GPU: Allow using the old NV04 values for the depth test function.

These seem to be just a valid as the GL token values. Thanks @ReinUsesLisp

This restores graphical output to Disgaea 5

CMakeModules/CopyYuzuQt5Deps.cmake

@@ -3,7 +3,9 @@ function(copy_yuzu_Qt5_deps target_dir)
     set(DLL_DEST "${CMAKE_BINARY_DIR}/bin/$<CONFIG>/")
     set(Qt5_DLL_DIR "${Qt5_DIR}/../../../bin")
     set(Qt5_PLATFORMS_DIR "${Qt5_DIR}/../../../plugins/platforms/")
+    set(Qt5_STYLES_DIR "${Qt5_DIR}/../../../plugins/styles/")
     set(PLATFORMS ${DLL_DEST}platforms/)
+    set(STYLES ${DLL_DEST}styles/)
     windows_copy_files(${target_dir} ${Qt5_DLL_DIR} ${DLL_DEST}
         icudt*.dll
         icuin*.dll
@@ -14,4 +16,5 @@ function(copy_yuzu_Qt5_deps target_dir)
         Qt5Widgets$<$<CONFIG:Debug>:d>.*
     )
     windows_copy_files(yuzu ${Qt5_PLATFORMS_DIR} ${PLATFORMS} qwindows$<$<CONFIG:Debug>:d>.*)
+    windows_copy_files(yuzu ${Qt5_STYLES_DIR} ${STYLES} qwindowsvistastyle$<$<CONFIG:Debug>:d>.*)
 endfunction(copy_yuzu_Qt5_deps)

appveyor.yml

@@ -116,6 +116,7 @@ after_build:
 
           mkdir $RELEASE_DIST
           mkdir $RELEASE_DIST/platforms
+          mkdir $RELEASE_DIST/styles
 
           # copy the compiled binaries and other release files to the release folder
           Get-ChildItem "$CMAKE_BINARY_DIR" -Filter "yuzu*.exe" | Copy-Item -destination $RELEASE_DIST
@@ -136,6 +137,9 @@ after_build:
           # copy the qt windows plugin dll to platforms
           Copy-Item -path "C:/msys64/mingw64/share/qt5/plugins/platforms/qwindows.dll" -force -destination "$RELEASE_DIST/platforms"
 
+          # copy the qt windows vista style dll to platforms
+          Copy-Item -path "C:/msys64/mingw64/share/qt5/plugins/styles/qwindowsvistastyle.dll" -force -destination "$RELEASE_DIST/styles"
+
           7z a -tzip $MINGW_BUILD_ZIP $RELEASE_DIST\*
           7z a $MINGW_SEVENZIP $RELEASE_DIST
         }

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/video_core/engines/maxwell_3d.cpp

@@ -126,6 +126,10 @@ void Maxwell3D::WriteReg(u32 method, u32 value, u32 remaining_params) {
         DrawArrays();
         break;
     }
+    case MAXWELL3D_REG_INDEX(clear_buffers): {
+        ProcessClearBuffers();
+        break;
+    }
     case MAXWELL3D_REG_INDEX(query.query_get): {
         ProcessQueryGet();
         break;
@@ -415,5 +419,13 @@ bool Maxwell3D::IsShaderStageEnabled(Regs::ShaderStage stage) const {
     UNREACHABLE();
 }
 
+void Maxwell3D::ProcessClearBuffers() {
+    ASSERT(regs.clear_buffers.R == regs.clear_buffers.G &&
+           regs.clear_buffers.R == regs.clear_buffers.B &&
+           regs.clear_buffers.R == regs.clear_buffers.A);
+
+    VideoCore::g_renderer->Rasterizer()->Clear();
+}
+
 } // namespace Engines
 } // namespace Tegra

src/video_core/engines/maxwell_3d.h

@@ -281,14 +281,26 @@ public:
         };
 
         enum class ComparisonOp : u32 {
-            Never = 0,
-            Less = 1,
-            Equal = 2,
-            LessEqual = 3,
-            Greater = 4,
-            NotEqual = 5,
-            GreaterEqual = 6,
-            Always = 7,
+            // These values are used by Nouveau and most games, they correspond to the OpenGL token
+            // values for these operations.
+            Never = 0x200,
+            Less = 0x201,
+            Equal = 0x202,
+            LessEqual = 0x203,
+            Greater = 0x204,
+            NotEqual = 0x205,
+            GreaterEqual = 0x206,
+            Always = 0x207,
+
+            // These values are used by some games, they seem to be NV04 values.
+            NeverOld = 1,
+            LessOld = 2,
+            EqualOld = 3,
+            LessEqualOld = 4,
+            GreaterOld = 5,
+            NotEqualOld = 6,
+            GreaterEqualOld = 7,
+            AlwaysOld = 8,
         };
 
         struct Cull {
@@ -436,7 +448,12 @@ public:
                     u32 count;
                 } vertex_buffer;
 
-                INSERT_PADDING_WORDS(0x99);
+                INSERT_PADDING_WORDS(1);
+
+                float clear_color[4];
+                float clear_depth;
+
+                INSERT_PADDING_WORDS(0x93);
 
                 struct {
                     u32 address_high;
@@ -473,9 +490,11 @@ public:
 
                 u32 depth_write_enabled;
 
-                INSERT_PADDING_WORDS(0x8);
+                INSERT_PADDING_WORDS(0x7);
 
-                BitField<0, 3, ComparisonOp> depth_test_func;
+                u32 d3d_cull_mode;
+
+                ComparisonOp depth_test_func;
 
                 INSERT_PADDING_WORDS(0xB);
 
@@ -493,7 +512,13 @@ public:
                     u32 enable[NumRenderTargets];
                 } blend;
 
-                INSERT_PADDING_WORDS(0x2D);
+                INSERT_PADDING_WORDS(0xB);
+
+                union {
+                    BitField<4, 1, u32> triangle_rast_flip;
+                } screen_y_control;
+
+                INSERT_PADDING_WORDS(0x21);
 
                 u32 vb_element_base;
 
@@ -523,7 +548,12 @@ public:
                     }
                 } tic;
 
-                INSERT_PADDING_WORDS(0x22);
+                INSERT_PADDING_WORDS(0x21);
+
+                union {
+                    BitField<2, 1, u32> coord_origin;
+                    BitField<3, 10, u32> enable;
+                } point_coord_replace;
 
                 struct {
                     u32 code_address_high;
@@ -584,7 +614,21 @@ public:
 
                 Cull cull;
 
-                INSERT_PADDING_WORDS(0x77);
+                INSERT_PADDING_WORDS(0x2B);
+
+                union {
+                    u32 raw;
+                    BitField<0, 1, u32> Z;
+                    BitField<1, 1, u32> S;
+                    BitField<2, 1, u32> R;
+                    BitField<3, 1, u32> G;
+                    BitField<4, 1, u32> B;
+                    BitField<5, 1, u32> A;
+                    BitField<6, 4, u32> RT;
+                    BitField<10, 11, u32> layer;
+                } clear_buffers;
+
+                INSERT_PADDING_WORDS(0x4B);
 
                 struct {
                     u32 query_address_high;
@@ -766,6 +810,9 @@ private:
     /// Handles writes to the macro uploading registers.
     void ProcessMacroUpload(u32 data);
 
+    /// Handles a write to the CLEAR_BUFFERS register.
+    void ProcessClearBuffers();
+
     /// Handles a write to the QUERY_GET register.
     void ProcessQueryGet();
 
@@ -788,21 +835,27 @@ ASSERT_REG_POSITION(rt, 0x200);
 ASSERT_REG_POSITION(viewport_transform[0], 0x280);
 ASSERT_REG_POSITION(viewport, 0x300);
 ASSERT_REG_POSITION(vertex_buffer, 0x35D);
+ASSERT_REG_POSITION(clear_color[0], 0x360);
+ASSERT_REG_POSITION(clear_depth, 0x364);
 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(d3d_cull_mode, 0x4C2);
 ASSERT_REG_POSITION(depth_test_func, 0x4C3);
 ASSERT_REG_POSITION(blend, 0x4CF);
+ASSERT_REG_POSITION(screen_y_control, 0x4EB);
 ASSERT_REG_POSITION(vb_element_base, 0x50D);
 ASSERT_REG_POSITION(tsc, 0x557);
 ASSERT_REG_POSITION(tic, 0x55D);
+ASSERT_REG_POSITION(point_coord_replace, 0x581);
 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(clear_buffers, 0x674);
 ASSERT_REG_POSITION(query, 0x6C0);
 ASSERT_REG_POSITION(vertex_array[0], 0x700);
 ASSERT_REG_POSITION(independent_blend, 0x780);

src/video_core/engines/shader_bytecode.h

@@ -328,6 +328,19 @@ union Instruction {
         BitField<49, 3, PredCondition> cond;
     } isetp;
 
+    union {
+        BitField<0, 3, u64> pred0;
+        BitField<3, 3, u64> pred3;
+        BitField<12, 3, u64> pred12;
+        BitField<15, 1, u64> neg_pred12;
+        BitField<24, 2, PredOperation> cond;
+        BitField<29, 3, u64> pred29;
+        BitField<32, 1, u64> neg_pred29;
+        BitField<39, 3, u64> pred39;
+        BitField<42, 1, u64> neg_pred39;
+        BitField<45, 2, PredOperation> op;
+    } psetp;
+
     union {
         BitField<39, 3, u64> pred39;
         BitField<42, 1, u64> neg_pred;
@@ -641,7 +654,7 @@ private:
             INST("1110111111011---", Id::LD_A, Type::Memory, "LD_A"),
             INST("1110111110010---", Id::LD_C, Type::Memory, "LD_C"),
             INST("1110111111110---", Id::ST_A, Type::Memory, "ST_A"),
-            INST("1100000000111---", Id::TEX, Type::Memory, "TEX"),
+            INST("110000----111---", Id::TEX, Type::Memory, "TEX"),
             INST("1101111101001---", Id::TEXQ, Type::Memory, "TEXQ"),
             INST("1101100---------", Id::TEXS, Type::Memory, "TEXS"),
             INST("1101101---------", Id::TLDS, Type::Memory, "TLDS"),

src/video_core/rasterizer_interface.h

@@ -19,6 +19,9 @@ public:
     /// Draw the current batch of vertex arrays
     virtual void DrawArrays() = 0;
 
+    /// Clear the current framebuffer
+    virtual void Clear() = 0;
+
     /// Notify rasterizer that the specified Maxwell register has been changed
     virtual void NotifyMaxwellRegisterChanged(u32 method) = 0;
 

src/video_core/renderer_opengl/gl_rasterizer.cpp

@@ -297,11 +297,8 @@ bool RasterizerOpenGL::AccelerateDrawBatch(bool is_indexed) {
     return true;
 }
 
-void RasterizerOpenGL::DrawArrays() {
-    if (accelerate_draw == AccelDraw::Disabled)
-        return;
-
-    MICROPROFILE_SCOPE(OpenGL_Drawing);
+std::pair<Surface, Surface> RasterizerOpenGL::ConfigureFramebuffers(bool using_color_fb,
+                                                                    bool using_depth_fb) {
     const auto& regs = Core::System().GetInstance().GPU().Maxwell3D().regs;
 
     // Sync the depth test state before configuring the framebuffer surfaces.
@@ -310,9 +307,6 @@ void RasterizerOpenGL::DrawArrays() {
     // TODO(bunnei): Implement this
     const bool has_stencil = false;
 
-    const bool using_color_fb = true;
-    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 =
@@ -344,11 +338,6 @@ void RasterizerOpenGL::DrawArrays() {
     BindFramebufferSurfaces(color_surface, depth_surface, has_stencil);
 
     SyncViewport(surfaces_rect);
-    SyncBlendState();
-    SyncCullMode();
-
-    // TODO(bunnei): Sync framebuffer_scale uniform here
-    // TODO(bunnei): Sync scissorbox uniform(s) here
 
     // Viewport can have negative offsets or larger dimensions than our framebuffer sub-rect. Enable
     // scissor test to prevent drawing outside of the framebuffer region
@@ -359,6 +348,66 @@ void RasterizerOpenGL::DrawArrays() {
     state.scissor.height = draw_rect.GetHeight();
     state.Apply();
 
+    // Only return the surface to be marked as dirty if writing to it is enabled.
+    return std::make_pair(write_color_fb ? color_surface : nullptr,
+                          write_depth_fb ? depth_surface : nullptr);
+}
+
+void RasterizerOpenGL::Clear() {
+    const auto& regs = Core::System().GetInstance().GPU().Maxwell3D().regs;
+
+    bool use_color_fb = false;
+    bool use_depth_fb = false;
+
+    GLbitfield clear_mask = 0;
+    if (regs.clear_buffers.R && regs.clear_buffers.G && regs.clear_buffers.B &&
+        regs.clear_buffers.A) {
+        clear_mask |= GL_COLOR_BUFFER_BIT;
+        use_color_fb = true;
+    }
+    if (regs.clear_buffers.Z) {
+        clear_mask |= GL_DEPTH_BUFFER_BIT;
+        use_depth_fb = true;
+    }
+
+    if (clear_mask == 0)
+        return;
+
+    auto [dirty_color_surface, dirty_depth_surface] =
+        ConfigureFramebuffers(use_color_fb, use_depth_fb);
+
+    // TODO(Subv): Support clearing only partial colors.
+    glClearColor(regs.clear_color[0], regs.clear_color[1], regs.clear_color[2],
+                 regs.clear_color[3]);
+    glClearDepth(regs.clear_depth);
+
+    glClear(clear_mask);
+
+    // Mark framebuffer surfaces as dirty
+    if (dirty_color_surface != nullptr) {
+        res_cache.MarkSurfaceAsDirty(dirty_color_surface);
+    }
+    if (dirty_depth_surface != nullptr) {
+        res_cache.MarkSurfaceAsDirty(dirty_depth_surface);
+    }
+}
+
+void RasterizerOpenGL::DrawArrays() {
+    if (accelerate_draw == AccelDraw::Disabled)
+        return;
+
+    MICROPROFILE_SCOPE(OpenGL_Drawing);
+    const auto& regs = Core::System().GetInstance().GPU().Maxwell3D().regs;
+
+    auto [dirty_color_surface, dirty_depth_surface] =
+        ConfigureFramebuffers(true, regs.zeta.Address() != 0);
+
+    SyncBlendState();
+    SyncCullMode();
+
+    // TODO(bunnei): Sync framebuffer_scale uniform here
+    // TODO(bunnei): Sync scissorbox uniform(s) here
+
     // Draw the vertex batch
     const bool is_indexed = accelerate_draw == AccelDraw::Indexed;
     const u64 index_buffer_size{regs.index_array.count * regs.index_array.FormatSizeInBytes()};
@@ -439,11 +488,11 @@ void RasterizerOpenGL::DrawArrays() {
     state.Apply();
 
     // Mark framebuffer surfaces as dirty
-    if (color_surface != nullptr && write_color_fb) {
-        res_cache.MarkSurfaceAsDirty(color_surface);
+    if (dirty_color_surface != nullptr) {
+        res_cache.MarkSurfaceAsDirty(dirty_color_surface);
     }
-    if (depth_surface != nullptr && write_depth_fb) {
-        res_cache.MarkSurfaceAsDirty(depth_surface);
+    if (dirty_depth_surface != nullptr) {
+        res_cache.MarkSurfaceAsDirty(dirty_depth_surface);
     }
 }
 
@@ -637,7 +686,10 @@ u32 RasterizerOpenGL::SetupTextures(Maxwell::ShaderStage stage, GLuint program,
 
         // Bind the uniform to the sampler.
         GLint uniform = glGetUniformLocation(program, entry.GetName().c_str());
-        ASSERT(uniform != -1);
+        if (uniform == -1) {
+            continue;
+        }
+
         glProgramUniform1i(program, uniform, current_bindpoint);
 
         const auto texture = maxwell3d.GetStageTexture(entry.GetStage(), entry.GetOffset());
@@ -722,6 +774,16 @@ void RasterizerOpenGL::SyncCullMode() {
     if (state.cull.enabled) {
         state.cull.front_face = MaxwellToGL::FrontFace(regs.cull.front_face);
         state.cull.mode = MaxwellToGL::CullFace(regs.cull.cull_face);
+
+        // If the GPU is configured to flip the rasterized triangles, then we need to flip the
+        // notion of front and back. Note: We flip the triangles when the value of the register is 0
+        // because OpenGL already does it for us.
+        if (regs.screen_y_control.triangle_rast_flip == 0) {
+            if (state.cull.front_face == GL_CCW)
+                state.cull.front_face = GL_CW;
+            else if (state.cull.front_face == GL_CW)
+                state.cull.front_face = GL_CCW;
+        }
     }
 }
 

src/video_core/renderer_opengl/gl_rasterizer.h

@@ -7,6 +7,7 @@
 #include <array>
 #include <cstddef>
 #include <memory>
+#include <utility>
 #include <vector>
 #include <glad/glad.h>
 #include "common/common_types.h"
@@ -28,6 +29,7 @@ public:
     ~RasterizerOpenGL() override;
 
     void DrawArrays() override;
+    void Clear() override;
     void NotifyMaxwellRegisterChanged(u32 method) override;
     void FlushAll() override;
     void FlushRegion(Tegra::GPUVAddr addr, u64 size) override;
@@ -81,6 +83,10 @@ private:
         u32 border_color_a;
     };
 
+    /// Configures the color and depth framebuffer states and returns the dirty <Color, Depth>
+    /// surfaces if writing was enabled.
+    std::pair<Surface, Surface> ConfigureFramebuffers(bool using_color_fb, bool using_depth_fb);
+
     /// Binds the framebuffer color and depth surface
     void BindFramebufferSurfaces(const Surface& color_surface, const Surface& depth_surface,
                                  bool has_stencil);

src/video_core/renderer_opengl/gl_rasterizer_cache.cpp

@@ -65,6 +65,25 @@ struct FormatTuple {
     return params;
 }
 
+/*static*/ SurfaceParams SurfaceParams::CreateForDepthBuffer(
+    const Tegra::Engines::Maxwell3D::Regs::RenderTargetConfig& config, Tegra::GPUVAddr zeta_address,
+    Tegra::DepthFormat format) {
+
+    SurfaceParams params{};
+    params.addr = zeta_address;
+    params.is_tiled = true;
+    params.block_height = Tegra::Texture::TICEntry::DefaultBlockHeight;
+    params.pixel_format = PixelFormatFromDepthFormat(format);
+    params.component_type = ComponentTypeFromDepthFormat(format);
+    params.type = GetFormatType(params.pixel_format);
+    params.size_in_bytes = params.SizeInBytes();
+    params.width = config.width;
+    params.height = config.height;
+    params.unaligned_height = config.height;
+    params.size_in_bytes = params.SizeInBytes();
+    return params;
+}
+
 static constexpr std::array<FormatTuple, SurfaceParams::MaxPixelFormat> tex_format_tuples = {{
     {GL_RGBA8, GL_RGBA, GL_UNSIGNED_INT_8_8_8_8_REV, ComponentType::UNorm, false}, // ABGR8
     {GL_RGB, GL_RGB, GL_UNSIGNED_SHORT_5_6_5_REV, ComponentType::UNorm, false},    // B5G6R5
@@ -88,6 +107,9 @@ static constexpr std::array<FormatTuple, SurfaceParams::MaxPixelFormat> tex_form
     // DepthStencil formats
     {GL_DEPTH24_STENCIL8, GL_DEPTH_STENCIL, GL_UNSIGNED_INT_24_8, ComponentType::UNorm,
      false}, // Z24S8
+    {GL_DEPTH24_STENCIL8, GL_DEPTH_STENCIL, GL_UNSIGNED_INT_24_8, ComponentType::UNorm,
+     false},                                                                            // S8Z24
+    {GL_DEPTH_COMPONENT32F, GL_DEPTH_COMPONENT, GL_FLOAT, ComponentType::Float, false}, // Z32F
 }};
 
 static const FormatTuple& GetFormatTuple(PixelFormat pixel_format, ComponentType component_type) {
@@ -131,13 +153,6 @@ MathUtil::Rectangle<u32> SurfaceParams::GetRect() const {
     return {0, actual_height, width, 0};
 }
 
-static void ConvertASTCToRGBA8(std::vector<u8>& data, PixelFormat format, u32 width, u32 height) {
-    u32 block_width{};
-    u32 block_height{};
-    std::tie(block_width, block_height) = GetASTCBlockSize(format);
-    data = Tegra::Texture::ASTC::Decompress(data, width, height, block_width, block_height);
-}
-
 template <bool morton_to_gl, PixelFormat format>
 void MortonCopy(u32 stride, u32 block_height, u32 height, u8* gl_buffer, Tegra::GPUVAddr addr) {
     constexpr u32 bytes_per_pixel = SurfaceParams::GetFormatBpp(format) / CHAR_BIT;
@@ -177,6 +192,7 @@ static constexpr std::array<void (*)(u32, u32, u32, u8*, Tegra::GPUVAddr),
         MortonCopy<true, PixelFormat::DXT1>,         MortonCopy<true, PixelFormat::DXT23>,
         MortonCopy<true, PixelFormat::DXT45>,        MortonCopy<true, PixelFormat::DXN1>,
         MortonCopy<true, PixelFormat::ASTC_2D_4X4>,  MortonCopy<true, PixelFormat::Z24S8>,
+        MortonCopy<true, PixelFormat::S8Z24>,        MortonCopy<true, PixelFormat::Z32F>,
 };
 
 static constexpr std::array<void (*)(u32, u32, u32, u8*, Tegra::GPUVAddr),
@@ -197,6 +213,8 @@ static constexpr std::array<void (*)(u32, u32, u32, u8*, Tegra::GPUVAddr),
         nullptr,
         MortonCopy<false, PixelFormat::ABGR8>,
         MortonCopy<false, PixelFormat::Z24S8>,
+        MortonCopy<false, PixelFormat::S8Z24>,
+        MortonCopy<false, PixelFormat::Z32F>,
 };
 
 // Allocate an uninitialized texture of appropriate size and format for the surface
@@ -234,6 +252,71 @@ CachedSurface::CachedSurface(const SurfaceParams& params) : params(params) {
                            rect.GetWidth(), rect.GetHeight());
 }
 
+static void ConvertS8Z24ToZ24S8(std::vector<u8>& data, u32 width, u32 height) {
+    union S8Z24 {
+        BitField<0, 24, u32> z24;
+        BitField<24, 8, u32> s8;
+    };
+    static_assert(sizeof(S8Z24) == 4, "S8Z24 is incorrect size");
+
+    union Z24S8 {
+        BitField<0, 8, u32> s8;
+        BitField<8, 24, u32> z24;
+    };
+    static_assert(sizeof(Z24S8) == 4, "Z24S8 is incorrect size");
+
+    S8Z24 input_pixel{};
+    Z24S8 output_pixel{};
+    for (size_t y = 0; y < height; ++y) {
+        for (size_t x = 0; x < width; ++x) {
+            const size_t offset{y * width + x};
+            std::memcpy(&input_pixel, &data[offset], sizeof(S8Z24));
+            output_pixel.s8.Assign(input_pixel.s8);
+            output_pixel.z24.Assign(input_pixel.z24);
+            std::memcpy(&data[offset], &output_pixel, sizeof(Z24S8));
+        }
+    }
+}
+/**
+ * Helper function to perform software conversion (as needed) when loading a buffer from Switch
+ * memory. This is for Maxwell pixel formats that cannot be represented as-is in OpenGL or with
+ * typical desktop GPUs.
+ */
+static void ConvertFormatAsNeeded_LoadGLBuffer(std::vector<u8>& data, PixelFormat pixel_format,
+                                               u32 width, u32 height) {
+    switch (pixel_format) {
+    case PixelFormat::ASTC_2D_4X4: {
+        // Convert ASTC pixel formats to RGBA8, as most desktop GPUs do not support ASTC.
+        u32 block_width{};
+        u32 block_height{};
+        std::tie(block_width, block_height) = GetASTCBlockSize(pixel_format);
+        data = Tegra::Texture::ASTC::Decompress(data, width, height, block_width, block_height);
+        break;
+    }
+    case PixelFormat::S8Z24:
+        // Convert the S8Z24 depth format to Z24S8, as OpenGL does not support S8Z24.
+        ConvertS8Z24ToZ24S8(data, width, height);
+        break;
+    }
+}
+
+/**
+ * Helper function to perform software conversion (as needed) when flushing a buffer to Switch
+ * memory. This is for Maxwell pixel formats that cannot be represented as-is in OpenGL or with
+ * typical desktop GPUs.
+ */
+static void ConvertFormatAsNeeded_FlushGLBuffer(std::vector<u8>& /*data*/, PixelFormat pixel_format,
+                                                u32 /*width*/, u32 /*height*/) {
+    switch (pixel_format) {
+    case PixelFormat::ASTC_2D_4X4:
+    case PixelFormat::S8Z24:
+        LOG_CRITICAL(Render_OpenGL, "Unimplemented pixel_format={}",
+                     static_cast<u32>(pixel_format));
+        UNREACHABLE();
+        break;
+    }
+}
+
 MICROPROFILE_DEFINE(OpenGL_SurfaceLoad, "OpenGL", "Surface Load", MP_RGB(128, 64, 192));
 void CachedSurface::LoadGLBuffer() {
     ASSERT(params.type != SurfaceType::Fill);
@@ -256,10 +339,7 @@ void CachedSurface::LoadGLBuffer() {
             params.width, params.block_height, params.height, gl_buffer.data(), params.addr);
     }
 
-    if (IsPixelFormatASTC(params.pixel_format)) {
-        // ASTC formats are converted to RGBA8 in software, as most PC GPUs do not support this
-        ConvertASTCToRGBA8(gl_buffer, params.pixel_format, params.width, params.height);
-    }
+    ConvertFormatAsNeeded_LoadGLBuffer(gl_buffer, params.pixel_format, params.width, params.height);
 }
 
 MICROPROFILE_DEFINE(OpenGL_SurfaceFlush, "OpenGL", "Surface Flush", MP_RGB(128, 192, 64));
@@ -272,6 +352,9 @@ void CachedSurface::FlushGLBuffer() {
 
     MICROPROFILE_SCOPE(OpenGL_SurfaceFlush);
 
+    ConvertFormatAsNeeded_FlushGLBuffer(gl_buffer, params.pixel_format, params.width,
+                                        params.height);
+
     if (!params.is_tiled) {
         std::memcpy(dst_buffer, gl_buffer.data(), params.size_in_bytes);
     } else {
@@ -399,15 +482,16 @@ SurfaceSurfaceRect_Tuple RasterizerCacheOpenGL::GetFramebufferSurfaces(
     LOG_WARNING(Render_OpenGL, "hard-coded for render target 0!");
 
     // get color and depth surfaces
-    const SurfaceParams color_params{SurfaceParams::CreateForFramebuffer(regs.rt[0])};
-    SurfaceParams depth_params{color_params};
+    SurfaceParams color_params{};
+    SurfaceParams depth_params{};
+
+    if (using_color_fb) {
+        color_params = SurfaceParams::CreateForFramebuffer(regs.rt[0]);
+    }
 
     if (using_depth_fb) {
-        depth_params.addr = regs.zeta.Address();
-        depth_params.pixel_format = SurfaceParams::PixelFormatFromDepthFormat(regs.zeta.format);
-        depth_params.component_type = SurfaceParams::ComponentTypeFromDepthFormat(regs.zeta.format);
-        depth_params.type = SurfaceParams::GetFormatType(depth_params.pixel_format);
-        depth_params.size_in_bytes = depth_params.SizeInBytes();
+        depth_params =
+            SurfaceParams::CreateForDepthBuffer(regs.rt[0], regs.zeta.Address(), regs.zeta.format);
     }
 
     MathUtil::Rectangle<u32> color_rect{};

src/video_core/renderer_opengl/gl_rasterizer_cache.h

@@ -41,6 +41,8 @@ struct SurfaceParams {
 
         // DepthStencil formats
         Z24S8 = 13,
+        S8Z24 = 14,
+        Z32F = 15,
 
         MaxDepthStencilFormat,
 
@@ -92,6 +94,8 @@ struct SurfaceParams {
             4, // DXN1
             4, // ASTC_2D_4X4
             1, // Z24S8
+            1, // S8Z24
+            1, // Z32F
         }};
 
         ASSERT(static_cast<size_t>(format) < compression_factor_table.size());
@@ -117,6 +121,8 @@ struct SurfaceParams {
             64,  // DXN1
             32,  // ASTC_2D_4X4
             32,  // Z24S8
+            32,  // S8Z24
+            32,  // Z32F
         }};
 
         ASSERT(static_cast<size_t>(format) < bpp_table.size());
@@ -128,8 +134,12 @@ struct SurfaceParams {
 
     static PixelFormat PixelFormatFromDepthFormat(Tegra::DepthFormat format) {
         switch (format) {
+        case Tegra::DepthFormat::S8_Z24_UNORM:
+            return PixelFormat::S8Z24;
         case Tegra::DepthFormat::Z24_S8_UNORM:
             return PixelFormat::Z24S8;
+        case Tegra::DepthFormat::Z32_FLOAT:
+            return PixelFormat::Z32F;
         default:
             LOG_CRITICAL(HW_GPU, "Unimplemented format={}", static_cast<u32>(format));
             UNREACHABLE();
@@ -226,8 +236,12 @@ struct SurfaceParams {
 
     static Tegra::DepthFormat DepthFormatFromPixelFormat(PixelFormat format) {
         switch (format) {
+        case PixelFormat::S8Z24:
+            return Tegra::DepthFormat::S8_Z24_UNORM;
         case PixelFormat::Z24S8:
             return Tegra::DepthFormat::Z24_S8_UNORM;
+        case PixelFormat::Z32F:
+            return Tegra::DepthFormat::Z32_FLOAT;
         default:
             UNREACHABLE();
         }
@@ -274,8 +288,11 @@ struct SurfaceParams {
 
     static ComponentType ComponentTypeFromDepthFormat(Tegra::DepthFormat format) {
         switch (format) {
+        case Tegra::DepthFormat::S8_Z24_UNORM:
         case Tegra::DepthFormat::Z24_S8_UNORM:
             return ComponentType::UNorm;
+        case Tegra::DepthFormat::Z32_FLOAT:
+            return ComponentType::Float;
         default:
             LOG_CRITICAL(HW_GPU, "Unimplemented format={}", static_cast<u32>(format));
             UNREACHABLE();
@@ -318,13 +335,18 @@ struct SurfaceParams {
         return addr <= (region_addr + region_size) && region_addr <= (addr + size_in_bytes);
     }
 
-    /// Creates SurfaceParams from a texture configation
+    /// Creates SurfaceParams from a texture configuration
     static SurfaceParams CreateForTexture(const Tegra::Texture::FullTextureInfo& config);
 
-    /// Creates SurfaceParams from a framebuffer configation
+    /// Creates SurfaceParams from a framebuffer configuration
     static SurfaceParams CreateForFramebuffer(
         const Tegra::Engines::Maxwell3D::Regs::RenderTargetConfig& config);
 
+    /// Creates SurfaceParams for a depth buffer configuration
+    static SurfaceParams CreateForDepthBuffer(
+        const Tegra::Engines::Maxwell3D::Regs::RenderTargetConfig& config,
+        Tegra::GPUVAddr zeta_address, Tegra::DepthFormat format);
+
     Tegra::GPUVAddr addr;
     bool is_tiled;
     u32 block_height;

src/video_core/renderer_opengl/gl_shader_decompiler.cpp

@@ -1213,6 +1213,9 @@ private:
                 switch (instr.conversion.f2f.rounding) {
                 case Tegra::Shader::F2fRoundingOp::None:
                     break;
+                case Tegra::Shader::F2fRoundingOp::Round:
+                    op_a = "roundEven(" + op_a + ')';
+                    break;
                 case Tegra::Shader::F2fRoundingOp::Floor:
                     op_a = "floor(" + op_a + ')';
                     break;
@@ -1477,6 +1480,36 @@ private:
             }
             break;
         }
+        case OpCode::Type::PredicateSetPredicate: {
+            std::string op_a =
+                GetPredicateCondition(instr.psetp.pred12, instr.psetp.neg_pred12 != 0);
+            std::string op_b =
+                GetPredicateCondition(instr.psetp.pred29, instr.psetp.neg_pred29 != 0);
+
+            using Tegra::Shader::Pred;
+            // We can't use the constant predicate as destination.
+            ASSERT(instr.psetp.pred3 != static_cast<u64>(Pred::UnusedIndex));
+
+            std::string second_pred =
+                GetPredicateCondition(instr.psetp.pred39, instr.psetp.neg_pred39 != 0);
+
+            std::string combiner = GetPredicateCombiner(instr.psetp.op);
+
+            std::string predicate =
+                '(' + op_a + ") " + GetPredicateCombiner(instr.psetp.cond) + " (" + op_b + ')';
+
+            // Set the primary predicate to the result of Predicate OP SecondPredicate
+            SetPredicate(instr.psetp.pred3,
+                         '(' + predicate + ") " + combiner + " (" + second_pred + ')');
+
+            if (instr.psetp.pred0 != static_cast<u64>(Pred::UnusedIndex)) {
+                // Set the secondary predicate to the result of !Predicate OP SecondPredicate,
+                // if enabled
+                SetPredicate(instr.psetp.pred0,
+                             "!(" + predicate + ") " + combiner + " (" + second_pred + ')');
+            }
+            break;
+        }
         case OpCode::Type::FloatSet: {
             std::string op_a = instr.fset.neg_a ? "-" : "";
             op_a += regs.GetRegisterAsFloat(instr.gpr8);

src/video_core/renderer_opengl/maxwell_to_gl.h

@@ -29,6 +29,10 @@ inline GLenum VertexType(Maxwell::VertexAttribute attrib) {
         switch (attrib.size) {
         case Maxwell::VertexAttribute::Size::Size_8_8_8_8:
             return GL_UNSIGNED_BYTE;
+        case Maxwell::VertexAttribute::Size::Size_16_16:
+            return GL_UNSIGNED_SHORT;
+        case Maxwell::VertexAttribute::Size::Size_10_10_10_2:
+            return GL_UNSIGNED_INT_2_10_10_10_REV;
         }
 
         LOG_CRITICAL(Render_OpenGL, "Unimplemented vertex size={}", attrib.SizeString());
@@ -41,6 +45,10 @@ inline GLenum VertexType(Maxwell::VertexAttribute attrib) {
         switch (attrib.size) {
         case Maxwell::VertexAttribute::Size::Size_8_8_8_8:
             return GL_BYTE;
+        case Maxwell::VertexAttribute::Size::Size_16_16:
+            return GL_SHORT;
+        case Maxwell::VertexAttribute::Size::Size_10_10_10_2:
+            return GL_INT_2_10_10_10_REV;
         }
 
         LOG_CRITICAL(Render_OpenGL, "Unimplemented vertex size={}", attrib.SizeString());
@@ -203,20 +211,28 @@ inline GLenum SwizzleSource(Tegra::Texture::SwizzleSource source) {
 inline GLenum ComparisonOp(Maxwell::ComparisonOp comparison) {
     switch (comparison) {
     case Maxwell::ComparisonOp::Never:
+    case Maxwell::ComparisonOp::NeverOld:
         return GL_NEVER;
     case Maxwell::ComparisonOp::Less:
+    case Maxwell::ComparisonOp::LessOld:
         return GL_LESS;
     case Maxwell::ComparisonOp::Equal:
+    case Maxwell::ComparisonOp::EqualOld:
         return GL_EQUAL;
     case Maxwell::ComparisonOp::LessEqual:
+    case Maxwell::ComparisonOp::LessEqualOld:
         return GL_LEQUAL;
     case Maxwell::ComparisonOp::Greater:
+    case Maxwell::ComparisonOp::GreaterOld:
         return GL_GREATER;
     case Maxwell::ComparisonOp::NotEqual:
+    case Maxwell::ComparisonOp::NotEqualOld:
         return GL_NOTEQUAL;
     case Maxwell::ComparisonOp::GreaterEqual:
+    case Maxwell::ComparisonOp::GreaterEqualOld:
         return GL_GEQUAL;
     case Maxwell::ComparisonOp::Always:
+    case Maxwell::ComparisonOp::AlwaysOld:
         return GL_ALWAYS;
     }
     LOG_CRITICAL(Render_OpenGL, "Unimplemented comparison op={}", static_cast<u32>(comparison));

src/video_core/textures/decoders.cpp

@@ -76,7 +76,9 @@ u32 BytesPerPixel(TextureFormat format) {
 
 static u32 DepthBytesPerPixel(DepthFormat format) {
     switch (format) {
+    case DepthFormat::S8_Z24_UNORM:
     case DepthFormat::Z24_S8_UNORM:
+    case DepthFormat::Z32_FLOAT:
         return 4;
     default:
         UNIMPLEMENTED_MSG("Format not implemented");
@@ -129,7 +131,9 @@ std::vector<u8> UnswizzleDepthTexture(VAddr address, DepthFormat format, u32 wid
     std::vector<u8> unswizzled_data(width * height * bytes_per_pixel);
 
     switch (format) {
+    case DepthFormat::S8_Z24_UNORM:
     case DepthFormat::Z24_S8_UNORM:
+    case DepthFormat::Z32_FLOAT:
         CopySwizzledData(width, height, bytes_per_pixel, bytes_per_pixel, data,
                          unswizzled_data.data(), true, block_height);
         break;