::H264> h264_decoder;
+    std::unique_ptr<Decoder::VP9> vp9_decoder;
+
+    NvdecCommon::NvdecRegisters state{};
+};
+
+} // namespace Tegra
diff --git a/src/video_core/command_classes/codecs/h264.cpp b/src/video_core/command_classes/codecs/h264.cpp
new file mode 100644
index 000000000..549a40f52
--- /dev/null
+++ b/src/video_core/command_classes/codecs/h264.cpp
@@ -0,0 +1,292 @@
+// MIT License
+//
+// Copyright (c) Ryujinx Team and Contributors
+//
+// 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.
+//
+
+#include <array>
+#include "common/bit_util.h"
+#include "video_core/command_classes/codecs/h264.h"
+#include "video_core/gpu.h"
+#include "video_core/memory_manager.h"
+
+namespace Tegra::Decoder {
+namespace {
+// ZigZag LUTs from libavcodec.
+constexpr std::array<u8, 64> zig_zag_direct{
+    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,
+};
+
+constexpr std::array<u8, 16> zig_zag_scan{
+    0 + 0 * 4, 1 + 0 * 4, 0 + 1 * 4, 0 + 2 * 4, 1 + 1 * 4, 2 + 0 * 4, 3 + 0 * 4, 2 + 1 * 4,
+    1 + 2 * 4, 0 + 3 * 4, 1 + 3 * 4, 2 + 2 * 4, 3 + 1 * 4, 3 + 2 * 4, 2 + 3 * 4, 3 + 3 * 4,
+};
+} // Anonymous namespace
+
+H264::H264(GPU& gpu_) : gpu(gpu_) {}
+
+H264::~H264() = default;
+
+std::vector<u8>& H264::ComposeFrameHeader(NvdecCommon::NvdecRegisters& state, bool is_first_frame) {
+    H264DecoderContext context{};
+    gpu.MemoryManager().ReadBlock(state.picture_info_offset, &context, sizeof(H264DecoderContext));
+
+    const s32 frame_number = static_cast<s32>((context.h264_parameter_set.flags >> 46) & 0x1ffff);
+    if (!is_first_frame && frame_number != 0) {
+        frame.resize(context.frame_data_size);
+
+        gpu.MemoryManager().ReadBlock(state.frame_bitstream_offset, frame.data(), frame.size());
+    } else {
+        /// Encode header
+        H264BitWriter writer{};
+        writer.WriteU(1, 24);
+        writer.WriteU(0, 1);
+        writer.WriteU(3, 2);
+        writer.WriteU(7, 5);
+        writer.WriteU(100, 8);
+        writer.WriteU(0, 8);
+        writer.WriteU(31, 8);
+        writer.WriteUe(0);
+        const auto chroma_format_idc =
+            static_cast<u32>((context.h264_parameter_set.flags >> 12) & 3);
+        writer.WriteUe(chroma_format_idc);
+        if (chroma_format_idc == 3) {
+            writer.WriteBit(false);
+        }
+
+        writer.WriteUe(0);
+        writer.WriteUe(0);
+        writer.WriteBit(false); // QpprimeYZeroTransformBypassFlag
+        writer.WriteBit(false); // Scaling matrix present flag
+
+        const auto order_cnt_type = static_cast<u32>((context.h264_parameter_set.flags >> 14) & 3);
+        writer.WriteUe(static_cast<u32>((context.h264_parameter_set.flags >> 8) & 0xf));
+        writer.WriteUe(order_cnt_type);
+        if (order_cnt_type == 0) {
+            writer.WriteUe(context.h264_parameter_set.log2_max_pic_order_cnt);
+        } else if (order_cnt_type == 1) {
+            writer.WriteBit(context.h264_parameter_set.delta_pic_order_always_zero_flag != 0);
+
+            writer.WriteSe(0);
+            writer.WriteSe(0);
+            writer.WriteUe(0);
+        }
+
+        const s32 pic_height = context.h264_parameter_set.pic_height_in_map_units /
+                               (context.h264_parameter_set.frame_mbs_only_flag ? 1 : 2);
+
+        writer.WriteUe(16);
+        writer.WriteBit(false);
+        writer.WriteUe(context.h264_parameter_set.pic_width_in_mbs - 1);
+        writer.WriteUe(pic_height - 1);
+        writer.WriteBit(context.h264_parameter_set.frame_mbs_only_flag != 0);
+
+        if (!context.h264_parameter_set.frame_mbs_only_flag) {
+            writer.WriteBit(((context.h264_parameter_set.flags >> 0) & 1) != 0);
+        }
+
+        writer.WriteBit(((context.h264_parameter_set.flags >> 1) & 1) != 0);
+        writer.WriteBit(false); // Frame cropping flag
+        writer.WriteBit(false); // VUI parameter present flag
+
+        writer.End();
+
+        // H264 PPS
+        writer.WriteU(1, 24);
+        writer.WriteU(0, 1);
+        writer.WriteU(3, 2);
+        writer.WriteU(8, 5);
+
+        writer.WriteUe(0);
+        writer.WriteUe(0);
+
+        writer.WriteBit(context.h264_parameter_set.entropy_coding_mode_flag != 0);
+        writer.WriteBit(false);
+        writer.WriteUe(0);
+        writer.WriteUe(context.h264_parameter_set.num_refidx_l0_default_active);
+        writer.WriteUe(context.h264_parameter_set.num_refidx_l1_default_active);
+        writer.WriteBit(((context.h264_parameter_set.flags >> 2) & 1) != 0);
+        writer.WriteU(static_cast<s32>((context.h264_parameter_set.flags >> 32) & 0x3), 2);
+        s32 pic_init_qp = static_cast<s32>((context.h264_parameter_set.flags >> 16) & 0x3f);
+        pic_init_qp = (pic_init_qp << 26) >> 26;
+        writer.WriteSe(pic_init_qp);
+        writer.WriteSe(0);
+        s32 chroma_qp_index_offset =
+            static_cast<s32>((context.h264_parameter_set.flags >> 22) & 0x1f);
+        chroma_qp_index_offset = (chroma_qp_index_offset << 27) >> 27;
+
+        writer.WriteSe(chroma_qp_index_offset);
+        writer.WriteBit(context.h264_parameter_set.deblocking_filter_control_flag != 0);
+        writer.WriteBit(((context.h264_parameter_set.flags >> 3) & 1) != 0);
+        writer.WriteBit(context.h264_parameter_set.redundant_pic_count_flag != 0);
+        writer.WriteBit(context.h264_parameter_set.transform_8x8_mode_flag != 0);
+
+        writer.WriteBit(true);
+
+        for (s32 index = 0; index < 6; index++) {
+            writer.WriteBit(true);
+            const auto matrix_x4 =
+                std::vector<u8>(context.scaling_matrix_4.begin(), context.scaling_matrix_4.end());
+            writer.WriteScalingList(matrix_x4, index * 16, 16);
+        }
+
+        if (context.h264_parameter_set.transform_8x8_mode_flag) {
+            for (s32 index = 0; index < 2; index++) {
+                writer.WriteBit(true);
+                const auto matrix_x8 = std::vector<u8>(context.scaling_matrix_8.begin(),
+                                                       context.scaling_matrix_8.end());
+
+                writer.WriteScalingList(matrix_x8, index * 64, 64);
+            }
+        }
+
+        s32 chroma_qp_index_offset2 =
+            static_cast<s32>((context.h264_parameter_set.flags >> 27) & 0x1f);
+        chroma_qp_index_offset2 = (chroma_qp_index_offset2 << 27) >> 27;
+
+        writer.WriteSe(chroma_qp_index_offset2);
+
+        writer.End();
+
+        const auto& encoded_header = writer.GetByteArray();
+        frame.resize(encoded_header.size() + context.frame_data_size);
+        std::memcpy(frame.data(), encoded_header.data(), encoded_header.size());
+
+        gpu.MemoryManager().ReadBlock(state.frame_bitstream_offset,
+                                      frame.data() + encoded_header.size(),
+                                      context.frame_data_size);
+    }
+
+    return frame;
+}
+
+H264BitWriter::H264BitWriter() = default;
+
+H264BitWriter::~H264BitWriter() = default;
+
+void H264BitWriter::WriteU(s32 value, s32 value_sz) {
+    WriteBits(value, value_sz);
+}
+
+void H264BitWriter::WriteSe(s32 value) {
+    WriteExpGolombCodedInt(value);
+}
+
+void H264BitWriter::WriteUe(u32 value) {
+    WriteExpGolombCodedUInt(value);
+}
+
+void H264BitWriter::End() {
+    WriteBit(true);
+    Flush();
+}
+
+void H264BitWriter::WriteBit(bool state) {
+    WriteBits(state ? 1 : 0, 1);
+}
+
+void H264BitWriter::WriteScalingList(const std::vector<u8>& list, s32 start, s32 count) {
+    std::vector<u8> scan(count);
+    if (count == 16) {
+        std::memcpy(scan.data(), zig_zag_scan.data(), scan.size());
+    } else {
+        std::memcpy(scan.data(), zig_zag_direct.data(), scan.size());
+    }
+    u8 last_scale = 8;
+
+    for (s32 index = 0; index < count; index++) {
+        const u8 value = list[start + scan[index]];
+        const s32 delta_scale = static_cast<s32>(value - last_scale);
+
+        WriteSe(delta_scale);
+
+        last_scale = value;
+    }
+}
+
+std::vector<u8>& H264BitWriter::GetByteArray() {
+    return byte_array;
+}
+
+const std::vector<u8>& H264BitWriter::GetByteArray() const {
+    return byte_array;
+}
+
+void H264BitWriter::WriteBits(s32 value, s32 bit_count) {
+    s32 value_pos = 0;
+
+    s32 remaining = bit_count;
+
+    while (remaining > 0) {
+        s32 copy_size = remaining;
+
+        const s32 free_bits = GetFreeBufferBits();
+
+        if (copy_size > free_bits) {
+            copy_size = free_bits;
+        }
+
+        const s32 mask = (1 << copy_size) - 1;
+
+        const s32 src_shift = (bit_count - value_pos) - copy_size;
+        const s32 dst_shift = (buffer_size - buffer_pos) - copy_size;
+
+        buffer |= ((value >> src_shift) & mask) << dst_shift;
+
+        value_pos += copy_size;
+        buffer_pos += copy_size;
+        remaining -= copy_size;
+    }
+}
+
+void H264BitWriter::WriteExpGolombCodedInt(s32 value) {
+    const s32 sign = value <= 0 ? 0 : 1;
+    if (value < 0) {
+        value = -value;
+    }
+    value = (value << 1) - sign;
+    WriteExpGolombCodedUInt(value);
+}
+
+void H264BitWriter::WriteExpGolombCodedUInt(u32 value) {
+    const s32 size = 32 - Common::CountLeadingZeroes32(static_cast<s32>(value + 1));
+    WriteBits(1, size);
+
+    value -= (1U << (size - 1)) - 1;
+    WriteBits(static_cast<s32>(value), size - 1);
+}
+
+s32 H264BitWriter::GetFreeBufferBits() {
+    if (buffer_pos == buffer_size) {
+        Flush();
+    }
+
+    return buffer_size - buffer_pos;
+}
+
+void H264BitWriter::Flush() {
+    if (buffer_pos == 0) {
+        return;
+    }
+    byte_array.push_back(static_cast<u8>(buffer));
+
+    buffer = 0;
+    buffer_pos = 0;
+}
+} // namespace Tegra::Decoder
diff --git a/src/video_core/command_classes/codecs/h264.h b/src/video_core/command_classes/codecs/h264.h
new file mode 100644
index 000000000..f2292fd2f
--- /dev/null
+++ b/src/video_core/command_classes/codecs/h264.h
@@ -0,0 +1,118 @@
+// MIT License
+//
+// Copyright (c) Ryujinx Team and Contributors
+//
+// 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.
+//
+
+#pragma once
+
+#include <vector>
+#include "common/common_funcs.h"
+#include "common/common_types.h"
+#include "video_core/command_classes/nvdec_common.h"
+
+namespace Tegra {
+class GPU;
+namespace Decoder {
+
+class H264BitWriter {
+public:
+    H264BitWriter();
+    ~H264BitWriter();
+
+    /// The following Write methods are based on clause 9.1 in the H.264 specification.
+    /// WriteSe and WriteUe write in the Exp-Golomb-coded syntax
+    void WriteU(s32 value, s32 value_sz);
+    void WriteSe(s32 value);
+    void WriteUe(u32 value);
+
+    /// Finalize the bitstream
+    void End();
+
+    /// append a bit to the stream, equivalent value to the state parameter
+    void WriteBit(bool state);
+
+    /// Based on section 7.3.2.1.1.1 and Table 7-4 in the H.264 specification
+    /// Writes the scaling matrices of the sream
+    void WriteScalingList(const std::vector<u8>& list, s32 start, s32 count);
+
+    /// Return the bitstream as a vector.
+    std::vector<u8>& GetByteArray();
+    const std::vector<u8>& GetByteArray() const;
+
+private:
+    void WriteBits(s32 value, s32 bit_count);
+    void WriteExpGolombCodedInt(s32 value);
+    void WriteExpGolombCodedUInt(u32 value);
+    s32 GetFreeBufferBits();
+    void Flush();
+
+    s32 buffer_size{8};
+
+    s32 buffer{};
+    s32 buffer_pos{};
+    std::vector<u8> byte_array;
+};
+
+class H264 {
+public:
+    explicit H264(GPU& gpu);
+    ~H264();
+
+    /// Compose the H264 header of the frame for FFmpeg decoding
+    std::vector<u8>& ComposeFrameHeader(NvdecCommon::NvdecRegisters& state,
+                                        bool is_first_frame = false);
+
+private:
+    struct H264ParameterSet {
+        u32 log2_max_pic_order_cnt{};
+        u32 delta_pic_order_always_zero_flag{};
+        u32 frame_mbs_only_flag{};
+        u32 pic_width_in_mbs{};
+        u32 pic_height_in_map_units{};
+        INSERT_PADDING_WORDS(1);
+        u32 entropy_coding_mode_flag{};
+        u32 bottom_field_pic_order_flag{};
+        u32 num_refidx_l0_default_active{};
+        u32 num_refidx_l1_default_active{};
+        u32 deblocking_filter_control_flag{};
+        u32 redundant_pic_count_flag{};
+        u32 transform_8x8_mode_flag{};
+        INSERT_PADDING_WORDS(9);
+        u64 flags{};
+        u32 frame_number{};
+        u32 frame_number2{};
+    };
+    static_assert(sizeof(H264ParameterSet) == 0x68, "H264ParameterSet is an invalid size");
+
+    struct H264DecoderContext {
+        INSERT_PADDING_BYTES(0x48);
+        u32 frame_data_size{};
+        INSERT_PADDING_BYTES(0xc);
+        H264ParameterSet h264_parameter_set{};
+        INSERT_PADDING_BYTES(0x100);
+        std::array<u8, 0x60> scaling_matrix_4;
+        std::array<u8, 0x80> scaling_matrix_8;
+    };
+    static_assert(sizeof(H264DecoderContext) == 0x2a0, "H264DecoderContext is an invalid size");
+
+    std::vector<u8> frame;
+    GPU& gpu;
+};
+
+} // namespace Decoder
+} // namespace Tegra
diff --git a/src/video_core/command_classes/codecs/vp9.cpp b/src/video_core/command_classes/codecs/vp9.cpp
new file mode 100644
index 000000000..b3e98aa9f
--- /dev/null
+++ b/src/video_core/command_classes/codecs/vp9.cpp
@@ -0,0 +1,1034 @@
+// Copyright 2020 yuzu Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#include <cstring> // for std::memcpy
+#include <numeric>
+#include "video_core/command_classes/codecs/vp9.h"
+#include "video_core/gpu.h"
+#include "video_core/memory_manager.h"
+
+namespace Tegra::Decoder {
+namespace {
+// Default compressed header probabilities once frame context resets
+constexpr Vp9EntropyProbs default_probs{
+    .y_mode_prob{
+        65,  32, 18, 144, 162, 194, 41, 51, 98, 132, 68,  18, 165, 217, 196, 45, 40, 78,
+        173, 80, 19, 176, 240, 193, 64, 35, 46, 221, 135, 38, 194, 248, 121, 96, 85, 29,
+    },
+    .partition_prob{
+        199, 122, 141, 0, 147, 63, 159, 0, 148, 133, 118, 0, 121, 104, 114, 0,
+        174, 73,  87,  0, 92,  41, 83,  0, 82,  99,  50,  0, 53,  39,  39,  0,
+        177, 58,  59,  0, 68,  26, 63,  0, 52,  79,  25,  0, 17,  14,  12,  0,
+        222, 34,  30,  0, 72,  16, 44,  0, 58,  32,  12,  0, 10,  7,   6,   0,
+    },
+    .coef_probs{
+        195, 29,  183, 0, 84,  49,  136, 0, 8,   42,  71,  0, 0,   0,   0,   0, 0,   0,   0,   0,
+        0,   0,   0,   0, 31,  107, 169, 0, 35,  99,  159, 0, 17,  82,  140, 0, 8,   66,  114, 0,
+        2,   44,  76,  0, 1,   19,  32,  0, 40,  132, 201, 0, 29,  114, 187, 0, 13,  91,  157, 0,
+        7,   75,  127, 0, 3,   58,  95,  0, 1,   28,  47,  0, 69,  142, 221, 0, 42,  122, 201, 0,
+        15,  91,  159, 0, 6,   67,  121, 0, 1,   42,  77,  0, 1,   17,  31,  0, 102, 148, 228, 0,
+        67,  117, 204, 0, 17,  82,  154, 0, 6,   59,  114, 0, 2,   39,  75,  0, 1,   15,  29,  0,
+        156, 57,  233, 0, 119, 57,  212, 0, 58,  48,  163, 0, 29,  40,  124, 0, 12,  30,  81,  0,
+        3,   12,  31,  0, 191, 107, 226, 0, 124, 117, 204, 0, 25,  99,  155, 0, 0,   0,   0,   0,
+        0,   0,   0,   0, 0,   0,   0,   0, 29,  148, 210, 0, 37,  126, 194, 0, 8,   93,  157, 0,
+        2,   68,  118, 0, 1,   39,  69,  0, 1,   17,  33,  0, 41,  151, 213, 0, 27,  123, 193, 0,
+        3,   82,  144, 0, 1,   58,  105, 0, 1,   32,  60,  0, 1,   13,  26,  0, 59,  159, 220, 0,
+        23,  126, 198, 0, 4,   88,  151, 0, 1,   66,  114, 0, 1,   38,  71,  0, 1,   18,  34,  0,
+        114, 136, 232, 0, 51,  114, 207, 0, 11,  83,  155, 0, 3,   56,  105, 0, 1,   33,  65,  0,
+        1,   17,  34,  0, 149, 65,  234, 0, 121, 57,  215, 0, 61,  49,  166, 0, 28,  36,  114, 0,
+        12,  25,  76,  0, 3,   16,  42,  0, 214, 49,  220, 0, 132, 63,  188, 0, 42,  65,  137, 0,
+        0,   0,   0,   0, 0,   0,   0,   0, 0,   0,   0,   0, 85,  137, 221, 0, 104, 131, 216, 0,
+        49,  111, 192, 0, 21,  87,  155, 0, 2,   49,  87,  0, 1,   16,  28,  0, 89,  163, 230, 0,
+        90,  137, 220, 0, 29,  100, 183, 0, 10,  70,  135, 0, 2,   42,  81,  0, 1,   17,  33,  0,
+        108, 167, 237, 0, 55,  133, 222, 0, 15,  97,  179, 0, 4,   72,  135, 0, 1,   45,  85,  0,
+        1,   19,  38,  0, 124, 146, 240, 0, 66,  124, 224, 0, 17,  88,  175, 0, 4,   58,  122, 0,
+        1,   36,  75,  0, 1,   18,  37,  0, 141, 79,  241, 0, 126, 70,  227, 0, 66,  58,  182, 0,
+        30,  44,  136, 0, 12,  34,  96,  0, 2,   20,  47,  0, 229, 99,  249, 0, 143, 111, 235, 0,
+        46,  109, 192, 0, 0,   0,   0,   0, 0,   0,   0,   0, 0,   0,   0,   0, 82,  158, 236, 0,
+        94,  146, 224, 0, 25,  117, 191, 0, 9,   87,  149, 0, 3,   56,  99,  0, 1,   33,  57,  0,
+        83,  167, 237, 0, 68,  145, 222, 0, 10,  103, 177, 0, 2,   72,  131, 0, 1,   41,  79,  0,
+        1,   20,  39,  0, 99,  167, 239, 0, 47,  141, 224, 0, 10,  104, 178, 0, 2,   73,  133, 0,
+        1,   44,  85,  0, 1,   22,  47,  0, 127, 145, 243, 0, 71,  129, 228, 0, 17,  93,  177, 0,
+        3,   61,  124, 0, 1,   41,  84,  0, 1,   21,  52,  0, 157, 78,  244, 0, 140, 72,  231, 0,
+        69,  58,  184, 0, 31,  44,  137, 0, 14,  38,  105, 0, 8,   23,  61,  0, 125, 34,  187, 0,
+        52,  41,  133, 0, 6,   31,  56,  0, 0,   0,   0,   0, 0,   0,   0,   0, 0,   0,   0,   0,
+        37,  109, 153, 0, 51,  102, 147, 0, 23,  87,  128, 0, 8,   67,  101, 0, 1,   41,  63,  0,
+        1,   19,  29,  0, 31,  154, 185, 0, 17,  127, 175, 0, 6,   96,  145, 0, 2,   73,  114, 0,
+        1,   51,  82,  0, 1,   28,  45,  0, 23,  163, 200, 0, 10,  131, 185, 0, 2,   93,  148, 0,
+        1,   67,  111, 0, 1,   41,  69,  0, 1,   14,  24,  0, 29,  176, 217, 0, 12,  145, 201, 0,
+        3,   101, 156, 0, 1,   69,  111, 0, 1,   39,  63,  0, 1,   14,  23,  0, 57,  192, 233, 0,
+        25,  154, 215, 0, 6,   109, 167, 0, 3,   78,  118, 0, 1,   48,  69,  0, 1,   21,  29,  0,
+        202, 105, 245, 0, 108, 106, 216, 0, 18,  90,  144, 0, 0,   0,   0,   0, 0,   0,   0,   0,
+        0,   0,   0,   0, 33,  172, 219, 0, 64,  149, 206, 0, 14,  117, 177, 0, 5,   90,  141, 0,
+        2,   61,  95,  0, 1,   37,  57,  0, 33,  179, 220, 0, 11,  140, 198, 0, 1,   89,  148, 0,
+        1,   60,  104, 0, 1,   33,  57,  0, 1,   12,  21,  0, 30,  181, 221, 0, 8,   141, 198, 0,
+        1,   87,  145, 0, 1,   58,  100, 0, 1,   31,  55,  0, 1,   12,  20,  0, 32,  186, 224, 0,
+        7,   142, 198, 0, 1,   86,  143, 0, 1,   58,  100, 0, 1,   31,  55,  0, 1,   12,  22,  0,
+        57,  192, 227, 0, 20,  143, 204, 0, 3,   96,  154, 0, 1,   68,  112, 0, 1,   42,  69,  0,
+        1,   19,  32,  0, 212, 35,  215, 0, 113, 47,  169, 0, 29,  48,  105, 0, 0,   0,   0,   0,
+        0,   0,   0,   0, 0,   0,   0,   0, 74,  129, 203, 0, 106, 120, 203, 0, 49,  107, 178, 0,
+        19,  84,  144, 0, 4,   50,  84,  0, 1,   15,  25,  0, 71,  172, 217, 0, 44,  141, 209, 0,
+        15,  102, 173, 0, 6,   76,  133, 0, 2,   51,  89,  0, 1,   24,  42,  0, 64,  185, 231, 0,
+        31,  148, 216, 0, 8,   103, 175, 0, 3,   74,  131, 0, 1,   46,  81,  0, 1,   18,  30,  0,
+        65,  196, 235, 0, 25,  157, 221, 0, 5,   105, 174, 0, 1,   67,  120, 0, 1,   38,  69,  0,
+        1,   15,  30,  0, 65,  204, 238, 0, 30,  156, 224, 0, 7,   107, 177, 0, 2,   70,  124, 0,
+        1,   42,  73,  0, 1,   18,  34,  0, 225, 86,  251, 0, 144, 104, 235, 0, 42,  99,  181, 0,
+        0,   0,   0,   0, 0,   0,   0,   0, 0,   0,   0,   0, 85,  175, 239, 0, 112, 165, 229, 0,
+        29,  136, 200, 0, 12,  103, 162, 0, 6,   77,  123, 0, 2,   53,  84,  0, 75,  183, 239, 0,
+        30,  155, 221, 0, 3,   106, 171, 0, 1,   74,  128, 0, 1,   44,  76,  0, 1,   17,  28,  0,
+        73,  185, 240, 0, 27,  159, 222, 0, 2,   107, 172, 0, 1,   75,  127, 0, 1,   42,  73,  0,
+        1,   17,  29,  0, 62,  190, 238, 0, 21,  159, 222, 0, 2,   107, 172, 0, 1,   72,  122, 0,
+        1,   40,  71,  0, 1,   18,  32,  0, 61,  199, 240, 0, 27,  161, 226, 0, 4,   113, 180, 0,
+        1,   76,  129, 0, 1,   46,  80,  0, 1,   23,  41,  0, 7,   27,  153, 0, 5,   30,  95,  0,
+        1,   16,  30,  0, 0,   0,   0,   0, 0,   0,   0,   0, 0,   0,   0,   0, 50,  75,  127, 0,
+        57,  75,  124, 0, 27,  67,  108, 0, 10,  54,  86,  0, 1,   33,  52,  0, 1,   12,  18,  0,
+        43,  125, 151, 0, 26,  108, 148, 0, 7,   83,  122, 0, 2,   59,  89,  0, 1,   38,  60,  0,
+        1,   17,  27,  0, 23,  144, 163, 0, 13,  112, 154, 0, 2,   75,  117, 0, 1,   50,  81,  0,
+        1,   31,  51,  0, 1,   14,  23,  0, 18,  162, 185, 0, 6,   123, 171, 0, 1,   78,  125, 0,
+        1,   51,  86,  0, 1,   31,  54,  0, 1,   14,  23,  0, 15,  199, 227, 0, 3,   150, 204, 0,
+        1,   91,  146, 0, 1,   55,  95,  0, 1,   30,  53,  0, 1,   11,  20,  0, 19,  55,  240, 0,
+        19,  59,  196, 0, 3,   52,  105, 0, 0,   0,   0,   0, 0,   0,   0,   0, 0,   0,   0,   0,
+        41,  166, 207, 0, 104, 153, 199, 0, 31,  123, 181, 0, 14,  101, 152, 0, 5,   72,  106, 0,
+        1,   36,  52,  0, 35,  176, 211, 0, 12,  131, 190, 0, 2,   88,  144, 0, 1,   60,  101, 0,
+        1,   36,  60,  0, 1,   16,  28,  0, 28,  183, 213, 0, 8,   134, 191, 0, 1,   86,  142, 0,
+        1,   56,  96,  0, 1,   30,  53,  0, 1,   12,  20,  0, 20,  190, 215, 0, 4,   135, 192, 0,
+        1,   84,  139, 0, 1,   53,  91,  0, 1,   28,  49,  0, 1,   11,  20,  0, 13,  196, 216, 0,
+        2,   137, 192, 0, 1,   86,  143, 0, 1,   57,  99,  0, 1,   32,  56,  0, 1,   13,  24,  0,
+        211, 29,  217, 0, 96,  47,  156, 0, 22,  43,  87,  0, 0,   0,   0,   0, 0,   0,   0,   0,
+        0,   0,   0,   0, 78,  120, 193, 0, 111, 116, 186, 0, 46,  102, 164, 0, 15,  80,  128, 0,
+        2,   49,  76,  0, 1,   18,  28,  0, 71,  161, 203, 0, 42,  132, 192, 0, 10,  98,  150, 0,
+        3,   69,  109, 0, 1,   44,  70,  0, 1,   18,  29,  0, 57,  186, 211, 0, 30,  140, 196, 0,
+        4,   93,  146, 0, 1,   62,  102, 0, 1,   38,  65,  0, 1,   16,  27,  0, 47,  199, 217, 0,
+        14,  145, 196, 0, 1,   88,  142, 0, 1,   57,  98,  0, 1,   36,  62,  0, 1,   15,  26,  0,
+        26,  219, 229, 0, 5,   155, 207, 0, 1,   94,  151, 0, 1,   60,  104, 0, 1,   36,  62,  0,
+        1,   16,  28,  0, 233, 29,  248, 0, 146, 47,  220, 0, 43,  52,  140, 0, 0,   0,   0,   0,
+        0,   0,   0,   0, 0,   0,   0,   0, 100, 163, 232, 0, 179, 161, 222, 0, 63,  142, 204, 0,
+        37,  113, 174, 0, 26,  89,  137, 0, 18,  68,  97,  0, 85,  181, 230, 0, 32,  146, 209, 0,
+        7,   100, 164, 0, 3,   71,  121, 0, 1,   45,  77,  0, 1,   18,  30,  0, 65,  187, 230, 0,
+        20,  148, 207, 0, 2,   97,  159, 0, 1,   68,  116, 0, 1,   40,  70,  0, 1,   14,  29,  0,
+        40,  194, 227, 0, 8,   147, 204, 0, 1,   94,  155, 0, 1,   65,  112, 0, 1,   39,  66,  0,
+        1,   14,  26,  0, 16,  208, 228, 0, 3,   151, 207, 0, 1,   98,  160, 0, 1,   67,  117, 0,
+        1,   41,  74,  0, 1,   17,  31,  0, 17,  38,  140, 0, 7,   34,  80,  0, 1,   17,  29,  0,
+        0,   0,   0,   0, 0,   0,   0,   0, 0,   0,   0,   0, 37,  75,  128, 0, 41,  76,  128, 0,
+        26,  66,  116, 0, 12,  52,  94,  0, 2,   32,  55,  0, 1,   10,  16,  0, 50,  127, 154, 0,
+        37,  109, 152, 0, 16,  82,  121, 0, 5,   59,  85,  0, 1,   35,  54,  0, 1,   13,  20,  0,
+        40,  142, 167, 0, 17,  110, 157, 0, 2,   71,  112, 0, 1,   44,  72,  0, 1,   27,  45,  0,
+        1,   11,  17,  0, 30,  175, 188, 0, 9,   124, 169, 0, 1,   74,  116, 0, 1,   48,  78,  0,
+        1,   30,  49,  0, 1,   11,  18,  0, 10,  222, 223, 0, 2,   150, 194, 0, 1,   83,  128, 0,
+        1,   48,  79,  0, 1,   27,  45,  0, 1,   11,  17,  0, 36,  41,  235, 0, 29,  36,  193, 0,
+        10,  27,  111, 0, 0,   0,   0,   0, 0,   0,   0,   0, 0,   0,   0,   0, 85,  165, 222, 0,
+        177, 162, 215, 0, 110, 135, 195, 0, 57,  113, 168, 0, 23,  83,  120, 0, 10,  49,  61,  0,
+        85,  190, 223, 0, 36,  139, 200, 0, 5,   90,  146, 0, 1,   60,  103, 0, 1,   38,  65,  0,
+        1,   18,  30,  0, 72,  202, 223, 0, 23,  141, 199, 0, 2,   86,  140, 0, 1,   56,  97,  0,
+        1,   36,  61,  0, 1,   16,  27,  0, 55,  218, 225, 0, 13,  145, 200, 0, 1,   86,  141, 0,
+        1,   57,  99,  0, 1,   35,  61,  0, 1,   13,  22,  0, 15,  235, 212, 0, 1,   132, 184, 0,
+        1,   84,  139, 0, 1,   57,  97,  0, 1,   34,  56,  0, 1,   14,  23,  0, 181, 21,  201, 0,
+        61,  37,  123, 0, 10,  38,  71,  0, 0,   0,   0,   0, 0,   0,   0,   0, 0,   0,   0,   0,
+        47,  106, 172, 0, 95,  104, 173, 0, 42,  93,  159, 0, 18,  77,  131, 0, 4,   50,  81,  0,
+        1,   17,  23,  0, 62,  147, 199, 0, 44,  130, 189, 0, 28,  102, 154, 0, 18,  75,  115, 0,
+        2,   44,  65,  0, 1,   12,  19,  0, 55,  153, 210, 0, 24,  130, 194, 0, 3,   93,  146, 0,
+        1,   61,  97,  0, 1,   31,  50,  0, 1,   10,  16,  0, 49,  186, 223, 0, 17,  148, 204, 0,
+        1,   96,  142, 0, 1,   53,  83,  0, 1,   26,  44,  0, 1,   11,  17,  0, 13,  217, 212, 0,
+        2,   136, 180, 0, 1,   78,  124, 0, 1,   50,  83,  0, 1,   29,  49,  0, 1,   14,  23,  0,
+        197, 13,  247, 0, 82,  17,  222, 0, 25,  17,  162, 0, 0,   0,   0,   0, 0,   0,   0,   0,
+        0,   0,   0,   0, 126, 186, 247, 0, 234, 191, 243, 0, 176, 177, 234, 0, 104, 158, 220, 0,
+        66,  128, 186, 0, 55,  90,  137, 0, 111, 197, 242, 0, 46,  158, 219, 0, 9,   104, 171, 0,
+        2,   65,  125, 0, 1,   44,  80,  0, 1,   17,  91,  0, 104, 208, 245, 0, 39,  168, 224, 0,
+        3,   109, 162, 0, 1,   79,  124, 0, 1,   50,  102, 0, 1,   43,  102, 0, 84,  220, 246, 0,
+        31,  177, 231, 0, 2,   115, 180, 0, 1,   79,  134, 0, 1,   55,  77,  0, 1,   60,  79,  0,
+        43,  243, 240, 0, 8,   180, 217, 0, 1,   115, 166, 0, 1,   84,  121, 0, 1,   51,  67,  0,
+        1,   16,  6,   0,
+    },
+    .switchable_interp_prob{235, 162, 36, 255, 34, 3, 149, 144},
+    .inter_mode_prob{
+        2,  173, 34, 0,  7,  145, 85, 0,  7,  166, 63, 0,  7,  94,
+        66, 0,   8,  64, 46, 0,   17, 81, 31, 0,   25, 29, 30, 0,
+    },
+    .intra_inter_prob{9, 102, 187, 225},
+    .comp_inter_prob{9, 102, 187, 225, 0},
+    .single_ref_prob{33, 16, 77, 74, 142, 142, 172, 170, 238, 247},
+    .comp_ref_prob{50, 126, 123, 221, 226},
+    .tx_32x32_prob{3, 136, 37, 5, 52, 13},
+    .tx_16x16_prob{20, 152, 15, 101},
+    .tx_8x8_prob{100, 66},
+    .skip_probs{192, 128, 64},
+    .joints{32, 64, 96},
+    .sign{128, 128},
+    .classes{
+        224, 144, 192, 168, 192, 176, 192, 198, 198, 245,
+        216, 128, 176, 160, 176, 176, 192, 198, 198, 208,
+    },
+    .class_0{216, 208},
+    .prob_bits{
+        136, 140, 148, 160, 176, 192, 224, 234, 234, 240,
+        136, 140, 148, 160, 176, 192, 224, 234, 234, 240,
+    },
+    .class_0_fr{128, 128, 64, 96, 112, 64, 128, 128, 64, 96, 112, 64},
+    .fr{64, 96, 64, 64, 96, 64},
+    .class_0_hp{160, 160},
+    .high_precision{128, 128},
+};
+
+constexpr std::array<s32, 256> norm_lut{
+    0, 7, 6, 6, 5, 5, 5, 5, 4, 4, 4, 4, 4, 4, 4, 4, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3,
+    2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2,
+    1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,
+    1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,
+    0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
+    0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
+    0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
+    0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
+};
+
+constexpr std::array<s32, 254> map_lut{
+    20,  21,  22,  23,  24,  25,  0,   26,  27,  28,  29,  30,  31,  32,  33,  34,  35,  36,  37,
+    1,   38,  39,  40,  41,  42,  43,  44,  45,  46,  47,  48,  49,  2,   50,  51,  52,  53,  54,
+    55,  56,  57,  58,  59,  60,  61,  3,   62,  63,  64,  65,  66,  67,  68,  69,  70,  71,  72,
+    73,  4,   74,  75,  76,  77,  78,  79,  80,  81,  82,  83,  84,  85,  5,   86,  87,  88,  89,
+    90,  91,  92,  93,  94,  95,  96,  97,  6,   98,  99,  100, 101, 102, 103, 104, 105, 106, 107,
+    108, 109, 7,   110, 111, 112, 113, 114, 115, 116, 117, 118, 119, 120, 121, 8,   122, 123, 124,
+    125, 126, 127, 128, 129, 130, 131, 132, 133, 9,   134, 135, 136, 137, 138, 139, 140, 141, 142,
+    143, 144, 145, 10,  146, 147, 148, 149, 150, 151, 152, 153, 154, 155, 156, 157, 11,  158, 159,
+    160, 161, 162, 163, 164, 165, 166, 167, 168, 169, 12,  170, 171, 172, 173, 174, 175, 176, 177,
+    178, 179, 180, 181, 13,  182, 183, 184, 185, 186, 187, 188, 189, 190, 191, 192, 193, 14,  194,
+    195, 196, 197, 198, 199, 200, 201, 202, 203, 204, 205, 15,  206, 207, 208, 209, 210, 211, 212,
+    213, 214, 215, 216, 217, 16,  218, 219, 220, 221, 222, 223, 224, 225, 226, 227, 228, 229, 17,
+    230, 231, 232, 233, 234, 235, 236, 237, 238, 239, 240, 241, 18,  242, 243, 244, 245, 246, 247,
+    248, 249, 250, 251, 252, 253, 19,
+};
+} // Anonymous namespace
+
+VP9::VP9(GPU& gpu) : gpu(gpu) {}
+
+VP9::~VP9() = default;
+
+void VP9::WriteProbabilityUpdate(VpxRangeEncoder& writer, u8 new_prob, u8 old_prob) {
+    const bool update = new_prob != old_prob;
+
+    writer.Write(update, diff_update_probability);
+
+    if (update) {
+        WriteProbabilityDelta(writer, new_prob, old_prob);
+    }
+}
+template <typename T, std::size_t N>
+void VP9::WriteProbabilityUpdate(VpxRangeEncoder& writer, const std::array<T, N>& new_prob,
+                                 const std::array<T, N>& old_prob) {
+    for (std::size_t offset = 0; offset < new_prob.size(); ++offset) {
+        WriteProbabilityUpdate(writer, new_prob[offset], old_prob[offset]);
+    }
+}
+
+template <typename T, std::size_t N>
+void VP9::WriteProbabilityUpdateAligned4(VpxRangeEncoder& writer, const std::array<T, N>& new_prob,
+                                         const std::array<T, N>& old_prob) {
+    for (std::size_t offset = 0; offset < new_prob.size(); offset += 4) {
+        WriteProbabilityUpdate(writer, new_prob[offset + 0], old_prob[offset + 0]);
+        WriteProbabilityUpdate(writer, new_prob[offset + 1], old_prob[offset + 1]);
+        WriteProbabilityUpdate(writer, new_prob[offset + 2], old_prob[offset + 2]);
+    }
+}
+
+void VP9::WriteProbabilityDelta(VpxRangeEncoder& writer, u8 new_prob, u8 old_prob) {
+    const int delta = RemapProbability(new_prob, old_prob);
+
+    EncodeTermSubExp(writer, delta);
+}
+
+s32 VP9::RemapProbability(s32 new_prob, s32 old_prob) {
+    new_prob--;
+    old_prob--;
+
+    std::size_t index{};
+
+    if (old_prob * 2 <= 0xff) {
+        index = static_cast<std::size_t>(std::max(0, RecenterNonNeg(new_prob, old_prob) - 1));
+    } else {
+        index = static_cast<std::size_t>(
+            std::max(0, RecenterNonNeg(0xff - 1 - new_prob, 0xff - 1 - old_prob) - 1));
+    }
+
+    return map_lut[index];
+}
+
+s32 VP9::RecenterNonNeg(s32 new_prob, s32 old_prob) {
+    if (new_prob > old_prob * 2) {
+        return new_prob;
+    } else if (new_prob >= old_prob) {
+        return (new_prob - old_prob) * 2;
+    } else {
+        return (old_prob - new_prob) * 2 - 1;
+    }
+}
+
+void VP9::EncodeTermSubExp(VpxRangeEncoder& writer, s32 value) {
+    if (WriteLessThan(writer, value, 16)) {
+        writer.Write(value, 4);
+    } else if (WriteLessThan(writer, value, 32)) {
+        writer.Write(value - 16, 4);
+    } else if (WriteLessThan(writer, value, 64)) {
+        writer.Write(value - 32, 5);
+    } else {
+        value -= 64;
+
+        constexpr s32 size = 8;
+
+        const s32 mask = (1 << size) - 191;
+
+        const s32 delta = value - mask;
+
+        if (delta < 0) {
+            writer.Write(value, size - 1);
+        } else {
+            writer.Write(delta / 2 + mask, size - 1);
+            writer.Write(delta & 1, 1);
+        }
+    }
+}
+
+bool VP9::WriteLessThan(VpxRangeEncoder& writer, s32 value, s32 test) {
+    const bool is_lt = value < test;
+    writer.Write(!is_lt);
+    return is_lt;
+}
+
+void VP9::WriteCoefProbabilityUpdate(VpxRangeEncoder& writer, s32 tx_mode,
+                                     const std::array<u8, 2304>& new_prob,
+                                     const std::array<u8, 2304>& old_prob) {
+    // Note: There's 1 byte added on each packet for alignment,
+    // this byte is ignored when doing updates.
+    constexpr s32 block_bytes = 2 * 2 * 6 * 6 * 4;
+
+    const auto needs_update = [&](s32 base_index) -> bool {
+        s32 index = base_index;
+        for (s32 i = 0; i < 2; i++) {
+            for (s32 j = 0; j < 2; j++) {
+                for (s32 k = 0; k < 6; k++) {
+                    for (s32 l = 0; l < 6; l++) {
+                        if (new_prob[index + 0] != old_prob[index + 0] ||
+                            new_prob[index + 1] != old_prob[index + 1] ||
+                            new_prob[index + 2] != old_prob[index + 2]) {
+                            return true;
+                        }
+
+                        index += 4;
+                    }
+                }
+            }
+        }
+        return false;
+    };
+
+    for (s32 block_index = 0; block_index < 4; block_index++) {
+        const s32 base_index = block_index * block_bytes;
+        const bool update = needs_update(base_index);
+        writer.Write(update);
+
+        if (update) {
+            s32 index = base_index;
+            for (s32 i = 0; i < 2; i++) {
+                for (s32 j = 0; j < 2; j++) {
+                    for (s32 k = 0; k < 6; k++) {
+                        for (s32 l = 0; l < 6; l++) {
+                            if (k != 0 || l < 3) {
+                                WriteProbabilityUpdate(writer, new_prob[index + 0],
+                                                       old_prob[index + 0]);
+                                WriteProbabilityUpdate(writer, new_prob[index + 1],
+                                                       old_prob[index + 1]);
+                                WriteProbabilityUpdate(writer, new_prob[index + 2],
+                                                       old_prob[index + 2]);
+                            }
+                            index += 4;
+                        }
+                    }
+                }
+            }
+        }
+
+        if (block_index == tx_mode) {
+            break;
+        }
+    }
+}
+
+void VP9::WriteMvProbabilityUpdate(VpxRangeEncoder& writer, u8 new_prob, u8 old_prob) {
+    const bool update = new_prob != old_prob;
+    writer.Write(update, diff_update_probability);
+
+    if (update) {
+        writer.Write(new_prob >> 1, 7);
+    }
+}
+
+s32 VP9::CalcMinLog2TileCols(s32 frame_width) {
+    const s32 sb64_cols = (frame_width + 63) / 64;
+    s32 min_log2 = 0;
+
+    while ((64 << min_log2) < sb64_cols) {
+        min_log2++;
+    }
+
+    return min_log2;
+}
+
+s32 VP9::CalcMaxLog2TileCols(s32 frameWidth) {
+    const s32 sb64_cols = (frameWidth + 63) / 64;
+    s32 max_log2 = 1;
+
+    while ((sb64_cols >> max_log2) >= 4) {
+        max_log2++;
+    }
+
+    return max_log2 - 1;
+}
+
+Vp9PictureInfo VP9::GetVp9PictureInfo(const NvdecCommon::NvdecRegisters& state) {
+    PictureInfo picture_info{};
+    gpu.MemoryManager().ReadBlock(state.picture_info_offset, &picture_info, sizeof(PictureInfo));
+    Vp9PictureInfo vp9_info = picture_info.Convert();
+
+    InsertEntropy(state.vp9_entropy_probs_offset, vp9_info.entropy);
+
+    // surface_luma_offset[0:3] contains the address of the reference frame offsets in the following
+    // order: last, golden, altref, current. It may be worthwhile to track the updates done here
+    // to avoid buffering frame data needed for reference frame updating in the header composition.
+    std::memcpy(vp9_info.frame_offsets.data(), state.surface_luma_offset.data(), 4 * sizeof(u64));
+
+    return vp9_info;
+}
+
+void VP9::InsertEntropy(u64 offset, Vp9EntropyProbs& dst) {
+    EntropyProbs entropy{};
+    gpu.MemoryManager().ReadBlock(offset, &entropy, sizeof(EntropyProbs));
+    entropy.Convert(dst);
+}
+
+Vp9FrameContainer VP9::GetCurrentFrame(const NvdecCommon::NvdecRegisters& state) {
+    Vp9FrameContainer frame{};
+    {
+        gpu.SyncGuestHost();
+        frame.info = GetVp9PictureInfo(state);
+
+        frame.bit_stream.resize(frame.info.bitstream_size);
+        gpu.MemoryManager().ReadBlock(state.frame_bitstream_offset, frame.bit_stream.data(),
+                                      frame.info.bitstream_size);
+    }
+    // Buffer two frames, saving the last show frame info
+    if (!next_next_frame.bit_stream.empty()) {
+        Vp9FrameContainer temp{
+            .info = frame.info,
+            .bit_stream = frame.bit_stream,
+        };
+        next_next_frame.info.show_frame = frame.info.last_frame_shown;
+        frame.info = next_next_frame.info;
+        frame.bit_stream = next_next_frame.bit_stream;
+        next_next_frame = std::move(temp);
+
+        if (!next_frame.bit_stream.empty()) {
+            Vp9FrameContainer temp2{
+                .info = frame.info,
+                .bit_stream = frame.bit_stream,
+            };
+            next_frame.info.show_frame = frame.info.last_frame_shown;
+            frame.info = next_frame.info;
+            frame.bit_stream = next_frame.bit_stream;
+            next_frame = std::move(temp2);
+        } else {
+            next_frame.info = frame.info;
+            next_frame.bit_stream = frame.bit_stream;
+        }
+    } else {
+        next_next_frame.info = frame.info;
+        next_next_frame.bit_stream = frame.bit_stream;
+    }
+    return frame;
+}
+
+std::vector<u8> VP9::ComposeCompressedHeader() {
+    VpxRangeEncoder writer{};
+
+    if (!current_frame_info.lossless) {
+        if (static_cast<u32>(current_frame_info.transform_mode) >= 3) {
+            writer.Write(3, 2);
+            writer.Write(current_frame_info.transform_mode == 4);
+        } else {
+            writer.Write(current_frame_info.transform_mode, 2);
+        }
+    }
+
+    if (current_frame_info.transform_mode == 4) {
+        // tx_mode_probs() in the spec
+        WriteProbabilityUpdate(writer, current_frame_info.entropy.tx_8x8_prob,
+                               prev_frame_probs.tx_8x8_prob);
+        WriteProbabilityUpdate(writer, current_frame_info.entropy.tx_16x16_prob,
+                               prev_frame_probs.tx_16x16_prob);
+        WriteProbabilityUpdate(writer, current_frame_info.entropy.tx_32x32_prob,
+                               prev_frame_probs.tx_32x32_prob);
+        if (current_frame_info.show_frame && !current_frame_info.is_key_frame) {
+            prev_frame_probs.tx_8x8_prob = current_frame_info.entropy.tx_8x8_prob;
+            prev_frame_probs.tx_16x16_prob = current_frame_info.entropy.tx_16x16_prob;
+            prev_frame_probs.tx_32x32_prob = current_frame_info.entropy.tx_32x32_prob;
+        }
+    }
+    // read_coef_probs()  in the spec
+    WriteCoefProbabilityUpdate(writer, current_frame_info.transform_mode,
+                               current_frame_info.entropy.coef_probs, prev_frame_probs.coef_probs);
+    // read_skip_probs()  in the spec
+    WriteProbabilityUpdate(writer, current_frame_info.entropy.skip_probs,
+                           prev_frame_probs.skip_probs);
+
+    if (current_frame_info.show_frame && !current_frame_info.is_key_frame) {
+        prev_frame_probs.coef_probs = current_frame_info.entropy.coef_probs;
+        prev_frame_probs.skip_probs = current_frame_info.entropy.skip_probs;
+    }
+
+    if (!current_frame_info.intra_only) {
+        // read_inter_probs() in the spec
+        WriteProbabilityUpdateAligned4(writer, current_frame_info.entropy.inter_mode_prob,
+                                       prev_frame_probs.inter_mode_prob);
+        if (current_frame_info.show_frame && !current_frame_info.is_key_frame) {
+            prev_frame_probs.inter_mode_prob = current_frame_info.entropy.inter_mode_prob;
+        }
+
+        if (current_frame_info.interp_filter == 4) {
+            // read_interp_filter_probs() in the spec
+            WriteProbabilityUpdate(writer, current_frame_info.entropy.switchable_interp_prob,
+                                   prev_frame_probs.switchable_interp_prob);
+            if (current_frame_info.show_frame && !current_frame_info.is_key_frame) {
+                prev_frame_probs.switchable_interp_prob =
+                    current_frame_info.entropy.switchable_interp_prob;
+            }
+        }
+
+        // read_is_inter_probs() in the spec
+        WriteProbabilityUpdate(writer, current_frame_info.entropy.intra_inter_prob,
+                               prev_frame_probs.intra_inter_prob);
+        if (current_frame_info.show_frame && !current_frame_info.is_key_frame) {
+            prev_frame_probs.intra_inter_prob = current_frame_info.entropy.intra_inter_prob;
+        }
+        // frame_reference_mode() in the spec
+        if ((current_frame_info.ref_frame_sign_bias[1] & 1) !=
+                (current_frame_info.ref_frame_sign_bias[2] & 1) ||
+            (current_frame_info.ref_frame_sign_bias[1] & 1) !=
+                (current_frame_info.ref_frame_sign_bias[3] & 1)) {
+            if (current_frame_info.reference_mode >= 1) {
+                writer.Write(1, 1);
+                writer.Write(current_frame_info.reference_mode == 2);
+            } else {
+                writer.Write(0, 1);
+            }
+        }
+
+        // frame_reference_mode_probs() in the spec
+        if (current_frame_info.reference_mode == 2) {
+            WriteProbabilityUpdate(writer, current_frame_info.entropy.comp_inter_prob,
+                                   prev_frame_probs.comp_inter_prob);
+            if (current_frame_info.show_frame && !current_frame_info.is_key_frame) {
+                prev_frame_probs.comp_inter_prob = current_frame_info.entropy.comp_inter_prob;
+            }
+        }
+
+        if (current_frame_info.reference_mode != 1) {
+            WriteProbabilityUpdate(writer, current_frame_info.entropy.single_ref_prob,
+                                   prev_frame_probs.single_ref_prob);
+            if (current_frame_info.show_frame && !current_frame_info.is_key_frame) {
+                prev_frame_probs.single_ref_prob = current_frame_info.entropy.single_ref_prob;
+            }
+        }
+
+        if (current_frame_info.reference_mode != 0) {
+            WriteProbabilityUpdate(writer, current_frame_info.entropy.comp_ref_prob,
+                                   prev_frame_probs.comp_ref_prob);
+            if (current_frame_info.show_frame && !current_frame_info.is_key_frame) {
+                prev_frame_probs.comp_ref_prob = current_frame_info.entropy.comp_ref_prob;
+            }
+        }
+
+        // read_y_mode_probs
+        for (std::size_t index = 0; index < current_frame_info.entropy.y_mode_prob.size();
+             ++index) {
+            WriteProbabilityUpdate(writer, current_frame_info.entropy.y_mode_prob[index],
+                                   prev_frame_probs.y_mode_prob[index]);
+        }
+        if (current_frame_info.show_frame && !current_frame_info.is_key_frame) {
+            prev_frame_probs.y_mode_prob = current_frame_info.entropy.y_mode_prob;
+        }
+        // read_partition_probs
+        WriteProbabilityUpdateAligned4(writer, current_frame_info.entropy.partition_prob,
+                                       prev_frame_probs.partition_prob);
+        if (current_frame_info.show_frame && !current_frame_info.is_key_frame) {
+            prev_frame_probs.partition_prob = current_frame_info.entropy.partition_prob;
+        }
+
+        // mv_probs
+        for (s32 i = 0; i < 3; i++) {
+            WriteMvProbabilityUpdate(writer, current_frame_info.entropy.joints[i],
+                                     prev_frame_probs.joints[i]);
+        }
+        if (current_frame_info.show_frame && !current_frame_info.is_key_frame) {
+            prev_frame_probs.joints = current_frame_info.entropy.joints;
+        }
+
+        for (s32 i = 0; i < 2; i++) {
+            WriteMvProbabilityUpdate(writer, current_frame_info.entropy.sign[i],
+                                     prev_frame_probs.sign[i]);
+
+            for (s32 j = 0; j < 10; j++) {
+                const int index = i * 10 + j;
+
+                WriteMvProbabilityUpdate(writer, current_frame_info.entropy.classes[index],
+                                         prev_frame_probs.classes[index]);
+            }
+
+            WriteMvProbabilityUpdate(writer, current_frame_info.entropy.class_0[i],
+                                     prev_frame_probs.class_0[i]);
+
+            for (s32 j = 0; j < 10; j++) {
+                const int index = i * 10 + j;
+
+                WriteMvProbabilityUpdate(writer, current_frame_info.entropy.prob_bits[index],
+                                         prev_frame_probs.prob_bits[index]);
+            }
+        }
+
+        for (s32 i = 0; i < 2; i++) {
+            for (s32 j = 0; j < 2; j++) {
+                for (s32 k = 0; k < 3; k++) {
+                    const int index = i * 2 * 3 + j * 3 + k;
+
+                    WriteMvProbabilityUpdate(writer, current_frame_info.entropy.class_0_fr[index],
+                                             prev_frame_probs.class_0_fr[index]);
+                }
+            }
+
+            for (s32 j = 0; j < 3; j++) {
+                const int index = i * 3 + j;
+
+                WriteMvProbabilityUpdate(writer, current_frame_info.entropy.fr[index],
+                                         prev_frame_probs.fr[index]);
+            }
+        }
+
+        if (current_frame_info.allow_high_precision_mv) {
+            for (s32 index = 0; index < 2; index++) {
+                WriteMvProbabilityUpdate(writer, current_frame_info.entropy.class_0_hp[index],
+                                         prev_frame_probs.class_0_hp[index]);
+                WriteMvProbabilityUpdate(writer, current_frame_info.entropy.high_precision[index],
+                                         prev_frame_probs.high_precision[index]);
+            }
+        }
+
+        // save previous probs
+        if (current_frame_info.show_frame && !current_frame_info.is_key_frame) {
+            prev_frame_probs.sign = current_frame_info.entropy.sign;
+            prev_frame_probs.classes = current_frame_info.entropy.classes;
+            prev_frame_probs.class_0 = current_frame_info.entropy.class_0;
+            prev_frame_probs.prob_bits = current_frame_info.entropy.prob_bits;
+            prev_frame_probs.class_0_fr = current_frame_info.entropy.class_0_fr;
+            prev_frame_probs.fr = current_frame_info.entropy.fr;
+            prev_frame_probs.class_0_hp = current_frame_info.entropy.class_0_hp;
+            prev_frame_probs.high_precision = current_frame_info.entropy.high_precision;
+        }
+    }
+
+    writer.End();
+    return writer.GetBuffer();
+}
+
+VpxBitStreamWriter VP9::ComposeUncompressedHeader() {
+    VpxBitStreamWriter uncomp_writer{};
+
+    uncomp_writer.WriteU(2, 2);                                      // Frame marker.
+    uncomp_writer.WriteU(0, 2);                                      // Profile.
+    uncomp_writer.WriteBit(false);                                   // Show existing frame.
+    uncomp_writer.WriteBit(!current_frame_info.is_key_frame);        // is key frame?
+    uncomp_writer.WriteBit(current_frame_info.show_frame);           // show frame?
+    uncomp_writer.WriteBit(current_frame_info.error_resilient_mode); // error reslience
+
+    if (current_frame_info.is_key_frame) {
+        uncomp_writer.WriteU(frame_sync_code, 24);
+        uncomp_writer.WriteU(0, 3); // Color space.
+        uncomp_writer.WriteU(0, 1); // Color range.
+        uncomp_writer.WriteU(current_frame_info.frame_size.width - 1, 16);
+        uncomp_writer.WriteU(current_frame_info.frame_size.height - 1, 16);
+        uncomp_writer.WriteBit(false); // Render and frame size different.
+
+        // Reset context
+        prev_frame_probs = default_probs;
+        swap_next_golden = false;
+        loop_filter_ref_deltas.fill(0);
+        loop_filter_mode_deltas.fill(0);
+
+        // allow frames offsets to stabilize before checking for golden frames
+        grace_period = 4;
+
+        // On key frames, all frame slots are set to the current frame,
+        // so the value of the selected slot doesn't really matter.
+        frame_ctxs.fill({current_frame_number, false, default_probs});
+
+        // intra only, meaning the frame can be recreated with no other references
+        current_frame_info.intra_only = true;
+
+    } else {
+
+        if (!current_frame_info.show_frame) {
+            uncomp_writer.WriteBit(current_frame_info.intra_only);
+            if (!current_frame_info.last_frame_was_key) {
+                swap_next_golden = !swap_next_golden;
+            }
+        } else {
+            current_frame_info.intra_only = false;
+        }
+        if (!current_frame_info.error_resilient_mode) {
+            uncomp_writer.WriteU(0, 2); // Reset frame context.
+        }
+
+        // Last, Golden, Altref frames
+        std::array<s32, 3> ref_frame_index{0, 1, 2};
+
+        // Set when next frame is hidden
+        // altref and golden references are swapped
+        if (swap_next_golden) {
+            ref_frame_index = std::array<s32, 3>{0, 2, 1};
+        }
+
+        // update Last Frame
+        u64 refresh_frame_flags = 1;
+
+        // golden frame may refresh, determined if the next golden frame offset is changed
+        bool golden_refresh = false;
+        if (grace_period <= 0) {
+            for (s32 index = 1; index < 3; ++index) {
+                if (current_frame_info.frame_offsets[index] !=
+                    next_frame.info.frame_offsets[index]) {
+                    current_frame_info.refresh_frame[index] = true;
+                    golden_refresh = true;
+                    grace_period = 3;
+                }
+            }
+        }
+
+        if (current_frame_info.show_frame &&
+            (!next_frame.info.show_frame || next_frame.info.is_key_frame)) {
+            // Update golden frame
+            refresh_frame_flags = swap_next_golden ? 2 : 4;
+        }
+
+        if (!current_frame_info.show_frame) {
+            // Update altref
+            refresh_frame_flags = swap_next_golden ? 2 : 4;
+        } else if (golden_refresh) {
+            refresh_frame_flags = 3;
+        }
+
+        if (current_frame_info.intra_only) {
+            uncomp_writer.WriteU(frame_sync_code, 24);
+            uncomp_writer.WriteU(static_cast<s32>(refresh_frame_flags), 8);
+            uncomp_writer.WriteU(current_frame_info.frame_size.width - 1, 16);
+            uncomp_writer.WriteU(current_frame_info.frame_size.height - 1, 16);
+            uncomp_writer.WriteBit(false); // Render and frame size different.
+        } else {
+            uncomp_writer.WriteU(static_cast<s32>(refresh_frame_flags), 8);
+
+            for (s32 index = 1; index < 4; index++) {
+                uncomp_writer.WriteU(ref_frame_index[index - 1], 3);
+                uncomp_writer.WriteU(current_frame_info.ref_frame_sign_bias[index], 1);
+            }
+
+            uncomp_writer.WriteBit(true);  // Frame size with refs.
+            uncomp_writer.WriteBit(false); // Render and frame size different.
+            uncomp_writer.WriteBit(current_frame_info.allow_high_precision_mv);
+            uncomp_writer.WriteBit(current_frame_info.interp_filter == 4);
+
+            if (current_frame_info.interp_filter != 4) {
+                uncomp_writer.WriteU(current_frame_info.interp_filter, 2);
+            }
+        }
+    }
+
+    if (!current_frame_info.error_resilient_mode) {
+        uncomp_writer.WriteBit(true); // Refresh frame context. where do i get this info from?
+        uncomp_writer.WriteBit(true); // Frame parallel decoding mode.
+    }
+
+    int frame_ctx_idx = 0;
+    if (!current_frame_info.show_frame) {
+        frame_ctx_idx = 1;
+    }
+
+    uncomp_writer.WriteU(frame_ctx_idx, 2); // Frame context index.
+    prev_frame_probs =
+        frame_ctxs[frame_ctx_idx].probs; // reference probabilities for compressed header
+    frame_ctxs[frame_ctx_idx] = {current_frame_number, false, current_frame_info.entropy};
+
+    uncomp_writer.WriteU(current_frame_info.first_level, 6);
+    uncomp_writer.WriteU(current_frame_info.sharpness_level, 3);
+    uncomp_writer.WriteBit(current_frame_info.mode_ref_delta_enabled);
+
+    if (current_frame_info.mode_ref_delta_enabled) {
+        // check if ref deltas are different, update accordingly
+        std::array<bool, 4> update_loop_filter_ref_deltas;
+        std::array<bool, 2> update_loop_filter_mode_deltas;
+
+        bool loop_filter_delta_update = false;
+
+        for (std::size_t index = 0; index < current_frame_info.ref_deltas.size(); index++) {
+            const s8 old_deltas = loop_filter_ref_deltas[index];
+            const s8 new_deltas = current_frame_info.ref_deltas[index];
+            const bool differing_delta = old_deltas != new_deltas;
+
+            update_loop_filter_ref_deltas[index] = differing_delta;
+            loop_filter_delta_update |= differing_delta;
+        }
+
+        for (std::size_t index = 0; index < current_frame_info.mode_deltas.size(); index++) {
+            const s8 old_deltas = loop_filter_mode_deltas[index];
+            const s8 new_deltas = current_frame_info.mode_deltas[index];
+            const bool differing_delta = old_deltas != new_deltas;
+
+            update_loop_filter_mode_deltas[index] = differing_delta;
+            loop_filter_delta_update |= differing_delta;
+        }
+
+        uncomp_writer.WriteBit(loop_filter_delta_update);
+
+        if (loop_filter_delta_update) {
+            for (std::size_t index = 0; index < current_frame_info.ref_deltas.size(); index++) {
+                uncomp_writer.WriteBit(update_loop_filter_ref_deltas[index]);
+
+                if (update_loop_filter_ref_deltas[index]) {
+                    uncomp_writer.WriteS(current_frame_info.ref_deltas[index], 6);
+                }
+            }
+
+            for (std::size_t index = 0; index < current_frame_info.mode_deltas.size(); index++) {
+                uncomp_writer.WriteBit(update_loop_filter_mode_deltas[index]);
+
+                if (update_loop_filter_mode_deltas[index]) {
+                    uncomp_writer.WriteS(current_frame_info.mode_deltas[index], 6);
+                }
+            }
+            // save new deltas
+            loop_filter_ref_deltas = current_frame_info.ref_deltas;
+            loop_filter_mode_deltas = current_frame_info.mode_deltas;
+        }
+    }
+
+    uncomp_writer.WriteU(current_frame_info.base_q_index, 8);
+
+    uncomp_writer.WriteDeltaQ(current_frame_info.y_dc_delta_q);
+    uncomp_writer.WriteDeltaQ(current_frame_info.uv_dc_delta_q);
+    uncomp_writer.WriteDeltaQ(current_frame_info.uv_ac_delta_q);
+
+    uncomp_writer.WriteBit(false); // Segmentation enabled (TODO).
+
+    const s32 min_tile_cols_log2 = CalcMinLog2TileCols(current_frame_info.frame_size.width);
+    const s32 max_tile_cols_log2 = CalcMaxLog2TileCols(current_frame_info.frame_size.width);
+
+    const s32 tile_cols_log2_diff = current_frame_info.log2_tile_cols - min_tile_cols_log2;
+    const s32 tile_cols_log2_inc_mask = (1 << tile_cols_log2_diff) - 1;
+
+    // If it's less than the maximum, we need to add an extra 0 on the bitstream
+    // to indicate that it should stop reading.
+    if (current_frame_info.log2_tile_cols < max_tile_cols_log2) {
+        uncomp_writer.WriteU(tile_cols_log2_inc_mask << 1, tile_cols_log2_diff + 1);
+    } else {
+        uncomp_writer.WriteU(tile_cols_log2_inc_mask, tile_cols_log2_diff);
+    }
+
+    const bool tile_rows_log2_is_nonzero = current_frame_info.log2_tile_rows != 0;
+
+    uncomp_writer.WriteBit(tile_rows_log2_is_nonzero);
+
+    if (tile_rows_log2_is_nonzero) {
+        uncomp_writer.WriteBit(current_frame_info.log2_tile_rows > 1);
+    }
+
+    return uncomp_writer;
+}
+
+std::vector<u8>& VP9::ComposeFrameHeader(NvdecCommon::NvdecRegisters& state) {
+    std::vector<u8> bitstream;
+    {
+        Vp9FrameContainer curr_frame = GetCurrentFrame(state);
+        current_frame_info = curr_frame.info;
+        bitstream = std::move(curr_frame.bit_stream);
+    }
+
+    // The uncompressed header routine sets PrevProb parameters needed for the compressed header
+    auto uncomp_writer = ComposeUncompressedHeader();
+    std::vector<u8> compressed_header = ComposeCompressedHeader();
+
+    uncomp_writer.WriteU(static_cast<s32>(compressed_header.size()), 16);
+    uncomp_writer.Flush();
+    std::vector<u8> uncompressed_header = uncomp_writer.GetByteArray();
+
+    // Write headers and frame to buffer
+    frame.resize(uncompressed_header.size() + compressed_header.size() + bitstream.size());
+    std::memcpy(frame.data(), uncompressed_header.data(), uncompressed_header.size());
+    std::memcpy(frame.data() + uncompressed_header.size(), compressed_header.data(),
+                compressed_header.size());
+    std::memcpy(frame.data() + uncompressed_header.size() + compressed_header.size(),
+                bitstream.data(), bitstream.size());
+
+    // keep track of frame number
+    current_frame_number++;
+    grace_period--;
+
+    // don't display hidden frames
+    hidden = !current_frame_info.show_frame;
+    return frame;
+}
+
+VpxRangeEncoder::VpxRangeEncoder() {
+    Write(false);
+}
+
+VpxRangeEncoder::~VpxRangeEncoder() = default;
+
+void VpxRangeEncoder::Write(s32 value, s32 value_size) {
+    for (s32 bit = value_size - 1; bit >= 0; bit--) {
+        Write(((value >> bit) & 1) != 0);
+    }
+}
+
+void VpxRangeEncoder::Write(bool bit) {
+    Write(bit, half_probability);
+}
+
+void VpxRangeEncoder::Write(bool bit, s32 probability) {
+    u32 local_range = range;
+    const u32 split = 1 + (((local_range - 1) * static_cast<u32>(probability)) >> 8);
+    local_range = split;
+
+    if (bit) {
+        low_value += split;
+        local_range = range - split;
+    }
+
+    s32 shift = norm_lut[local_range];
+    local_range <<= shift;
+    count += shift;
+
+    if (count >= 0) {
+        const s32 offset = shift - count;
+
+        if (((low_value << (offset - 1)) >> 31) != 0) {
+            const s32 current_pos = static_cast<s32>(base_stream.GetPosition());
+            base_stream.Seek(-1, Common::SeekOrigin::FromCurrentPos);
+            while (PeekByte() == 0xff) {
+                base_stream.WriteByte(0);
+
+                base_stream.Seek(-2, Common::SeekOrigin::FromCurrentPos);
+            }
+            base_stream.WriteByte(static_cast<u8>((PeekByte() + 1)));
+            base_stream.Seek(current_pos, Common::SeekOrigin::SetOrigin);
+        }
+        base_stream.WriteByte(static_cast<u8>((low_value >> (24 - offset))));
+
+        low_value <<= offset;
+        shift = count;
+        low_value &= 0xffffff;
+        count -= 8;
+    }
+
+    low_value <<= shift;
+    range = local_range;
+}
+
+void VpxRangeEncoder::End() {
+    for (std::size_t index = 0; index < 32; ++index) {
+        Write(false);
+    }
+}
+
+u8 VpxRangeEncoder::PeekByte() {
+    const u8 value = base_stream.ReadByte();
+    base_stream.Seek(-1, Common::SeekOrigin::FromCurrentPos);
+
+    return value;
+}
+
+VpxBitStreamWriter::VpxBitStreamWriter() = default;
+
+VpxBitStreamWriter::~VpxBitStreamWriter() = default;
+
+void VpxBitStreamWriter::WriteU(u32 value, u32 value_size) {
+    WriteBits(value, value_size);
+}
+
+void VpxBitStreamWriter::WriteS(s32 value, u32 value_size) {
+    const bool sign = value < 0;
+    if (sign) {
+        value = -value;
+    }
+
+    WriteBits(static_cast<u32>(value << 1) | (sign ? 1 : 0), value_size + 1);
+}
+
+void VpxBitStreamWriter::WriteDeltaQ(u32 value) {
+    const bool delta_coded = value != 0;
+    WriteBit(delta_coded);
+
+    if (delta_coded) {
+        WriteBits(value, 4);
+    }
+}
+
+void VpxBitStreamWriter::WriteBits(u32 value, u32 bit_count) {
+    s32 value_pos = 0;
+    s32 remaining = bit_count;
+
+    while (remaining > 0) {
+        s32 copy_size = remaining;
+
+        const s32 free = GetFreeBufferBits();
+
+        if (copy_size > free) {
+            copy_size = free;
+        }
+
+        const s32 mask = (1 << copy_size) - 1;
+
+        const s32 src_shift = (bit_count - value_pos) - copy_size;
+        const s32 dst_shift = (buffer_size - buffer_pos) - copy_size;
+
+        buffer |= ((value >> src_shift) & mask) << dst_shift;
+
+        value_pos += copy_size;
+        buffer_pos += copy_size;
+        remaining -= copy_size;
+    }
+}
+
+void VpxBitStreamWriter::WriteBit(bool state) {
+    WriteBits(state ? 1 : 0, 1);
+}
+
+s32 VpxBitStreamWriter::GetFreeBufferBits() {
+    if (buffer_pos == buffer_size) {
+        Flush();
+    }
+
+    return buffer_size - buffer_pos;
+}
+
+void VpxBitStreamWriter::Flush() {
+    if (buffer_pos == 0) {
+        return;
+    }
+    byte_array.push_back(static_cast<u8>(buffer));
+    buffer = 0;
+    buffer_pos = 0;
+}
+
+std::vector<u8>& VpxBitStreamWriter::GetByteArray() {
+    return byte_array;
+}
+
+const std::vector<u8>& VpxBitStreamWriter::GetByteArray() const {
+    return byte_array;
+}
+
+} // namespace Tegra::Decoder
diff --git a/src/video_core/command_classes/codecs/vp9.h b/src/video_core/command_classes/codecs/vp9.h
new file mode 100644
index 000000000..dc52ddbde
--- /dev/null
+++ b/src/video_core/command_classes/codecs/vp9.h
@@ -0,0 +1,188 @@
+// Copyright 2020 yuzu Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#pragma once
+
+#include <array>
+#include <vector>
+
+#include "common/common_types.h"
+#include "common/stream.h"
+#include "video_core/command_classes/codecs/vp9_types.h"
+#include "video_core/command_classes/nvdec_common.h"
+
+namespace Tegra {
+class GPU;
+enum class FrameType { KeyFrame = 0, InterFrame = 1 };
+namespace Decoder {
+
+/// The VpxRangeEncoder, and VpxBitStreamWriter classes are used to compose the
+/// VP9 header bitstreams.
+
+class VpxRangeEncoder {
+public:
+    VpxRangeEncoder();
+    ~VpxRangeEncoder();
+
+    /// Writes the rightmost value_size bits from value into the stream
+    void Write(s32 value, s32 value_size);
+
+    /// Writes a single bit with half probability
+    void Write(bool bit);
+
+    /// Writes a bit to the base_stream encoded with probability
+    void Write(bool bit, s32 probability);
+
+    /// Signal the end of the bitstream
+    void End();
+
+    std::vector<u8>& GetBuffer() {
+        return base_stream.GetBuffer();
+    }
+
+    const std::vector<u8>& GetBuffer() const {
+        return base_stream.GetBuffer();
+    }
+
+private:
+    u8 PeekByte();
+    Common::Stream base_stream{};
+    u32 low_value{};
+    u32 range{0xff};
+    s32 count{-24};
+    s32 half_probability{128};
+};
+
+class VpxBitStreamWriter {
+public:
+    VpxBitStreamWriter();
+    ~VpxBitStreamWriter();
+
+    /// Write an unsigned integer value
+    void WriteU(u32 value, u32 value_size);
+
+    /// Write a signed integer value
+    void WriteS(s32 value, u32 value_size);
+
+    /// Based on 6.2.10 of VP9 Spec, writes a delta coded value
+    void WriteDeltaQ(u32 value);
+
+    /// Write a single bit.
+    void WriteBit(bool state);
+
+    /// Pushes current buffer into buffer_array, resets buffer
+    void Flush();
+
+    /// Returns byte_array
+    std::vector<u8>& GetByteArray();
+
+    /// Returns const byte_array
+    const std::vector<u8>& GetByteArray() const;
+
+private:
+    /// Write bit_count bits from value into buffer
+    void WriteBits(u32 value, u32 bit_count);
+
+    /// Gets next available position in buffer, invokes Flush() if buffer is full
+    s32 GetFreeBufferBits();
+
+    s32 buffer_size{8};
+
+    s32 buffer{};
+    s32 buffer_pos{};
+    std::vector<u8> byte_array;
+};
+
+class VP9 {
+public:
+    explicit VP9(GPU& gpu);
+    ~VP9();
+
+    /// Composes the VP9 frame from the GPU state information. Based on the official VP9 spec
+    /// documentation
+    std::vector<u8>& ComposeFrameHeader(NvdecCommon::NvdecRegisters& state);
+
+    /// Returns true if the most recent frame was a hidden frame.
+    bool WasFrameHidden() const {
+        return hidden;
+    }
+
+private:
+    /// Generates compressed header probability updates in the bitstream writer
+    template <typename T, std::size_t N>
+    void WriteProbabilityUpdate(VpxRangeEncoder& writer, const std::array<T, N>& new_prob,
+                                const std::array<T, N>& old_prob);
+
+    /// Generates compressed header probability updates in the bitstream writer
+    /// If probs are not equal, WriteProbabilityDelta is invoked
+    void WriteProbabilityUpdate(VpxRangeEncoder& writer, u8 new_prob, u8 old_prob);
+
+    /// Generates compressed header probability deltas in the bitstream writer
+    void WriteProbabilityDelta(VpxRangeEncoder& writer, u8 new_prob, u8 old_prob);
+
+    /// Adjusts old_prob depending on new_prob. Based on section 6.3.5 of VP9 Specification
+    s32 RemapProbability(s32 new_prob, s32 old_prob);
+
+    /// Recenters probability. Based on section 6.3.6 of VP9 Specification
+    s32 RecenterNonNeg(s32 new_prob, s32 old_prob);
+
+    /// Inverse of 6.3.4 Decode term subexp
+    void EncodeTermSubExp(VpxRangeEncoder& writer, s32 value);
+
+    /// Writes if the value is less than the test value
+    bool WriteLessThan(VpxRangeEncoder& writer, s32 value, s32 test);
+
+    /// Writes probability updates for the Coef probabilities
+    void WriteCoefProbabilityUpdate(VpxRangeEncoder& writer, s32 tx_mode,
+                                    const std::array<u8, 2304>& new_prob,
+                                    const std::array<u8, 2304>& old_prob);
+
+    /// Write probabilities for 4-byte aligned structures
+    template <typename T, std::size_t N>
+    void WriteProbabilityUpdateAligned4(VpxRangeEncoder& writer, const std::array<T, N>& new_prob,
+                                        const std::array<T, N>& old_prob);
+
+    /// Write motion vector probability updates. 6.3.17 in the spec
+    void WriteMvProbabilityUpdate(VpxRangeEncoder& writer, u8 new_prob, u8 old_prob);
+
+    /// 6.2.14 Tile size calculation
+    s32 CalcMinLog2TileCols(s32 frame_width);
+    s32 CalcMaxLog2TileCols(s32 frame_width);
+
+    /// Returns VP9 information from NVDEC provided offset and size
+    Vp9PictureInfo GetVp9PictureInfo(const NvdecCommon::NvdecRegisters& state);
+
+    /// Read and convert NVDEC provided entropy probs to Vp9EntropyProbs struct
+    void InsertEntropy(u64 offset, Vp9EntropyProbs& dst);
+
+    /// Returns frame to be decoded after buffering
+    Vp9FrameContainer GetCurrentFrame(const NvdecCommon::NvdecRegisters& state);
+
+    /// Use NVDEC providied information to compose the headers for the current frame
+    std::vector<u8> ComposeCompressedHeader();
+    VpxBitStreamWriter ComposeUncompressedHeader();
+
+    GPU& gpu;
+    std::vector<u8> frame;
+
+    std::array<s8, 4> loop_filter_ref_deltas{};
+    std::array<s8, 2> loop_filter_mode_deltas{};
+
+    bool hidden;
+    s64 current_frame_number = -2; // since we buffer 2 frames
+    s32 grace_period = 6;          // frame offsets need to stabilize
+    std::array<FrameContexts, 4> frame_ctxs{};
+    Vp9FrameContainer next_frame{};
+    Vp9FrameContainer next_next_frame{};
+    bool swap_next_golden{};
+
+    Vp9PictureInfo current_frame_info{};
+    Vp9EntropyProbs prev_frame_probs{};
+
+    s32 diff_update_probability = 252;
+    s32 frame_sync_code = 0x498342;
+};
+
+} // namespace Decoder
+} // namespace Tegra
diff --git a/src/video_core/command_classes/codecs/vp9_types.h b/src/video_core/command_classes/codecs/vp9_types.h
new file mode 100644
index 000000000..a50acf6e8
--- /dev/null
+++ b/src/video_core/command_classes/codecs/vp9_types.h
@@ -0,0 +1,367 @@
+// Copyright 2020 yuzu Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#pragma once
+
+#include <array>
+#include <cstring>
+#include <vector>
+#include "common/common_funcs.h"
+#include "common/common_types.h"
+
+namespace Tegra {
+class GPU;
+
+namespace Decoder {
+struct Vp9FrameDimensions {
+    s16 width{};
+    s16 height{};
+    s16 luma_pitch{};
+    s16 chroma_pitch{};
+};
+static_assert(sizeof(Vp9FrameDimensions) == 0x8, "Vp9 Vp9FrameDimensions is an invalid size");
+
+enum FrameFlags : u32 {
+    IsKeyFrame = 1 << 0,
+    LastFrameIsKeyFrame = 1 << 1,
+    FrameSizeChanged = 1 << 2,
+    ErrorResilientMode = 1 << 3,
+    LastShowFrame = 1 << 4,
+    IntraOnly = 1 << 5,
+};
+
+enum class MvJointType {
+    MvJointZero = 0,   /* Zero vector */
+    MvJointHnzvz = 1,  /* Vert zero, hor nonzero */
+    MvJointHzvnz = 2,  /* Hor zero, vert nonzero */
+    MvJointHnzvnz = 3, /* Both components nonzero */
+};
+enum class MvClassType {
+    MvClass0 = 0,   /* (0, 2]     integer pel */
+    MvClass1 = 1,   /* (2, 4]     integer pel */
+    MvClass2 = 2,   /* (4, 8]     integer pel */
+    MvClass3 = 3,   /* (8, 16]    integer pel */
+    MvClass4 = 4,   /* (16, 32]   integer pel */
+    MvClass5 = 5,   /* (32, 64]   integer pel */
+    MvClass6 = 6,   /* (64, 128]  integer pel */
+    MvClass7 = 7,   /* (128, 256] integer pel */
+    MvClass8 = 8,   /* (256, 512] integer pel */
+    MvClass9 = 9,   /* (512, 1024] integer pel */
+    MvClass10 = 10, /* (1024,2048] integer pel */
+};
+
+enum class BlockSize {
+    Block4x4 = 0,
+    Block4x8 = 1,
+    Block8x4 = 2,
+    Block8x8 = 3,
+    Block8x16 = 4,
+    Block16x8 = 5,
+    Block16x16 = 6,
+    Block16x32 = 7,
+    Block32x16 = 8,
+    Block32x32 = 9,
+    Block32x64 = 10,
+    Block64x32 = 11,
+    Block64x64 = 12,
+    BlockSizes = 13,
+    BlockInvalid = BlockSizes
+};
+
+enum class PredictionMode {
+    DcPred = 0,   // Average of above and left pixels
+    VPred = 1,    // Vertical
+    HPred = 2,    // Horizontal
+    D45Pred = 3,  // Directional 45  deg = round(arctan(1 / 1) * 180 / pi)
+    D135Pred = 4, // Directional 135 deg = 180 - 45
+    D117Pred = 5, // Directional 117 deg = 180 - 63
+    D153Pred = 6, // Directional 153 deg = 180 - 27
+    D207Pred = 7, // Directional 207 deg = 180 + 27
+    D63Pred = 8,  // Directional 63  deg = round(arctan(2 / 1) * 180 / pi)
+    TmPred = 9,   // True-motion
+    NearestMv = 10,
+    NearMv = 11,
+    ZeroMv = 12,
+    NewMv = 13,
+    MbModeCount = 14
+};
+
+enum class TxSize {
+    Tx4x4 = 0,   // 4x4 transform
+    Tx8x8 = 1,   // 8x8 transform
+    Tx16x16 = 2, // 16x16 transform
+    Tx32x32 = 3, // 32x32 transform
+    TxSizes = 4
+};
+
+enum class TxMode {
+    Only4X4 = 0,      // Only 4x4 transform used
+    Allow8X8 = 1,     // Allow block transform size up to 8x8
+    Allow16X16 = 2,   // Allow block transform size up to 16x16
+    Allow32X32 = 3,   // Allow block transform size up to 32x32
+    TxModeSelect = 4, // Transform specified for each block
+    TxModes = 5
+};
+
+enum class reference_mode {
+    SingleReference = 0,
+    CompoundReference = 1,
+    ReferenceModeSelect = 2,
+    ReferenceModes = 3
+};
+
+struct Segmentation {
+    u8 enabled{};
+    u8 update_map{};
+    u8 temporal_update{};
+    u8 abs_delta{};
+    std::array<u32, 8> feature_mask{};
+    std::array<std::array<s16, 4>, 8> feature_data{};
+};
+static_assert(sizeof(Segmentation) == 0x64, "Segmentation is an invalid size");
+
+struct LoopFilter {
+    u8 mode_ref_delta_enabled{};
+    std::array<s8, 4> ref_deltas{};
+    std::array<s8, 2> mode_deltas{};
+};
+static_assert(sizeof(LoopFilter) == 0x7, "LoopFilter is an invalid size");
+
+struct Vp9EntropyProbs {
+    std::array<u8, 36> y_mode_prob{};
+    std::array<u8, 64> partition_prob{};
+    std::array<u8, 2304> coef_probs{};
+    std::array<u8, 8> switchable_interp_prob{};
+    std::array<u8, 28> inter_mode_prob{};
+    std::array<u8, 4> intra_inter_prob{};
+    std::array<u8, 5> comp_inter_prob{};
+    std::array<u8, 10> single_ref_prob{};
+    std::array<u8, 5> comp_ref_prob{};
+    std::array<u8, 6> tx_32x32_prob{};
+    std::array<u8, 4> tx_16x16_prob{};
+    std::array<u8, 2> tx_8x8_prob{};
+    std::array<u8, 3> skip_probs{};
+    std::array<u8, 3> joints{};
+    std::array<u8, 2> sign{};
+    std::array<u8, 20> classes{};
+    std::array<u8, 2> class_0{};
+    std::array<u8, 20> prob_bits{};
+    std::array<u8, 12> class_0_fr{};
+    std::array<u8, 6> fr{};
+    std::array<u8, 2> class_0_hp{};
+    std::array<u8, 2> high_precision{};
+};
+static_assert(sizeof(Vp9EntropyProbs) == 0x9F4, "Vp9EntropyProbs is an invalid size");
+
+struct Vp9PictureInfo {
+    bool is_key_frame{};
+    bool intra_only{};
+    bool last_frame_was_key{};
+    bool frame_size_changed{};
+    bool error_resilient_mode{};
+    bool last_frame_shown{};
+    bool show_frame{};
+    std::array<s8, 4> ref_frame_sign_bias{};
+    s32 base_q_index{};
+    s32 y_dc_delta_q{};
+    s32 uv_dc_delta_q{};
+    s32 uv_ac_delta_q{};
+    bool lossless{};
+    s32 transform_mode{};
+    bool allow_high_precision_mv{};
+    s32 interp_filter{};
+    s32 reference_mode{};
+    s8 comp_fixed_ref{};
+    std::array<s8, 2> comp_var_ref{};
+    s32 log2_tile_cols{};
+    s32 log2_tile_rows{};
+    bool segment_enabled{};
+    bool segment_map_update{};
+    bool segment_map_temporal_update{};
+    s32 segment_abs_delta{};
+    std::array<u32, 8> segment_feature_enable{};
+    std::array<std::array<s16, 4>, 8> segment_feature_data{};
+    bool mode_ref_delta_enabled{};
+    bool use_prev_in_find_mv_refs{};
+    std::array<s8, 4> ref_deltas{};
+    std::array<s8, 2> mode_deltas{};
+    Vp9EntropyProbs entropy{};
+    Vp9FrameDimensions frame_size{};
+    u8 first_level{};
+    u8 sharpness_level{};
+    u32 bitstream_size{};
+    std::array<u64, 4> frame_offsets{};
+    std::array<bool, 4> refresh_frame{};
+};
+
+struct Vp9FrameContainer {
+    Vp9PictureInfo info{};
+    std::vector<u8> bit_stream;
+};
+
+struct PictureInfo {
+    INSERT_PADDING_WORDS(12);
+    u32 bitstream_size{};
+    INSERT_PADDING_WORDS(5);
+    Vp9FrameDimensions last_frame_size{};
+    Vp9FrameDimensions golden_frame_size{};
+    Vp9FrameDimensions alt_frame_size{};
+    Vp9FrameDimensions current_frame_size{};
+    u32 vp9_flags{};
+    std::array<s8, 4> ref_frame_sign_bias{};
+    u8 first_level{};
+    u8 sharpness_level{};
+    u8 base_q_index{};
+    u8 y_dc_delta_q{};
+    u8 uv_ac_delta_q{};
+    u8 uv_dc_delta_q{};
+    u8 lossless{};
+    u8 tx_mode{};
+    u8 allow_high_precision_mv{};
+    u8 interp_filter{};
+    u8 reference_mode{};
+    s8 comp_fixed_ref{};
+    std::array<s8, 2> comp_var_ref{};
+    u8 log2_tile_cols{};
+    u8 log2_tile_rows{};
+    Segmentation segmentation{};
+    LoopFilter loop_filter{};
+    INSERT_PADDING_BYTES(5);
+    u32 surface_params{};
+    INSERT_PADDING_WORDS(3);
+
+    Vp9PictureInfo Convert() const {
+
+        return Vp9PictureInfo{
+            .is_key_frame = (vp9_flags & FrameFlags::IsKeyFrame) != 0,
+            .intra_only = (vp9_flags & FrameFlags::IntraOnly) != 0,
+            .last_frame_was_key = (vp9_flags & FrameFlags::LastFrameIsKeyFrame) != 0,
+            .frame_size_changed = (vp9_flags & FrameFlags::FrameSizeChanged) != 0,
+            .error_resilient_mode = (vp9_flags & FrameFlags::ErrorResilientMode) != 0,
+            .last_frame_shown = (vp9_flags & FrameFlags::LastShowFrame) != 0,
+            .ref_frame_sign_bias = ref_frame_sign_bias,
+            .base_q_index = base_q_index,
+            .y_dc_delta_q = y_dc_delta_q,
+            .uv_dc_delta_q = uv_dc_delta_q,
+            .uv_ac_delta_q = uv_ac_delta_q,
+            .lossless = lossless != 0,
+            .transform_mode = tx_mode,
+            .allow_high_precision_mv = allow_high_precision_mv != 0,
+            .interp_filter = interp_filter,
+            .reference_mode = reference_mode,
+            .comp_fixed_ref = comp_fixed_ref,
+            .comp_var_ref = comp_var_ref,
+            .log2_tile_cols = log2_tile_cols,
+            .log2_tile_rows = log2_tile_rows,
+            .segment_enabled = segmentation.enabled != 0,
+            .segment_map_update = segmentation.update_map != 0,
+            .segment_map_temporal_update = segmentation.temporal_update != 0,
+            .segment_abs_delta = segmentation.abs_delta,
+            .segment_feature_enable = segmentation.feature_mask,
+            .segment_feature_data = segmentation.feature_data,
+            .mode_ref_delta_enabled = loop_filter.mode_ref_delta_enabled != 0,
+            .use_prev_in_find_mv_refs = !(vp9_flags == (FrameFlags::ErrorResilientMode)) &&
+                                        !(vp9_flags == (FrameFlags::FrameSizeChanged)) &&
+                                        !(vp9_flags == (FrameFlags::IntraOnly)) &&
+                                        (vp9_flags == (FrameFlags::LastShowFrame)) &&
+                                        !(vp9_flags == (FrameFlags::LastFrameIsKeyFrame)),
+            .ref_deltas = loop_filter.ref_deltas,
+            .mode_deltas = loop_filter.mode_deltas,
+            .frame_size = current_frame_size,
+            .first_level = first_level,
+            .sharpness_level = sharpness_level,
+            .bitstream_size = bitstream_size,
+        };
+    }
+};
+static_assert(sizeof(PictureInfo) == 0x100, "PictureInfo is an invalid size");
+
+struct EntropyProbs {
+    INSERT_PADDING_BYTES(1024);
+    std::array<std::array<u8, 4>, 7> inter_mode_prob{};
+    std::array<u8, 4> intra_inter_prob{};
+    INSERT_PADDING_BYTES(80);
+    std::array<std::array<u8, 1>, 2> tx_8x8_prob{};
+    std::array<std::array<u8, 2>, 2> tx_16x16_prob{};
+    std::array<std::array<u8, 3>, 2> tx_32x32_prob{};
+    std::array<u8, 4> y_mode_prob_e8{};
+    std::array<std::array<u8, 8>, 4> y_mode_prob_e0e7{};
+    INSERT_PADDING_BYTES(64);
+    std::array<std::array<u8, 4>, 16> partition_prob{};
+    INSERT_PADDING_BYTES(10);
+    std::array<std::array<u8, 2>, 4> switchable_interp_prob{};
+    std::array<u8, 5> comp_inter_prob{};
+    std::array<u8, 4> skip_probs{};
+    std::array<u8, 3> joints{};
+    std::array<u8, 2> sign{};
+    std::array<std::array<u8, 1>, 2> class_0{};
+    std::array<std::array<u8, 3>, 2> fr{};
+    std::array<u8, 2> class_0_hp{};
+    std::array<u8, 2> high_precision{};
+    std::array<std::array<u8, 10>, 2> classes{};
+    std::array<std::array<std::array<u8, 3>, 2>, 2> class_0_fr{};
+    std::array<std::array<u8, 10>, 2> pred_bits{};
+    std::array<std::array<u8, 2>, 5> single_ref_prob{};
+    std::array<u8, 5> comp_ref_prob{};
+    INSERT_PADDING_BYTES(17);
+    std::array<std::array<std::array<std::array<std::array<std::array<u8, 4>, 6>, 6>, 2>, 2>, 4>
+        coef_probs{};
+
+    void Convert(Vp9EntropyProbs& fc) {
+        std::memcpy(fc.inter_mode_prob.data(), inter_mode_prob.data(), fc.inter_mode_prob.size());
+
+        std::memcpy(fc.intra_inter_prob.data(), intra_inter_prob.data(),
+                    fc.intra_inter_prob.size());
+
+        std::memcpy(fc.tx_8x8_prob.data(), tx_8x8_prob.data(), fc.tx_8x8_prob.size());
+        std::memcpy(fc.tx_16x16_prob.data(), tx_16x16_prob.data(), fc.tx_16x16_prob.size());
+        std::memcpy(fc.tx_32x32_prob.data(), tx_32x32_prob.data(), fc.tx_32x32_prob.size());
+
+        for (s32 i = 0; i < 4; i++) {
+            for (s32 j = 0; j < 9; j++) {
+                fc.y_mode_prob[j + 9 * i] = j < 8 ? y_mode_prob_e0e7[i][j] : y_mode_prob_e8[i];
+            }
+        }
+
+        std::memcpy(fc.partition_prob.data(), partition_prob.data(), fc.partition_prob.size());
+
+        std::memcpy(fc.switchable_interp_prob.data(), switchable_interp_prob.data(),
+                    fc.switchable_interp_prob.size());
+        std::memcpy(fc.comp_inter_prob.data(), comp_inter_prob.data(), fc.comp_inter_prob.size());
+        std::memcpy(fc.skip_probs.data(), skip_probs.data(), fc.skip_probs.size());
+
+        std::memcpy(fc.joints.data(), joints.data(), fc.joints.size());
+
+        std::memcpy(fc.sign.data(), sign.data(), fc.sign.size());
+        std::memcpy(fc.class_0.data(), class_0.data(), fc.class_0.size());
+        std::memcpy(fc.fr.data(), fr.data(), fc.fr.size());
+        std::memcpy(fc.class_0_hp.data(), class_0_hp.data(), fc.class_0_hp.size());
+        std::memcpy(fc.high_precision.data(), high_precision.data(), fc.high_precision.size());
+        std::memcpy(fc.classes.data(), classes.data(), fc.classes.size());
+        std::memcpy(fc.class_0_fr.data(), class_0_fr.data(), fc.class_0_fr.size());
+        std::memcpy(fc.prob_bits.data(), pred_bits.data(), fc.prob_bits.size());
+        std::memcpy(fc.single_ref_prob.data(), single_ref_prob.data(), fc.single_ref_prob.size());
+        std::memcpy(fc.comp_ref_prob.data(), comp_ref_prob.data(), fc.comp_ref_prob.size());
+
+        std::memcpy(fc.coef_probs.data(), coef_probs.data(), fc.coef_probs.size());
+    }
+};
+static_assert(sizeof(EntropyProbs) == 0xEA0, "EntropyProbs is an invalid size");
+
+enum class Ref { Last, Golden, AltRef };
+
+struct RefPoolElement {
+    s64 frame{};
+    Ref ref{};
+    bool refresh{};
+};
+
+struct FrameContexts {
+    s64 from{};
+    bool adapted{};
+    Vp9EntropyProbs probs{};
+};
+
+}; // namespace Decoder
+}; // namespace Tegra
diff --git a/src/video_core/command_classes/host1x.cpp b/src/video_core/command_classes/host1x.cpp
new file mode 100644
index 000000000..c4dd4881a
--- /dev/null
+++ b/src/video_core/command_classes/host1x.cpp
@@ -0,0 +1,39 @@
+// Copyright 2020 yuzu Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#include "common/assert.h"
+#include "video_core/command_classes/host1x.h"
+#include "video_core/gpu.h"
+
+Tegra::Host1x::Host1x(GPU& gpu_) : gpu(gpu_) {}
+
+Tegra::Host1x::~Host1x() = default;
+
+void Tegra::Host1x::StateWrite(u32 offset, u32 arguments) {
+    u8* const state_offset = reinterpret_cast<u8*>(&state) + offset * sizeof(u32);
+    std::memcpy(state_offset, &arguments, sizeof(u32));
+}
+
+void Tegra::Host1x::ProcessMethod(Method method, const std::vector<u32>& arguments) {
+    StateWrite(static_cast<u32>(method), arguments[0]);
+    switch (method) {
+    case Method::WaitSyncpt:
+        Execute(arguments[0]);
+        break;
+    case Method::LoadSyncptPayload32:
+        syncpoint_value = arguments[0];
+        break;
+    case Method::WaitSyncpt32:
+        Execute(arguments[0]);
+        break;
+    default:
+        UNIMPLEMENTED_MSG("Host1x method 0x{:X}", static_cast<u32>(method));
+        break;
+    }
+}
+
+void Tegra::Host1x::Execute(u32 data) {
+    // This method waits on a valid syncpoint.
+    // TODO: Implement when proper Async is in place
+}
diff --git a/src/video_core/command_classes/host1x.h b/src/video_core/command_classes/host1x.h
new file mode 100644
index 000000000..013eaa0c1
--- /dev/null
+++ b/src/video_core/command_classes/host1x.h
@@ -0,0 +1,78 @@
+// Copyright 2020 yuzu Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#pragma once
+
+#include <vector>
+#include "common/common_funcs.h"
+#include "common/common_types.h"
+
+namespace Tegra {
+class GPU;
+class Nvdec;
+
+class Host1x {
+public:
+    struct Host1xClassRegisters {
+        u32 incr_syncpt{};
+        u32 incr_syncpt_ctrl{};
+        u32 incr_syncpt_error{};
+        INSERT_PADDING_WORDS(5);
+        u32 wait_syncpt{};
+        u32 wait_syncpt_base{};
+        u32 wait_syncpt_incr{};
+        u32 load_syncpt_base{};
+        u32 incr_syncpt_base{};
+        u32 clear{};
+        u32 wait{};
+        u32 wait_with_interrupt{};
+        u32 delay_use{};
+        u32 tick_count_high{};
+        u32 tick_count_low{};
+        u32 tick_ctrl{};
+        INSERT_PADDING_WORDS(23);
+        u32 ind_ctrl{};
+        u32 ind_off2{};
+        u32 ind_off{};
+        std::array<u32, 31> ind_data{};
+        INSERT_PADDING_WORDS(1);
+        u32 load_syncpoint_payload32{};
+        u32 stall_ctrl{};
+        u32 wait_syncpt32{};
+        u32 wait_syncpt_base32{};
+        u32 load_syncpt_base32{};
+        u32 incr_syncpt_base32{};
+        u32 stall_count_high{};
+        u32 stall_count_low{};
+        u32 xref_ctrl{};
+        u32 channel_xref_high{};
+        u32 channel_xref_low{};
+    };
+    static_assert(sizeof(Host1xClassRegisters) == 0x164, "Host1xClassRegisters is an invalid size");
+
+    enum class Method : u32 {
+        WaitSyncpt = offsetof(Host1xClassRegisters, wait_syncpt) / 4,
+        LoadSyncptPayload32 = offsetof(Host1xClassRegisters, load_syncpoint_payload32) / 4,
+        WaitSyncpt32 = offsetof(Host1xClassRegisters, wait_syncpt32) / 4,
+    };
+
+    explicit Host1x(GPU& gpu);
+    ~Host1x();
+
+    /// Writes the method into the state, Invoke Execute() if encountered
+    void ProcessMethod(Method method, const std::vector<u32>& arguments);
+
+private:
+    /// For Host1x, execute is waiting on a syncpoint previously written into the state
+    void Execute(u32 data);
+
+    /// Write argument into the provided offset
+    void StateWrite(u32 offset, u32 arguments);
+
+    u32 syncpoint_value{};
+    Host1xClassRegisters state{};
+    GPU& gpu;
+};
+
+} // namespace Tegra
diff --git a/src/video_core/command_classes/nvdec.cpp b/src/video_core/command_classes/nvdec.cpp
new file mode 100644
index 000000000..8ca7a7b06
--- /dev/null
+++ b/src/video_core/command_classes/nvdec.cpp
@@ -0,0 +1,52 @@
+// Copyright 2020 yuzu Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#include "common/assert.h"
+#include "video_core/command_classes/nvdec.h"
+#include "video_core/gpu.h"
+
+namespace Tegra {
+
+Nvdec::Nvdec(GPU& gpu_) : gpu(gpu_), codec(std::make_unique<Codec>(gpu)) {}
+
+Nvdec::~Nvdec() = default;
+
+void Nvdec::ProcessMethod(Method method, const std::vector<u32>& arguments) {
+    if (method == Method::SetVideoCodec) {
+        codec->StateWrite(static_cast<u32>(method), arguments[0]);
+    } else {
+        codec->StateWrite(static_cast<u32>(method), static_cast<u64>(arguments[0]) << 8);
+    }
+
+    switch (method) {
+    case Method::SetVideoCodec:
+        codec->SetTargetCodec(static_cast<NvdecCommon::VideoCodec>(arguments[0]));
+        break;
+    case Method::Execute:
+        Execute();
+        break;
+    }
+}
+
+AVFrame* Nvdec::GetFrame() {
+    return codec->GetCurrentFrame();
+}
+
+const AVFrame* Nvdec::GetFrame() const {
+    return codec->GetCurrentFrame();
+}
+
+void Nvdec::Execute() {
+    switch (codec->GetCurrentCodec()) {
+    case NvdecCommon::VideoCodec::H264:
+    case NvdecCommon::VideoCodec::Vp9:
+        codec->Decode();
+        break;
+    default:
+        UNIMPLEMENTED_MSG("Unknown codec {}", static_cast<u32>(codec->GetCurrentCodec()));
+        break;
+    }
+}
+
+} // namespace Tegra
diff --git a/src/video_core/command_classes/nvdec.h b/src/video_core/command_classes/nvdec.h
new file mode 100644
index 000000000..af14f9857
--- /dev/null
+++ b/src/video_core/command_classes/nvdec.h
@@ -0,0 +1,39 @@
+// Copyright 2020 yuzu Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#pragma once
+
+#include <memory>
+#include <vector>
+#include "common/common_types.h"
+#include "video_core/command_classes/codecs/codec.h"
+
+namespace Tegra {
+class GPU;
+
+class Nvdec {
+public:
+    enum class Method : u32 {
+        SetVideoCodec = 0x80,
+        Execute = 0xc0,
+    };
+
+    explicit Nvdec(GPU& gpu);
+    ~Nvdec();
+
+    /// Writes the method into the state, Invoke Execute() if encountered
+    void ProcessMethod(Method method, const std::vector<u32>& arguments);
+
+    /// Return most recently decoded frame
+    AVFrame* GetFrame();
+    const AVFrame* GetFrame() const;
+
+private:
+    /// Invoke codec to decode a frame
+    void Execute();
+
+    GPU& gpu;
+    std::unique_ptr<Codec> codec;
+};
+} // namespace Tegra
diff --git a/src/video_core/command_classes/nvdec_common.h b/src/video_core/command_classes/nvdec_common.h
new file mode 100644
index 000000000..01b5e086d
--- /dev/null
+++ b/src/video_core/command_classes/nvdec_common.h
@@ -0,0 +1,48 @@
+// Copyright 2020 yuzu Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#pragma once
+
+#include "common/common_funcs.h"
+#include "common/common_types.h"
+
+namespace Tegra::NvdecCommon {
+
+struct NvdecRegisters {
+    INSERT_PADDING_WORDS(256);
+    u64 set_codec_id{};
+    INSERT_PADDING_WORDS(254);
+    u64 set_platform_id{};
+    u64 picture_info_offset{};
+    u64 frame_bitstream_offset{};
+    u64 frame_number{};
+    u64 h264_slice_data_offsets{};
+    u64 h264_mv_dump_offset{};
+    INSERT_PADDING_WORDS(6);
+    u64 frame_stats_offset{};
+    u64 h264_last_surface_luma_offset{};
+    u64 h264_last_surface_chroma_offset{};
+    std::array<u64, 17> surface_luma_offset{};
+    std::array<u64, 17> surface_chroma_offset{};
+    INSERT_PADDING_WORDS(132);
+    u64 vp9_entropy_probs_offset{};
+    u64 vp9_backward_updates_offset{};
+    u64 vp9_last_frame_segmap_offset{};
+    u64 vp9_curr_frame_segmap_offset{};
+    INSERT_PADDING_WORDS(2);
+    u64 vp9_last_frame_mvs_offset{};
+    u64 vp9_curr_frame_mvs_offset{};
+    INSERT_PADDING_WORDS(2);
+};
+static_assert(sizeof(NvdecRegisters) == (0xBC0), "NvdecRegisters is incorrect size");
+
+enum class VideoCodec : u32 {
+    None = 0x0,
+    H264 = 0x3,
+    Vp8 = 0x5,
+    H265 = 0x7,
+    Vp9 = 0x9,
+};
+
+} // namespace Tegra::NvdecCommon
diff --git a/src/video_core/command_classes/sync_manager.cpp b/src/video_core/command_classes/sync_manager.cpp
new file mode 100644
index 000000000..19dc9e0ab
--- /dev/null
+++ b/src/video_core/command_classes/sync_manager.cpp
@@ -0,0 +1,60 @@
+// MIT License
+//
+// Copyright (c) Ryujinx Team and Contributors
+//
+// 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.
+//
+
+#include <algorithm>
+#include "sync_manager.h"
+#include "video_core/gpu.h"
+
+namespace Tegra {
+SyncptIncrManager::SyncptIncrManager(GPU& gpu_) : gpu(gpu_) {}
+SyncptIncrManager::~SyncptIncrManager() = default;
+
+void SyncptIncrManager::Increment(u32 id) {
+    increments.emplace_back(0, 0, id, true);
+    IncrementAllDone();
+}
+
+u32 SyncptIncrManager::IncrementWhenDone(u32 class_id, u32 id) {
+    const u32 handle = current_id++;
+    increments.emplace_back(handle, class_id, id);
+    return handle;
+}
+
+void SyncptIncrManager::SignalDone(u32 handle) {
+    const auto done_incr =
+        std::find_if(increments.begin(), increments.end(),
+                     [handle](const SyncptIncr& incr) { return incr.id == handle; });
+    if (done_incr != increments.cend()) {
+        done_incr->complete = true;
+    }
+    IncrementAllDone();
+}
+
+void SyncptIncrManager::IncrementAllDone() {
+    std::size_t done_count = 0;
+    for (; done_count < increments.size(); ++done_count) {
+        if (!increments[done_count].complete) {
+            break;
+        }
+        gpu.IncrementSyncPoint(increments[done_count].syncpt_id);
+    }
+    increments.erase(increments.begin(), increments.begin() + done_count);
+}
+} // namespace Tegra
diff --git a/src/video_core/command_classes/sync_manager.h b/src/video_core/command_classes/sync_manager.h
new file mode 100644
index 000000000..2c321ec58
--- /dev/null
+++ b/src/video_core/command_classes/sync_manager.h
@@ -0,0 +1,64 @@
+// MIT License
+//
+// Copyright (c) Ryujinx Team and Contributors
+//
+// 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.
+//
+
+#pragma once
+
+#include <mutex>
+#include <vector>
+#include "common/common_types.h"
+
+namespace Tegra {
+class GPU;
+struct SyncptIncr {
+    u32 id;
+    u32 class_id;
+    u32 syncpt_id;
+    bool complete;
+
+    SyncptIncr(u32 id_, u32 class_id_, u32 syncpt_id_, bool done = false)
+        : id(id_), class_id(class_id_), syncpt_id(syncpt_id_), complete(done) {}
+};
+
+class SyncptIncrManager {
+public:
+    explicit SyncptIncrManager(GPU& gpu);
+    ~SyncptIncrManager();
+
+    /// Add syncpoint id and increment all
+    void Increment(u32 id);
+
+    /// Returns a handle to increment later
+    u32 IncrementWhenDone(u32 class_id, u32 id);
+
+    /// IncrememntAllDone, including handle
+    void SignalDone(u32 handle);
+
+    /// Increment all sequential pending increments that are already done.
+    void IncrementAllDone();
+
+private:
+    std::vector<SyncptIncr> increments;
+    std::mutex increment_lock;
+    u32 current_id{};
+
+    GPU& gpu;
+};
+
+} // namespace Tegra
diff --git a/src/video_core/command_classes/vic.cpp b/src/video_core/command_classes/vic.cpp
new file mode 100644
index 000000000..5b52da277
--- /dev/null
+++ b/src/video_core/command_classes/vic.cpp
@@ -0,0 +1,180 @@
+// Copyright 2020 yuzu Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#include <array>
+#include "common/assert.h"
+#include "video_core/command_classes/nvdec.h"
+#include "video_core/command_classes/vic.h"
+#include "video_core/engines/maxwell_3d.h"
+#include "video_core/gpu.h"
+#include "video_core/memory_manager.h"
+#include "video_core/texture_cache/surface_params.h"
+
+extern "C" {
+#include <libswscale/swscale.h>
+}
+
+namespace Tegra {
+
+Vic::Vic(GPU& gpu_, std::shared_ptr<Nvdec> nvdec_processor_)
+    : gpu(gpu_), nvdec_processor(std::move(nvdec_processor_)) {}
+Vic::~Vic() = default;
+
+void Vic::VicStateWrite(u32 offset, u32 arguments) {
+    u8* const state_offset = reinterpret_cast<u8*>(&vic_state) + offset * sizeof(u32);
+    std::memcpy(state_offset, &arguments, sizeof(u32));
+}
+
+void Vic::ProcessMethod(Method method, const std::vector<u32>& arguments) {
+    LOG_DEBUG(HW_GPU, "Vic method 0x{:X}", static_cast<u32>(method));
+    VicStateWrite(static_cast<u32>(method), arguments[0]);
+    const u64 arg = static_cast<u64>(arguments[0]) << 8;
+    switch (method) {
+    case Method::Execute:
+        Execute();
+        break;
+    case Method::SetConfigStructOffset:
+        config_struct_address = arg;
+        break;
+    case Method::SetOutputSurfaceLumaOffset:
+        output_surface_luma_address = arg;
+        break;
+    case Method::SetOutputSurfaceChromaUOffset:
+        output_surface_chroma_u_address = arg;
+        break;
+    case Method::SetOutputSurfaceChromaVOffset:
+        output_surface_chroma_v_address = arg;
+        break;
+    default:
+        break;
+    }
+}
+
+void Vic::Execute() {
+    if (output_surface_luma_address == 0) {
+        LOG_ERROR(Service_NVDRV, "VIC Luma address not set. Recieved 0x{:X}",
+                  vic_state.output_surface.luma_offset);
+        return;
+    }
+    const VicConfig config{gpu.MemoryManager().Read<u64>(config_struct_address + 0x20)};
+    const VideoPixelFormat pixel_format =
+        static_cast<VideoPixelFormat>(config.pixel_format.Value());
+    switch (pixel_format) {
+    case VideoPixelFormat::BGRA8:
+    case VideoPixelFormat::RGBA8: {
+        LOG_TRACE(Service_NVDRV, "Writing RGB Frame");
+        const auto* frame = nvdec_processor->GetFrame();
+
+        if (!frame || frame->width == 0 || frame->height == 0) {
+            return;
+        }
+        if (scaler_ctx == nullptr || frame->width != scaler_width ||
+            frame->height != scaler_height) {
+            const AVPixelFormat target_format =
+                (pixel_format == VideoPixelFormat::RGBA8) ? AV_PIX_FMT_RGBA : AV_PIX_FMT_BGRA;
+
+            sws_freeContext(scaler_ctx);
+            scaler_ctx = nullptr;
+
+            // FFmpeg returns all frames in YUV420, convert it into expected format
+            scaler_ctx =
+                sws_getContext(frame->width, frame->height, AV_PIX_FMT_YUV420P, frame->width,
+                               frame->height, target_format, 0, nullptr, nullptr, nullptr);
+
+            scaler_width = frame->width;
+            scaler_height = frame->height;
+        }
+        // Get Converted frame
+        const std::size_t linear_size = frame->width * frame->height * 4;
+
+        using AVMallocPtr = std::unique_ptr<u8, decltype(&av_free)>;
+        AVMallocPtr converted_frame_buffer{static_cast<u8*>(av_malloc(linear_size)), av_free};
+
+        const int converted_stride{frame->width * 4};
+        u8* const converted_frame_buf_addr{converted_frame_buffer.get()};
+
+        sws_scale(scaler_ctx, frame->data, frame->linesize, 0, frame->height,
+                  &converted_frame_buf_addr, &converted_stride);
+
+        const u32 blk_kind = static_cast<u32>(config.block_linear_kind);
+        if (blk_kind != 0) {
+            // swizzle pitch linear to block linear
+            const u32 block_height = static_cast<u32>(config.block_linear_height_log2);
+            const auto size = Tegra::Texture::CalculateSize(true, 4, frame->width, frame->height, 1,
+                                                            block_height, 0);
+            std::vector<u8> swizzled_data(size);
+            Tegra::Texture::CopySwizzledData(frame->width, frame->height, 1, 4, 4,
+                                             swizzled_data.data(), converted_frame_buffer.get(),
+                                             false, block_height, 0, 1);
+
+            gpu.MemoryManager().WriteBlock(output_surface_luma_address, swizzled_data.data(), size);
+            gpu.Maxwell3D().OnMemoryWrite();
+        } else {
+            // send pitch linear frame
+            gpu.MemoryManager().WriteBlock(output_surface_luma_address, converted_frame_buf_addr,
+                                           linear_size);
+            gpu.Maxwell3D().OnMemoryWrite();
+        }
+        break;
+    }
+    case VideoPixelFormat::Yuv420: {
+        LOG_TRACE(Service_NVDRV, "Writing YUV420 Frame");
+
+        const auto* frame = nvdec_processor->GetFrame();
+
+        if (!frame || frame->width == 0 || frame->height == 0) {
+            return;
+        }
+
+        const std::size_t surface_width = config.surface_width_minus1 + 1;
+        const std::size_t surface_height = config.surface_height_minus1 + 1;
+        const std::size_t half_width = surface_width / 2;
+        const std::size_t half_height = config.surface_height_minus1 / 2;
+        const std::size_t aligned_width = (surface_width + 0xff) & ~0xff;
+
+        const auto* luma_ptr = frame->data[0];
+        const auto* chroma_b_ptr = frame->data[1];
+        const auto* chroma_r_ptr = frame->data[2];
+        const auto stride = frame->linesize[0];
+        const auto half_stride = frame->linesize[1];
+
+        std::vector<u8> luma_buffer(aligned_width * surface_height);
+        std::vector<u8> chroma_buffer(aligned_width * half_height);
+
+        // Populate luma buffer
+        for (std::size_t y = 0; y < surface_height - 1; ++y) {
+            std::size_t src = y * stride;
+            std::size_t dst = y * aligned_width;
+
+            std::size_t size = surface_width;
+
+            for (std::size_t offset = 0; offset < size; ++offset) {
+                luma_buffer[dst + offset] = luma_ptr[src + offset];
+            }
+        }
+        gpu.MemoryManager().WriteBlock(output_surface_luma_address, luma_buffer.data(),
+                                       luma_buffer.size());
+
+        // Populate chroma buffer from both channels with interleaving.
+        for (std::size_t y = 0; y < half_height; ++y) {
+            std::size_t src = y * half_stride;
+            std::size_t dst = y * aligned_width;
+
+            for (std::size_t x = 0; x < half_width; ++x) {
+                chroma_buffer[dst + x * 2] = chroma_b_ptr[src + x];
+                chroma_buffer[dst + x * 2 + 1] = chroma_r_ptr[src + x];
+            }
+        }
+        gpu.MemoryManager().WriteBlock(output_surface_chroma_u_address, chroma_buffer.data(),
+                                       chroma_buffer.size());
+        gpu.Maxwell3D().OnMemoryWrite();
+        break;
+    }
+    default:
+        UNIMPLEMENTED_MSG("Unknown video pixel format {}", config.pixel_format.Value());
+        break;
+    }
+}
+
+} // namespace Tegra
diff --git a/src/video_core/command_classes/vic.h b/src/video_core/command_classes/vic.h
new file mode 100644
index 000000000..8c4e284a1
--- /dev/null
+++ b/src/video_core/command_classes/vic.h
@@ -0,0 +1,110 @@
+// Copyright 2020 yuzu Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#pragma once
+
+#include <memory>
+#include <vector>
+#include "common/bit_field.h"
+#include "common/common_types.h"
+
+struct SwsContext;
+
+namespace Tegra {
+class GPU;
+class Nvdec;
+
+struct PlaneOffsets {
+    u32 luma_offset{};
+    u32 chroma_u_offset{};
+    u32 chroma_v_offset{};
+};
+
+struct VicRegisters {
+    INSERT_PADDING_WORDS(64);
+    u32 nop{};
+    INSERT_PADDING_WORDS(15);
+    u32 pm_trigger{};
+    INSERT_PADDING_WORDS(47);
+    u32 set_application_id{};
+    u32 set_watchdog_timer{};
+    INSERT_PADDING_WORDS(17);
+    u32 context_save_area{};
+    u32 context_switch{};
+    INSERT_PADDING_WORDS(43);
+    u32 execute{};
+    INSERT_PADDING_WORDS(63);
+    std::array<std::array<PlaneOffsets, 8>, 8> surfacex_slots{};
+    u32 picture_index{};
+    u32 control_params{};
+    u32 config_struct_offset{};
+    u32 filter_struct_offset{};
+    u32 palette_offset{};
+    u32 hist_offset{};
+    u32 context_id{};
+    u32 fce_ucode_size{};
+    PlaneOffsets output_surface{};
+    u32 fce_ucode_offset{};
+    INSERT_PADDING_WORDS(4);
+    std::array<u32, 8> slot_context_id{};
+    INSERT_PADDING_WORDS(16);
+};
+static_assert(sizeof(VicRegisters) == 0x7A0, "VicRegisters is an invalid size");
+
+class Vic {
+public:
+    enum class Method : u32 {
+        Execute = 0xc0,
+        SetControlParams = 0x1c1,
+        SetConfigStructOffset = 0x1c2,
+        SetOutputSurfaceLumaOffset = 0x1c8,
+        SetOutputSurfaceChromaUOffset = 0x1c9,
+        SetOutputSurfaceChromaVOffset = 0x1ca
+    };
+
+    explicit Vic(GPU& gpu, std::shared_ptr<Nvdec> nvdec_processor);
+    ~Vic();
+
+    /// Write to the device state.
+    void ProcessMethod(Method method, const std::vector<u32>& arguments);
+
+private:
+    void Execute();
+
+    void VicStateWrite(u32 offset, u32 arguments);
+    VicRegisters vic_state{};
+
+    enum class VideoPixelFormat : u64_le {
+        RGBA8 = 0x1f,
+        BGRA8 = 0x20,
+        Yuv420 = 0x44,
+    };
+
+    union VicConfig {
+        u64_le raw{};
+        BitField<0, 7, u64_le> pixel_format;
+        BitField<7, 2, u64_le> chroma_loc_horiz;
+        BitField<9, 2, u64_le> chroma_loc_vert;
+        BitField<11, 4, u64_le> block_linear_kind;
+        BitField<15, 4, u64_le> block_linear_height_log2;
+        BitField<19, 3, u64_le> reserved0;
+        BitField<22, 10, u64_le> reserved1;
+        BitField<32, 14, u64_le> surface_width_minus1;
+        BitField<46, 14, u64_le> surface_height_minus1;
+    };
+
+    GPU& gpu;
+    std::shared_ptr<Tegra::Nvdec> nvdec_processor;
+
+    GPUVAddr config_struct_address{};
+    GPUVAddr output_surface_luma_address{};
+    GPUVAddr output_surface_chroma_u_address{};
+    GPUVAddr output_surface_chroma_v_address{};
+
+    SwsContext* scaler_ctx{};
+    s32 scaler_width{};
+    s32 scaler_height{};
+};
+
+} // namespace Tegra
diff --git a/src/video_core/gpu.cpp b/src/video_core/gpu.cpp
index 4bb9256e9..171f78183 100644
--- a/src/video_core/gpu.cpp
+++ b/src/video_core/gpu.cpp
@@ -27,9 +27,10 @@ namespace Tegra {
 
 MICROPROFILE_DEFINE(GPU_wait, "GPU", "Wait for the GPU", MP_RGB(128, 128, 192));
 
-GPU::GPU(Core::System& system_, bool is_async_)
+GPU::GPU(Core::System& system_, bool is_async_, bool use_nvdec_)
     : system{system_}, memory_manager{std::make_unique<Tegra::MemoryManager>(system)},
       dma_pusher{std::make_unique<Tegra::DmaPusher>(system, *this)},
+      cdma_pusher{std::make_unique<Tegra::CDmaPusher>(*this)}, use_nvdec{use_nvdec_},
       maxwell_3d{std::make_unique<Engines::Maxwell3D>(system, *memory_manager)},
       fermi_2d{std::make_unique<Engines::Fermi2D>()},
       kepler_compute{std::make_unique<Engines::KeplerCompute>(system, *memory_manager)},
@@ -77,10 +78,18 @@ DmaPusher& GPU::DmaPusher() {
     return *dma_pusher;
 }
 
+Tegra::CDmaPusher& GPU::CDmaPusher() {
+    return *cdma_pusher;
+}
+
 const DmaPusher& GPU::DmaPusher() const {
     return *dma_pusher;
 }
 
+const Tegra::CDmaPusher& GPU::CDmaPusher() const {
+    return *cdma_pusher;
+}
+
 void GPU::WaitFence(u32 syncpoint_id, u32 value) {
     // Synced GPU, is always in sync
     if (!is_async) {
diff --git a/src/video_core/gpu.h b/src/video_core/gpu.h
index 2d15d1c6f..b8c613b11 100644
--- a/src/video_core/gpu.h
+++ b/src/video_core/gpu.h
@@ -13,6 +13,7 @@
 #include "common/common_types.h"
 #include "core/hle/service/nvdrv/nvdata.h"
 #include "core/hle/service/nvflinger/buffer_queue.h"
+#include "video_core/cdma_pusher.h"
 #include "video_core/dma_pusher.h"
 
 using CacheAddr = std::uintptr_t;
@@ -157,7 +158,7 @@ public:
               method_count(method_count) {}
     };
 
-    explicit GPU(Core::System& system, bool is_async);
+    explicit GPU(Core::System& system, bool is_async, bool use_nvdec);
     virtual ~GPU();
 
     /// Binds a renderer to the GPU.
@@ -209,6 +210,15 @@ public:
     /// Returns a reference to the GPU DMA pusher.
     Tegra::DmaPusher& DmaPusher();
 
+    /// Returns a const reference to the GPU DMA pusher.
+    const Tegra::DmaPusher& DmaPusher() const;
+
+    /// Returns a reference to the GPU CDMA pusher.
+    Tegra::CDmaPusher& CDmaPusher();
+
+    /// Returns a const reference to the GPU CDMA pusher.
+    const Tegra::CDmaPusher& CDmaPusher() const;
+
     VideoCore::RendererBase& Renderer() {
         return *renderer;
     }
@@ -249,8 +259,9 @@ public:
         return is_async;
     }
 
-    /// Returns a const reference to the GPU DMA pusher.
-    const Tegra::DmaPusher& DmaPusher() const;
+    bool UseNvdec() const {
+        return use_nvdec;
+    }
 
     struct Regs {
         static constexpr size_t NUM_REGS = 0x40;
@@ -311,6 +322,9 @@ public:
     /// Push GPU command entries to be processed
     virtual void PushGPUEntries(Tegra::CommandList&& entries) = 0;
 
+    /// Push GPU command buffer entries to be processed
+    virtual void PushCommandBuffer(Tegra::ChCommandHeaderList& entries) = 0;
+
     /// Swap buffers (render frame)
     virtual void SwapBuffers(const Tegra::FramebufferConfig* framebuffer) = 0;
 
@@ -349,7 +363,9 @@ protected:
     Core::System& system;
     std::unique_ptr<Tegra::MemoryManager> memory_manager;
     std::unique_ptr<Tegra::DmaPusher> dma_pusher;
+    std::unique_ptr<Tegra::CDmaPusher> cdma_pusher;
     std::unique_ptr<VideoCore::RendererBase> renderer;
+    const bool use_nvdec;
 
 private:
     /// Mapping of command subchannels to their bound engine ids
@@ -372,6 +388,7 @@ private:
     std::array<std::list<u32>, Service::Nvidia::MaxSyncPoints> syncpt_interrupts;
 
     std::mutex sync_mutex;
+    std::mutex device_mutex;
 
     std::condition_variable sync_cv;
 
diff --git a/src/video_core/gpu_asynch.cpp b/src/video_core/gpu_asynch.cpp
index 70a3d5738..a9baaf7ef 100644
--- a/src/video_core/gpu_asynch.cpp
+++ b/src/video_core/gpu_asynch.cpp
@@ -10,12 +10,13 @@
 
 namespace VideoCommon {
 
-GPUAsynch::GPUAsynch(Core::System& system) : GPU{system, true}, gpu_thread{system} {}
+GPUAsynch::GPUAsynch(Core::System& system, bool use_nvdec)
+    : GPU{system, true, use_nvdec}, gpu_thread{system} {}
 
 GPUAsynch::~GPUAsynch() = default;
 
 void GPUAsynch::Start() {
-    gpu_thread.StartThread(*renderer, renderer->Context(), *dma_pusher);
+    gpu_thread.StartThread(*renderer, renderer->Context(), *dma_pusher, *cdma_pusher);
     cpu_context = renderer->GetRenderWindow().CreateSharedContext();
     cpu_context->MakeCurrent();
 }
@@ -32,6 +33,27 @@ void GPUAsynch::PushGPUEntries(Tegra::CommandList&& entries) {
     gpu_thread.SubmitList(std::move(entries));
 }
 
+void GPUAsynch::PushCommandBuffer(Tegra::ChCommandHeaderList& entries) {
+    if (!use_nvdec) {
+        return;
+    }
+    // This condition fires when a video stream ends, clear all intermediary data
+    if (entries[0].raw == 0xDEADB33F) {
+        cdma_pusher.reset();
+        return;
+    }
+    if (!cdma_pusher) {
+        cdma_pusher = std::make_unique<Tegra::CDmaPusher>(*this);
+    }
+
+    // SubmitCommandBuffer would make the nvdec operations async, this is not currently working
+    // TODO(ameerj): RE proper async nvdec operation
+    // gpu_thread.SubmitCommandBuffer(std::move(entries));
+
+    cdma_pusher->Push(std::move(entries));
+    cdma_pusher->DispatchCalls();
+}
+
 void GPUAsynch::SwapBuffers(const Tegra::FramebufferConfig* framebuffer) {
     gpu_thread.SwapBuffers(framebuffer);
 }
diff --git a/src/video_core/gpu_asynch.h b/src/video_core/gpu_asynch.h
index f89c855a5..0c0872e73 100644
--- a/src/video_core/gpu_asynch.h
+++ b/src/video_core/gpu_asynch.h
@@ -20,13 +20,14 @@ namespace VideoCommon {
 /// Implementation of GPU interface that runs the GPU asynchronously
 class GPUAsynch final : public Tegra::GPU {
 public:
-    explicit GPUAsynch(Core::System& system);
+    explicit GPUAsynch(Core::System& system, bool use_nvdec);
     ~GPUAsynch() override;
 
     void Start() override;
     void ObtainContext() override;
     void ReleaseContext() override;
     void PushGPUEntries(Tegra::CommandList&& entries) override;
+    void PushCommandBuffer(Tegra::ChCommandHeaderList& entries) override;
     void SwapBuffers(const Tegra::FramebufferConfig* framebuffer) override;
     void FlushRegion(VAddr addr, u64 size) override;
     void InvalidateRegion(VAddr addr, u64 size) override;
diff --git a/src/video_core/gpu_synch.cpp b/src/video_core/gpu_synch.cpp
index 1ca47ddef..ecf7bbdf3 100644
--- a/src/video_core/gpu_synch.cpp
+++ b/src/video_core/gpu_synch.cpp
@@ -7,7 +7,7 @@
 
 namespace VideoCommon {
 
-GPUSynch::GPUSynch(Core::System& system) : GPU{system, false} {}
+GPUSynch::GPUSynch(Core::System& system, bool use_nvdec) : GPU{system, false, use_nvdec} {}
 
 GPUSynch::~GPUSynch() = default;
 
@@ -26,6 +26,22 @@ void GPUSynch::PushGPUEntries(Tegra::CommandList&& entries) {
     dma_pusher->DispatchCalls();
 }
 
+void GPUSynch::PushCommandBuffer(Tegra::ChCommandHeaderList& entries) {
+    if (!use_nvdec) {
+        return;
+    }
+    // This condition fires when a video stream ends, clears all intermediary data
+    if (entries[0].raw == 0xDEADB33F) {
+        cdma_pusher.reset();
+        return;
+    }
+    if (!cdma_pusher) {
+        cdma_pusher = std::make_unique<Tegra::CDmaPusher>(*this);
+    }
+    cdma_pusher->Push(std::move(entries));
+    cdma_pusher->DispatchCalls();
+}
+
 void GPUSynch::SwapBuffers(const Tegra::FramebufferConfig* framebuffer) {
     renderer->SwapBuffers(framebuffer);
 }
diff --git a/src/video_core/gpu_synch.h b/src/video_core/gpu_synch.h
index 297258cb1..9d778c71a 100644
--- a/src/video_core/gpu_synch.h
+++ b/src/video_core/gpu_synch.h
@@ -19,13 +19,14 @@ namespace VideoCommon {
 /// Implementation of GPU interface that runs the GPU synchronously
 class GPUSynch final : public Tegra::GPU {
 public:
-    explicit GPUSynch(Core::System& system);
+    explicit GPUSynch(Core::System& system, bool use_nvdec);
     ~GPUSynch() override;
 
     void Start() override;
     void ObtainContext() override;
     void ReleaseContext() override;
     void PushGPUEntries(Tegra::CommandList&& entries) override;
+    void PushCommandBuffer(Tegra::ChCommandHeaderList& entries) override;
     void SwapBuffers(const Tegra::FramebufferConfig* framebuffer) override;
     void FlushRegion(VAddr addr, u64 size) override;
     void InvalidateRegion(VAddr addr, u64 size) override;
diff --git a/src/video_core/gpu_thread.cpp b/src/video_core/gpu_thread.cpp
index bf761abf2..4b8f58283 100644
--- a/src/video_core/gpu_thread.cpp
+++ b/src/video_core/gpu_thread.cpp
@@ -18,7 +18,7 @@ namespace VideoCommon::GPUThread {
 /// Runs the GPU thread
 static void RunThread(Core::System& system, VideoCore::RendererBase& renderer,
                       Core::Frontend::GraphicsContext& context, Tegra::DmaPusher& dma_pusher,
-                      SynchState& state) {
+                      SynchState& state, Tegra::CDmaPusher& cdma_pusher) {
     std::string name = "yuzu:GPU";
     MicroProfileOnThreadCreate(name.c_str());
     Common::SetCurrentThreadName(name.c_str());
@@ -42,6 +42,10 @@ static void RunThread(Core::System& system, VideoCore::RendererBase& renderer,
         if (const auto submit_list = std::get_if<SubmitListCommand>(&next.data)) {
             dma_pusher.Push(std::move(submit_list->entries));
             dma_pusher.DispatchCalls();
+        } else if (const auto command_list = std::get_if<SubmitChCommandEntries>(&next.data)) {
+            // NVDEC
+            cdma_pusher.Push(std::move(command_list->entries));
+            cdma_pusher.DispatchCalls();
         } else if (const auto data = std::get_if<SwapBuffersCommand>(&next.data)) {
             renderer.SwapBuffers(data->framebuffer ? &*data->framebuffer : nullptr);
         } else if (std::holds_alternative<OnCommandListEndCommand>(next.data)) {
@@ -75,15 +79,19 @@ ThreadManager::~ThreadManager() {
 
 void ThreadManager::StartThread(VideoCore::RendererBase& renderer,
                                 Core::Frontend::GraphicsContext& context,
-                                Tegra::DmaPusher& dma_pusher) {
-    thread = std::thread{RunThread,         std::ref(system),     std::ref(renderer),
-                         std::ref(context), std::ref(dma_pusher), std::ref(state)};
+                                Tegra::DmaPusher& dma_pusher, Tegra::CDmaPusher& cdma_pusher) {
+    thread = std::thread(RunThread, std::ref(system), std::ref(renderer), std::ref(context),
+                         std::ref(dma_pusher), std::ref(state), std::ref(cdma_pusher));
 }
 
 void ThreadManager::SubmitList(Tegra::CommandList&& entries) {
     PushCommand(SubmitListCommand(std::move(entries)));
 }
 
+void ThreadManager::SubmitCommandBuffer(Tegra::ChCommandHeaderList&& entries) {
+    PushCommand(SubmitChCommandEntries(std::move(entries)));
+}
+
 void ThreadManager::SwapBuffers(const Tegra::FramebufferConfig* framebuffer) {
     PushCommand(SwapBuffersCommand(framebuffer ? std::make_optional(*framebuffer) : std::nullopt));
 }
diff --git a/src/video_core/gpu_thread.h b/src/video_core/gpu_thread.h
index 5a28335d6..32a34e3a7 100644
--- a/src/video_core/gpu_thread.h
+++ b/src/video_core/gpu_thread.h
@@ -37,6 +37,14 @@ struct SubmitListCommand final {
     Tegra::CommandList entries;
 };
 
+/// Command to signal to the GPU thread that a cdma command list is ready for processing
+struct SubmitChCommandEntries final {
+    explicit SubmitChCommandEntries(Tegra::ChCommandHeaderList&& entries)
+        : entries{std::move(entries)} {}
+
+    Tegra::ChCommandHeaderList entries;
+};
+
 /// Command to signal to the GPU thread that a swap buffers is pending
 struct SwapBuffersCommand final {
     explicit SwapBuffersCommand(std::optional<const Tegra::FramebufferConfig> framebuffer)
@@ -77,9 +85,9 @@ struct OnCommandListEndCommand final {};
 struct GPUTickCommand final {};
 
 using CommandData =
-    std::variant<EndProcessingCommand, SubmitListCommand, SwapBuffersCommand, FlushRegionCommand,
-                 InvalidateRegionCommand, FlushAndInvalidateRegionCommand, OnCommandListEndCommand,
-                 GPUTickCommand>;
+    std::variant<EndProcessingCommand, SubmitListCommand, SubmitChCommandEntries,
+                 SwapBuffersCommand, FlushRegionCommand, InvalidateRegionCommand,
+                 FlushAndInvalidateRegionCommand, OnCommandListEndCommand, GPUTickCommand>;
 
 struct CommandDataContainer {
     CommandDataContainer() = default;
@@ -109,11 +117,14 @@ public:
 
     /// Creates and starts the GPU thread.
     void StartThread(VideoCore::RendererBase& renderer, Core::Frontend::GraphicsContext& context,
-                     Tegra::DmaPusher& dma_pusher);
+                     Tegra::DmaPusher& dma_pusher, Tegra::CDmaPusher& cdma_pusher);
 
     /// Push GPU command entries to be processed
     void SubmitList(Tegra::CommandList&& entries);
 
+    /// Push GPU CDMA command buffer entries to be processed
+    void SubmitCommandBuffer(Tegra::ChCommandHeaderList&& entries);
+
     /// Swap buffers (render frame)
     void SwapBuffers(const Tegra::FramebufferConfig* framebuffer);
 
diff --git a/src/video_core/host_shaders/CMakeLists.txt b/src/video_core/host_shaders/CMakeLists.txt
index aa62363a7..c157724a9 100644
--- a/src/video_core/host_shaders/CMakeLists.txt
+++ b/src/video_core/host_shaders/CMakeLists.txt
@@ -1,23 +1,16 @@
-set(SHADER_FILES
+set(SHADER_SOURCES
     opengl_present.frag
     opengl_present.vert
 )
 
 set(SHADER_INCLUDE ${CMAKE_CURRENT_BINARY_DIR}/include)
-set(HOST_SHADERS_INCLUDE ${SHADER_INCLUDE} PARENT_SCOPE)
-
 set(SHADER_DIR ${SHADER_INCLUDE}/video_core/host_shaders)
-add_custom_command(
-    OUTPUT
-        ${SHADER_DIR}
-    COMMAND
-        ${CMAKE_COMMAND} -E make_directory ${SHADER_DIR}
-)
+set(HOST_SHADERS_INCLUDE ${SHADER_INCLUDE} PARENT_SCOPE)
 
 set(INPUT_FILE ${CMAKE_CURRENT_SOURCE_DIR}/source_shader.h.in)
 set(HEADER_GENERATOR ${CMAKE_CURRENT_SOURCE_DIR}/StringShaderHeader.cmake)
 
-foreach(FILENAME IN ITEMS ${SHADER_FILES})
+foreach(FILENAME IN ITEMS ${SHADER_SOURCES})
     string(REPLACE "." "_" SHADER_NAME ${FILENAME})
     set(SOURCE_FILE ${CMAKE_CURRENT_SOURCE_DIR}/${FILENAME})
     set(HEADER_FILE ${SHADER_DIR}/${SHADER_NAME}.h)
@@ -29,8 +22,8 @@ foreach(FILENAME IN ITEMS ${SHADER_FILES})
         MAIN_DEPENDENCY
             ${SOURCE_FILE}
         DEPENDS
-            ${HEADER_GENERATOR}
             ${INPUT_FILE}
+            # HEADER_GENERATOR should be included here but msbuild seems to assume it's always modified
     )
     set(SHADER_HEADERS ${SHADER_HEADERS} ${HEADER_FILE})
 endforeach()
@@ -39,5 +32,5 @@ add_custom_target(host_shaders
     DEPENDS
         ${SHADER_HEADERS}
     SOURCES
-        ${SHADER_FILES}
+        ${SHADER_SOURCES}
 )
diff --git a/src/video_core/host_shaders/StringShaderHeader.cmake b/src/video_core/host_shaders/StringShaderHeader.cmake
index 368bce0ed..c0fc49768 100644
--- a/src/video_core/host_shaders/StringShaderHeader.cmake
+++ b/src/video_core/host_shaders/StringShaderHeader.cmake
@@ -8,4 +8,6 @@ string(TOUPPER ${CONTENTS_NAME} CONTENTS_NAME)
 
 file(READ ${SOURCE_FILE} CONTENTS)
 
+get_filename_component(OUTPUT_DIR ${HEADER_FILE} DIRECTORY)
+make_directory(${OUTPUT_DIR})
 configure_file(${INPUT_FILE} ${HEADER_FILE} @ONLY)
diff --git a/src/video_core/memory_manager.cpp b/src/video_core/memory_manager.cpp
index 02cf53d15..6e70bd362 100644
--- a/src/video_core/memory_manager.cpp
+++ b/src/video_core/memory_manager.cpp
@@ -11,6 +11,7 @@
 #include "video_core/gpu.h"
 #include "video_core/memory_manager.h"
 #include "video_core/rasterizer_interface.h"
+#include "video_core/renderer_base.h"
 
 namespace Tegra {
 
@@ -44,6 +45,12 @@ GPUVAddr MemoryManager::MapAllocate(VAddr cpu_addr, std::size_t size, std::size_
     return Map(cpu_addr, *FindFreeRange(size, align), size);
 }
 
+GPUVAddr MemoryManager::MapAllocate32(VAddr cpu_addr, std::size_t size) {
+    const std::optional<GPUVAddr> gpu_addr = FindFreeRange(size, 1, true);
+    ASSERT(gpu_addr);
+    return Map(cpu_addr, *gpu_addr, size);
+}
+
 void MemoryManager::Unmap(GPUVAddr gpu_addr, std::size_t size) {
     if (!size) {
         return;
@@ -108,7 +115,8 @@ void MemoryManager::SetPageEntry(GPUVAddr gpu_addr, PageEntry page_entry, std::s
     page_table[PageEntryIndex(gpu_addr)] = page_entry;
 }
 
-std::optional<GPUVAddr> MemoryManager::FindFreeRange(std::size_t size, std::size_t align) const {
+std::optional<GPUVAddr> MemoryManager::FindFreeRange(std::size_t size, std::size_t align,
+                                                     bool start_32bit_address) const {
     if (!align) {
         align = page_size;
     } else {
@@ -116,7 +124,7 @@ std::optional<GPUVAddr> MemoryManager::FindFreeRange(std::size_t size, std::size
     }
 
     u64 available_size{};
-    GPUVAddr gpu_addr{address_space_start};
+    GPUVAddr gpu_addr{start_32bit_address ? address_space_start_low : address_space_start};
     while (gpu_addr + available_size < address_space_size) {
         if (GetPageEntry(gpu_addr + available_size).IsUnmapped()) {
             available_size += page_size;
diff --git a/src/video_core/memory_manager.h b/src/video_core/memory_manager.h
index 53c8d122a..c078193d9 100644
--- a/src/video_core/memory_manager.h
+++ b/src/video_core/memory_manager.h
@@ -116,6 +116,7 @@ public:
 
     [[nodiscard]] GPUVAddr Map(VAddr cpu_addr, GPUVAddr gpu_addr, std::size_t size);
     [[nodiscard]] GPUVAddr MapAllocate(VAddr cpu_addr, std::size_t size, std::size_t align);
+    [[nodiscard]] GPUVAddr MapAllocate32(VAddr cpu_addr, std::size_t size);
     [[nodiscard]] std::optional<GPUVAddr> AllocateFixed(GPUVAddr gpu_addr, std::size_t size);
     [[nodiscard]] GPUVAddr Allocate(std::size_t size, std::size_t align);
     void Unmap(GPUVAddr gpu_addr, std::size_t size);
@@ -124,7 +125,8 @@ private:
     [[nodiscard]] PageEntry GetPageEntry(GPUVAddr gpu_addr) const;
     void SetPageEntry(GPUVAddr gpu_addr, PageEntry page_entry, std::size_t size = page_size);
     GPUVAddr UpdateRange(GPUVAddr gpu_addr, PageEntry page_entry, std::size_t size);
-    [[nodiscard]] std::optional<GPUVAddr> FindFreeRange(std::size_t size, std::size_t align) const;
+    [[nodiscard]] std::optional<GPUVAddr> FindFreeRange(std::size_t size, std::size_t align,
+                                                        bool start_32bit_address = false) const;
 
     void TryLockPage(PageEntry page_entry, std::size_t size);
     void TryUnlockPage(PageEntry page_entry, std::size_t size);
@@ -135,6 +137,7 @@ private:
 
     static constexpr u64 address_space_size = 1ULL << 40;
     static constexpr u64 address_space_start = 1ULL << 32;
+    static constexpr u64 address_space_start_low = 1ULL << 16;
     static constexpr u64 page_bits{16};
     static constexpr u64 page_size{1 << page_bits};
     static constexpr u64 page_mask{page_size - 1};
diff --git a/src/video_core/renderer_opengl/gl_shader_disk_cache.cpp b/src/video_core/renderer_opengl/gl_shader_disk_cache.cpp
index 166ee34e1..70dd0c3c6 100644
--- a/src/video_core/renderer_opengl/gl_shader_disk_cache.cpp
+++ b/src/video_core/renderer_opengl/gl_shader_disk_cache.cpp
@@ -317,8 +317,7 @@ std::optional<std::vector<ShaderDiskCachePrecompiled>> ShaderDiskCacheOpenGL::Lo
             return std::nullopt;
         }
     }
-
-    return std::move(entries);
+    return entries;
 }
 
 void ShaderDiskCacheOpenGL::InvalidateTransferable() {
diff --git a/src/video_core/renderer_vulkan/wrapper.cpp b/src/video_core/renderer_vulkan/wrapper.cpp
index c034558a3..4e83303d8 100644
--- a/src/video_core/renderer_vulkan/wrapper.cpp
+++ b/src/video_core/renderer_vulkan/wrapper.cpp
@@ -844,7 +844,7 @@ std::optional<std::vector<VkExtensionProperties>> EnumerateInstanceExtensionProp
         VK_SUCCESS) {
         return std::nullopt;
     }
-    return std::move(properties);
+    return properties;
 }
 
 std::optional<std::vector<VkLayerProperties>> EnumerateInstanceLayerProperties(
diff --git a/src/video_core/video_core.cpp b/src/video_core/video_core.cpp
index a14df06a3..dd5cee4a1 100644
--- a/src/video_core/video_core.cpp
+++ b/src/video_core/video_core.cpp
@@ -44,10 +44,11 @@ namespace VideoCore {
 
 std::unique_ptr<Tegra::GPU> CreateGPU(Core::Frontend::EmuWindow& emu_window, Core::System& system) {
     std::unique_ptr<Tegra::GPU> gpu;
+    const bool use_nvdec = Settings::values.use_nvdec_emulation.GetValue();
     if (Settings::values.use_asynchronous_gpu_emulation.GetValue()) {
-        gpu = std::make_unique<VideoCommon::GPUAsynch>(system);
+        gpu = std::make_unique<VideoCommon::GPUAsynch>(system, use_nvdec);
     } else {
-        gpu = std::make_unique<VideoCommon::GPUSynch>(system);
+        gpu = std::make_unique<VideoCommon::GPUSynch>(system, use_nvdec);
     }
 
     auto context = emu_window.CreateSharedContext();
diff --git a/src/web_service/CMakeLists.txt b/src/web_service/CMakeLists.txt
index 7e484b906..ae85a72ea 100644
--- a/src/web_service/CMakeLists.txt
+++ b/src/web_service/CMakeLists.txt
@@ -9,4 +9,4 @@ add_library(web_service STATIC
 )
 
 create_target_directory_groups(web_service)
-target_link_libraries(web_service PRIVATE common nlohmann_json::nlohmann_json httplib lurlparser)
+target_link_libraries(web_service PRIVATE common nlohmann_json::nlohmann_json httplib)
diff --git a/src/web_service/web_backend.cpp b/src/web_service/web_backend.cpp
index 534960d09..67183e64c 100644
--- a/src/web_service/web_backend.cpp
+++ b/src/web_service/web_backend.cpp
@@ -7,7 +7,6 @@
 #include <mutex>
 #include <string>
 
-#include <LUrlParser.h>
 #include <fmt/format.h>
 #include <httplib.h>
 
@@ -19,9 +18,6 @@ namespace WebService {
 
 constexpr std::array<const char, 1> API_VERSION{'1'};
 
-constexpr int HTTP_PORT = 80;
-constexpr int HTTPS_PORT = 443;
-
 constexpr std::size_t TIMEOUT_SECONDS = 30;
 
 struct Client::Impl {
@@ -67,22 +63,14 @@ struct Client::Impl {
                              const std::string& jwt = "", const std::string& username = "",
                              const std::string& token = "") {
         if (cli == nullptr) {
-            const auto parsedUrl = LUrlParser::clParseURL::ParseURL(host);
-            int port{};
-            if (parsedUrl.m_Scheme == "http") {
-                if (!parsedUrl.GetPort(&port)) {
-                    port = HTTP_PORT;
-                }
-            } else if (parsedUrl.m_Scheme == "https") {
-                if (!parsedUrl.GetPort(&port)) {
-                    port = HTTPS_PORT;
-                }
-            } else {
-                LOG_ERROR(WebService, "Bad URL scheme {}", parsedUrl.m_Scheme);
-                return WebResult{WebResult::Code::InvalidURL, "Bad URL scheme", ""};
-            }
-            cli = std::make_unique<httplib::Client>(parsedUrl.m_Host.c_str(), port);
+            cli = std::make_unique<httplib::Client>(host.c_str());
+        }
+
+        if (!cli->is_valid()) {
+            LOG_ERROR(WebService, "Client is invalid, skipping request!");
+            return {};
         }
+
         cli->set_connection_timeout(TIMEOUT_SECONDS);
         cli->set_read_timeout(TIMEOUT_SECONDS);
         cli->set_write_timeout(TIMEOUT_SECONDS);
diff --git a/src/yuzu/CMakeLists.txt b/src/yuzu/CMakeLists.txt
index cc0291b15..4659e1f89 100644
--- a/src/yuzu/CMakeLists.txt
+++ b/src/yuzu/CMakeLists.txt
@@ -265,9 +265,11 @@ if (MSVC)
     include(CopyYuzuQt5Deps)
     include(CopyYuzuSDLDeps)
     include(CopyYuzuUnicornDeps)
+    include(CopyYuzuFFmpegDeps)
     copy_yuzu_Qt5_deps(yuzu)
     copy_yuzu_SDL_deps(yuzu)
     copy_yuzu_unicorn_deps(yuzu)
+    copy_yuzu_FFmpeg_deps(yuzu)
 endif()
 
 if (NOT APPLE)
diff --git a/src/yuzu/applets/controller.cpp b/src/yuzu/applets/controller.cpp
index 2760487a3..c6fa3e4f6 100644
--- a/src/yuzu/applets/controller.cpp
+++ b/src/yuzu/applets/controller.cpp
@@ -18,15 +18,15 @@
 
 namespace {
 
-constexpr std::array<std::array<bool, 4>, 8> led_patterns = {{
-    {1, 0, 0, 0},
-    {1, 1, 0, 0},
-    {1, 1, 1, 0},
-    {1, 1, 1, 1},
-    {1, 0, 0, 1},
-    {1, 0, 1, 0},
-    {1, 0, 1, 1},
-    {0, 1, 1, 0},
+constexpr std::array<std::array<bool, 4>, 8> led_patterns{{
+    {true, false, false, false},
+    {true, true, false, false},
+    {true, true, true, false},
+    {true, true, true, true},
+    {true, false, false, true},
+    {true, false, true, false},
+    {true, false, true, true},
+    {false, true, true, false},
 }};
 
 void UpdateController(Settings::ControllerType controller_type, std::size_t npad_index,
@@ -589,7 +589,7 @@ QtControllerSelector::QtControllerSelector(GMainWindow& parent) {
 QtControllerSelector::~QtControllerSelector() = default;
 
 void QtControllerSelector::ReconfigureControllers(
-    std::function<void()> callback, Core::Frontend::ControllerParameters parameters) const {
+    std::function<void()> callback, const Core::Frontend::ControllerParameters& parameters) const {
     this->callback = std::move(callback);
     emit MainWindowReconfigureControllers(parameters);
 }
diff --git a/src/yuzu/applets/controller.h b/src/yuzu/applets/controller.h
index 2d6d588c6..729ecc831 100644
--- a/src/yuzu/applets/controller.h
+++ b/src/yuzu/applets/controller.h
@@ -120,11 +120,13 @@ public:
     explicit QtControllerSelector(GMainWindow& parent);
     ~QtControllerSelector() override;
 
-    void ReconfigureControllers(std::function<void()> callback,
-                                Core::Frontend::ControllerParameters parameters) const override;
+    void ReconfigureControllers(
+        std::function<void()> callback,
+        const Core::Frontend::ControllerParameters& parameters) const override;
 
 signals:
-    void MainWindowReconfigureControllers(Core::Frontend::ControllerParameters parameters) const;
+    void MainWindowReconfigureControllers(
+        const Core::Frontend::ControllerParameters& parameters) const;
 
 private:
     void MainWindowReconfigureFinished();
diff --git a/src/yuzu/applets/profile_select.cpp b/src/yuzu/applets/profile_select.cpp
index dca8835ed..c9a2f8601 100644
--- a/src/yuzu/applets/profile_select.cpp
+++ b/src/yuzu/applets/profile_select.cpp
@@ -114,6 +114,15 @@ QtProfileSelectionDialog::QtProfileSelectionDialog(QWidget* parent)
 
 QtProfileSelectionDialog::~QtProfileSelectionDialog() = default;
 
+int QtProfileSelectionDialog::exec() {
+    // Skip profile selection when there's only one.
+    if (profile_manager->GetUserCount() == 1) {
+        user_index = 0;
+        return QDialog::Accepted;
+    }
+    return QDialog::exec();
+}
+
 void QtProfileSelectionDialog::accept() {
     QDialog::accept();
 }
diff --git a/src/yuzu/applets/profile_select.h b/src/yuzu/applets/profile_select.h
index cee886a77..29c33cca0 100644
--- a/src/yuzu/applets/profile_select.h
+++ b/src/yuzu/applets/profile_select.h
@@ -27,6 +27,7 @@ public:
     explicit QtProfileSelectionDialog(QWidget* parent);
     ~QtProfileSelectionDialog() override;
 
+    int exec() override;
     void accept() override;
     void reject() override;
 
diff --git a/src/yuzu/configuration/config.cpp b/src/yuzu/configuration/config.cpp
index d2913d613..1ce62e4a6 100644
--- a/src/yuzu/configuration/config.cpp
+++ b/src/yuzu/configuration/config.cpp
@@ -515,7 +515,7 @@ void Config::ReadMotionTouchValues() {
 void Config::ReadCoreValues() {
     qt_config->beginGroup(QStringLiteral("Core"));
 
-    ReadSettingGlobal(Settings::values.use_multi_core, QStringLiteral("use_multi_core"), false);
+    ReadSettingGlobal(Settings::values.use_multi_core, QStringLiteral("use_multi_core"), true);
 
     qt_config->endGroup();
 }
@@ -716,10 +716,12 @@ void Config::ReadRendererValues() {
                       QStringLiteral("use_disk_shader_cache"), true);
     ReadSettingGlobal(Settings::values.gpu_accuracy, QStringLiteral("gpu_accuracy"), 0);
     ReadSettingGlobal(Settings::values.use_asynchronous_gpu_emulation,
-                      QStringLiteral("use_asynchronous_gpu_emulation"), false);
+                      QStringLiteral("use_asynchronous_gpu_emulation"), true);
+    ReadSettingGlobal(Settings::values.use_nvdec_emulation, QStringLiteral("use_nvdec_emulation"),
+                      true);
     ReadSettingGlobal(Settings::values.use_vsync, QStringLiteral("use_vsync"), true);
     ReadSettingGlobal(Settings::values.use_assembly_shaders, QStringLiteral("use_assembly_shaders"),
-                      false);
+                      true);
     ReadSettingGlobal(Settings::values.use_asynchronous_shaders,
                       QStringLiteral("use_asynchronous_shaders"), false);
     ReadSettingGlobal(Settings::values.use_fast_gpu_time, QStringLiteral("use_fast_gpu_time"),
@@ -1108,7 +1110,7 @@ void Config::SaveControlValues() {
 void Config::SaveCoreValues() {
     qt_config->beginGroup(QStringLiteral("Core"));
 
-    WriteSettingGlobal(QStringLiteral("use_multi_core"), Settings::values.use_multi_core, false);
+    WriteSettingGlobal(QStringLiteral("use_multi_core"), Settings::values.use_multi_core, true);
 
     qt_config->endGroup();
 }
@@ -1264,10 +1266,12 @@ void Config::SaveRendererValues() {
                        static_cast<int>(Settings::values.gpu_accuracy.GetValue(global)),
                        Settings::values.gpu_accuracy.UsingGlobal(), 0);
     WriteSettingGlobal(QStringLiteral("use_asynchronous_gpu_emulation"),
-                       Settings::values.use_asynchronous_gpu_emulation, false);
+                       Settings::values.use_asynchronous_gpu_emulation, true);
+    WriteSettingGlobal(QStringLiteral("use_nvdec_emulation"), Settings::values.use_nvdec_emulation,
+                       true);
     WriteSettingGlobal(QStringLiteral("use_vsync"), Settings::values.use_vsync, true);
     WriteSettingGlobal(QStringLiteral("use_assembly_shaders"),
-                       Settings::values.use_assembly_shaders, false);
+                       Settings::values.use_assembly_shaders, true);
     WriteSettingGlobal(QStringLiteral("use_asynchronous_shaders"),
                        Settings::values.use_asynchronous_shaders, false);
     WriteSettingGlobal(QStringLiteral("use_fast_gpu_time"), Settings::values.use_fast_gpu_time,
diff --git a/src/yuzu/configuration/configure_graphics.cpp b/src/yuzu/configuration/configure_graphics.cpp
index 07d818548..4f083ecda 100644
--- a/src/yuzu/configuration/configure_graphics.cpp
+++ b/src/yuzu/configuration/configure_graphics.cpp
@@ -70,9 +70,11 @@ void ConfigureGraphics::SetConfiguration() {
     ui->api->setEnabled(runtime_lock);
     ui->use_asynchronous_gpu_emulation->setEnabled(runtime_lock);
     ui->use_disk_shader_cache->setEnabled(runtime_lock);
+    ui->use_nvdec_emulation->setEnabled(runtime_lock);
     ui->use_disk_shader_cache->setChecked(Settings::values.use_disk_shader_cache.GetValue());
     ui->use_asynchronous_gpu_emulation->setChecked(
         Settings::values.use_asynchronous_gpu_emulation.GetValue());
+    ui->use_nvdec_emulation->setChecked(Settings::values.use_nvdec_emulation.GetValue());
 
     if (Settings::configuring_global) {
         ui->api->setCurrentIndex(static_cast<int>(Settings::values.renderer_backend.GetValue()));
@@ -116,6 +118,9 @@ void ConfigureGraphics::ApplyConfiguration() {
             Settings::values.use_asynchronous_gpu_emulation.SetValue(
                 ui->use_asynchronous_gpu_emulation->isChecked());
         }
+        if (Settings::values.use_nvdec_emulation.UsingGlobal()) {
+            Settings::values.use_nvdec_emulation.SetValue(ui->use_nvdec_emulation->isChecked());
+        }
         if (Settings::values.bg_red.UsingGlobal()) {
             Settings::values.bg_red.SetValue(static_cast<float>(bg_color.redF()));
             Settings::values.bg_green.SetValue(static_cast<float>(bg_color.greenF()));
@@ -144,6 +149,8 @@ void ConfigureGraphics::ApplyConfiguration() {
         ConfigurationShared::ApplyPerGameSetting(&Settings::values.use_asynchronous_gpu_emulation,
                                                  ui->use_asynchronous_gpu_emulation,
                                                  use_asynchronous_gpu_emulation);
+        ConfigurationShared::ApplyPerGameSetting(&Settings::values.use_nvdec_emulation,
+                                                 ui->use_nvdec_emulation, use_nvdec_emulation);
 
         if (ui->bg_combobox->currentIndex() == ConfigurationShared::USE_GLOBAL_INDEX) {
             Settings::values.bg_red.SetGlobal(true);
@@ -240,6 +247,7 @@ void ConfigureGraphics::SetupPerGameUI() {
         ui->aspect_ratio_combobox->setEnabled(Settings::values.aspect_ratio.UsingGlobal());
         ui->use_asynchronous_gpu_emulation->setEnabled(
             Settings::values.use_asynchronous_gpu_emulation.UsingGlobal());
+        ui->use_nvdec_emulation->setEnabled(Settings::values.use_nvdec_emulation.UsingGlobal());
         ui->use_disk_shader_cache->setEnabled(Settings::values.use_disk_shader_cache.UsingGlobal());
         ui->bg_button->setEnabled(Settings::values.bg_red.UsingGlobal());
 
@@ -253,6 +261,8 @@ void ConfigureGraphics::SetupPerGameUI() {
 
     ConfigurationShared::SetColoredTristate(
         ui->use_disk_shader_cache, Settings::values.use_disk_shader_cache, use_disk_shader_cache);
+    ConfigurationShared::SetColoredTristate(
+        ui->use_nvdec_emulation, Settings::values.use_nvdec_emulation, use_nvdec_emulation);
     ConfigurationShared::SetColoredTristate(ui->use_asynchronous_gpu_emulation,
                                             Settings::values.use_asynchronous_gpu_emulation,
                                             use_asynchronous_gpu_emulation);
diff --git a/src/yuzu/configuration/configure_graphics.h b/src/yuzu/configuration/configure_graphics.h
index b4961f719..1fefc88eb 100644
--- a/src/yuzu/configuration/configure_graphics.h
+++ b/src/yuzu/configuration/configure_graphics.h
@@ -46,6 +46,7 @@ private:
     std::unique_ptr<Ui::ConfigureGraphics> ui;
     QColor bg_color;
 
+    ConfigurationShared::CheckState use_nvdec_emulation;
     ConfigurationShared::CheckState use_disk_shader_cache;
     ConfigurationShared::CheckState use_asynchronous_gpu_emulation;
 
diff --git a/src/yuzu/configuration/configure_graphics.ui b/src/yuzu/configuration/configure_graphics.ui
index 62aa337e7..58486eb1e 100644
--- a/src/yuzu/configuration/configure_graphics.ui
+++ b/src/yuzu/configuration/configure_graphics.ui
@@ -98,6 +98,13 @@
          </widget>
         </item>
         <item>
+         <widget class="QCheckBox" name="use_nvdec_emulation">
+          <property name="text">
+           <string>Use NVDEC emulation</string>
+          </property>
+         </widget>
+        </item>
+        <item>
          <widget class="QWidget" name="aspect_ratio_layout" native="true">
           <layout class="QHBoxLayout" name="horizontalLayout_6">
            <property name="leftMargin">
diff --git a/src/yuzu/configuration/configure_system.cpp b/src/yuzu/configuration/configure_system.cpp
index 9ad43ed8f..5e8e201dc 100644
--- a/src/yuzu/configuration/configure_system.cpp
+++ b/src/yuzu/configuration/configure_system.cpp
@@ -12,6 +12,7 @@
 #include "common/assert.h"
 #include "common/file_util.h"
 #include "core/core.h"
+#include "core/hle/service/time/time.h"
 #include "core/settings.h"
 #include "ui_configure_system.h"
 #include "yuzu/configuration/configuration_shared.h"
@@ -104,6 +105,22 @@ void ConfigureSystem::SetConfiguration() {
 void ConfigureSystem::ReadSystemSettings() {}
 
 void ConfigureSystem::ApplyConfiguration() {
+    // Allow setting custom RTC even if system is powered on, to allow in-game time to be fast
+    // forwared
+    if (Settings::values.custom_rtc.UsingGlobal()) {
+        if (ui->custom_rtc_checkbox->isChecked()) {
+            Settings::values.custom_rtc.SetValue(
+                std::chrono::seconds(ui->custom_rtc_edit->dateTime().toSecsSinceEpoch()));
+            if (Core::System::GetInstance().IsPoweredOn()) {
+                const s64 posix_time{Settings::values.custom_rtc.GetValue()->count() +
+                                     Service::Time::TimeManager::GetExternalTimeZoneOffset()};
+                Core::System::GetInstance().GetTimeManager().UpdateLocalSystemClockTime(posix_time);
+            }
+        } else {
+            Settings::values.custom_rtc.SetValue(std::nullopt);
+        }
+    }
+
     if (!enabled) {
         return;
     }
@@ -131,15 +148,6 @@ void ConfigureSystem::ApplyConfiguration() {
                 Settings::values.rng_seed.SetValue(std::nullopt);
             }
         }
-
-        if (Settings::values.custom_rtc.UsingGlobal()) {
-            if (ui->custom_rtc_checkbox->isChecked()) {
-                Settings::values.custom_rtc.SetValue(
-                    std::chrono::seconds(ui->custom_rtc_edit->dateTime().toSecsSinceEpoch()));
-            } else {
-                Settings::values.custom_rtc.SetValue(std::nullopt);
-            }
-        }
     } else {
         ConfigurationShared::ApplyPerGameSetting(&Settings::values.language_index,
                                                  ui->combo_language);
diff --git a/src/yuzu/main.cpp b/src/yuzu/main.cpp
index e3de0f0e1..18e68e590 100644
--- a/src/yuzu/main.cpp
+++ b/src/yuzu/main.cpp
@@ -303,24 +303,18 @@ void GMainWindow::ControllerSelectorReconfigureControllers(
 }
 
 void GMainWindow::ProfileSelectorSelectProfile() {
-    const Service::Account::ProfileManager manager;
-    int index = 0;
-    if (manager.GetUserCount() != 1) {
-        QtProfileSelectionDialog dialog(this);
-        dialog.setWindowFlags(Qt::Dialog | Qt::CustomizeWindowHint | Qt::WindowStaysOnTopHint |
-                              Qt::WindowTitleHint | Qt::WindowSystemMenuHint |
-                              Qt::WindowCloseButtonHint);
-        dialog.setWindowModality(Qt::WindowModal);
-
-        if (dialog.exec() == QDialog::Rejected) {
-            emit ProfileSelectorFinishedSelection(std::nullopt);
-            return;
-        }
-
-        index = dialog.GetIndex();
+    QtProfileSelectionDialog dialog(this);
+    dialog.setWindowFlags(Qt::Dialog | Qt::CustomizeWindowHint | Qt::WindowStaysOnTopHint |
+                          Qt::WindowTitleHint | Qt::WindowSystemMenuHint |
+                          Qt::WindowCloseButtonHint);
+    dialog.setWindowModality(Qt::WindowModal);
+    if (dialog.exec() == QDialog::Rejected) {
+        emit ProfileSelectorFinishedSelection(std::nullopt);
+        return;
     }
 
-    const auto uuid = manager.GetUser(static_cast<std::size_t>(index));
+    const Service::Account::ProfileManager manager;
+    const auto uuid = manager.GetUser(static_cast<std::size_t>(dialog.GetIndex()));
     if (!uuid.has_value()) {
         emit ProfileSelectorFinishedSelection(std::nullopt);
         return;
diff --git a/src/yuzu_cmd/config.cpp b/src/yuzu_cmd/config.cpp
index 23448e747..334038ef9 100644
--- a/src/yuzu_cmd/config.cpp
+++ b/src/yuzu_cmd/config.cpp
@@ -371,7 +371,7 @@ void Config::ReadValues() {
 
     // Core
     Settings::values.use_multi_core.SetValue(
-        sdl2_config->GetBoolean("Core", "use_multi_core", false));
+        sdl2_config->GetBoolean("Core", "use_multi_core", true));
 
     // Renderer
     const int renderer_backend = sdl2_config->GetInteger(
@@ -395,11 +395,11 @@ void Config::ReadValues() {
     const int gpu_accuracy_level = sdl2_config->GetInteger("Renderer", "gpu_accuracy", 0);
     Settings::values.gpu_accuracy.SetValue(static_cast<Settings::GPUAccuracy>(gpu_accuracy_level));
     Settings::values.use_asynchronous_gpu_emulation.SetValue(
-        sdl2_config->GetBoolean("Renderer", "use_asynchronous_gpu_emulation", false));
+        sdl2_config->GetBoolean("Renderer", "use_asynchronous_gpu_emulation", true));
     Settings::values.use_vsync.SetValue(
         static_cast<u16>(sdl2_config->GetInteger("Renderer", "use_vsync", 1)));
     Settings::values.use_assembly_shaders.SetValue(
-        sdl2_config->GetBoolean("Renderer", "use_assembly_shaders", false));
+        sdl2_config->GetBoolean("Renderer", "use_assembly_shaders", true));
     Settings::values.use_asynchronous_shaders.SetValue(
         sdl2_config->GetBoolean("Renderer", "use_asynchronous_shaders", false));
     Settings::values.use_asynchronous_shaders.SetValue(
diff --git a/src/yuzu_cmd/default_ini.h b/src/yuzu_cmd/default_ini.h
index aa9e40380..796e27df4 100644
--- a/src/yuzu_cmd/default_ini.h
+++ b/src/yuzu_cmd/default_ini.h
@@ -94,7 +94,7 @@ udp_pad_index=
 
 [Core]
 # Whether to use multi-core for CPU emulation
-# 0 (default): Disabled, 1: Enabled
+# 0: Disabled, 1 (default): Enabled
 use_multi_core=
 
 [Cpu]
@@ -163,7 +163,7 @@ max_anisotropy =
 use_vsync =
 
 # Whether to use OpenGL assembly shaders or not. NV_gpu_program5 is required.
-# 0 (default): Off, 1: On
+# 0: Off, 1 (default): On
 use_assembly_shaders =
 
 # Whether to allow asynchronous shader building.