15 files changed, 320 insertions, 224 deletions

diff --git a/src/video_core/engines/shader_bytecode.h b/src/video_core/engines/shader_bytecode.h
index c9bc83cd7..eba42deb4 100644
--- a/src/video_core/engines/shader_bytecode.h
+++ b/src/video_core/engines/shader_bytecode.h
@@ -911,14 +911,9 @@ union Instruction {
     } fadd32i;
 
     union {
-        BitField<20, 8, u64> shift_position;
-        BitField<28, 8, u64> shift_length;
-        BitField<48, 1, u64> negate_b;
-        BitField<49, 1, u64> negate_a;
-
-        u64 GetLeftShiftValue() const {
-            return 32 - (shift_position + shift_length);
-        }
+        BitField<40, 1, u64> brev;
+        BitField<47, 1, u64> rd_cc;
+        BitField<48, 1, u64> is_signed;
     } bfe;
 
     union {
diff --git a/src/video_core/gpu.h b/src/video_core/gpu.h
index ba8c9d665..64acb17df 100644
--- a/src/video_core/gpu.h
+++ b/src/video_core/gpu.h
@@ -39,6 +39,7 @@ enum class RenderTargetFormat : u32 {
     RGBA32_FLOAT = 0xC0,
     RGBA32_UINT = 0xC2,
     RGBA16_UNORM = 0xC6,
+    RGBA16_SNORM = 0xC7,
     RGBA16_UINT = 0xC9,
     RGBA16_FLOAT = 0xCA,
     RG32_FLOAT = 0xCB,
diff --git a/src/video_core/morton.cpp b/src/video_core/morton.cpp
index f2c83266e..6d522c318 100644
--- a/src/video_core/morton.cpp
+++ b/src/video_core/morton.cpp
@@ -51,6 +51,7 @@ static constexpr ConversionArray morton_to_linear_fns = {
     MortonCopy<true, PixelFormat::R8UI>,
     MortonCopy<true, PixelFormat::RGBA16F>,
     MortonCopy<true, PixelFormat::RGBA16U>,
+    MortonCopy<true, PixelFormat::RGBA16S>,
     MortonCopy<true, PixelFormat::RGBA16UI>,
     MortonCopy<true, PixelFormat::R11FG11FB10F>,
     MortonCopy<true, PixelFormat::RGBA32UI>,
@@ -131,6 +132,7 @@ static constexpr ConversionArray linear_to_morton_fns = {
     MortonCopy<false, PixelFormat::R8U>,
     MortonCopy<false, PixelFormat::R8UI>,
     MortonCopy<false, PixelFormat::RGBA16F>,
+    MortonCopy<false, PixelFormat::RGBA16S>,
     MortonCopy<false, PixelFormat::RGBA16U>,
     MortonCopy<false, PixelFormat::RGBA16UI>,
     MortonCopy<false, PixelFormat::R11FG11FB10F>,
diff --git a/src/video_core/renderer_opengl/gl_shader_decompiler.cpp b/src/video_core/renderer_opengl/gl_shader_decompiler.cpp
index 3adf7f0cb..849839fe3 100644
--- a/src/video_core/renderer_opengl/gl_shader_decompiler.cpp
+++ b/src/video_core/renderer_opengl/gl_shader_decompiler.cpp
@@ -2009,16 +2009,19 @@ private:
         expr += GetSampler(meta->sampler);
         expr += ", ";
 
-        expr += constructors.at(operation.GetOperandsCount() - 1);
+        expr += constructors.at(operation.GetOperandsCount() + (meta->array ? 1 : 0) - 1);
         expr += '(';
         for (std::size_t i = 0; i < count; ++i) {
-            expr += VisitOperand(operation, i).AsInt();
-            const std::size_t next = i + 1;
-            if (next == count)
-                expr += ')';
-            else if (next < count)
+            if (i > 0) {
                 expr += ", ";
+            }
+            expr += VisitOperand(operation, i).AsInt();
         }
+        if (meta->array) {
+            expr += ", ";
+            expr += Visit(meta->array).AsInt();
+        }
+        expr += ')';
 
         if (meta->lod && !meta->sampler.IsBuffer()) {
             expr += ", ";
diff --git a/src/video_core/renderer_opengl/gl_texture_cache.cpp b/src/video_core/renderer_opengl/gl_texture_cache.cpp
index 2d3838a7a..f424e3000 100644
--- a/src/video_core/renderer_opengl/gl_texture_cache.cpp
+++ b/src/video_core/renderer_opengl/gl_texture_cache.cpp
@@ -53,6 +53,7 @@ constexpr std::array<FormatTuple, VideoCore::Surface::MaxPixelFormat> tex_format
     {GL_R8UI, GL_RED_INTEGER, GL_UNSIGNED_BYTE, false},                             // R8UI
     {GL_RGBA16F, GL_RGBA, GL_HALF_FLOAT, false},                                    // RGBA16F
     {GL_RGBA16, GL_RGBA, GL_UNSIGNED_SHORT, false},                                 // RGBA16U
+    {GL_RGBA16_SNORM, GL_RGBA, GL_SHORT, false},                                    // RGBA16S
     {GL_RGBA16UI, GL_RGBA_INTEGER, GL_UNSIGNED_SHORT, false},                       // RGBA16UI
     {GL_R11F_G11F_B10F, GL_RGB, GL_UNSIGNED_INT_10F_11F_11F_REV, false},            // R11FG11FB10F
     {GL_RGBA32UI, GL_RGBA_INTEGER, GL_UNSIGNED_INT, false},                         // RGBA32UI
diff --git a/src/video_core/renderer_opengl/renderer_opengl.cpp b/src/video_core/renderer_opengl/renderer_opengl.cpp
index 12333e8c9..fca5e3ec0 100644
--- a/src/video_core/renderer_opengl/renderer_opengl.cpp
+++ b/src/video_core/renderer_opengl/renderer_opengl.cpp
@@ -5,8 +5,11 @@
 #include <algorithm>
 #include <cstddef>
 #include <cstdlib>
+#include <cstring>
 #include <memory>
+
 #include <glad/glad.h>
+
 #include "common/assert.h"
 #include "common/logging/log.h"
 #include "common/microprofile.h"
@@ -25,6 +28,8 @@
 
 namespace OpenGL {
 
+namespace {
+
 // If the size of this is too small, it ends up creating a soft cap on FPS as the renderer will have
 // to wait on available presentation frames.
 constexpr std::size_t SWAP_CHAIN_SIZE = 3;
@@ -41,124 +46,6 @@ struct Frame {
     bool is_srgb{};                   /// Framebuffer is sRGB or RGB
 };
 
-/**
- * For smooth Vsync rendering, we want to always present the latest frame that the core generates,
- * but also make sure that rendering happens at the pace that the frontend dictates. This is a
- * helper class that the renderer uses to sync frames between the render thread and the presentation
- * thread
- */
-class FrameMailbox {
-public:
-    std::mutex swap_chain_lock;
-    std::condition_variable present_cv;
-    std::array<Frame, SWAP_CHAIN_SIZE> swap_chain{};
-    std::queue<Frame*> free_queue;
-    std::deque<Frame*> present_queue;
-    Frame* previous_frame{};
-
-    FrameMailbox() {
-        for (auto& frame : swap_chain) {
-            free_queue.push(&frame);
-        }
-    }
-
-    ~FrameMailbox() {
-        // lock the mutex and clear out the present and free_queues and notify any people who are
-        // blocked to prevent deadlock on shutdown
-        std::scoped_lock lock{swap_chain_lock};
-        std::queue<Frame*>().swap(free_queue);
-        present_queue.clear();
-        present_cv.notify_all();
-    }
-
-    void ReloadPresentFrame(Frame* frame, u32 height, u32 width) {
-        frame->present.Release();
-        frame->present.Create();
-        GLint previous_draw_fbo{};
-        glGetIntegerv(GL_DRAW_FRAMEBUFFER_BINDING, &previous_draw_fbo);
-        glBindFramebuffer(GL_FRAMEBUFFER, frame->present.handle);
-        glFramebufferRenderbuffer(GL_FRAMEBUFFER, GL_COLOR_ATTACHMENT0, GL_RENDERBUFFER,
-                                  frame->color.handle);
-        if (glCheckFramebufferStatus(GL_FRAMEBUFFER) != GL_FRAMEBUFFER_COMPLETE) {
-            LOG_CRITICAL(Render_OpenGL, "Failed to recreate present FBO!");
-        }
-        glBindFramebuffer(GL_DRAW_FRAMEBUFFER, previous_draw_fbo);
-        frame->color_reloaded = false;
-    }
-
-    void ReloadRenderFrame(Frame* frame, u32 width, u32 height) {
-        // Recreate the color texture attachment
-        frame->color.Release();
-        frame->color.Create();
-        const GLenum internal_format = frame->is_srgb ? GL_SRGB8 : GL_RGB8;
-        glNamedRenderbufferStorage(frame->color.handle, internal_format, width, height);
-
-        // Recreate the FBO for the render target
-        frame->render.Release();
-        frame->render.Create();
-        glBindFramebuffer(GL_FRAMEBUFFER, frame->render.handle);
-        glFramebufferRenderbuffer(GL_FRAMEBUFFER, GL_COLOR_ATTACHMENT0, GL_RENDERBUFFER,
-                                  frame->color.handle);
-        if (glCheckFramebufferStatus(GL_FRAMEBUFFER) != GL_FRAMEBUFFER_COMPLETE) {
-            LOG_CRITICAL(Render_OpenGL, "Failed to recreate render FBO!");
-        }
-
-        frame->width = width;
-        frame->height = height;
-        frame->color_reloaded = true;
-    }
-
-    Frame* GetRenderFrame() {
-        std::unique_lock lock{swap_chain_lock};
-
-        // If theres no free frames, we will reuse the oldest render frame
-        if (free_queue.empty()) {
-            auto frame = present_queue.back();
-            present_queue.pop_back();
-            return frame;
-        }
-
-        Frame* frame = free_queue.front();
-        free_queue.pop();
-        return frame;
-    }
-
-    void ReleaseRenderFrame(Frame* frame) {
-        std::unique_lock lock{swap_chain_lock};
-        present_queue.push_front(frame);
-        present_cv.notify_one();
-    }
-
-    Frame* TryGetPresentFrame(int timeout_ms) {
-        std::unique_lock lock{swap_chain_lock};
-        // wait for new entries in the present_queue
-        present_cv.wait_for(lock, std::chrono::milliseconds(timeout_ms),
-                            [&] { return !present_queue.empty(); });
-        if (present_queue.empty()) {
-            // timed out waiting for a frame to draw so return the previous frame
-            return previous_frame;
-        }
-
-        // free the previous frame and add it back to the free queue
-        if (previous_frame) {
-            free_queue.push(previous_frame);
-        }
-
-        // the newest entries are pushed to the front of the queue
-        Frame* frame = present_queue.front();
-        present_queue.pop_front();
-        // remove all old entries from the present queue and move them back to the free_queue
-        for (auto f : present_queue) {
-            free_queue.push(f);
-        }
-        present_queue.clear();
-        previous_frame = frame;
-        return frame;
-    }
-};
-
-namespace {
-
 constexpr char VERTEX_SHADER[] = R"(
 #version 430 core
 
@@ -211,6 +98,24 @@ struct ScreenRectVertex {
     std::array<GLfloat, 2> tex_coord;
 };
 
+/// Returns true if any debug tool is attached
+bool HasDebugTool() {
+    const bool nsight = std::getenv("NVTX_INJECTION64_PATH") || std::getenv("NSIGHT_LAUNCHED");
+    if (nsight) {
+        return true;
+    }
+
+    GLint num_extensions;
+    glGetIntegerv(GL_NUM_EXTENSIONS, &num_extensions);
+    for (GLuint index = 0; index < static_cast<GLuint>(num_extensions); ++index) {
+        const auto name = reinterpret_cast<const char*>(glGetStringi(GL_EXTENSIONS, index));
+        if (!std::strcmp(name, "GL_EXT_debug_tool")) {
+            return true;
+        }
+    }
+    return false;
+}
+
 /**
  * Defines a 1:1 pixel ortographic projection matrix with (0,0) on the top-left
  * corner and (width, height) on the lower-bottom.
@@ -294,6 +199,153 @@ void APIENTRY DebugHandler(GLenum source, GLenum type, GLuint id, GLenum severit
 
 } // Anonymous namespace
 
+/**
+ * For smooth Vsync rendering, we want to always present the latest frame that the core generates,
+ * but also make sure that rendering happens at the pace that the frontend dictates. This is a
+ * helper class that the renderer uses to sync frames between the render thread and the presentation
+ * thread
+ */
+class FrameMailbox {
+public:
+    std::mutex swap_chain_lock;
+    std::condition_variable present_cv;
+    std::array<Frame, SWAP_CHAIN_SIZE> swap_chain{};
+    std::queue<Frame*> free_queue;
+    std::deque<Frame*> present_queue;
+    Frame* previous_frame{};
+
+    FrameMailbox() : has_debug_tool{HasDebugTool()} {
+        for (auto& frame : swap_chain) {
+            free_queue.push(&frame);
+        }
+    }
+
+    ~FrameMailbox() {
+        // lock the mutex and clear out the present and free_queues and notify any people who are
+        // blocked to prevent deadlock on shutdown
+        std::scoped_lock lock{swap_chain_lock};
+        std::queue<Frame*>().swap(free_queue);
+        present_queue.clear();
+        present_cv.notify_all();
+    }
+
+    void ReloadPresentFrame(Frame* frame, u32 height, u32 width) {
+        frame->present.Release();
+        frame->present.Create();
+        GLint previous_draw_fbo{};
+        glGetIntegerv(GL_DRAW_FRAMEBUFFER_BINDING, &previous_draw_fbo);
+        glBindFramebuffer(GL_FRAMEBUFFER, frame->present.handle);
+        glFramebufferRenderbuffer(GL_FRAMEBUFFER, GL_COLOR_ATTACHMENT0, GL_RENDERBUFFER,
+                                  frame->color.handle);
+        if (glCheckFramebufferStatus(GL_FRAMEBUFFER) != GL_FRAMEBUFFER_COMPLETE) {
+            LOG_CRITICAL(Render_OpenGL, "Failed to recreate present FBO!");
+        }
+        glBindFramebuffer(GL_DRAW_FRAMEBUFFER, previous_draw_fbo);
+        frame->color_reloaded = false;
+    }
+
+    void ReloadRenderFrame(Frame* frame, u32 width, u32 height) {
+        // Recreate the color texture attachment
+        frame->color.Release();
+        frame->color.Create();
+        const GLenum internal_format = frame->is_srgb ? GL_SRGB8 : GL_RGB8;
+        glNamedRenderbufferStorage(frame->color.handle, internal_format, width, height);
+
+        // Recreate the FBO for the render target
+        frame->render.Release();
+        frame->render.Create();
+        glBindFramebuffer(GL_FRAMEBUFFER, frame->render.handle);
+        glFramebufferRenderbuffer(GL_FRAMEBUFFER, GL_COLOR_ATTACHMENT0, GL_RENDERBUFFER,
+                                  frame->color.handle);
+        if (glCheckFramebufferStatus(GL_FRAMEBUFFER) != GL_FRAMEBUFFER_COMPLETE) {
+            LOG_CRITICAL(Render_OpenGL, "Failed to recreate render FBO!");
+        }
+
+        frame->width = width;
+        frame->height = height;
+        frame->color_reloaded = true;
+    }
+
+    Frame* GetRenderFrame() {
+        std::unique_lock lock{swap_chain_lock};
+
+        // If theres no free frames, we will reuse the oldest render frame
+        if (free_queue.empty()) {
+            auto frame = present_queue.back();
+            present_queue.pop_back();
+            return frame;
+        }
+
+        Frame* frame = free_queue.front();
+        free_queue.pop();
+        return frame;
+    }
+
+    void ReleaseRenderFrame(Frame* frame) {
+        std::unique_lock lock{swap_chain_lock};
+        present_queue.push_front(frame);
+        present_cv.notify_one();
+
+        DebugNotifyNextFrame();
+    }
+
+    Frame* TryGetPresentFrame(int timeout_ms) {
+        DebugWaitForNextFrame();
+
+        std::unique_lock lock{swap_chain_lock};
+        // wait for new entries in the present_queue
+        present_cv.wait_for(lock, std::chrono::milliseconds(timeout_ms),
+                            [&] { return !present_queue.empty(); });
+        if (present_queue.empty()) {
+            // timed out waiting for a frame to draw so return the previous frame
+            return previous_frame;
+        }
+
+        // free the previous frame and add it back to the free queue
+        if (previous_frame) {
+            free_queue.push(previous_frame);
+        }
+
+        // the newest entries are pushed to the front of the queue
+        Frame* frame = present_queue.front();
+        present_queue.pop_front();
+        // remove all old entries from the present queue and move them back to the free_queue
+        for (auto f : present_queue) {
+            free_queue.push(f);
+        }
+        present_queue.clear();
+        previous_frame = frame;
+        return frame;
+    }
+
+private:
+    std::mutex debug_synch_mutex;
+    std::condition_variable debug_synch_condition;
+    std::atomic_int frame_for_debug{};
+    const bool has_debug_tool; // When true, using a GPU debugger, so keep frames in lock-step
+
+    /// Signal that a new frame is available (called from GPU thread)
+    void DebugNotifyNextFrame() {
+        if (!has_debug_tool) {
+            return;
+        }
+        frame_for_debug++;
+        std::lock_guard lock{debug_synch_mutex};
+        debug_synch_condition.notify_one();
+    }
+
+    /// Wait for a new frame to be available (called from presentation thread)
+    void DebugWaitForNextFrame() {
+        if (!has_debug_tool) {
+            return;
+        }
+        const int last_frame = frame_for_debug;
+        std::unique_lock lock{debug_synch_mutex};
+        debug_synch_condition.wait(lock,
+                                   [this, last_frame] { return frame_for_debug > last_frame; });
+    }
+};
+
 RendererOpenGL::RendererOpenGL(Core::Frontend::EmuWindow& emu_window, Core::System& system)
     : VideoCore::RendererBase{emu_window}, emu_window{emu_window}, system{system},
       frame_mailbox{std::make_unique<FrameMailbox>()} {}
diff --git a/src/video_core/renderer_vulkan/maxwell_to_vk.cpp b/src/video_core/renderer_vulkan/maxwell_to_vk.cpp
index 72880d7ea..0e2e5e6c7 100644
--- a/src/video_core/renderer_vulkan/maxwell_to_vk.cpp
+++ b/src/video_core/renderer_vulkan/maxwell_to_vk.cpp
@@ -125,6 +125,7 @@ struct FormatTuple {
     {vk::Format::eR8Uint, Attachable | Storage},                 // R8UI
     {vk::Format::eR16G16B16A16Sfloat, Attachable | Storage},     // RGBA16F
     {vk::Format::eR16G16B16A16Unorm, Attachable | Storage},      // RGBA16U
+    {vk::Format::eR16G16B16A16Snorm, Attachable | Storage},      // RGBA16S
     {vk::Format::eR16G16B16A16Uint, Attachable | Storage},       // RGBA16UI
     {vk::Format::eB10G11R11UfloatPack32, Attachable | Storage},  // R11FG11FB10F
     {vk::Format::eR32G32B32A32Uint, Attachable | Storage},       // RGBA32UI
@@ -331,6 +332,8 @@ vk::Format VertexFormat(Maxwell::VertexAttribute::Type type, Maxwell::VertexAttr
             return vk::Format::eR16G16B16Unorm;
         case Maxwell::VertexAttribute::Size::Size_16_16_16_16:
             return vk::Format::eR16G16B16A16Unorm;
+        case Maxwell::VertexAttribute::Size::Size_10_10_10_2:
+            return vk::Format::eA2B10G10R10UnormPack32;
         default:
             break;
         }
diff --git a/src/video_core/renderer_vulkan/vk_device.cpp b/src/video_core/renderer_vulkan/vk_device.cpp
index 3847bd722..28d2fbc4f 100644
--- a/src/video_core/renderer_vulkan/vk_device.cpp
+++ b/src/video_core/renderer_vulkan/vk_device.cpp
@@ -535,6 +535,7 @@ std::unordered_map<vk::Format, vk::FormatProperties> VKDevice::GetFormatProperti
                                         vk::Format::eR32G32Sfloat,
                                         vk::Format::eR32G32Uint,
                                         vk::Format::eR16G16B16A16Uint,
+                                        vk::Format::eR16G16B16A16Snorm,
                                         vk::Format::eR16G16B16A16Unorm,
                                         vk::Format::eR16G16Unorm,
                                         vk::Format::eR16G16Snorm,
diff --git a/src/video_core/renderer_vulkan/vk_rasterizer.cpp b/src/video_core/renderer_vulkan/vk_rasterizer.cpp
index f889019c1..c9886cc16 100644
--- a/src/video_core/renderer_vulkan/vk_rasterizer.cpp
+++ b/src/video_core/renderer_vulkan/vk_rasterizer.cpp
@@ -1151,7 +1151,7 @@ std::size_t RasterizerVulkan::CalculateVertexArraysSize() const {
         // This implementation assumes that all attributes are used in the shader.
         const GPUVAddr start{regs.vertex_array[index].StartAddress()};
         const GPUVAddr end{regs.vertex_array_limit[index].LimitAddress()};
-        DEBUG_ASSERT(end > start);
+        DEBUG_ASSERT(end >= start);
 
         size += (end - start + 1) * regs.vertex_array[index].enable;
     }
diff --git a/src/video_core/renderer_vulkan/vk_shader_decompiler.cpp b/src/video_core/renderer_vulkan/vk_shader_decompiler.cpp
index b2c298051..51ecb5567 100644
--- a/src/video_core/renderer_vulkan/vk_shader_decompiler.cpp
+++ b/src/video_core/renderer_vulkan/vk_shader_decompiler.cpp
@@ -837,7 +837,7 @@ private:
                 Decorate(id, spv::Decoration::Offset, static_cast<u32>(tfb->offset));
             }
 
-            element += static_cast<u8>(num_components);
+            element = static_cast<u8>(static_cast<std::size_t>(element) + num_components);
         }
     }
 
diff --git a/src/video_core/shader/decode/bfe.cpp b/src/video_core/shader/decode/bfe.cpp
index e02bcd097..8e3b46e8e 100644
--- a/src/video_core/shader/decode/bfe.cpp
+++ b/src/video_core/shader/decode/bfe.cpp
@@ -17,33 +17,60 @@ u32 ShaderIR::DecodeBfe(NodeBlock& bb, u32 pc) {
     const Instruction instr = {program_code[pc]};
     const auto opcode = OpCode::Decode(instr);
 
-    UNIMPLEMENTED_IF(instr.bfe.negate_b);
-
     Node op_a = GetRegister(instr.gpr8);
-    op_a = GetOperandAbsNegInteger(op_a, false, instr.bfe.negate_a, false);
-
-    switch (opcode->get().GetId()) {
-    case OpCode::Id::BFE_IMM: {
-        UNIMPLEMENTED_IF_MSG(instr.generates_cc,
-                             "Condition codes generation in BFE is not implemented");
+    Node op_b = [&] {
+        switch (opcode->get().GetId()) {
+        case OpCode::Id::BFE_R:
+            return GetRegister(instr.gpr20);
+        case OpCode::Id::BFE_C:
+            return GetConstBuffer(instr.cbuf34.index, instr.cbuf34.GetOffset());
+        case OpCode::Id::BFE_IMM:
+            return Immediate(instr.alu.GetSignedImm20_20());
+        default:
+            UNREACHABLE();
+            return Immediate(0);
+        }
+    }();
 
-        const Node inner_shift_imm = Immediate(static_cast<u32>(instr.bfe.GetLeftShiftValue()));
-        const Node outer_shift_imm =
-            Immediate(static_cast<u32>(instr.bfe.GetLeftShiftValue() + instr.bfe.shift_position));
+    UNIMPLEMENTED_IF_MSG(instr.bfe.rd_cc, "Condition codes in BFE is not implemented");
 
-        const Node inner_shift =
-            Operation(OperationCode::ILogicalShiftLeft, NO_PRECISE, op_a, inner_shift_imm);
-        const Node outer_shift =
-            Operation(OperationCode::ILogicalShiftRight, NO_PRECISE, inner_shift, outer_shift_imm);
+    const bool is_signed = instr.bfe.is_signed;
 
-        SetInternalFlagsFromInteger(bb, outer_shift, instr.generates_cc);
-        SetRegister(bb, instr.gpr0, outer_shift);
-        break;
-    }
-    default:
-        UNIMPLEMENTED_MSG("Unhandled BFE instruction: {}", opcode->get().GetName());
+    // using reverse parallel method in
+    // https://graphics.stanford.edu/~seander/bithacks.html#ReverseParallel
+    // note for later if possible to implement faster method.
+    if (instr.bfe.brev) {
+        const auto swap = [&](u32 s, u32 mask) {
+            Node v1 =
+                SignedOperation(OperationCode::ILogicalShiftRight, is_signed, op_a, Immediate(s));
+            if (mask != 0) {
+                v1 = SignedOperation(OperationCode::IBitwiseAnd, is_signed, std::move(v1),
+                                     Immediate(mask));
+            }
+            Node v2 = op_a;
+            if (mask != 0) {
+                v2 = SignedOperation(OperationCode::IBitwiseAnd, is_signed, std::move(v2),
+                                     Immediate(mask));
+            }
+            v2 = SignedOperation(OperationCode::ILogicalShiftLeft, is_signed, std::move(v2),
+                                 Immediate(s));
+            return SignedOperation(OperationCode::IBitwiseOr, is_signed, std::move(v1),
+                                   std::move(v2));
+        };
+        op_a = swap(1, 0x55555555U);
+        op_a = swap(2, 0x33333333U);
+        op_a = swap(4, 0x0F0F0F0FU);
+        op_a = swap(8, 0x00FF00FFU);
+        op_a = swap(16, 0);
     }
 
+    const auto offset = SignedOperation(OperationCode::IBitfieldExtract, is_signed, op_b,
+                                        Immediate(0), Immediate(8));
+    const auto bits = SignedOperation(OperationCode::IBitfieldExtract, is_signed, op_b,
+                                      Immediate(8), Immediate(8));
+    auto result = SignedOperation(OperationCode::IBitfieldExtract, is_signed, op_a, offset, bits);
+    SetRegister(bb, instr.gpr0, std::move(result));
+
     return pc;
 }
 
diff --git a/src/video_core/shader/node_helper.cpp b/src/video_core/shader/node_helper.cpp
index b3dcd291c..76c56abb5 100644
--- a/src/video_core/shader/node_helper.cpp
+++ b/src/video_core/shader/node_helper.cpp
@@ -68,6 +68,8 @@ OperationCode SignedToUnsignedCode(OperationCode operation_code, bool is_signed)
         return OperationCode::UBitwiseXor;
     case OperationCode::IBitwiseNot:
         return OperationCode::UBitwiseNot;
+    case OperationCode::IBitfieldExtract:
+        return OperationCode::UBitfieldExtract;
     case OperationCode::IBitfieldInsert:
         return OperationCode::UBitfieldInsert;
     case OperationCode::IBitCount:
diff --git a/src/video_core/surface.cpp b/src/video_core/surface.cpp
index 9707c353d..cc7181229 100644
--- a/src/video_core/surface.cpp
+++ b/src/video_core/surface.cpp
@@ -111,6 +111,8 @@ PixelFormat PixelFormatFromRenderTargetFormat(Tegra::RenderTargetFormat format)
         return PixelFormat::RGBA16F;
     case Tegra::RenderTargetFormat::RGBA16_UNORM:
         return PixelFormat::RGBA16U;
+    case Tegra::RenderTargetFormat::RGBA16_SNORM:
+        return PixelFormat::RGBA16S;
     case Tegra::RenderTargetFormat::RGBA16_UINT:
         return PixelFormat::RGBA16UI;
     case Tegra::RenderTargetFormat::RGBA32_FLOAT:
diff --git a/src/video_core/surface.h b/src/video_core/surface.h
index d88109e5a..ae8817465 100644
--- a/src/video_core/surface.h
+++ b/src/video_core/surface.h
@@ -25,82 +25,83 @@ enum class PixelFormat {
     R8UI = 7,
     RGBA16F = 8,
     RGBA16U = 9,
-    RGBA16UI = 10,
-    R11FG11FB10F = 11,
-    RGBA32UI = 12,
-    DXT1 = 13,
-    DXT23 = 14,
-    DXT45 = 15,
-    DXN1 = 16, // This is also known as BC4
-    DXN2UNORM = 17,
-    DXN2SNORM = 18,
-    BC7U = 19,
-    BC6H_UF16 = 20,
-    BC6H_SF16 = 21,
-    ASTC_2D_4X4 = 22,
-    BGRA8 = 23,
-    RGBA32F = 24,
-    RG32F = 25,
-    R32F = 26,
-    R16F = 27,
-    R16U = 28,
-    R16S = 29,
-    R16UI = 30,
-    R16I = 31,
-    RG16 = 32,
-    RG16F = 33,
-    RG16UI = 34,
-    RG16I = 35,
-    RG16S = 36,
-    RGB32F = 37,
-    RGBA8_SRGB = 38,
-    RG8U = 39,
-    RG8S = 40,
-    RG32UI = 41,
-    RGBX16F = 42,
-    R32UI = 43,
-    R32I = 44,
-    ASTC_2D_8X8 = 45,
-    ASTC_2D_8X5 = 46,
-    ASTC_2D_5X4 = 47,
-    BGRA8_SRGB = 48,
-    DXT1_SRGB = 49,
-    DXT23_SRGB = 50,
-    DXT45_SRGB = 51,
-    BC7U_SRGB = 52,
-    R4G4B4A4U = 53,
-    ASTC_2D_4X4_SRGB = 54,
-    ASTC_2D_8X8_SRGB = 55,
-    ASTC_2D_8X5_SRGB = 56,
-    ASTC_2D_5X4_SRGB = 57,
-    ASTC_2D_5X5 = 58,
-    ASTC_2D_5X5_SRGB = 59,
-    ASTC_2D_10X8 = 60,
-    ASTC_2D_10X8_SRGB = 61,
-    ASTC_2D_6X6 = 62,
-    ASTC_2D_6X6_SRGB = 63,
-    ASTC_2D_10X10 = 64,
-    ASTC_2D_10X10_SRGB = 65,
-    ASTC_2D_12X12 = 66,
-    ASTC_2D_12X12_SRGB = 67,
-    ASTC_2D_8X6 = 68,
-    ASTC_2D_8X6_SRGB = 69,
-    ASTC_2D_6X5 = 70,
-    ASTC_2D_6X5_SRGB = 71,
-    E5B9G9R9F = 72,
+    RGBA16S = 10,
+    RGBA16UI = 11,
+    R11FG11FB10F = 12,
+    RGBA32UI = 13,
+    DXT1 = 14,
+    DXT23 = 15,
+    DXT45 = 16,
+    DXN1 = 17, // This is also known as BC4
+    DXN2UNORM = 18,
+    DXN2SNORM = 19,
+    BC7U = 20,
+    BC6H_UF16 = 21,
+    BC6H_SF16 = 22,
+    ASTC_2D_4X4 = 23,
+    BGRA8 = 24,
+    RGBA32F = 25,
+    RG32F = 26,
+    R32F = 27,
+    R16F = 28,
+    R16U = 29,
+    R16S = 30,
+    R16UI = 31,
+    R16I = 32,
+    RG16 = 33,
+    RG16F = 34,
+    RG16UI = 35,
+    RG16I = 36,
+    RG16S = 37,
+    RGB32F = 38,
+    RGBA8_SRGB = 39,
+    RG8U = 40,
+    RG8S = 41,
+    RG32UI = 42,
+    RGBX16F = 43,
+    R32UI = 44,
+    R32I = 45,
+    ASTC_2D_8X8 = 46,
+    ASTC_2D_8X5 = 47,
+    ASTC_2D_5X4 = 48,
+    BGRA8_SRGB = 49,
+    DXT1_SRGB = 50,
+    DXT23_SRGB = 51,
+    DXT45_SRGB = 52,
+    BC7U_SRGB = 53,
+    R4G4B4A4U = 54,
+    ASTC_2D_4X4_SRGB = 55,
+    ASTC_2D_8X8_SRGB = 56,
+    ASTC_2D_8X5_SRGB = 57,
+    ASTC_2D_5X4_SRGB = 58,
+    ASTC_2D_5X5 = 59,
+    ASTC_2D_5X5_SRGB = 60,
+    ASTC_2D_10X8 = 61,
+    ASTC_2D_10X8_SRGB = 62,
+    ASTC_2D_6X6 = 63,
+    ASTC_2D_6X6_SRGB = 64,
+    ASTC_2D_10X10 = 65,
+    ASTC_2D_10X10_SRGB = 66,
+    ASTC_2D_12X12 = 67,
+    ASTC_2D_12X12_SRGB = 68,
+    ASTC_2D_8X6 = 69,
+    ASTC_2D_8X6_SRGB = 70,
+    ASTC_2D_6X5 = 71,
+    ASTC_2D_6X5_SRGB = 72,
+    E5B9G9R9F = 73,
 
     MaxColorFormat,
 
     // Depth formats
-    Z32F = 73,
-    Z16 = 74,
+    Z32F = 74,
+    Z16 = 75,
 
     MaxDepthFormat,
 
     // DepthStencil formats
-    Z24S8 = 75,
-    S8Z24 = 76,
-    Z32FS8 = 77,
+    Z24S8 = 76,
+    S8Z24 = 77,
+    Z32FS8 = 78,
 
     MaxDepthStencilFormat,
 
@@ -138,6 +139,7 @@ constexpr std::array<u32, MaxPixelFormat> compression_factor_shift_table = {{
     0, // R8UI
     0, // RGBA16F
     0, // RGBA16U
+    0, // RGBA16S
     0, // RGBA16UI
     0, // R11FG11FB10F
     0, // RGBA32UI
@@ -235,6 +237,7 @@ constexpr std::array<u32, MaxPixelFormat> block_width_table = {{
     1,  // R8UI
     1,  // RGBA16F
     1,  // RGBA16U
+    1,  // RGBA16S
     1,  // RGBA16UI
     1,  // R11FG11FB10F
     1,  // RGBA32UI
@@ -324,6 +327,7 @@ constexpr std::array<u32, MaxPixelFormat> block_height_table = {{
     1,  // R8UI
     1,  // RGBA16F
     1,  // RGBA16U
+    1,  // RGBA16S
     1,  // RGBA16UI
     1,  // R11FG11FB10F
     1,  // RGBA32UI
@@ -413,6 +417,7 @@ constexpr std::array<u32, MaxPixelFormat> bpp_table = {{
     8,   // R8UI
     64,  // RGBA16F
     64,  // RGBA16U
+    64,  // RGBA16S
     64,  // RGBA16UI
     32,  // R11FG11FB10F
     128, // RGBA32UI
@@ -517,6 +522,7 @@ constexpr std::array<SurfaceCompression, MaxPixelFormat> compression_type_table
     SurfaceCompression::None,       // R8UI
     SurfaceCompression::None,       // RGBA16F
     SurfaceCompression::None,       // RGBA16U
+    SurfaceCompression::None,       // RGBA16S
     SurfaceCompression::None,       // RGBA16UI
     SurfaceCompression::None,       // R11FG11FB10F
     SurfaceCompression::None,       // RGBA32UI
diff --git a/src/video_core/texture_cache/format_lookup_table.cpp b/src/video_core/texture_cache/format_lookup_table.cpp
index cc3ad8417..e151c26c4 100644
--- a/src/video_core/texture_cache/format_lookup_table.cpp
+++ b/src/video_core/texture_cache/format_lookup_table.cpp
@@ -41,7 +41,7 @@ struct Table {
     ComponentType alpha_component;
     bool is_srgb;
 };
-constexpr std::array<Table, 75> DefinitionTable = {{
+constexpr std::array<Table, 76> DefinitionTable = {{
     {TextureFormat::A8R8G8B8, C, UNORM, UNORM, UNORM, UNORM, PixelFormat::ABGR8U},
     {TextureFormat::A8R8G8B8, C, SNORM, SNORM, SNORM, SNORM, PixelFormat::ABGR8S},
     {TextureFormat::A8R8G8B8, C, UINT, UINT, UINT, UINT, PixelFormat::ABGR8UI},
@@ -61,6 +61,7 @@ constexpr std::array<Table, 75> DefinitionTable = {{
     {TextureFormat::G8R8, C, UNORM, UNORM, UNORM, UNORM, PixelFormat::RG8U},
     {TextureFormat::G8R8, C, SNORM, SNORM, SNORM, SNORM, PixelFormat::RG8S},
 
+    {TextureFormat::R16_G16_B16_A16, C, SNORM, SNORM, SNORM, SNORM, PixelFormat::RGBA16S},
     {TextureFormat::R16_G16_B16_A16, C, UNORM, UNORM, UNORM, UNORM, PixelFormat::RGBA16U},
     {TextureFormat::R16_G16_B16_A16, C, FLOAT, FLOAT, FLOAT, FLOAT, PixelFormat::RGBA16F},
     {TextureFormat::R16_G16_B16_A16, C, UINT, UINT, UINT, UINT, PixelFormat::RGBA16UI},