fixed overflow

GPU: Implemented MUFU suboperation 8, sqrt.

src/core/CMakeLists.txt

@@ -126,6 +126,8 @@ add_library(core STATIC
     hle/service/audio/audren_u.h
     hle/service/audio/codecctl.cpp
     hle/service/audio/codecctl.h
+    hle/service/audio/hwopus.cpp
+    hle/service/audio/hwopus.h
     hle/service/bcat/module.cpp
     hle/service/bcat/module.h
     hle/service/bcat/bcat.cpp

src/core/hle/service/audio/audio.cpp

@@ -8,6 +8,7 @@
 #include "core/hle/service/audio/audrec_u.h"
 #include "core/hle/service/audio/audren_u.h"
 #include "core/hle/service/audio/codecctl.h"
+#include "core/hle/service/audio/hwopus.h"
 
 namespace Service::Audio {
 
@@ -17,6 +18,7 @@ void InstallInterfaces(SM::ServiceManager& service_manager) {
     std::make_shared<AudRecU>()->InstallAsService(service_manager);
     std::make_shared<AudRenU>()->InstallAsService(service_manager);
     std::make_shared<CodecCtl>()->InstallAsService(service_manager);
+    std::make_shared<HwOpus>()->InstallAsService(service_manager);
 }
 
 } // namespace Service::Audio

src/core/hle/service/audio/audren_u.cpp

@@ -47,6 +47,7 @@ public:
 
         // Start the audio event
         CoreTiming::ScheduleEvent(audio_ticks, audio_event);
+        voice_status_list.reserve(worker_params.voice_count);
     }
     ~IAudioRenderer() {
         CoreTiming::UnscheduleEvent(audio_event, 0);
@@ -68,6 +69,12 @@ private:
                     buf.data() + sizeof(UpdateDataHeader) + config.behavior_size,
                     memory_pool_count * sizeof(MemoryPoolInfo));
 
+        std::vector<VoiceInfo> voice_info(worker_params.voice_count);
+        std::memcpy(voice_info.data(),
+                    buf.data() + sizeof(UpdateDataHeader) + config.behavior_size +
+                        config.memory_pools_size + config.voice_resource_size,
+                    worker_params.voice_count * sizeof(VoiceInfo));
+
         UpdateDataHeader response_data{worker_params};
 
         ASSERT(ctx.GetWriteBufferSize() == response_data.total_size);
@@ -86,6 +93,23 @@ private:
         std::memcpy(output.data() + sizeof(UpdateDataHeader), memory_pool.data(),
                     response_data.memory_pools_size);
 
+        for (unsigned i = 0; i < voice_info.size(); i++) {
+            if (voice_info[i].is_new) {
+                voice_status_list[i].played_sample_count = 0;
+                voice_status_list[i].wave_buffer_consumed = 0;
+            } else if (voice_info[i].play_state == (u8)PlayStates::Started) {
+                for (u32 buff_idx = 0; buff_idx < voice_info[i].wave_buffer_count; buff_idx++) {
+                    voice_status_list[i].played_sample_count +=
+                        (voice_info[i].wave_buffer[buff_idx].end_sample_offset -
+                         voice_info[i].wave_buffer[buff_idx].start_sample_offset) /
+                        2;
+                    voice_status_list[i].wave_buffer_consumed++;
+                }
+            }
+        }
+        std::memcpy(output.data() + sizeof(UpdateDataHeader) + response_data.memory_pools_size,
+                    voice_status_list.data(), response_data.voices_size);
+
         ctx.WriteBuffer(output);
 
         IPC::ResponseBuilder rb{ctx, 2};
@@ -130,6 +154,11 @@ private:
         Released = 0x6,
     };
 
+    enum class PlayStates : u8 {
+        Started = 0,
+        Stopped = 1,
+    };
+
     struct MemoryPoolEntry {
         MemoryPoolStates state;
         u32_le unknown_4;
@@ -175,11 +204,69 @@ private:
     };
     static_assert(sizeof(UpdateDataHeader) == 0x40, "UpdateDataHeader has wrong size");
 
+    struct BiquadFilter {
+        u8 enable;
+        INSERT_PADDING_BYTES(1);
+        s16_le numerator[3];
+        s16_le denominator[2];
+    };
+    static_assert(sizeof(BiquadFilter) == 0xc, "BiquadFilter has wrong size");
+
+    struct WaveBuffer {
+        u64_le buffer_addr;
+        u64_le buffer_sz;
+        s32_le start_sample_offset;
+        s32_le end_sample_offset;
+        u8 loop;
+        u8 end_of_stream;
+        u8 sent_to_server;
+        INSERT_PADDING_BYTES(5);
+        u64 context_addr;
+        u64 context_sz;
+        INSERT_PADDING_BYTES(8);
+    };
+    static_assert(sizeof(WaveBuffer) == 0x38, "WaveBuffer has wrong size");
+
+    struct VoiceInfo {
+        u32_le id;
+        u32_le node_id;
+        u8 is_new;
+        u8 is_in_use;
+        u8 play_state;
+        u8 sample_format;
+        u32_le sample_rate;
+        u32_le priority;
+        u32_le sorting_order;
+        u32_le channel_count;
+        float_le pitch;
+        float_le volume;
+        BiquadFilter biquad_filter[2];
+        u32_le wave_buffer_count;
+        u16_le wave_buffer_head;
+        INSERT_PADDING_BYTES(6);
+        u64_le additional_params_addr;
+        u64_le additional_params_sz;
+        u32_le mix_id;
+        u32_le splitter_info_id;
+        WaveBuffer wave_buffer[4];
+        u32_le voice_channel_resource_ids[6];
+        INSERT_PADDING_BYTES(24);
+    };
+    static_assert(sizeof(VoiceInfo) == 0x170, "VoiceInfo is wrong size");
+
+    struct VoiceOutStatus {
+        u64_le played_sample_count;
+        u32_le wave_buffer_consumed;
+        INSERT_PADDING_WORDS(1);
+    };
+    static_assert(sizeof(VoiceOutStatus) == 0x10, "VoiceOutStatus has wrong size");
+
     /// This is used to trigger the audio event callback.
     CoreTiming::EventType* audio_event;
 
     Kernel::SharedPtr<Kernel::Event> system_event;
     AudioRendererParameter worker_params;
+    std::vector<VoiceOutStatus> voice_status_list;
 };
 
 class IAudioDevice final : public ServiceFramework<IAudioDevice> {

src/core/hle/service/audio/hwopus.cpp

@@ -0,0 +1,29 @@
+// Copyright 2018 yuzu emulator team
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#include "common/logging/log.h"
+#include "core/hle/ipc_helpers.h"
+#include "core/hle/kernel/hle_ipc.h"
+#include "core/hle/service/audio/hwopus.h"
+
+namespace Service::Audio {
+
+void HwOpus::GetWorkBufferSize(Kernel::HLERequestContext& ctx) {
+    NGLOG_WARNING(Service_Audio, "(STUBBED) called");
+    IPC::ResponseBuilder rb{ctx, 3};
+    rb.Push(RESULT_SUCCESS);
+    rb.Push<u32>(0x4000);
+}
+
+HwOpus::HwOpus() : ServiceFramework("hwopus") {
+    static const FunctionInfo functions[] = {
+        {0, nullptr, "Initialize"},
+        {1, &HwOpus::GetWorkBufferSize, "GetWorkBufferSize"},
+        {2, nullptr, "InitializeMultiStream"},
+        {3, nullptr, "GetWorkBufferSizeMultiStream"},
+    };
+    RegisterHandlers(functions);
+}
+
+} // namespace Service::Audio

src/core/hle/service/audio/hwopus.h

@@ -0,0 +1,20 @@
+// Copyright 2018 yuzu emulator team
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#pragma once
+
+#include "core/hle/service/service.h"
+
+namespace Service::Audio {
+
+class HwOpus final : public ServiceFramework<HwOpus> {
+public:
+    explicit HwOpus();
+    ~HwOpus() = default;
+
+private:
+    void GetWorkBufferSize(Kernel::HLERequestContext& ctx);
+};
+
+} // namespace Service::Audio

src/core/hle/service/nvdrv/devices/nvhost_as_gpu.cpp

@@ -8,6 +8,8 @@
 #include "core/core.h"
 #include "core/hle/service/nvdrv/devices/nvhost_as_gpu.h"
 #include "core/hle/service/nvdrv/devices/nvmap.h"
+#include "video_core/renderer_base.h"
+#include "video_core/video_core.h"
 
 namespace Service::Nvidia::Devices {
 
@@ -154,6 +156,9 @@ u32 nvhost_as_gpu::UnmapBuffer(const std::vector<u8>& input, std::vector<u8>& ou
 
     ASSERT_MSG(itr != buffer_mappings.end(), "Tried to unmap invalid mapping");
 
+    // Remove this memory region from the rasterizer cache.
+    VideoCore::g_renderer->Rasterizer()->FlushAndInvalidateRegion(params.offset, itr->second.size);
+
     params.offset = gpu.memory_manager->UnmapBuffer(params.offset, itr->second.size);
 
     buffer_mappings.erase(itr->second.offset);

src/core/hle/service/nvdrv/devices/nvmap.cpp

@@ -148,6 +148,7 @@ u32 nvmap::IocParam(const std::vector<u8>& input, std::vector<u8>& output) {
 }
 
 u32 nvmap::IocFree(const std::vector<u8>& input, std::vector<u8>& output) {
+    // TODO(Subv): These flags are unconfirmed.
     enum FreeFlags {
         Freed = 0,
         NotFreedYet = 1,
@@ -161,15 +162,21 @@ u32 nvmap::IocFree(const std::vector<u8>& input, std::vector<u8>& output) {
     auto itr = handles.find(params.handle);
     ASSERT(itr != handles.end());
 
+    ASSERT(itr->second->refcount > 0);
+
     itr->second->refcount--;
 
-    params.refcount = itr->second->refcount;
     params.size = itr->second->size;
 
-    if (itr->second->refcount == 0)
+    if (itr->second->refcount == 0) {
         params.flags = Freed;
-    else
+        // The address of the nvmap is written to the output if we're finally freeing it, otherwise
+        // 0 is written.
+        params.address = itr->second->addr;
+    } else {
         params.flags = NotFreedYet;
+        params.address = 0;
+    }
 
     handles.erase(params.handle);
 

src/core/hle/service/nvdrv/devices/nvmap.h

@@ -94,7 +94,7 @@ private:
     struct IocFreeParams {
         u32_le handle;
         INSERT_PADDING_BYTES(4);
-        u64_le refcount;
+        u64_le address;
         u32_le size;
         u32_le flags;
     };

src/core/settings.h

@@ -129,6 +129,7 @@ struct Values {
     // Renderer
     float resolution_factor;
     bool toggle_framelimit;
+    bool use_accurate_framebuffers;
 
     float bg_red;
     float bg_green;

src/core/telemetry_session.cpp

@@ -161,6 +161,8 @@ TelemetrySession::TelemetrySession() {
              Settings::values.resolution_factor);
     AddField(Telemetry::FieldType::UserConfig, "Renderer_ToggleFramelimit",
              Settings::values.toggle_framelimit);
+    AddField(Telemetry::FieldType::UserConfig, "Renderer_UseAccurateFramebuffers",
+             Settings::values.use_accurate_framebuffers);
     AddField(Telemetry::FieldType::UserConfig, "System_UseDockedMode",
              Settings::values.use_docked_mode);
 }

src/video_core/engines/maxwell_3d.h

@@ -280,6 +280,34 @@ public:
             UnsignedInt = 0x2,
         };
 
+        enum class ComparisonOp : u32 {
+            Never = 0x200,
+            Less = 0x201,
+            Equal = 0x202,
+            LessEqual = 0x203,
+            Greater = 0x204,
+            NotEqual = 0x205,
+            GreaterEqual = 0x206,
+            Always = 0x207,
+        };
+
+        struct Cull {
+            enum class FrontFace : u32 {
+                ClockWise = 0x0900,
+                CounterClockWise = 0x0901,
+            };
+
+            enum class CullFace : u32 {
+                Front = 0x0404,
+                Back = 0x0405,
+                FrontAndBack = 0x0408,
+            };
+
+            u32 enabled;
+            FrontFace front_face;
+            CullFace cull_face;
+        };
+
         struct Blend {
             enum class Equation : u32 {
                 Add = 1,
@@ -321,6 +349,24 @@ public:
             INSERT_PADDING_WORDS(1);
         };
 
+        struct RenderTargetConfig {
+            u32 address_high;
+            u32 address_low;
+            u32 width;
+            u32 height;
+            Tegra::RenderTargetFormat format;
+            u32 block_dimensions;
+            u32 array_mode;
+            u32 layer_stride;
+            u32 base_layer;
+            INSERT_PADDING_WORDS(7);
+
+            GPUVAddr Address() const {
+                return static_cast<GPUVAddr>((static_cast<GPUVAddr>(address_high) << 32) |
+                                             address_low);
+            }
+        };
+
         union {
             struct {
                 INSERT_PADDING_WORDS(0x45);
@@ -333,23 +379,7 @@ public:
 
                 INSERT_PADDING_WORDS(0x1B8);
 
-                struct {
-                    u32 address_high;
-                    u32 address_low;
-                    u32 width;
-                    u32 height;
-                    Tegra::RenderTargetFormat format;
-                    u32 block_dimensions;
-                    u32 array_mode;
-                    u32 layer_stride;
-                    u32 base_layer;
-                    INSERT_PADDING_WORDS(7);
-
-                    GPUVAddr Address() const {
-                        return static_cast<GPUVAddr>((static_cast<GPUVAddr>(address_high) << 32) |
-                                                     address_low);
-                    }
-                } rt[NumRenderTargets];
+                RenderTargetConfig rt[NumRenderTargets];
 
                 struct {
                     f32 scale_x;
@@ -411,7 +441,7 @@ public:
                 struct {
                     u32 address_high;
                     u32 address_low;
-                    u32 format;
+                    Tegra::DepthFormat format;
                     u32 block_dimensions;
                     u32 layer_stride;
 
@@ -433,11 +463,21 @@ public:
                     };
                 } rt_control;
 
-                INSERT_PADDING_WORDS(0x31);
+                INSERT_PADDING_WORDS(0x2B);
+
+                u32 depth_test_enable;
+
+                INSERT_PADDING_WORDS(0x5);
 
                 u32 independent_blend_enable;
 
-                INSERT_PADDING_WORDS(0x15);
+                u32 depth_write_enabled;
+
+                INSERT_PADDING_WORDS(0x8);
+
+                ComparisonOp depth_test_func;
+
+                INSERT_PADDING_WORDS(0xB);
 
                 struct {
                     u32 separate_alpha;
@@ -453,7 +493,11 @@ public:
                     u32 enable[NumRenderTargets];
                 } blend;
 
-                INSERT_PADDING_WORDS(0x77);
+                INSERT_PADDING_WORDS(0x2D);
+
+                u32 vb_element_base;
+
+                INSERT_PADDING_WORDS(0x49);
 
                 struct {
                     u32 tsc_address_high;
@@ -534,7 +578,13 @@ public:
                     }
                 } index_array;
 
-                INSERT_PADDING_WORDS(0xC7);
+                INSERT_PADDING_WORDS(0x7);
+
+                INSERT_PADDING_WORDS(0x46);
+
+                Cull cull;
+
+                INSERT_PADDING_WORDS(0x77);
 
                 struct {
                     u32 query_address_high;
@@ -741,13 +791,18 @@ ASSERT_REG_POSITION(vertex_buffer, 0x35D);
 ASSERT_REG_POSITION(zeta, 0x3F8);
 ASSERT_REG_POSITION(vertex_attrib_format[0], 0x458);
 ASSERT_REG_POSITION(rt_control, 0x487);
+ASSERT_REG_POSITION(depth_test_enable, 0x4B3);
 ASSERT_REG_POSITION(independent_blend_enable, 0x4B9);
+ASSERT_REG_POSITION(depth_write_enabled, 0x4BA);
+ASSERT_REG_POSITION(depth_test_func, 0x4C3);
 ASSERT_REG_POSITION(blend, 0x4CF);
+ASSERT_REG_POSITION(vb_element_base, 0x50D);
 ASSERT_REG_POSITION(tsc, 0x557);
 ASSERT_REG_POSITION(tic, 0x55D);
 ASSERT_REG_POSITION(code_address, 0x582);
 ASSERT_REG_POSITION(draw, 0x585);
 ASSERT_REG_POSITION(index_array, 0x5F2);
+ASSERT_REG_POSITION(cull, 0x646);
 ASSERT_REG_POSITION(query, 0x6C0);
 ASSERT_REG_POSITION(vertex_array[0], 0x700);
 ASSERT_REG_POSITION(independent_blend, 0x780);

src/video_core/engines/maxwell_dma.cpp

@@ -49,7 +49,11 @@ void MaxwellDMA::HandleCopy() {
     ASSERT(regs.src_params.pos_y == 0);
     ASSERT(regs.dst_params.pos_x == 0);
     ASSERT(regs.dst_params.pos_y == 0);
-    ASSERT(regs.exec.is_dst_linear != regs.exec.is_src_linear);
+
+    if (regs.exec.is_dst_linear == regs.exec.is_src_linear) {
+        Memory::CopyBlock(dest_cpu, source_cpu, regs.x_count * regs.y_count);
+        return;
+    }
 
     u8* src_buffer = Memory::GetPointer(source_cpu);
     u8* dst_buffer = Memory::GetPointer(dest_cpu);

src/video_core/engines/shader_bytecode.h

@@ -142,6 +142,7 @@ enum class PredCondition : u64 {
     GreaterThan = 4,
     NotEqual = 5,
     GreaterEqual = 6,
+    NotEqualWithNan = 13,
     // TODO(Subv): Other condition types
 };
 
@@ -165,7 +166,7 @@ enum class SubOp : u64 {
     Lg2 = 0x3,
     Rcp = 0x4,
     Rsq = 0x5,
-    Min = 0x8,
+    Sqrt = 0x8,
 };
 
 enum class F2iRoundingOp : u64 {
@@ -209,7 +210,7 @@ union Instruction {
     } pred;
     BitField<19, 1, u64> negate_pred;
     BitField<20, 8, Register> gpr20;
-    BitField<20, 7, SubOp> sub_op;
+    BitField<20, 4, SubOp> sub_op;
     BitField<28, 8, Register> gpr28;
     BitField<39, 8, Register> gpr39;
     BitField<48, 16, u64> opcode;

src/video_core/gpu.h

@@ -16,6 +16,7 @@ namespace Tegra {
 enum class RenderTargetFormat : u32 {
     NONE = 0x0,
     RGBA32_FLOAT = 0xC0,
+    RGBA32_UINT = 0xC2,
     RGBA16_FLOAT = 0xCA,
     RGB10_A2_UNORM = 0xD1,
     RGBA8_UNORM = 0xD5,
@@ -23,6 +24,15 @@ enum class RenderTargetFormat : u32 {
     R11G11B10_FLOAT = 0xE0,
 };
 
+enum class DepthFormat : u32 {
+    Z32_FLOAT = 0xA,
+    Z16_UNORM = 0x13,
+    S8_Z24_UNORM = 0x14,
+    Z24_X8_UNORM = 0x15,
+    Z24_S8_UNORM = 0x16,
+    Z24_C8_UNORM = 0x18,
+};
+
 /// Returns the number of bytes per pixel of each rendertarget format.
 u32 RenderTargetBytesPerPixel(RenderTargetFormat format);
 

src/video_core/memory_manager.cpp

@@ -100,9 +100,9 @@ boost::optional<GPUVAddr> MemoryManager::FindFreeBlock(u64 size, u64 align) {
 
 boost::optional<VAddr> MemoryManager::GpuToCpuAddress(GPUVAddr gpu_addr) {
     VAddr base_addr = PageSlot(gpu_addr);
-    ASSERT(base_addr != static_cast<u64>(PageStatus::Unmapped));
 
-    if (base_addr == static_cast<u64>(PageStatus::Allocated)) {
+    if (base_addr == static_cast<u64>(PageStatus::Allocated) ||
+        base_addr == static_cast<u64>(PageStatus::Unmapped)) {
         return {};
     }
 

src/video_core/rasterizer_interface.h

@@ -51,9 +51,8 @@ public:
     }
 
     /// Attempt to use a faster method to display the framebuffer to screen
-    virtual bool AccelerateDisplay(const Tegra::FramebufferConfig& framebuffer,
-                                   VAddr framebuffer_addr, u32 pixel_stride,
-                                   ScreenInfo& screen_info) {
+    virtual bool AccelerateDisplay(const Tegra::FramebufferConfig& config, VAddr framebuffer_addr,
+                                   u32 pixel_stride, ScreenInfo& screen_info) {
         return false;
     }
 

src/video_core/renderer_opengl/gl_rasterizer.cpp

@@ -146,7 +146,6 @@ std::pair<u8*, GLintptr> RasterizerOpenGL::SetupVertexArrays(u8* array_ptr,
         u64 size = end - start + 1;
 
         // Copy vertex array data
-        res_cache.FlushRegion(start, size, nullptr);
         Memory::ReadBlock(*memory_manager->GpuToCpuAddress(start), array_ptr, size);
 
         // Bind the vertex array to the buffer at the current offset.
@@ -305,10 +304,15 @@ void RasterizerOpenGL::DrawArrays() {
     MICROPROFILE_SCOPE(OpenGL_Drawing);
     const auto& regs = Core::System().GetInstance().GPU().Maxwell3D().regs;
 
-    // TODO(bunnei): Implement these
+    // Sync the depth test state before configuring the framebuffer surfaces.
+    SyncDepthTestState();
+
+    // TODO(bunnei): Implement this
     const bool has_stencil = false;
+
     const bool using_color_fb = true;
-    const bool using_depth_fb = false;
+    const bool using_depth_fb = regs.zeta.Address() != 0;
+
     const MathUtil::Rectangle<s32> viewport_rect{regs.viewport_transform[0].GetRect()};
 
     const bool write_color_fb =
@@ -325,32 +329,23 @@ void RasterizerOpenGL::DrawArrays() {
     std::tie(color_surface, depth_surface, surfaces_rect) =
         res_cache.GetFramebufferSurfaces(using_color_fb, using_depth_fb, viewport_rect);
 
-    const u16 res_scale = color_surface != nullptr
-                              ? color_surface->res_scale
-                              : (depth_surface == nullptr ? 1u : depth_surface->res_scale);
-
     MathUtil::Rectangle<u32> draw_rect{
+        static_cast<u32>(std::clamp<s32>(static_cast<s32>(surfaces_rect.left) + viewport_rect.left,
+                                         surfaces_rect.left, surfaces_rect.right)), // Left
+        static_cast<u32>(std::clamp<s32>(static_cast<s32>(surfaces_rect.bottom) + viewport_rect.top,
+                                         surfaces_rect.bottom, surfaces_rect.top)), // Top
+        static_cast<u32>(std::clamp<s32>(static_cast<s32>(surfaces_rect.left) + viewport_rect.right,
+                                         surfaces_rect.left, surfaces_rect.right)), // Right
         static_cast<u32>(
-            std::clamp<s32>(static_cast<s32>(surfaces_rect.left) + viewport_rect.left * res_scale,
-                            surfaces_rect.left, surfaces_rect.right)), // Left
-        static_cast<u32>(
-            std::clamp<s32>(static_cast<s32>(surfaces_rect.bottom) + viewport_rect.top * res_scale,
-                            surfaces_rect.bottom, surfaces_rect.top)), // Top
-        static_cast<u32>(
-            std::clamp<s32>(static_cast<s32>(surfaces_rect.left) + viewport_rect.right * res_scale,
-                            surfaces_rect.left, surfaces_rect.right)), // Right
-        static_cast<u32>(std::clamp<s32>(static_cast<s32>(surfaces_rect.bottom) +
-                                             viewport_rect.bottom * res_scale,
-                                         surfaces_rect.bottom, surfaces_rect.top))}; // Bottom
+            std::clamp<s32>(static_cast<s32>(surfaces_rect.bottom) + viewport_rect.bottom,
+                            surfaces_rect.bottom, surfaces_rect.top))}; // Bottom
 
     // Bind the framebuffer surfaces
     BindFramebufferSurfaces(color_surface, depth_surface, has_stencil);
 
-    // Sync the viewport
-    SyncViewport(surfaces_rect, res_scale);
-
-    // Sync the blend state registers
+    SyncViewport(surfaces_rect);
     SyncBlendState();
+    SyncCullMode();
 
     // TODO(bunnei): Sync framebuffer_scale uniform here
     // TODO(bunnei): Sync scissorbox uniform(s) here
@@ -420,14 +415,16 @@ void RasterizerOpenGL::DrawArrays() {
 
     const GLenum primitive_mode{MaxwellToGL::PrimitiveTopology(regs.draw.topology)};
     if (is_indexed) {
-        const GLint index_min{static_cast<GLint>(regs.index_array.first)};
-        const GLint index_max{static_cast<GLint>(regs.index_array.first + regs.index_array.count)};
-        glDrawRangeElementsBaseVertex(primitive_mode, index_min, index_max, regs.index_array.count,
-                                      MaxwellToGL::IndexFormat(regs.index_array.format),
-                                      reinterpret_cast<const void*>(index_buffer_offset),
-                                      -index_min);
+        const GLint base_vertex{static_cast<GLint>(regs.vb_element_base)};
+
+        // Adjust the index buffer offset so it points to the first desired index.
+        index_buffer_offset += regs.index_array.first * regs.index_array.FormatSizeInBytes();
+
+        glDrawElementsBaseVertex(primitive_mode, regs.index_array.count,
+                                 MaxwellToGL::IndexFormat(regs.index_array.format),
+                                 reinterpret_cast<const void*>(index_buffer_offset), base_vertex);
     } else {
-        glDrawArrays(primitive_mode, 0, regs.vertex_buffer.count);
+        glDrawArrays(primitive_mode, regs.vertex_buffer.first, regs.vertex_buffer.count);
     }
 
     // Disable scissor test
@@ -442,19 +439,11 @@ void RasterizerOpenGL::DrawArrays() {
     state.Apply();
 
     // Mark framebuffer surfaces as dirty
-    MathUtil::Rectangle<u32> draw_rect_unscaled{
-        draw_rect.left / res_scale, draw_rect.top / res_scale, draw_rect.right / res_scale,
-        draw_rect.bottom / res_scale};
-
     if (color_surface != nullptr && write_color_fb) {
-        auto interval = color_surface->GetSubRectInterval(draw_rect_unscaled);
-        res_cache.InvalidateRegion(boost::icl::first(interval), boost::icl::length(interval),
-                                   color_surface);
+        res_cache.MarkSurfaceAsDirty(color_surface);
     }
     if (depth_surface != nullptr && write_depth_fb) {
-        auto interval = depth_surface->GetSubRectInterval(draw_rect_unscaled);
-        res_cache.InvalidateRegion(boost::icl::first(interval), boost::icl::length(interval),
-                                   depth_surface);
+        res_cache.MarkSurfaceAsDirty(depth_surface);
     }
 }
 
@@ -462,7 +451,7 @@ void RasterizerOpenGL::NotifyMaxwellRegisterChanged(u32 method) {}
 
 void RasterizerOpenGL::FlushAll() {
     MICROPROFILE_SCOPE(OpenGL_CacheManagement);
-    res_cache.FlushAll();
+    res_cache.FlushRegion(0, Kernel::VMManager::MAX_ADDRESS);
 }
 
 void RasterizerOpenGL::FlushRegion(Tegra::GPUVAddr addr, u64 size) {
@@ -472,13 +461,13 @@ void RasterizerOpenGL::FlushRegion(Tegra::GPUVAddr addr, u64 size) {
 
 void RasterizerOpenGL::InvalidateRegion(Tegra::GPUVAddr addr, u64 size) {
     MICROPROFILE_SCOPE(OpenGL_CacheManagement);
-    res_cache.InvalidateRegion(addr, size, nullptr);
+    res_cache.InvalidateRegion(addr, size);
 }
 
 void RasterizerOpenGL::FlushAndInvalidateRegion(Tegra::GPUVAddr addr, u64 size) {
     MICROPROFILE_SCOPE(OpenGL_CacheManagement);
     res_cache.FlushRegion(addr, size);
-    res_cache.InvalidateRegion(addr, size, nullptr);
+    res_cache.InvalidateRegion(addr, size);
 }
 
 bool RasterizerOpenGL::AccelerateDisplayTransfer(const void* config) {
@@ -497,45 +486,28 @@ bool RasterizerOpenGL::AccelerateFill(const void* config) {
     return true;
 }
 
-bool RasterizerOpenGL::AccelerateDisplay(const Tegra::FramebufferConfig& framebuffer,
+bool RasterizerOpenGL::AccelerateDisplay(const Tegra::FramebufferConfig& config,
                                          VAddr framebuffer_addr, u32 pixel_stride,
                                          ScreenInfo& screen_info) {
-    if (framebuffer_addr == 0) {
-        return false;
+    if (!framebuffer_addr) {
+        return {};
     }
+
     MICROPROFILE_SCOPE(OpenGL_CacheManagement);
 
-    SurfaceParams src_params;
-    src_params.cpu_addr = framebuffer_addr;
-    src_params.addr = res_cache.TryFindFramebufferGpuAddress(framebuffer_addr).get_value_or(0);
-    src_params.width = std::min(framebuffer.width, pixel_stride);
-    src_params.height = framebuffer.height;
-    src_params.stride = pixel_stride;
-    src_params.is_tiled = true;
-    src_params.block_height = Tegra::Texture::TICEntry::DefaultBlockHeight;
-    src_params.pixel_format =
-        SurfaceParams::PixelFormatFromGPUPixelFormat(framebuffer.pixel_format);
-    src_params.component_type =
-        SurfaceParams::ComponentTypeFromGPUPixelFormat(framebuffer.pixel_format);
-    src_params.UpdateParams();
-
-    MathUtil::Rectangle<u32> src_rect;
-    Surface src_surface;
-    std::tie(src_surface, src_rect) =
-        res_cache.GetSurfaceSubRect(src_params, ScaleMatch::Ignore, true);
-
-    if (src_surface == nullptr) {
-        return false;
+    const auto& surface{res_cache.TryFindFramebufferSurface(framebuffer_addr)};
+    if (!surface) {
+        return {};
     }
 
-    u32 scaled_width = src_surface->GetScaledWidth();
-    u32 scaled_height = src_surface->GetScaledHeight();
+    // Verify that the cached surface is the same size and format as the requested framebuffer
+    const auto& params{surface->GetSurfaceParams()};
+    const auto& pixel_format{SurfaceParams::PixelFormatFromGPUPixelFormat(config.pixel_format)};
+    ASSERT_MSG(params.width == config.width, "Framebuffer width is different");
+    ASSERT_MSG(params.height == config.height, "Framebuffer height is different");
+    ASSERT_MSG(params.pixel_format == pixel_format, "Framebuffer pixel_format is different");
 
-    screen_info.display_texcoords = MathUtil::Rectangle<float>(
-        (float)src_rect.bottom / (float)scaled_height, (float)src_rect.left / (float)scaled_width,
-        (float)src_rect.top / (float)scaled_height, (float)src_rect.right / (float)scaled_width);
-
-    screen_info.display_texture = src_surface->texture.handle;
+    screen_info.display_texture = surface->Texture().handle;
 
     return true;
 }
@@ -613,6 +585,12 @@ u32 RasterizerOpenGL::SetupConstBuffers(Maxwell::ShaderStage stage, GLuint progr
         if (used_buffer.IsIndirect()) {
             // Buffer is accessed indirectly, so upload the entire thing
             size = buffer.size * sizeof(float);
+
+            if (size > MaxConstbufferSize) {
+                NGLOG_ERROR(HW_GPU, "indirect constbuffer size {} exceeds maximum {}", size,
+                            MaxConstbufferSize);
+                size = MaxConstbufferSize;
+            }
         } else {
             // Buffer is accessed directly, upload just what we use
             size = used_buffer.GetSize() * sizeof(float);
@@ -663,12 +641,16 @@ u32 RasterizerOpenGL::SetupTextures(Maxwell::ShaderStage stage, GLuint program,
         glProgramUniform1i(program, uniform, current_bindpoint);
 
         const auto texture = maxwell3d.GetStageTexture(entry.GetStage(), entry.GetOffset());
-        ASSERT(texture.enabled);
+
+        if (!texture.enabled) {
+            state.texture_units[current_bindpoint].texture_2d = 0;
+            continue;
+        }
 
         texture_samplers[current_bindpoint].SyncWithConfig(texture.tsc);
         Surface surface = res_cache.GetTextureSurface(texture);
         if (surface != nullptr) {
-            state.texture_units[current_bindpoint].texture_2d = surface->texture.handle;
+            state.texture_units[current_bindpoint].texture_2d = surface->Texture().handle;
             state.texture_units[current_bindpoint].swizzle.r =
                 MaxwellToGL::SwizzleSource(texture.tic.x_source);
             state.texture_units[current_bindpoint].swizzle.g =
@@ -694,16 +676,16 @@ void RasterizerOpenGL::BindFramebufferSurfaces(const Surface& color_surface,
     state.Apply();
 
     glFramebufferTexture2D(GL_DRAW_FRAMEBUFFER, GL_COLOR_ATTACHMENT0, GL_TEXTURE_2D,
-                           color_surface != nullptr ? color_surface->texture.handle : 0, 0);
+                           color_surface != nullptr ? color_surface->Texture().handle : 0, 0);
     if (depth_surface != nullptr) {
         if (has_stencil) {
             // attach both depth and stencil
             glFramebufferTexture2D(GL_DRAW_FRAMEBUFFER, GL_DEPTH_STENCIL_ATTACHMENT, GL_TEXTURE_2D,
-                                   depth_surface->texture.handle, 0);
+                                   depth_surface->Texture().handle, 0);
         } else {
             // attach depth
             glFramebufferTexture2D(GL_DRAW_FRAMEBUFFER, GL_DEPTH_ATTACHMENT, GL_TEXTURE_2D,
-                                   depth_surface->texture.handle, 0);
+                                   depth_surface->Texture().handle, 0);
             // clear stencil attachment
             glFramebufferTexture2D(GL_DRAW_FRAMEBUFFER, GL_STENCIL_ATTACHMENT, GL_TEXTURE_2D, 0, 0);
         }
@@ -714,14 +696,14 @@ void RasterizerOpenGL::BindFramebufferSurfaces(const Surface& color_surface,
     }
 }
 
-void RasterizerOpenGL::SyncViewport(const MathUtil::Rectangle<u32>& surfaces_rect, u16 res_scale) {
+void RasterizerOpenGL::SyncViewport(const MathUtil::Rectangle<u32>& surfaces_rect) {
     const auto& regs = Core::System().GetInstance().GPU().Maxwell3D().regs;
     const MathUtil::Rectangle<s32> viewport_rect{regs.viewport_transform[0].GetRect()};
 
-    state.viewport.x = static_cast<GLint>(surfaces_rect.left) + viewport_rect.left * res_scale;
-    state.viewport.y = static_cast<GLint>(surfaces_rect.bottom) + viewport_rect.bottom * res_scale;
-    state.viewport.width = static_cast<GLsizei>(viewport_rect.GetWidth() * res_scale);
-    state.viewport.height = static_cast<GLsizei>(viewport_rect.GetHeight() * res_scale);
+    state.viewport.x = static_cast<GLint>(surfaces_rect.left) + viewport_rect.left;
+    state.viewport.y = static_cast<GLint>(surfaces_rect.bottom) + viewport_rect.bottom;
+    state.viewport.width = static_cast<GLsizei>(viewport_rect.GetWidth());
+    state.viewport.height = static_cast<GLsizei>(viewport_rect.GetHeight());
 }
 
 void RasterizerOpenGL::SyncClipEnabled() {
@@ -733,7 +715,11 @@ void RasterizerOpenGL::SyncClipCoef() {
 }
 
 void RasterizerOpenGL::SyncCullMode() {
-    UNREACHABLE();
+    const auto& regs = Core::System().GetInstance().GPU().Maxwell3D().regs;
+
+    state.cull.enabled = regs.cull.enabled != 0;
+    state.cull.front_face = MaxwellToGL::FrontFace(regs.cull.front_face);
+    state.cull.mode = MaxwellToGL::CullFace(regs.cull.cull_face);
 }
 
 void RasterizerOpenGL::SyncDepthScale() {
@@ -744,6 +730,14 @@ void RasterizerOpenGL::SyncDepthOffset() {
     UNREACHABLE();
 }
 
+void RasterizerOpenGL::SyncDepthTestState() {
+    const auto& regs = Core::System().GetInstance().GPU().Maxwell3D().regs;
+
+    state.depth.test_enabled = regs.depth_test_enable != 0;
+    state.depth.write_mask = regs.depth_write_enabled ? GL_TRUE : GL_FALSE;
+    state.depth.test_func = MaxwellToGL::ComparisonOp(regs.depth_test_func);
+}
+
 void RasterizerOpenGL::SyncBlendState() {
     const auto& regs = Core::System().GetInstance().GPU().Maxwell3D().regs;
 

src/video_core/renderer_opengl/gl_rasterizer.h

@@ -55,7 +55,7 @@ public:
     };
 
     /// Maximum supported size that a constbuffer can have in bytes.
-    static constexpr size_t MaxConstbufferSize = 0x1000;
+    static constexpr size_t MaxConstbufferSize = 0x10000;
     static_assert(MaxConstbufferSize % sizeof(GLvec4) == 0,
                   "The maximum size of a constbuffer must be a multiple of the size of GLvec4");
 
@@ -109,7 +109,7 @@ private:
                       u32 current_unit, const std::vector<GLShader::SamplerEntry>& entries);
 
     /// Syncs the viewport to match the guest state
-    void SyncViewport(const MathUtil::Rectangle<u32>& surfaces_rect, u16 res_scale);
+    void SyncViewport(const MathUtil::Rectangle<u32>& surfaces_rect);
 
     /// Syncs the clip enabled status to match the guest state
     void SyncClipEnabled();
@@ -126,6 +126,9 @@ private:
     /// Syncs the depth offset to match the guest state
     void SyncDepthOffset();
 
+    /// Syncs the depth test state to match the guest state
+    void SyncDepthTestState();
+
     /// Syncs the blend state to match the guest state
     void SyncBlendState();
 

src/video_core/renderer_opengl/gl_rasterizer_cache.h

@@ -1,57 +1,26 @@
-// Copyright 2015 Citra Emulator Project
+// Copyright 2018 yuzu Emulator Project
 // Licensed under GPLv2 or any later version
 // Refer to the license.txt file included.
 
 #pragma once
 
 #include <array>
+#include <map>
 #include <memory>
-#include <set>
-#include <tuple>
-#ifdef __GNUC__
-#pragma GCC diagnostic push
-#pragma GCC diagnostic ignored "-Wunused-local-typedefs"
-#endif
+#include <vector>
 #include <boost/icl/interval_map.hpp>
-#include <boost/icl/interval_set.hpp>
-#ifdef __GNUC__
-#pragma GCC diagnostic pop
-#endif
-#include <boost/optional.hpp>
-#include <glad/glad.h>
-#include "common/assert.h"
-#include "common/common_funcs.h"
 #include "common/common_types.h"
+#include "common/hash.h"
 #include "common/math_util.h"
-#include "video_core/gpu.h"
-#include "video_core/memory_manager.h"
+#include "video_core/engines/maxwell_3d.h"
 #include "video_core/renderer_opengl/gl_resource_manager.h"
 #include "video_core/textures/texture.h"
 
-struct CachedSurface;
+class CachedSurface;
 using Surface = std::shared_ptr<CachedSurface>;
-using SurfaceSet = std::set<Surface>;
-
-using SurfaceRegions = boost::icl::interval_set<Tegra::GPUVAddr>;
-using SurfaceMap = boost::icl::interval_map<Tegra::GPUVAddr, Surface>;
-using SurfaceCache = boost::icl::interval_map<Tegra::GPUVAddr, SurfaceSet>;
-
-using SurfaceInterval = SurfaceCache::interval_type;
-static_assert(std::is_same<SurfaceRegions::interval_type, SurfaceCache::interval_type>() &&
-                  std::is_same<SurfaceMap::interval_type, SurfaceCache::interval_type>(),
-              "incorrect interval types");
-
-using SurfaceRect_Tuple = std::tuple<Surface, MathUtil::Rectangle<u32>>;
 using SurfaceSurfaceRect_Tuple = std::tuple<Surface, Surface, MathUtil::Rectangle<u32>>;
-
 using PageMap = boost::icl::interval_map<u64, int>;
 
-enum class ScaleMatch {
-    Exact,   // only accept same res scale
-    Upscale, // only allow higher scale than params
-    Ignore   // accept every scaled res
-};
-
 struct SurfaceParams {
     enum class PixelFormat {
         ABGR8 = 0,
@@ -61,13 +30,21 @@ struct SurfaceParams {
         R8 = 4,
         RGBA16F = 5,
         R11FG11FB10F = 6,
-        DXT1 = 7,
-        DXT23 = 8,
-        DXT45 = 9,
-        DXN1 = 10, // This is also known as BC4
-        ASTC_2D_4X4 = 11,
+        RGBA32UI = 7,
+        DXT1 = 8,
+        DXT23 = 9,
+        DXT45 = 10,
+        DXN1 = 11, // This is also known as BC4
+        ASTC_2D_4X4 = 12,
 
-        Max,
+        MaxColorFormat,
+
+        // DepthStencil formats
+        Z24S8 = 13,
+
+        MaxDepthStencilFormat,
+
+        Max = MaxDepthStencilFormat,
         Invalid = 255,
     };
 
@@ -93,10 +70,10 @@ struct SurfaceParams {
     /**
      * Gets the compression factor for the specified PixelFormat. This applies to just the
      * "compressed width" and "compressed height", not the overall compression factor of a
-     * compressed image. This is used for maintaining proper surface sizes for compressed texture
-     * formats.
+     * compressed image. This is used for maintaining proper surface sizes for compressed
+     * texture formats.
      */
-    static constexpr u32 GetCompresssionFactor(PixelFormat format) {
+    static constexpr u32 GetCompressionFactor(PixelFormat format) {
         if (format == PixelFormat::Invalid)
             return 0;
 
@@ -108,19 +85,18 @@ struct SurfaceParams {
             1, // R8
             1, // RGBA16F
             1, // R11FG11FB10F
+            1, // RGBA32UI
             4, // DXT1
             4, // DXT23
             4, // DXT45
             4, // DXN1
-            1, // ASTC_2D_4X4
+            4, // ASTC_2D_4X4
+            1, // Z24S8
         }};
 
         ASSERT(static_cast<size_t>(format) < compression_factor_table.size());
         return compression_factor_table[static_cast<size_t>(format)];
     }
-    u32 GetCompresssionFactor() const {
-        return GetCompresssionFactor(pixel_format);
-    }
 
     static constexpr u32 GetFormatBpp(PixelFormat format) {
         if (format == PixelFormat::Invalid)
@@ -134,11 +110,13 @@ struct SurfaceParams {
             8,   // R8
             64,  // RGBA16F
             32,  // R11FG11FB10F
+            128, // RGBA32UI
             64,  // DXT1
             128, // DXT23
             128, // DXT45
             64,  // DXN1
             32,  // ASTC_2D_4X4
+            32,  // Z24S8
         }};
 
         ASSERT(static_cast<size_t>(format) < bpp_table.size());
@@ -148,6 +126,16 @@ struct SurfaceParams {
         return GetFormatBpp(pixel_format);
     }
 
+    static PixelFormat PixelFormatFromDepthFormat(Tegra::DepthFormat format) {
+        switch (format) {
+        case Tegra::DepthFormat::Z24_S8_UNORM:
+            return PixelFormat::Z24S8;
+        default:
+            NGLOG_CRITICAL(HW_GPU, "Unimplemented format={}", static_cast<u32>(format));
+            UNREACHABLE();
+        }
+    }
+
     static PixelFormat PixelFormatFromRenderTargetFormat(Tegra::RenderTargetFormat format) {
         switch (format) {
         case Tegra::RenderTargetFormat::RGBA8_UNORM:
@@ -159,25 +147,8 @@ struct SurfaceParams {
             return PixelFormat::RGBA16F;
         case Tegra::RenderTargetFormat::R11G11B10_FLOAT:
             return PixelFormat::R11FG11FB10F;
-        default:
-            NGLOG_CRITICAL(HW_GPU, "Unimplemented format={}", static_cast<u32>(format));
-            UNREACHABLE();
-        }
-    }
-
-    static bool IsFormatASTC(PixelFormat format) {
-        switch (format) {
-        case PixelFormat::ASTC_2D_4X4:
-            return true;
-        default:
-            return false;
-        }
-    }
-
-    static PixelFormat PixelFormatFromGPUPixelFormat(Tegra::FramebufferConfig::PixelFormat format) {
-        switch (format) {
-        case Tegra::FramebufferConfig::PixelFormat::ABGR8:
-            return PixelFormat::ABGR8;
+        case Tegra::RenderTargetFormat::RGBA32_UINT:
+            return PixelFormat::RGBA32UI;
         default:
             NGLOG_CRITICAL(HW_GPU, "Unimplemented format={}", static_cast<u32>(format));
             UNREACHABLE();
@@ -201,6 +172,8 @@ struct SurfaceParams {
             return PixelFormat::RGBA16F;
         case Tegra::Texture::TextureFormat::BF10GF11RF11:
             return PixelFormat::R11FG11FB10F;
+        case Tegra::Texture::TextureFormat::R32_G32_B32_A32:
+            return PixelFormat::RGBA32UI;
         case Tegra::Texture::TextureFormat::DXT1:
             return PixelFormat::DXT1;
         case Tegra::Texture::TextureFormat::DXT23:
@@ -234,6 +207,8 @@ struct SurfaceParams {
             return Tegra::Texture::TextureFormat::R16_G16_B16_A16;
         case PixelFormat::R11FG11FB10F:
             return Tegra::Texture::TextureFormat::BF10GF11RF11;
+        case PixelFormat::RGBA32UI:
+            return Tegra::Texture::TextureFormat::R32_G32_B32_A32;
         case PixelFormat::DXT1:
             return Tegra::Texture::TextureFormat::DXT1;
         case PixelFormat::DXT23:
@@ -249,6 +224,15 @@ struct SurfaceParams {
         }
     }
 
+    static Tegra::DepthFormat DepthFormatFromPixelFormat(PixelFormat format) {
+        switch (format) {
+        case PixelFormat::Z24S8:
+            return Tegra::DepthFormat::Z24_S8_UNORM;
+        default:
+            UNREACHABLE();
+        }
+    }
+
     static ComponentType ComponentTypeFromTexture(Tegra::Texture::ComponentType type) {
         // TODO(Subv): Implement more component types
         switch (type) {
@@ -270,16 +254,27 @@ struct SurfaceParams {
         case Tegra::RenderTargetFormat::RGBA16_FLOAT:
         case Tegra::RenderTargetFormat::R11G11B10_FLOAT:
             return ComponentType::Float;
+        case Tegra::RenderTargetFormat::RGBA32_UINT:
+            return ComponentType::UInt;
         default:
             NGLOG_CRITICAL(HW_GPU, "Unimplemented format={}", static_cast<u32>(format));
             UNREACHABLE();
         }
     }
 
-    static ComponentType ComponentTypeFromGPUPixelFormat(
-        Tegra::FramebufferConfig::PixelFormat format) {
+    static PixelFormat PixelFormatFromGPUPixelFormat(Tegra::FramebufferConfig::PixelFormat format) {
         switch (format) {
         case Tegra::FramebufferConfig::PixelFormat::ABGR8:
+            return PixelFormat::ABGR8;
+        default:
+            NGLOG_CRITICAL(HW_GPU, "Unimplemented format={}", static_cast<u32>(format));
+            UNREACHABLE();
+        }
+    }
+
+    static ComponentType ComponentTypeFromDepthFormat(Tegra::DepthFormat format) {
+        switch (format) {
+        case Tegra::DepthFormat::Z24_S8_UNORM:
             return ComponentType::UNorm;
         default:
             NGLOG_CRITICAL(HW_GPU, "Unimplemented format={}", static_cast<u32>(format));
@@ -287,198 +282,117 @@ struct SurfaceParams {
         }
     }
 
-    static bool CheckFormatsBlittable(PixelFormat pixel_format_a, PixelFormat pixel_format_b) {
-        SurfaceType a_type = GetFormatType(pixel_format_a);
-        SurfaceType b_type = GetFormatType(pixel_format_b);
-
-        if (a_type == SurfaceType::ColorTexture && b_type == SurfaceType::ColorTexture) {
-            return true;
-        }
-
-        if (a_type == SurfaceType::Depth && b_type == SurfaceType::Depth) {
-            return true;
-        }
-
-        if (a_type == SurfaceType::DepthStencil && b_type == SurfaceType::DepthStencil) {
-            return true;
-        }
-
-        return false;
-    }
-
     static SurfaceType GetFormatType(PixelFormat pixel_format) {
-        if (static_cast<size_t>(pixel_format) < MaxPixelFormat) {
+        if (static_cast<size_t>(pixel_format) < static_cast<size_t>(PixelFormat::MaxColorFormat)) {
             return SurfaceType::ColorTexture;
         }
 
+        if (static_cast<size_t>(pixel_format) <
+            static_cast<size_t>(PixelFormat::MaxDepthStencilFormat)) {
+            return SurfaceType::DepthStencil;
+        }
+
         // TODO(Subv): Implement the other formats
         ASSERT(false);
 
         return SurfaceType::Invalid;
     }
 
-    /// Update the params "size", "end" and "type" from the already set "addr", "width", "height"
-    /// and "pixel_format"
-    void UpdateParams() {
-        if (stride == 0) {
-            stride = width;
-        }
-        type = GetFormatType(pixel_format);
-        size = !is_tiled ? BytesInPixels(stride * (height - 1) + width)
-                         : BytesInPixels(stride * 8 * (height / 8 - 1) + width * 8);
-        end = addr + size;
-    }
-
-    SurfaceInterval GetInterval() const {
-        return SurfaceInterval::right_open(addr, end);
-    }
-
-    // Returns the outer rectangle containing "interval"
-    SurfaceParams FromInterval(SurfaceInterval interval) const;
-
-    SurfaceInterval GetSubRectInterval(MathUtil::Rectangle<u32> unscaled_rect) const;
-
-    // Returns the region of the biggest valid rectange within interval
-    SurfaceInterval GetCopyableInterval(const Surface& src_surface) const;
-
-    /**
-     * Gets the actual width (in pixels) of the surface. This is provided because `width` is used
-     * for tracking the surface region in memory, which may be compressed for certain formats. In
-     * this scenario, `width` is actually the compressed width.
-     */
-    u32 GetActualWidth() const {
-        return width * GetCompresssionFactor();
-    }
-
-    /**
-     * Gets the actual height (in pixels) of the surface. This is provided because `height` is used
-     * for tracking the surface region in memory, which may be compressed for certain formats. In
-     * this scenario, `height` is actually the compressed height.
-     */
-    u32 GetActualHeight() const {
-        return height * GetCompresssionFactor();
-    }
-
-    u32 GetScaledWidth() const {
-        return width * res_scale;
-    }
-
-    u32 GetScaledHeight() const {
-        return height * res_scale;
-    }
-
-    MathUtil::Rectangle<u32> GetRect() const {
-        return {0, height, width, 0};
-    }
-
-    MathUtil::Rectangle<u32> GetScaledRect() const {
-        return {0, GetScaledHeight(), GetScaledWidth(), 0};
-    }
-
-    u64 PixelsInBytes(u64 size) const {
-        return size * CHAR_BIT / GetFormatBpp(pixel_format);
-    }
-
-    u64 BytesInPixels(u64 pixels) const {
-        return pixels * GetFormatBpp(pixel_format) / CHAR_BIT;
+    /// Returns the rectangle corresponding to this surface
+    MathUtil::Rectangle<u32> GetRect() const;
+
+    /// Returns the size of this surface in bytes, adjusted for compression
+    size_t SizeInBytes() const {
+        const u32 compression_factor{GetCompressionFactor(pixel_format)};
+        ASSERT(width % compression_factor == 0);
+        ASSERT(height % compression_factor == 0);
+        return (width / compression_factor) * (height / compression_factor) *
+               GetFormatBpp(pixel_format) / CHAR_BIT;
     }
 
+    /// Returns the CPU virtual address for this surface
     VAddr GetCpuAddr() const;
 
-    bool ExactMatch(const SurfaceParams& other_surface) const;
-    bool CanSubRect(const SurfaceParams& sub_surface) const;
-    bool CanExpand(const SurfaceParams& expanded_surface) const;
-    bool CanTexCopy(const SurfaceParams& texcopy_params) const;
+    /// Returns true if the specified region overlaps with this surface's region in Switch memory
+    bool IsOverlappingRegion(Tegra::GPUVAddr region_addr, size_t region_size) const {
+        return addr <= (region_addr + region_size) && region_addr <= (addr + size_in_bytes);
+    }
 
-    MathUtil::Rectangle<u32> GetSubRect(const SurfaceParams& sub_surface) const;
-    MathUtil::Rectangle<u32> GetScaledSubRect(const SurfaceParams& sub_surface) const;
+    /// Creates SurfaceParams from a texture configation
+    static SurfaceParams CreateForTexture(const Tegra::Texture::FullTextureInfo& config);
 
-    Tegra::GPUVAddr addr = 0;
-    Tegra::GPUVAddr end = 0;
-    boost::optional<VAddr> cpu_addr;
-    u64 size = 0;
+    /// Creates SurfaceParams from a framebuffer configation
+    static SurfaceParams CreateForFramebuffer(
+        const Tegra::Engines::Maxwell3D::Regs::RenderTargetConfig& config);
 
-    u32 width = 0;
-    u32 height = 0;
-    u32 stride = 0;
-    u32 block_height = 0;
-    u16 res_scale = 1;
-
-    bool is_tiled = false;
-    PixelFormat pixel_format = PixelFormat::Invalid;
-    SurfaceType type = SurfaceType::Invalid;
-    ComponentType component_type = ComponentType::Invalid;
+    Tegra::GPUVAddr addr;
+    bool is_tiled;
+    u32 block_height;
+    PixelFormat pixel_format;
+    ComponentType component_type;
+    SurfaceType type;
+    u32 width;
+    u32 height;
+    u32 unaligned_height;
+    size_t size_in_bytes;
 };
 
-struct CachedSurface : SurfaceParams {
-    bool CanFill(const SurfaceParams& dest_surface, SurfaceInterval fill_interval) const;
-    bool CanCopy(const SurfaceParams& dest_surface, SurfaceInterval copy_interval) const;
+/// Hashable variation of SurfaceParams, used for a key in the surface cache
+struct SurfaceKey : Common::HashableStruct<SurfaceParams> {
+    static SurfaceKey Create(const SurfaceParams& params) {
+        SurfaceKey res;
+        res.state = params;
+        return res;
+    }
+};
 
-    bool IsRegionValid(SurfaceInterval interval) const {
-        return (invalid_regions.find(interval) == invalid_regions.end());
+namespace std {
+template <>
+struct hash<SurfaceKey> {
+    size_t operator()(const SurfaceKey& k) const {
+        return k.Hash();
+    }
+};
+} // namespace std
+
+class CachedSurface final {
+public:
+    CachedSurface(const SurfaceParams& params);
+
+    const OGLTexture& Texture() const {
+        return texture;
     }
 
-    bool IsSurfaceFullyInvalid() const {
-        return (invalid_regions & GetInterval()) == SurfaceRegions(GetInterval());
-    }
-
-    bool registered = false;
-    SurfaceRegions invalid_regions;
-
-    u64 fill_size = 0; /// Number of bytes to read from fill_data
-    std::array<u8, 4> fill_data;
-
-    OGLTexture texture;
-
-    static constexpr unsigned int GetGLBytesPerPixel(PixelFormat format) {
-        if (format == PixelFormat::Invalid)
+    static constexpr unsigned int GetGLBytesPerPixel(SurfaceParams::PixelFormat format) {
+        if (format == SurfaceParams::PixelFormat::Invalid)
             return 0;
 
         return SurfaceParams::GetFormatBpp(format) / CHAR_BIT;
     }
 
-    std::unique_ptr<u8[]> gl_buffer;
-    size_t gl_buffer_size = 0;
+    const SurfaceParams& GetSurfaceParams() const {
+        return params;
+    }
 
     // Read/Write data in Switch memory to/from gl_buffer
-    void LoadGLBuffer(Tegra::GPUVAddr load_start, Tegra::GPUVAddr load_end);
-    void FlushGLBuffer(Tegra::GPUVAddr flush_start, Tegra::GPUVAddr flush_end);
+    void LoadGLBuffer();
+    void FlushGLBuffer();
 
     // Upload/Download data in gl_buffer in/to this surface's texture
-    void UploadGLTexture(const MathUtil::Rectangle<u32>& rect, GLuint read_fb_handle,
-                         GLuint draw_fb_handle);
-    void DownloadGLTexture(const MathUtil::Rectangle<u32>& rect, GLuint read_fb_handle,
-                           GLuint draw_fb_handle);
+    void UploadGLTexture(GLuint read_fb_handle, GLuint draw_fb_handle);
+    void DownloadGLTexture(GLuint read_fb_handle, GLuint draw_fb_handle);
+
+private:
+    OGLTexture texture;
+    std::vector<u8> gl_buffer;
+    SurfaceParams params;
 };
 
-class RasterizerCacheOpenGL : NonCopyable {
+class RasterizerCacheOpenGL final : NonCopyable {
 public:
     RasterizerCacheOpenGL();
     ~RasterizerCacheOpenGL();
 
-    /// Blit one surface's texture to another
-    bool BlitSurfaces(const Surface& src_surface, const MathUtil::Rectangle<u32>& src_rect,
-                      const Surface& dst_surface, const MathUtil::Rectangle<u32>& dst_rect);
-
-    void ConvertD24S8toABGR(GLuint src_tex, const MathUtil::Rectangle<u32>& src_rect,
-                            GLuint dst_tex, const MathUtil::Rectangle<u32>& dst_rect);
-
-    /// Copy one surface's region to another
-    void CopySurface(const Surface& src_surface, const Surface& dst_surface,
-                     SurfaceInterval copy_interval);
-
-    /// Load a texture from Switch memory to OpenGL and cache it (if not already cached)
-    Surface GetSurface(const SurfaceParams& params, ScaleMatch match_res_scale,
-                       bool load_if_create);
-
-    /// Tries to find a framebuffer GPU address based on the provided CPU address
-    boost::optional<Tegra::GPUVAddr> TryFindFramebufferGpuAddress(VAddr cpu_addr) const;
-
-    /// Attempt to find a subrect (resolution scaled) of a surface, otherwise loads a texture from
-    /// Switch memory to OpenGL and caches it (if not already cached)
-    SurfaceRect_Tuple GetSurfaceSubRect(const SurfaceParams& params, ScaleMatch match_res_scale,
-                                        bool load_if_create);
-
     /// Get a surface based on the texture configuration
     Surface GetTextureSurface(const Tegra::Texture::FullTextureInfo& config);
 
@@ -486,29 +400,21 @@ public:
     SurfaceSurfaceRect_Tuple GetFramebufferSurfaces(bool using_color_fb, bool using_depth_fb,
                                                     const MathUtil::Rectangle<s32>& viewport);
 
-    /// Get a surface that matches the fill config
-    Surface GetFillSurface(const void* config);
+    /// Marks the specified surface as "dirty", in that it is out of sync with Switch memory
+    void MarkSurfaceAsDirty(const Surface& surface);
 
-    /// Get a surface that matches a "texture copy" display transfer config
-    SurfaceRect_Tuple GetTexCopySurface(const SurfaceParams& params);
+    /// Tries to find a framebuffer GPU address based on the provided CPU address
+    Surface TryFindFramebufferSurface(VAddr cpu_addr) const;
 
     /// Write any cached resources overlapping the region back to memory (if dirty)
-    void FlushRegion(Tegra::GPUVAddr addr, u64 size, Surface flush_surface = nullptr);
+    void FlushRegion(Tegra::GPUVAddr addr, size_t size);
 
-    /// Mark region as being invalidated by region_owner (nullptr if Switch memory)
-    void InvalidateRegion(Tegra::GPUVAddr addr, u64 size, const Surface& region_owner);
-
-    /// Flush all cached resources tracked by this cache manager
-    void FlushAll();
+    /// Mark the specified region as being invalidated
+    void InvalidateRegion(Tegra::GPUVAddr addr, size_t size);
 
 private:
-    void DuplicateSurface(const Surface& src_surface, const Surface& dest_surface);
-
-    /// Update surface's texture for given region when necessary
-    void ValidateSurface(const Surface& surface, Tegra::GPUVAddr addr, u64 size);
-
-    /// Create a new surface
-    Surface CreateSurface(const SurfaceParams& params);
+    void LoadSurface(const Surface& surface);
+    Surface GetSurface(const SurfaceParams& params);
 
     /// Register surface into the cache
     void RegisterSurface(const Surface& surface);
@@ -519,18 +425,9 @@ private:
     /// Increase/decrease the number of surface in pages touching the specified region
     void UpdatePagesCachedCount(Tegra::GPUVAddr addr, u64 size, int delta);
 
-    SurfaceCache surface_cache;
+    std::unordered_map<SurfaceKey, Surface> surface_cache;
     PageMap cached_pages;
-    SurfaceMap dirty_regions;
-    SurfaceSet remove_surfaces;
 
     OGLFramebuffer read_framebuffer;
     OGLFramebuffer draw_framebuffer;
-
-    OGLVertexArray attributeless_vao;
-    OGLBuffer d24s8_abgr_buffer;
-    GLsizeiptr d24s8_abgr_buffer_size;
-    OGLProgram d24s8_abgr_shader;
-    GLint d24s8_abgr_tbo_size_u_id;
-    GLint d24s8_abgr_viewport_u_id;
 };

src/video_core/renderer_opengl/gl_shader_decompiler.cpp

@@ -719,21 +719,31 @@ private:
     /**
      * Returns the comparison string to use to compare two values in the 'set' family of
      * instructions.
-     * @params condition The condition used in the 'set'-family instruction.
+     * @param condition The condition used in the 'set'-family instruction.
+     * @param op_a First operand to use for the comparison.
+     * @param op_b Second operand to use for the comparison.
      * @returns String corresponding to the GLSL operator that matches the desired comparison.
      */
-    std::string GetPredicateComparison(Tegra::Shader::PredCondition condition) const {
+    std::string GetPredicateComparison(Tegra::Shader::PredCondition condition,
+                                       const std::string& op_a, const std::string& op_b) const {
         using Tegra::Shader::PredCondition;
         static const std::unordered_map<PredCondition, const char*> PredicateComparisonStrings = {
-            {PredCondition::LessThan, "<"},   {PredCondition::Equal, "=="},
-            {PredCondition::LessEqual, "<="}, {PredCondition::GreaterThan, ">"},
-            {PredCondition::NotEqual, "!="},  {PredCondition::GreaterEqual, ">="},
+            {PredCondition::LessThan, "<"},         {PredCondition::Equal, "=="},
+            {PredCondition::LessEqual, "<="},       {PredCondition::GreaterThan, ">"},
+            {PredCondition::NotEqual, "!="},        {PredCondition::GreaterEqual, ">="},
+            {PredCondition::NotEqualWithNan, "!="},
         };
 
-        auto comparison = PredicateComparisonStrings.find(condition);
+        const auto& comparison{PredicateComparisonStrings.find(condition)};
         ASSERT_MSG(comparison != PredicateComparisonStrings.end(),
                    "Unknown predicate comparison operation");
-        return comparison->second;
+
+        std::string predicate{'(' + op_a + ") " + comparison->second + " (" + op_b + ')'};
+        if (condition == PredCondition::NotEqualWithNan) {
+            predicate += " || isnan(" + op_a + ") || isnan(" + op_b + ')';
+        }
+
+        return predicate;
     }
 
     /**
@@ -811,6 +821,7 @@ private:
         if (!opcode) {
             NGLOG_CRITICAL(HW_GPU, "Unhandled instruction: {0:x}", instr.value);
             UNREACHABLE();
+            return offset + 1;
         }
 
         shader.AddLine("// " + std::to_string(offset) + ": " + opcode->GetName());
@@ -906,8 +917,8 @@ private:
                     regs.SetRegisterToFloat(instr.gpr0, 0, "inversesqrt(" + op_a + ')', 1, 1,
                                             instr.alu.saturate_d);
                     break;
-                case SubOp::Min:
-                    regs.SetRegisterToFloat(instr.gpr0, 0, "min(" + op_a + "," + op_b + ')', 1, 1,
+                case SubOp::Sqrt:
+                    regs.SetRegisterToFloat(instr.gpr0, 0, "sqrt(" + op_a + ')', 1, 1,
                                             instr.alu.saturate_d);
                     break;
                 default:
@@ -1414,10 +1425,9 @@ private:
             std::string second_pred =
                 GetPredicateCondition(instr.fsetp.pred39, instr.fsetp.neg_pred != 0);
 
-            std::string comparator = GetPredicateComparison(instr.fsetp.cond);
             std::string combiner = GetPredicateCombiner(instr.fsetp.op);
 
-            std::string predicate = '(' + op_a + ") " + comparator + " (" + op_b + ')';
+            std::string predicate = GetPredicateComparison(instr.fsetp.cond, op_a, op_b);
             // Set the primary predicate to the result of Predicate OP SecondPredicate
             SetPredicate(instr.fsetp.pred3,
                          '(' + predicate + ") " + combiner + " (" + second_pred + ')');
@@ -1452,10 +1462,9 @@ private:
             std::string second_pred =
                 GetPredicateCondition(instr.isetp.pred39, instr.isetp.neg_pred != 0);
 
-            std::string comparator = GetPredicateComparison(instr.isetp.cond);
             std::string combiner = GetPredicateCombiner(instr.isetp.op);
 
-            std::string predicate = '(' + op_a + ") " + comparator + " (" + op_b + ')';
+            std::string predicate = GetPredicateComparison(instr.isetp.cond, op_a, op_b);
             // Set the primary predicate to the result of Predicate OP SecondPredicate
             SetPredicate(instr.isetp.pred3,
                          '(' + predicate + ") " + combiner + " (" + second_pred + ')');
@@ -1502,11 +1511,10 @@ private:
             std::string second_pred =
                 GetPredicateCondition(instr.fset.pred39, instr.fset.neg_pred != 0);
 
-            std::string comparator = GetPredicateComparison(instr.fset.cond);
             std::string combiner = GetPredicateCombiner(instr.fset.op);
 
-            std::string predicate = "(((" + op_a + ") " + comparator + " (" + op_b + ")) " +
-                                    combiner + " (" + second_pred + "))";
+            std::string predicate = "((" + GetPredicateComparison(instr.fset.cond, op_a, op_b) +
+                                    ") " + combiner + " (" + second_pred + "))";
 
             if (instr.fset.bf) {
                 regs.SetRegisterToFloat(instr.gpr0, 0, predicate + " ? 1.0 : 0.0", 1, 1);
@@ -1537,11 +1545,10 @@ private:
             std::string second_pred =
                 GetPredicateCondition(instr.iset.pred39, instr.iset.neg_pred != 0);
 
-            std::string comparator = GetPredicateComparison(instr.iset.cond);
             std::string combiner = GetPredicateCombiner(instr.iset.op);
 
-            std::string predicate = "(((" + op_a + ") " + comparator + " (" + op_b + ")) " +
-                                    combiner + " (" + second_pred + "))";
+            std::string predicate = "((" + GetPredicateComparison(instr.iset.cond, op_a, op_b) +
+                                    ") " + combiner + " (" + second_pred + "))";
 
             if (instr.iset.bf) {
                 regs.SetRegisterToFloat(instr.gpr0, 0, predicate + " ? 1.0 : 0.0", 1, 1);

src/video_core/renderer_opengl/gl_state.cpp

@@ -51,16 +51,6 @@ OpenGLState::OpenGLState() {
         texture_unit.Reset();
     }
 
-    lighting_lut.texture_buffer = 0;
-
-    fog_lut.texture_buffer = 0;
-
-    proctex_lut.texture_buffer = 0;
-    proctex_diff_lut.texture_buffer = 0;
-    proctex_color_map.texture_buffer = 0;
-    proctex_alpha_map.texture_buffer = 0;
-    proctex_noise_lut.texture_buffer = 0;
-
     draw.read_framebuffer = 0;
     draw.draw_framebuffer = 0;
     draw.vertex_array = 0;
@@ -224,48 +214,6 @@ void OpenGLState::Apply() const {
         }
     }
 
-    // Lighting LUTs
-    if (lighting_lut.texture_buffer != cur_state.lighting_lut.texture_buffer) {
-        glActiveTexture(TextureUnits::LightingLUT.Enum());
-        glBindTexture(GL_TEXTURE_BUFFER, lighting_lut.texture_buffer);
-    }
-
-    // Fog LUT
-    if (fog_lut.texture_buffer != cur_state.fog_lut.texture_buffer) {
-        glActiveTexture(TextureUnits::FogLUT.Enum());
-        glBindTexture(GL_TEXTURE_BUFFER, fog_lut.texture_buffer);
-    }
-
-    // ProcTex Noise LUT
-    if (proctex_noise_lut.texture_buffer != cur_state.proctex_noise_lut.texture_buffer) {
-        glActiveTexture(TextureUnits::ProcTexNoiseLUT.Enum());
-        glBindTexture(GL_TEXTURE_BUFFER, proctex_noise_lut.texture_buffer);
-    }
-
-    // ProcTex Color Map
-    if (proctex_color_map.texture_buffer != cur_state.proctex_color_map.texture_buffer) {
-        glActiveTexture(TextureUnits::ProcTexColorMap.Enum());
-        glBindTexture(GL_TEXTURE_BUFFER, proctex_color_map.texture_buffer);
-    }
-
-    // ProcTex Alpha Map
-    if (proctex_alpha_map.texture_buffer != cur_state.proctex_alpha_map.texture_buffer) {
-        glActiveTexture(TextureUnits::ProcTexAlphaMap.Enum());
-        glBindTexture(GL_TEXTURE_BUFFER, proctex_alpha_map.texture_buffer);
-    }
-
-    // ProcTex LUT
-    if (proctex_lut.texture_buffer != cur_state.proctex_lut.texture_buffer) {
-        glActiveTexture(TextureUnits::ProcTexLUT.Enum());
-        glBindTexture(GL_TEXTURE_BUFFER, proctex_lut.texture_buffer);
-    }
-
-    // ProcTex Diff LUT
-    if (proctex_diff_lut.texture_buffer != cur_state.proctex_diff_lut.texture_buffer) {
-        glActiveTexture(TextureUnits::ProcTexDiffLUT.Enum());
-        glBindTexture(GL_TEXTURE_BUFFER, proctex_diff_lut.texture_buffer);
-    }
-
     // Framebuffer
     if (draw.read_framebuffer != cur_state.draw.read_framebuffer) {
         glBindFramebuffer(GL_READ_FRAMEBUFFER, draw.read_framebuffer);
@@ -339,20 +287,6 @@ OpenGLState& OpenGLState::UnbindTexture(GLuint handle) {
             unit.Unbind();
         }
     }
-    if (lighting_lut.texture_buffer == handle)
-        lighting_lut.texture_buffer = 0;
-    if (fog_lut.texture_buffer == handle)
-        fog_lut.texture_buffer = 0;
-    if (proctex_noise_lut.texture_buffer == handle)
-        proctex_noise_lut.texture_buffer = 0;
-    if (proctex_color_map.texture_buffer == handle)
-        proctex_color_map.texture_buffer = 0;
-    if (proctex_alpha_map.texture_buffer == handle)
-        proctex_alpha_map.texture_buffer = 0;
-    if (proctex_lut.texture_buffer == handle)
-        proctex_lut.texture_buffer = 0;
-    if (proctex_diff_lut.texture_buffer == handle)
-        proctex_diff_lut.texture_buffer = 0;
     return *this;
 }
 

src/video_core/renderer_opengl/gl_state.h

@@ -106,34 +106,6 @@ public:
         }
     } texture_units[32];
 
-    struct {
-        GLuint texture_buffer; // GL_TEXTURE_BINDING_BUFFER
-    } lighting_lut;
-
-    struct {
-        GLuint texture_buffer; // GL_TEXTURE_BINDING_BUFFER
-    } fog_lut;
-
-    struct {
-        GLuint texture_buffer; // GL_TEXTURE_BINDING_BUFFER
-    } proctex_noise_lut;
-
-    struct {
-        GLuint texture_buffer; // GL_TEXTURE_BINDING_BUFFER
-    } proctex_color_map;
-
-    struct {
-        GLuint texture_buffer; // GL_TEXTURE_BINDING_BUFFER
-    } proctex_alpha_map;
-
-    struct {
-        GLuint texture_buffer; // GL_TEXTURE_BINDING_BUFFER
-    } proctex_lut;
-
-    struct {
-        GLuint texture_buffer; // GL_TEXTURE_BINDING_BUFFER
-    } proctex_diff_lut;
-
     struct {
         GLuint read_framebuffer; // GL_READ_FRAMEBUFFER_BINDING
         GLuint draw_framebuffer; // GL_DRAW_FRAMEBUFFER_BINDING

src/video_core/renderer_opengl/maxwell_to_gl.h

@@ -201,4 +201,54 @@ inline GLenum SwizzleSource(Tegra::Texture::SwizzleSource source) {
     return {};
 }
 
+inline GLenum ComparisonOp(Maxwell::ComparisonOp comparison) {
+    switch (comparison) {
+    case Maxwell::ComparisonOp::Never:
+        return GL_NEVER;
+    case Maxwell::ComparisonOp::Less:
+        return GL_LESS;
+    case Maxwell::ComparisonOp::Equal:
+        return GL_EQUAL;
+    case Maxwell::ComparisonOp::LessEqual:
+        return GL_LEQUAL;
+    case Maxwell::ComparisonOp::Greater:
+        return GL_GREATER;
+    case Maxwell::ComparisonOp::NotEqual:
+        return GL_NOTEQUAL;
+    case Maxwell::ComparisonOp::GreaterEqual:
+        return GL_GEQUAL;
+    case Maxwell::ComparisonOp::Always:
+        return GL_ALWAYS;
+    }
+    NGLOG_CRITICAL(Render_OpenGL, "Unimplemented comparison op={}", static_cast<u32>(comparison));
+    UNREACHABLE();
+    return {};
+}
+
+inline GLenum FrontFace(Maxwell::Cull::FrontFace front_face) {
+    switch (front_face) {
+    case Maxwell::Cull::FrontFace::ClockWise:
+        return GL_CW;
+    case Maxwell::Cull::FrontFace::CounterClockWise:
+        return GL_CCW;
+    }
+    NGLOG_CRITICAL(Render_OpenGL, "Unimplemented front face cull={}", static_cast<u32>(front_face));
+    UNREACHABLE();
+    return {};
+}
+
+inline GLenum CullFace(Maxwell::Cull::CullFace cull_face) {
+    switch (cull_face) {
+    case Maxwell::Cull::CullFace::Front:
+        return GL_FRONT;
+    case Maxwell::Cull::CullFace::Back:
+        return GL_BACK;
+    case Maxwell::Cull::CullFace::FrontAndBack:
+        return GL_FRONT_AND_BACK;
+    }
+    NGLOG_CRITICAL(Render_OpenGL, "Unimplemented cull face={}", static_cast<u32>(cull_face));
+    UNREACHABLE();
+    return {};
+}
+
 } // namespace MaxwellToGL

src/video_core/renderer_opengl/renderer_opengl.cpp

@@ -150,7 +150,6 @@ void RendererOpenGL::LoadFBToScreenInfo(const Tegra::FramebufferConfig& framebuf
                                          screen_info)) {
         // Reset the screen info's display texture to its own permanent texture
         screen_info.display_texture = screen_info.texture.resource.handle;
-        screen_info.display_texcoords = MathUtil::Rectangle<float>(0.f, 0.f, 1.f, 1.f);
 
         Memory::RasterizerFlushVirtualRegion(framebuffer_addr, size_in_bytes,
                                              Memory::FlushMode::Flush);

src/video_core/renderer_opengl/renderer_opengl.h

@@ -27,7 +27,7 @@ struct TextureInfo {
 /// Structure used for storing information about the display target for the Switch screen
 struct ScreenInfo {
     GLuint display_texture;
-    MathUtil::Rectangle<float> display_texcoords;
+    const MathUtil::Rectangle<float> display_texcoords{0.0f, 0.0f, 1.0f, 1.0f};
     TextureInfo texture;
 };
 

src/video_core/textures/decoders.cpp

@@ -5,6 +5,7 @@
 #include <cstring>
 #include "common/assert.h"
 #include "core/memory.h"
+#include "video_core/gpu.h"
 #include "video_core/textures/decoders.h"
 #include "video_core/textures/texture.h"
 
@@ -65,6 +66,18 @@ u32 BytesPerPixel(TextureFormat format) {
         return 1;
     case TextureFormat::R16_G16_B16_A16:
         return 8;
+    case TextureFormat::R32_G32_B32_A32:
+        return 16;
+    default:
+        UNIMPLEMENTED_MSG("Format not implemented");
+        break;
+    }
+}
+
+static u32 DepthBytesPerPixel(DepthFormat format) {
+    switch (format) {
+    case DepthFormat::Z24_S8_UNORM:
+        return 4;
     default:
         UNIMPLEMENTED_MSG("Format not implemented");
         break;
@@ -94,6 +107,7 @@ std::vector<u8> UnswizzleTexture(VAddr address, TextureFormat format, u32 width,
     case TextureFormat::B5G6R5:
     case TextureFormat::R8:
     case TextureFormat::R16_G16_B16_A16:
+    case TextureFormat::R32_G32_B32_A32:
     case TextureFormat::BF10GF11RF11:
     case TextureFormat::ASTC_2D_4X4:
         CopySwizzledData(width, height, bytes_per_pixel, bytes_per_pixel, data,
@@ -107,6 +121,26 @@ std::vector<u8> UnswizzleTexture(VAddr address, TextureFormat format, u32 width,
     return unswizzled_data;
 }
 
+std::vector<u8> UnswizzleDepthTexture(VAddr address, DepthFormat format, u32 width, u32 height,
+                                      u32 block_height) {
+    u8* data = Memory::GetPointer(address);
+    u32 bytes_per_pixel = DepthBytesPerPixel(format);
+
+    std::vector<u8> unswizzled_data(width * height * bytes_per_pixel);
+
+    switch (format) {
+    case DepthFormat::Z24_S8_UNORM:
+        CopySwizzledData(width, height, bytes_per_pixel, bytes_per_pixel, data,
+                         unswizzled_data.data(), true, block_height);
+        break;
+    default:
+        UNIMPLEMENTED_MSG("Format not implemented");
+        break;
+    }
+
+    return unswizzled_data;
+}
+
 std::vector<u8> DecodeTexture(const std::vector<u8>& texture_data, TextureFormat format, u32 width,
                               u32 height) {
     std::vector<u8> rgba_data;
@@ -124,6 +158,7 @@ std::vector<u8> DecodeTexture(const std::vector<u8>& texture_data, TextureFormat
     case TextureFormat::B5G6R5:
     case TextureFormat::R8:
     case TextureFormat::BF10GF11RF11:
+    case TextureFormat::R32_G32_B32_A32:
         // TODO(Subv): For the time being just forward the same data without any decoding.
         rgba_data = texture_data;
         break;

src/video_core/textures/decoders.h

@@ -17,6 +17,12 @@ namespace Texture {
 std::vector<u8> UnswizzleTexture(VAddr address, TextureFormat format, u32 width, u32 height,
                                  u32 block_height = TICEntry::DefaultBlockHeight);
 
+/**
+ * Unswizzles a swizzled depth texture without changing its format.
+ */
+std::vector<u8> UnswizzleDepthTexture(VAddr address, DepthFormat format, u32 width, u32 height,
+                                      u32 block_height = TICEntry::DefaultBlockHeight);
+
 /// Copies texture data from a buffer and performs swizzling/unswizzling as necessary.
 void CopySwizzledData(u32 width, u32 height, u32 bytes_per_pixel, u32 out_bytes_per_pixel,
                       u8* swizzled_data, u8* unswizzled_data, bool unswizzle, u32 block_height);

src/yuzu/configuration/config.cpp

@@ -84,6 +84,8 @@ void Config::ReadValues() {
     qt_config->beginGroup("Renderer");
     Settings::values.resolution_factor = qt_config->value("resolution_factor", 1.0).toFloat();
     Settings::values.toggle_framelimit = qt_config->value("toggle_framelimit", true).toBool();
+    Settings::values.use_accurate_framebuffers =
+        qt_config->value("use_accurate_framebuffers", false).toBool();
 
     Settings::values.bg_red = qt_config->value("bg_red", 0.0).toFloat();
     Settings::values.bg_green = qt_config->value("bg_green", 0.0).toFloat();
@@ -184,6 +186,7 @@ void Config::SaveValues() {
     qt_config->beginGroup("Renderer");
     qt_config->setValue("resolution_factor", (double)Settings::values.resolution_factor);
     qt_config->setValue("toggle_framelimit", Settings::values.toggle_framelimit);
+    qt_config->setValue("use_accurate_framebuffers", Settings::values.use_accurate_framebuffers);
 
     // Cast to double because Qt's written float values are not human-readable
     qt_config->setValue("bg_red", (double)Settings::values.bg_red);

src/yuzu/configuration/configure_graphics.cpp

@@ -59,11 +59,13 @@ void ConfigureGraphics::setConfiguration() {
     ui->resolution_factor_combobox->setCurrentIndex(
         static_cast<int>(FromResolutionFactor(Settings::values.resolution_factor)));
     ui->toggle_framelimit->setChecked(Settings::values.toggle_framelimit);
+    ui->use_accurate_framebuffers->setChecked(Settings::values.use_accurate_framebuffers);
 }
 
 void ConfigureGraphics::applyConfiguration() {
     Settings::values.resolution_factor =
         ToResolutionFactor(static_cast<Resolution>(ui->resolution_factor_combobox->currentIndex()));
     Settings::values.toggle_framelimit = ui->toggle_framelimit->isChecked();
+    Settings::values.use_accurate_framebuffers = ui->use_accurate_framebuffers->isChecked();
     Settings::Apply();
 }

src/yuzu/configuration/configure_graphics.ui

@@ -29,6 +29,13 @@
           </property>
          </widget>
         </item>
+        <item>
+         <widget class="QCheckBox" name="use_accurate_framebuffers">
+          <property name="text">
+           <string>Use accurate framebuffers (slow)</string>
+          </property>
+         </widget>
+        </item>
         <item>
          <layout class="QHBoxLayout" name="horizontalLayout">
           <item>

src/yuzu_cmd/config.cpp

@@ -98,6 +98,8 @@ void Config::ReadValues() {
         (float)sdl2_config->GetReal("Renderer", "resolution_factor", 1.0);
     Settings::values.toggle_framelimit =
         sdl2_config->GetBoolean("Renderer", "toggle_framelimit", true);
+    Settings::values.use_accurate_framebuffers =
+        sdl2_config->GetBoolean("Renderer", "use_accurate_framebuffers", false);
 
     Settings::values.bg_red = (float)sdl2_config->GetReal("Renderer", "bg_red", 0.0);
     Settings::values.bg_green = (float)sdl2_config->GetReal("Renderer", "bg_green", 0.0);

src/yuzu_cmd/default_ini.h

@@ -102,6 +102,10 @@ resolution_factor =
 # 0 (default): Off, 1: On
 use_vsync =
 
+# Whether to use accurate framebuffers
+# 0 (default): Off (fast), 1 : On (slow)
+use_accurate_framebuffers =
+
 # The clear color for the renderer. What shows up on the sides of the bottom screen.
 # Must be in range of 0.0-1.0. Defaults to 1.0 for all.
 bg_red =