Texture Cache: Speedup texture views invalidation

OpenGL: Add GL_PRIMITIVES_GENERATED and GL_TRANSFORM_FEEDBACK_PRIMITIVES_WRITTEN queries

CMakeLists.txt

@@ -305,7 +305,7 @@ find_package(ZLIB 1.2 REQUIRED)
 find_package(zstd 1.5 REQUIRED)
 
 if (NOT YUZU_USE_EXTERNAL_VULKAN_HEADERS)
-    find_package(Vulkan 1.3.256 REQUIRED)
+    find_package(Vulkan 1.3.274 REQUIRED)
 endif()
 
 if (ENABLE_LIBUSB)

src/android/app/src/main/java/org/yuzu/yuzu_emu/ui/GamesFragment.kt

@@ -91,18 +91,20 @@ class GamesFragment : Fragment() {
         viewLifecycleOwner.lifecycleScope.apply {
             launch {
                 repeatOnLifecycle(Lifecycle.State.RESUMED) {
-                    gamesViewModel.isReloading.collect { binding.swipeRefresh.isRefreshing = it }
+                    gamesViewModel.isReloading.collect {
+                        binding.swipeRefresh.isRefreshing = it
+                        if (gamesViewModel.games.value.isEmpty() && !it) {
+                            binding.noticeText.visibility = View.VISIBLE
+                        } else {
+                            binding.noticeText.visibility = View.INVISIBLE
+                        }
+                    }
                 }
             }
             launch {
                 repeatOnLifecycle(Lifecycle.State.RESUMED) {
                     gamesViewModel.games.collectLatest {
                         (binding.gridGames.adapter as GameAdapter).submitList(it)
-                        if (it.isEmpty()) {
-                            binding.noticeText.visibility = View.VISIBLE
-                        } else {
-                            binding.noticeText.visibility = View.GONE
-                        }
                     }
                 }
             }

src/core/arm/nce/arm_nce.cpp

@@ -39,7 +39,7 @@ fpsimd_context* GetFloatingPointState(mcontext_t& host_ctx) {
 }
 
 using namespace Common::Literals;
-constexpr u32 StackSize = 32_KiB;
+constexpr u32 StackSize = 128_KiB;
 
 } // namespace
 

src/core/arm/nce/interpreter_visitor.cpp

@@ -5,8 +5,6 @@
 #include "common/bit_cast.h"
 #include "core/arm/nce/interpreter_visitor.h"
 
-#include <dynarmic/frontend/A64/decoder/a64.h>
-
 namespace Core {
 
 template <u32 BitSize>
@@ -249,6 +247,7 @@ bool InterpreterVisitor::LDR_lit_fpsimd(Imm<2> opc, Imm<19> imm19, Vec Vt) {
         return false;
     }
 
+    // Size in bytes
     const u64 size = 4 << opc.ZeroExtend();
     const u64 offset = imm19.SignExtend<u64>() << 2;
     const u64 address = this->GetPc() + offset;
@@ -530,7 +529,7 @@ bool InterpreterVisitor::SIMDImmediate(bool wback, bool postindex, size_t scale,
     }
     case MemOp::Load: {
         u128 data{};
-        m_memory.ReadBlock(address, &data, datasize);
+        m_memory.ReadBlock(address, &data, datasize / 8);
         this->SetVec(Vt, data);
         break;
     }

src/core/arm/nce/visitor_base.h

@@ -4,9 +4,15 @@
 
 #pragma once
 
+#pragma GCC diagnostic push
+#pragma GCC diagnostic ignored "-Wshadow"
+
 #include <dynarmic/frontend/A64/a64_types.h>
+#include <dynarmic/frontend/A64/decoder/a64.h>
 #include <dynarmic/frontend/imm.h>
 
+#pragma GCC diagnostic pop
+
 namespace Core {
 
 class VisitorBase {

src/core/hle/service/hid/hid_server.cpp

@@ -51,7 +51,7 @@ private:
         IPC::RequestParser rp{ctx};
         const auto vibration_device_handle{rp.PopRaw<Core::HID::VibrationDeviceHandle>()};
 
-        if (resource_manager != nullptr) {
+        if (resource_manager != nullptr && resource_manager->GetNpad()) {
             resource_manager->GetNpad()->InitializeVibrationDevice(vibration_device_handle);
         }
 

src/shader_recompiler/backend/spirv/emit_spirv_context_get_set.cpp

@@ -74,6 +74,11 @@ std::optional<OutAttr> OutputAttrPointer(EmitContext& ctx, IR::Attribute attr) {
     case IR::Attribute::ClipDistance7: {
         const u32 base{static_cast<u32>(IR::Attribute::ClipDistance0)};
         const u32 index{static_cast<u32>(attr) - base};
+        if (index >= ctx.profile.max_user_clip_distances) {
+            LOG_WARNING(Shader, "Ignoring clip distance store {} >= {} supported", index,
+                        ctx.profile.max_user_clip_distances);
+            return std::nullopt;
+        }
         const Id clip_num{ctx.Const(index)};
         return OutputAccessChain(ctx, ctx.output_f32, ctx.clip_distances, clip_num);
     }

src/shader_recompiler/backend/spirv/emit_spirv_image.cpp

@@ -214,16 +214,16 @@ Id TextureImage(EmitContext& ctx, IR::TextureInstInfo info, const IR::Value& ind
     }
 }
 
-Id Image(EmitContext& ctx, const IR::Value& index, IR::TextureInstInfo info) {
+std::pair<Id, bool> Image(EmitContext& ctx, const IR::Value& index, IR::TextureInstInfo info) {
     if (!index.IsImmediate() || index.U32() != 0) {
         throw NotImplementedException("Indirect image indexing");
     }
     if (info.type == TextureType::Buffer) {
         const ImageBufferDefinition def{ctx.image_buffers.at(info.descriptor_index)};
-        return ctx.OpLoad(def.image_type, def.id);
+        return {ctx.OpLoad(def.image_type, def.id), def.is_integer};
     } else {
         const ImageDefinition def{ctx.images.at(info.descriptor_index)};
-        return ctx.OpLoad(def.image_type, def.id);
+        return {ctx.OpLoad(def.image_type, def.id), def.is_integer};
     }
 }
 
@@ -566,13 +566,23 @@ Id EmitImageRead(EmitContext& ctx, IR::Inst* inst, const IR::Value& index, Id co
         LOG_WARNING(Shader_SPIRV, "Typeless image read not supported by host");
         return ctx.ConstantNull(ctx.U32[4]);
     }
-    return Emit(&EmitContext::OpImageSparseRead, &EmitContext::OpImageRead, ctx, inst, ctx.U32[4],
-                Image(ctx, index, info), coords, std::nullopt, std::span<const Id>{});
+    const auto [image, is_integer] = Image(ctx, index, info);
+    const Id result_type{is_integer ? ctx.U32[4] : ctx.F32[4]};
+    Id color{Emit(&EmitContext::OpImageSparseRead, &EmitContext::OpImageRead, ctx, inst,
+                  result_type, image, coords, std::nullopt, std::span<const Id>{})};
+    if (!is_integer) {
+        color = ctx.OpBitcast(ctx.U32[4], color);
+    }
+    return color;
 }
 
 void EmitImageWrite(EmitContext& ctx, IR::Inst* inst, const IR::Value& index, Id coords, Id color) {
     const auto info{inst->Flags<IR::TextureInstInfo>()};
-    ctx.OpImageWrite(Image(ctx, index, info), coords, color);
+    const auto [image, is_integer] = Image(ctx, index, info);
+    if (!is_integer) {
+        color = ctx.OpBitcast(ctx.F32[4], color);
+    }
+    ctx.OpImageWrite(image, coords, color);
 }
 
 Id EmitIsTextureScaled(EmitContext& ctx, const IR::Value& index) {

src/shader_recompiler/backend/spirv/spirv_emit_context.cpp

@@ -74,20 +74,19 @@ spv::ImageFormat GetImageFormat(ImageFormat format) {
     throw InvalidArgument("Invalid image format {}", format);
 }
 
-Id ImageType(EmitContext& ctx, const ImageDescriptor& desc) {
+Id ImageType(EmitContext& ctx, const ImageDescriptor& desc, Id sampled_type) {
     const spv::ImageFormat format{GetImageFormat(desc.format)};
-    const Id type{ctx.U32[1]};
     switch (desc.type) {
     case TextureType::Color1D:
-        return ctx.TypeImage(type, spv::Dim::Dim1D, false, false, false, 2, format);
+        return ctx.TypeImage(sampled_type, spv::Dim::Dim1D, false, false, false, 2, format);
     case TextureType::ColorArray1D:
-        return ctx.TypeImage(type, spv::Dim::Dim1D, false, true, false, 2, format);
+        return ctx.TypeImage(sampled_type, spv::Dim::Dim1D, false, true, false, 2, format);
     case TextureType::Color2D:
-        return ctx.TypeImage(type, spv::Dim::Dim2D, false, false, false, 2, format);
+        return ctx.TypeImage(sampled_type, spv::Dim::Dim2D, false, false, false, 2, format);
     case TextureType::ColorArray2D:
-        return ctx.TypeImage(type, spv::Dim::Dim2D, false, true, false, 2, format);
+        return ctx.TypeImage(sampled_type, spv::Dim::Dim2D, false, true, false, 2, format);
     case TextureType::Color3D:
-        return ctx.TypeImage(type, spv::Dim::Dim3D, false, false, false, 2, format);
+        return ctx.TypeImage(sampled_type, spv::Dim::Dim3D, false, false, false, 2, format);
     case TextureType::Buffer:
         throw NotImplementedException("Image buffer");
     default:
@@ -1273,7 +1272,9 @@ void EmitContext::DefineImageBuffers(const Info& info, u32& binding) {
             throw NotImplementedException("Array of image buffers");
         }
         const spv::ImageFormat format{GetImageFormat(desc.format)};
-        const Id image_type{TypeImage(U32[1], spv::Dim::Buffer, false, false, false, 2, format)};
+        const Id sampled_type{desc.is_integer ? U32[1] : F32[1]};
+        const Id image_type{
+            TypeImage(sampled_type, spv::Dim::Buffer, false, false, false, 2, format)};
         const Id pointer_type{TypePointer(spv::StorageClass::UniformConstant, image_type)};
         const Id id{AddGlobalVariable(pointer_type, spv::StorageClass::UniformConstant)};
         Decorate(id, spv::Decoration::Binding, binding);
@@ -1283,6 +1284,7 @@ void EmitContext::DefineImageBuffers(const Info& info, u32& binding) {
             .id = id,
             .image_type = image_type,
             .count = desc.count,
+            .is_integer = desc.is_integer,
         });
         if (profile.supported_spirv >= 0x00010400) {
             interfaces.push_back(id);
@@ -1327,7 +1329,8 @@ void EmitContext::DefineImages(const Info& info, u32& binding, u32& scaling_inde
         if (desc.count != 1) {
             throw NotImplementedException("Array of images");
         }
-        const Id image_type{ImageType(*this, desc)};
+        const Id sampled_type{desc.is_integer ? U32[1] : F32[1]};
+        const Id image_type{ImageType(*this, desc, sampled_type)};
         const Id pointer_type{TypePointer(spv::StorageClass::UniformConstant, image_type)};
         const Id id{AddGlobalVariable(pointer_type, spv::StorageClass::UniformConstant)};
         Decorate(id, spv::Decoration::Binding, binding);
@@ -1337,6 +1340,7 @@ void EmitContext::DefineImages(const Info& info, u32& binding, u32& scaling_inde
             .id = id,
             .image_type = image_type,
             .count = desc.count,
+            .is_integer = desc.is_integer,
         });
         if (profile.supported_spirv >= 0x00010400) {
             interfaces.push_back(id);
@@ -1528,7 +1532,8 @@ void EmitContext::DefineOutputs(const IR::Program& program) {
         if (stage == Stage::Fragment) {
             throw NotImplementedException("Storing ClipDistance in fragment stage");
         }
-        const Id type{TypeArray(F32[1], Const(8U))};
+        const Id type{TypeArray(
+            F32[1], Const(std::min(info.used_clip_distances, profile.max_user_clip_distances)))};
         clip_distances = DefineOutput(*this, type, invocations, spv::BuiltIn::ClipDistance);
     }
     if (info.stores[IR::Attribute::Layer] &&

src/shader_recompiler/backend/spirv/spirv_emit_context.h

@@ -47,12 +47,14 @@ struct ImageBufferDefinition {
     Id id;
     Id image_type;
     u32 count;
+    bool is_integer;
 };
 
 struct ImageDefinition {
     Id id;
     Id image_type;
     u32 count;
+    bool is_integer;
 };
 
 struct UniformDefinitions {

src/shader_recompiler/environment.h

@@ -24,6 +24,8 @@ public:
 
     [[nodiscard]] virtual TexturePixelFormat ReadTexturePixelFormat(u32 raw_handle) = 0;
 
+    [[nodiscard]] virtual bool IsTexturePixelFormatInteger(u32 raw_handle) = 0;
+
     [[nodiscard]] virtual u32 ReadViewportTransformState() = 0;
 
     [[nodiscard]] virtual u32 TextureBoundBuffer() const = 0;

src/shader_recompiler/ir_opt/collect_shader_info_pass.cpp

@@ -913,7 +913,11 @@ void GatherInfoFromHeader(Environment& env, Info& info) {
         }
         for (size_t index = 0; index < 8; ++index) {
             const u16 mask{header.vtg.omap_systemc.clip_distances};
-            info.stores.Set(IR::Attribute::ClipDistance0 + index, ((mask >> index) & 1) != 0);
+            const bool used{((mask >> index) & 1) != 0};
+            info.stores.Set(IR::Attribute::ClipDistance0 + index, used);
+            if (used) {
+                info.used_clip_distances = static_cast<u32>(index) + 1;
+            }
         }
         info.stores.Set(IR::Attribute::PrimitiveId,
                         header.vtg.omap_systemb.primitive_array_id != 0);

src/shader_recompiler/ir_opt/constant_propagation_pass.cpp

@@ -815,6 +815,15 @@ bool FindGradient3DDerivatives(std::array<IR::Value, 3>& results, IR::Value coor
     return true;
 }
 
+void ConvertDerivatives(std::array<IR::Value, 3>& results, IR::IREmitter& ir) {
+    for (size_t i = 0; i < 3; i++) {
+        if (results[i].Type() == IR::Type::U32) {
+            results[i] = results[i].IsImmediate() ? ir.Imm32(Common::BitCast<f32>(results[i].U32()))
+                                                  : ir.BitCast<IR::F32>(IR::U32(results[i]));
+        }
+    }
+}
+
 void FoldImageSampleImplicitLod(IR::Block& block, IR::Inst& inst) {
     IR::TextureInstInfo info = inst.Flags<IR::TextureInstInfo>();
     auto orig_opcode = inst.GetOpcode();
@@ -831,12 +840,14 @@ void FoldImageSampleImplicitLod(IR::Block& block, IR::Inst& inst) {
     if (!offset.IsImmediate()) {
         return;
     }
+    IR::IREmitter ir{block, IR::Block::InstructionList::s_iterator_to(inst)};
     IR::Inst* const inst2 = coords.InstRecursive();
     std::array<std::array<IR::Value, 3>, 3> results_matrix;
     for (size_t i = 0; i < 3; i++) {
         if (!FindGradient3DDerivatives(results_matrix[i], inst2->Arg(i).Resolve())) {
             return;
         }
+        ConvertDerivatives(results_matrix[i], ir);
     }
     IR::F32 lod_clamp{};
     if (info.has_lod_clamp != 0) {
@@ -846,7 +857,6 @@ void FoldImageSampleImplicitLod(IR::Block& block, IR::Inst& inst) {
             lod_clamp = IR::F32{bias_lc};
         }
     }
-    IR::IREmitter ir{block, IR::Block::InstructionList::s_iterator_to(inst)};
     IR::Value new_coords =
         ir.CompositeConstruct(results_matrix[0][0], results_matrix[1][0], results_matrix[2][0]);
     IR::Value derivatives_1 = ir.CompositeConstruct(results_matrix[0][1], results_matrix[0][2],

src/shader_recompiler/ir_opt/texture_pass.cpp

@@ -372,6 +372,10 @@ TexturePixelFormat ReadTexturePixelFormat(Environment& env, const ConstBufferAdd
     return env.ReadTexturePixelFormat(GetTextureHandle(env, cbuf));
 }
 
+bool IsTexturePixelFormatInteger(Environment& env, const ConstBufferAddr& cbuf) {
+    return env.IsTexturePixelFormatInteger(GetTextureHandle(env, cbuf));
+}
+
 class Descriptors {
 public:
     explicit Descriptors(TextureBufferDescriptors& texture_buffer_descriptors_,
@@ -403,6 +407,7 @@ public:
         })};
         image_buffer_descriptors[index].is_written |= desc.is_written;
         image_buffer_descriptors[index].is_read |= desc.is_read;
+        image_buffer_descriptors[index].is_integer |= desc.is_integer;
         return index;
     }
 
@@ -432,6 +437,7 @@ public:
         })};
         image_descriptors[index].is_written |= desc.is_written;
         image_descriptors[index].is_read |= desc.is_read;
+        image_descriptors[index].is_integer |= desc.is_integer;
         return index;
     }
 
@@ -469,6 +475,20 @@ void PatchImageSampleImplicitLod(IR::Block& block, IR::Inst& inst) {
                         ir.FPRecip(ir.ConvertUToF(32, 32, ir.CompositeExtract(texture_size, 1))))));
 }
 
+bool IsPixelFormatSNorm(TexturePixelFormat pixel_format) {
+    switch (pixel_format) {
+    case TexturePixelFormat::A8B8G8R8_SNORM:
+    case TexturePixelFormat::R8G8_SNORM:
+    case TexturePixelFormat::R8_SNORM:
+    case TexturePixelFormat::R16G16B16A16_SNORM:
+    case TexturePixelFormat::R16G16_SNORM:
+    case TexturePixelFormat::R16_SNORM:
+        return true;
+    default:
+        return false;
+    }
+}
+
 void PatchTexelFetch(IR::Block& block, IR::Inst& inst, TexturePixelFormat pixel_format) {
     const auto it{IR::Block::InstructionList::s_iterator_to(inst)};
     IR::IREmitter ir{block, IR::Block::InstructionList::s_iterator_to(inst)};
@@ -587,11 +607,13 @@ void TexturePass(Environment& env, IR::Program& program, const HostTranslateInfo
             }
             const bool is_written{inst->GetOpcode() != IR::Opcode::ImageRead};
             const bool is_read{inst->GetOpcode() != IR::Opcode::ImageWrite};
+            const bool is_integer{IsTexturePixelFormatInteger(env, cbuf)};
             if (flags.type == TextureType::Buffer) {
                 index = descriptors.Add(ImageBufferDescriptor{
                     .format = flags.image_format,
                     .is_written = is_written,
                     .is_read = is_read,
+                    .is_integer = is_integer,
                     .cbuf_index = cbuf.index,
                     .cbuf_offset = cbuf.offset,
                     .count = cbuf.count,
@@ -603,6 +625,7 @@ void TexturePass(Environment& env, IR::Program& program, const HostTranslateInfo
                     .format = flags.image_format,
                     .is_written = is_written,
                     .is_read = is_read,
+                    .is_integer = is_integer,
                     .cbuf_index = cbuf.index,
                     .cbuf_offset = cbuf.offset,
                     .count = cbuf.count,
@@ -658,7 +681,7 @@ void TexturePass(Environment& env, IR::Program& program, const HostTranslateInfo
         if (!host_info.support_snorm_render_buffer && inst->GetOpcode() == IR::Opcode::ImageFetch &&
             flags.type == TextureType::Buffer) {
             const auto pixel_format = ReadTexturePixelFormat(env, cbuf);
-            if (pixel_format != TexturePixelFormat::OTHER) {
+            if (IsPixelFormatSNorm(pixel_format)) {
                 PatchTexelFetch(*texture_inst.block, *texture_inst.inst, pixel_format);
             }
         }

src/shader_recompiler/profile.h

@@ -87,6 +87,8 @@ struct Profile {
     bool has_broken_robust{};
 
     u64 min_ssbo_alignment{};
+
+    u32 max_user_clip_distances{};
 };
 
 } // namespace Shader

src/shader_recompiler/shader_info.h

@@ -35,14 +35,109 @@ enum class TextureType : u32 {
 };
 constexpr u32 NUM_TEXTURE_TYPES = 9;
 
-enum class TexturePixelFormat : u32 {
+enum class TexturePixelFormat {
+    A8B8G8R8_UNORM,
     A8B8G8R8_SNORM,
+    A8B8G8R8_SINT,
+    A8B8G8R8_UINT,
+    R5G6B5_UNORM,
+    B5G6R5_UNORM,
+    A1R5G5B5_UNORM,
+    A2B10G10R10_UNORM,
+    A2B10G10R10_UINT,
+    A2R10G10B10_UNORM,
+    A1B5G5R5_UNORM,
+    A5B5G5R1_UNORM,
+    R8_UNORM,
     R8_SNORM,
-    R8G8_SNORM,
+    R8_SINT,
+    R8_UINT,
+    R16G16B16A16_FLOAT,
+    R16G16B16A16_UNORM,
     R16G16B16A16_SNORM,
-    R16G16_SNORM,
+    R16G16B16A16_SINT,
+    R16G16B16A16_UINT,
+    B10G11R11_FLOAT,
+    R32G32B32A32_UINT,
+    BC1_RGBA_UNORM,
+    BC2_UNORM,
+    BC3_UNORM,
+    BC4_UNORM,
+    BC4_SNORM,
+    BC5_UNORM,
+    BC5_SNORM,
+    BC7_UNORM,
+    BC6H_UFLOAT,
+    BC6H_SFLOAT,
+    ASTC_2D_4X4_UNORM,
+    B8G8R8A8_UNORM,
+    R32G32B32A32_FLOAT,
+    R32G32B32A32_SINT,
+    R32G32_FLOAT,
+    R32G32_SINT,
+    R32_FLOAT,
+    R16_FLOAT,
+    R16_UNORM,
     R16_SNORM,
-    OTHER
+    R16_UINT,
+    R16_SINT,
+    R16G16_UNORM,
+    R16G16_FLOAT,
+    R16G16_UINT,
+    R16G16_SINT,
+    R16G16_SNORM,
+    R32G32B32_FLOAT,
+    A8B8G8R8_SRGB,
+    R8G8_UNORM,
+    R8G8_SNORM,
+    R8G8_SINT,
+    R8G8_UINT,
+    R32G32_UINT,
+    R16G16B16X16_FLOAT,
+    R32_UINT,
+    R32_SINT,
+    ASTC_2D_8X8_UNORM,
+    ASTC_2D_8X5_UNORM,
+    ASTC_2D_5X4_UNORM,
+    B8G8R8A8_SRGB,
+    BC1_RGBA_SRGB,
+    BC2_SRGB,
+    BC3_SRGB,
+    BC7_SRGB,
+    A4B4G4R4_UNORM,
+    G4R4_UNORM,
+    ASTC_2D_4X4_SRGB,
+    ASTC_2D_8X8_SRGB,
+    ASTC_2D_8X5_SRGB,
+    ASTC_2D_5X4_SRGB,
+    ASTC_2D_5X5_UNORM,
+    ASTC_2D_5X5_SRGB,
+    ASTC_2D_10X8_UNORM,
+    ASTC_2D_10X8_SRGB,
+    ASTC_2D_6X6_UNORM,
+    ASTC_2D_6X6_SRGB,
+    ASTC_2D_10X6_UNORM,
+    ASTC_2D_10X6_SRGB,
+    ASTC_2D_10X5_UNORM,
+    ASTC_2D_10X5_SRGB,
+    ASTC_2D_10X10_UNORM,
+    ASTC_2D_10X10_SRGB,
+    ASTC_2D_12X10_UNORM,
+    ASTC_2D_12X10_SRGB,
+    ASTC_2D_12X12_UNORM,
+    ASTC_2D_12X12_SRGB,
+    ASTC_2D_8X6_UNORM,
+    ASTC_2D_8X6_SRGB,
+    ASTC_2D_6X5_UNORM,
+    ASTC_2D_6X5_SRGB,
+    E5B9G9R9_FLOAT,
+    D32_FLOAT,
+    D16_UNORM,
+    X8_D24_UNORM,
+    S8_UINT,
+    D24_UNORM_S8_UINT,
+    S8_UINT_D24_UNORM,
+    D32_FLOAT_S8_UINT,
 };
 
 enum class ImageFormat : u32 {
@@ -97,6 +192,7 @@ struct ImageBufferDescriptor {
     ImageFormat format;
     bool is_written;
     bool is_read;
+    bool is_integer;
     u32 cbuf_index;
     u32 cbuf_offset;
     u32 count;
@@ -129,6 +225,7 @@ struct ImageDescriptor {
     ImageFormat format;
     bool is_written;
     bool is_read;
+    bool is_integer;
     u32 cbuf_index;
     u32 cbuf_offset;
     u32 count;
@@ -227,6 +324,8 @@ struct Info {
     bool requires_layer_emulation{};
     IR::Attribute emulated_layer{};
 
+    u32 used_clip_distances{};
+
     boost::container::static_vector<ConstantBufferDescriptor, MAX_CBUFS>
         constant_buffer_descriptors;
     boost::container::static_vector<StorageBufferDescriptor, MAX_SSBOS> storage_buffers_descriptors;

src/video_core/engines/maxwell_3d.cpp

@@ -586,14 +586,22 @@ void Maxwell3D::ProcessQueryCondition() {
 }
 
 void Maxwell3D::ProcessCounterReset() {
-    switch (regs.clear_report_value) {
-    case Regs::ClearReport::ZPassPixelCount:
-        rasterizer->ResetCounter(VideoCommon::QueryType::ZPassPixelCount64);
-        break;
-    default:
-        LOG_DEBUG(Render_OpenGL, "Unimplemented counter reset={}", regs.clear_report_value);
-        break;
-    }
+    const auto query_type = [clear_report = regs.clear_report_value]() {
+        switch (clear_report) {
+        case Tegra::Engines::Maxwell3D::Regs::ClearReport::ZPassPixelCount:
+            return VideoCommon::QueryType::ZPassPixelCount64;
+        case Tegra::Engines::Maxwell3D::Regs::ClearReport::StreamingPrimitivesSucceeded:
+            return VideoCommon::QueryType::StreamingPrimitivesSucceeded;
+        case Tegra::Engines::Maxwell3D::Regs::ClearReport::PrimitivesGenerated:
+            return VideoCommon::QueryType::PrimitivesGenerated;
+        case Tegra::Engines::Maxwell3D::Regs::ClearReport::VtgPrimitivesOut:
+            return VideoCommon::QueryType::VtgPrimitivesOut;
+        default:
+            LOG_DEBUG(HW_GPU, "Unimplemented counter reset={}", clear_report);
+            return VideoCommon::QueryType::Payload;
+        }
+    }();
+    rasterizer->ResetCounter(query_type);
 }
 
 void Maxwell3D::ProcessSyncPoint() {

src/video_core/query_cache.h

@@ -28,8 +28,11 @@
 namespace VideoCore {
 enum class QueryType {
     SamplesPassed,
+    PrimitivesGenerated,
+    TfbPrimitivesWritten,
+    Count,
 };
-constexpr std::size_t NumQueryTypes = 1;
+constexpr std::size_t NumQueryTypes = static_cast<size_t>(QueryType::Count);
 } // namespace VideoCore
 
 namespace VideoCommon {
@@ -44,15 +47,6 @@ public:
     explicit CounterStreamBase(QueryCache& cache_, VideoCore::QueryType type_)
         : cache{cache_}, type{type_} {}
 
-    /// Updates the state of the stream, enabling or disabling as needed.
-    void Update(bool enabled) {
-        if (enabled) {
-            Enable();
-        } else {
-            Disable();
-        }
-    }
-
     /// Resets the stream to zero. It doesn't disable the query after resetting.
     void Reset() {
         if (current) {
@@ -80,7 +74,6 @@ public:
         return current != nullptr;
     }
 
-private:
     /// Enables the stream.
     void Enable() {
         if (current) {
@@ -97,6 +90,7 @@ private:
         last = std::exchange(current, nullptr);
     }
 
+private:
     QueryCache& cache;
     const VideoCore::QueryType type;
 
@@ -112,8 +106,14 @@ public:
         : rasterizer{rasterizer_},
           // Use reinterpret_cast instead of static_cast as workaround for
           // UBSan bug (https://github.com/llvm/llvm-project/issues/59060)
-          cpu_memory{cpu_memory_}, streams{{CounterStream{reinterpret_cast<QueryCache&>(*this),
-                                                          VideoCore::QueryType::SamplesPassed}}} {
+          cpu_memory{cpu_memory_}, streams{{
+                                       {CounterStream{reinterpret_cast<QueryCache&>(*this),
+                                                      VideoCore::QueryType::SamplesPassed}},
+                                       {CounterStream{reinterpret_cast<QueryCache&>(*this),
+                                                      VideoCore::QueryType::PrimitivesGenerated}},
+                                       {CounterStream{reinterpret_cast<QueryCache&>(*this),
+                                                      VideoCore::QueryType::TfbPrimitivesWritten}},
+                                   }} {
         (void)slot_async_jobs.insert(); // Null value
     }
 
@@ -157,12 +157,11 @@ public:
         AsyncFlushQuery(query, timestamp, lock);
     }
 
-    /// Updates counters from GPU state. Expected to be called once per draw, clear or dispatch.
-    void UpdateCounters() {
+    /// Enables all available GPU counters
+    void EnableCounters() {
         std::unique_lock lock{mutex};
-        if (maxwell3d) {
-            const auto& regs = maxwell3d->regs;
-            Stream(VideoCore::QueryType::SamplesPassed).Update(regs.zpass_pixel_count_enable);
+        for (auto& stream : streams) {
+            stream.Enable();
         }
     }
 
@@ -176,7 +175,7 @@ public:
     void DisableStreams() {
         std::unique_lock lock{mutex};
         for (auto& stream : streams) {
-            stream.Update(false);
+            stream.Disable();
         }
     }
 
@@ -353,7 +352,7 @@ private:
 
     std::shared_ptr<std::vector<AsyncJobId>> uncommitted_flushes{};
     std::list<std::shared_ptr<std::vector<AsyncJobId>>> committed_flushes;
-};
+}; // namespace VideoCommon
 
 template <class QueryCache, class HostCounter>
 class HostCounterBase {

src/video_core/renderer_opengl/gl_query_cache.cpp

@@ -18,16 +18,27 @@ namespace OpenGL {
 
 namespace {
 
-constexpr std::array<GLenum, VideoCore::NumQueryTypes> QueryTargets = {GL_SAMPLES_PASSED};
-
 constexpr GLenum GetTarget(VideoCore::QueryType type) {
-    return QueryTargets[static_cast<std::size_t>(type)];
+    switch (type) {
+    case VideoCore::QueryType::SamplesPassed:
+        return GL_SAMPLES_PASSED;
+    case VideoCore::QueryType::PrimitivesGenerated:
+        return GL_PRIMITIVES_GENERATED;
+    case VideoCore::QueryType::TfbPrimitivesWritten:
+        return GL_TRANSFORM_FEEDBACK_PRIMITIVES_WRITTEN;
+    default:
+        break;
+    }
+    UNIMPLEMENTED_MSG("Query type {}", type);
+    return 0;
 }
 
 } // Anonymous namespace
 
 QueryCache::QueryCache(RasterizerOpenGL& rasterizer_, Core::Memory::Memory& cpu_memory_)
-    : QueryCacheLegacy(rasterizer_, cpu_memory_), gl_rasterizer{rasterizer_} {}
+    : QueryCacheLegacy(rasterizer_, cpu_memory_), gl_rasterizer{rasterizer_} {
+    EnableCounters();
+}
 
 QueryCache::~QueryCache() = default;
 
@@ -103,13 +114,13 @@ u64 CachedQuery::Flush([[maybe_unused]] bool async) {
     auto& stream = cache->Stream(type);
     const bool slice_counter = WaitPending() && stream.IsEnabled();
     if (slice_counter) {
-        stream.Update(false);
+        stream.Disable();
     }
 
     auto result = VideoCommon::CachedQueryBase<HostCounter>::Flush();
 
     if (slice_counter) {
-        stream.Update(true);
+        stream.Enable();
     }
 
     return result;

src/video_core/renderer_opengl/gl_rasterizer.cpp

@@ -51,6 +51,22 @@ constexpr size_t NUM_SUPPORTED_VERTEX_ATTRIBUTES = 16;
 void oglEnable(GLenum cap, bool state) {
     (state ? glEnable : glDisable)(cap);
 }
+
+std::optional<VideoCore::QueryType> MaxwellToVideoCoreQuery(VideoCommon::QueryType type) {
+    switch (type) {
+    case VideoCommon::QueryType::PrimitivesGenerated:
+    case VideoCommon::QueryType::VtgPrimitivesOut:
+        return VideoCore::QueryType::PrimitivesGenerated;
+    case VideoCommon::QueryType::ZPassPixelCount64:
+        return VideoCore::QueryType::SamplesPassed;
+    case VideoCommon::QueryType::StreamingPrimitivesSucceeded:
+        // case VideoCommon::QueryType::StreamingByteCount:
+        // TODO: StreamingByteCount = StreamingPrimitivesSucceeded * num_verts * vert_stride
+        return VideoCore::QueryType::TfbPrimitivesWritten;
+    default:
+        return std::nullopt;
+    }
+}
 } // Anonymous namespace
 
 RasterizerOpenGL::RasterizerOpenGL(Core::Frontend::EmuWindow& emu_window_, Tegra::GPU& gpu_,
@@ -216,7 +232,6 @@ void RasterizerOpenGL::PrepareDraw(bool is_indexed, Func&& draw_func) {
 
     SCOPE_EXIT({ gpu.TickWork(); });
     gpu_memory->FlushCaching();
-    query_cache.UpdateCounters();
 
     GraphicsPipeline* const pipeline{shader_cache.CurrentGraphicsPipeline()};
     if (!pipeline) {
@@ -334,7 +349,6 @@ void RasterizerOpenGL::DrawTexture() {
     MICROPROFILE_SCOPE(OpenGL_Drawing);
 
     SCOPE_EXIT({ gpu.TickWork(); });
-    query_cache.UpdateCounters();
 
     texture_cache.SynchronizeGraphicsDescriptors();
     texture_cache.UpdateRenderTargets(false);
@@ -401,21 +415,28 @@ void RasterizerOpenGL::DispatchCompute() {
 }
 
 void RasterizerOpenGL::ResetCounter(VideoCommon::QueryType type) {
-    if (type == VideoCommon::QueryType::ZPassPixelCount64) {
-        query_cache.ResetCounter(VideoCore::QueryType::SamplesPassed);
+    const auto query_cache_type = MaxwellToVideoCoreQuery(type);
+    if (!query_cache_type.has_value()) {
+        UNIMPLEMENTED_IF_MSG(type != VideoCommon::QueryType::Payload, "Reset query type: {}", type);
+        return;
     }
+    query_cache.ResetCounter(*query_cache_type);
 }
 
 void RasterizerOpenGL::Query(GPUVAddr gpu_addr, VideoCommon::QueryType type,
                              VideoCommon::QueryPropertiesFlags flags, u32 payload, u32 subreport) {
-    if (type == VideoCommon::QueryType::ZPassPixelCount64) {
-        if (True(flags & VideoCommon::QueryPropertiesFlags::HasTimeout)) {
-            query_cache.Query(gpu_addr, VideoCore::QueryType::SamplesPassed, {gpu.GetTicks()});
-        } else {
-            query_cache.Query(gpu_addr, VideoCore::QueryType::SamplesPassed, std::nullopt);
-        }
-        return;
+    const auto query_cache_type = MaxwellToVideoCoreQuery(type);
+    if (!query_cache_type.has_value()) {
+        return QueryFallback(gpu_addr, type, flags, payload, subreport);
     }
+    const bool has_timeout = True(flags & VideoCommon::QueryPropertiesFlags::HasTimeout);
+    const auto timestamp = has_timeout ? std::optional<u64>{gpu.GetTicks()} : std::nullopt;
+    query_cache.Query(gpu_addr, *query_cache_type, timestamp);
+}
+
+void RasterizerOpenGL::QueryFallback(GPUVAddr gpu_addr, VideoCommon::QueryType type,
+                                     VideoCommon::QueryPropertiesFlags flags, u32 payload,
+                                     u32 subreport) {
     if (type != VideoCommon::QueryType::Payload) {
         payload = 1u;
     }

src/video_core/renderer_opengl/gl_rasterizer.h

@@ -225,6 +225,9 @@ private:
     /// End a transform feedback
     void EndTransformFeedback();
 
+    void QueryFallback(GPUVAddr gpu_addr, VideoCommon::QueryType type,
+                       VideoCommon::QueryPropertiesFlags flags, u32 payload, u32 subreport);
+
     Tegra::GPU& gpu;
 
     const Device& device;

src/video_core/renderer_opengl/gl_shader_cache.cpp

@@ -51,7 +51,7 @@ using VideoCommon::LoadPipelines;
 using VideoCommon::SerializePipeline;
 using Context = ShaderContext::Context;
 
-constexpr u32 CACHE_VERSION = 9;
+constexpr u32 CACHE_VERSION = 10;
 
 template <typename Container>
 auto MakeSpan(Container& container) {
@@ -233,6 +233,7 @@ ShaderCache::ShaderCache(RasterizerOpenGL& rasterizer_, Core::Frontend::EmuWindo
           .ignore_nan_fp_comparisons = true,
           .gl_max_compute_smem_size = device.GetMaxComputeSharedMemorySize(),
           .min_ssbo_alignment = device.GetShaderStorageBufferAlignment(),
+          .max_user_clip_distances = 8,
       },
       host_info{
           .support_float64 = true,

src/video_core/renderer_vulkan/vk_buffer_cache.cpp

@@ -78,8 +78,15 @@ vk::Buffer CreateBuffer(const Device& device, const MemoryAllocator& memory_allo
 }
 } // Anonymous namespace
 
-Buffer::Buffer(BufferCacheRuntime&, VideoCommon::NullBufferParams null_params)
-    : VideoCommon::BufferBase<VideoCore::RasterizerInterface>(null_params), tracker{4096} {}
+Buffer::Buffer(BufferCacheRuntime& runtime, VideoCommon::NullBufferParams null_params)
+    : VideoCommon::BufferBase<VideoCore::RasterizerInterface>(null_params), tracker{4096} {
+    if (runtime.device.HasNullDescriptor()) {
+        return;
+    }
+    device = &runtime.device;
+    buffer = runtime.CreateNullBuffer();
+    is_null = true;
+}
 
 Buffer::Buffer(BufferCacheRuntime& runtime, VideoCore::RasterizerInterface& rasterizer_,
                VAddr cpu_addr_, u64 size_bytes_)
@@ -93,8 +100,12 @@ Buffer::Buffer(BufferCacheRuntime& runtime, VideoCore::RasterizerInterface& rast
 
 VkBufferView Buffer::View(u32 offset, u32 size, VideoCore::Surface::PixelFormat format) {
     if (!device) {
-        // Null buffer, return a null descriptor
+        // Null buffer supported, return a null descriptor
         return VK_NULL_HANDLE;
+    } else if (is_null) {
+        // Null buffer not supported, adjust offset and size
+        offset = 0;
+        size = 0;
     }
     const auto it{std::ranges::find_if(views, [offset, size, format](const BufferView& view) {
         return offset == view.offset && size == view.size && format == view.format;
@@ -563,22 +574,27 @@ void BufferCacheRuntime::BindVertexBuffers(VideoCommon::HostBindings<Buffer>& bi
         }
         buffer_handles.push_back(handle);
     }
+    const u32 device_max = device.GetMaxVertexInputBindings();
+    const u32 min_binding = std::min(bindings.min_index, device_max);
+    const u32 max_binding = std::min(bindings.max_index, device_max);
+    const u32 binding_count = max_binding - min_binding;
+    if (binding_count == 0) {
+        return;
+    }
     if (device.IsExtExtendedDynamicStateSupported()) {
-        scheduler.Record([this, bindings_ = std::move(bindings),
-                          buffer_handles_ = std::move(buffer_handles)](vk::CommandBuffer cmdbuf) {
-            cmdbuf.BindVertexBuffers2EXT(bindings_.min_index,
-                                         std::min(bindings_.max_index - bindings_.min_index,
-                                                  device.GetMaxVertexInputBindings()),
-                                         buffer_handles_.data(), bindings_.offsets.data(),
-                                         bindings_.sizes.data(), bindings_.strides.data());
+        scheduler.Record([bindings_ = std::move(bindings),
+                          buffer_handles_ = std::move(buffer_handles),
+                          binding_count](vk::CommandBuffer cmdbuf) {
+            cmdbuf.BindVertexBuffers2EXT(bindings_.min_index, binding_count, buffer_handles_.data(),
+                                         bindings_.offsets.data(), bindings_.sizes.data(),
+                                         bindings_.strides.data());
         });
     } else {
-        scheduler.Record([this, bindings_ = std::move(bindings),
-                          buffer_handles_ = std::move(buffer_handles)](vk::CommandBuffer cmdbuf) {
-            cmdbuf.BindVertexBuffers(bindings_.min_index,
-                                     std::min(bindings_.max_index - bindings_.min_index,
-                                              device.GetMaxVertexInputBindings()),
-                                     buffer_handles_.data(), bindings_.offsets.data());
+        scheduler.Record([bindings_ = std::move(bindings),
+                          buffer_handles_ = std::move(buffer_handles),
+                          binding_count](vk::CommandBuffer cmdbuf) {
+            cmdbuf.BindVertexBuffers(bindings_.min_index, binding_count, buffer_handles_.data(),
+                                     bindings_.offsets.data());
         });
     }
 }
@@ -622,9 +638,12 @@ void BufferCacheRuntime::BindTransformFeedbackBuffers(VideoCommon::HostBindings<
 }
 
 void BufferCacheRuntime::ReserveNullBuffer() {
-    if (null_buffer) {
-        return;
+    if (!null_buffer) {
+        null_buffer = CreateNullBuffer();
     }
+}
+
+vk::Buffer BufferCacheRuntime::CreateNullBuffer() {
     VkBufferCreateInfo create_info{
         .sType = VK_STRUCTURE_TYPE_BUFFER_CREATE_INFO,
         .pNext = nullptr,
@@ -639,15 +658,17 @@ void BufferCacheRuntime::ReserveNullBuffer() {
     if (device.IsExtTransformFeedbackSupported()) {
         create_info.usage |= VK_BUFFER_USAGE_TRANSFORM_FEEDBACK_BUFFER_BIT_EXT;
     }
-    null_buffer = memory_allocator.CreateBuffer(create_info, MemoryUsage::DeviceLocal);
+    vk::Buffer ret = memory_allocator.CreateBuffer(create_info, MemoryUsage::DeviceLocal);
     if (device.HasDebuggingToolAttached()) {
-        null_buffer.SetObjectNameEXT("Null buffer");
+        ret.SetObjectNameEXT("Null buffer");
     }
 
     scheduler.RequestOutsideRenderPassOperationContext();
-    scheduler.Record([buffer = *null_buffer](vk::CommandBuffer cmdbuf) {
+    scheduler.Record([buffer = *ret](vk::CommandBuffer cmdbuf) {
         cmdbuf.FillBuffer(buffer, 0, VK_WHOLE_SIZE, 0);
     });
+
+    return ret;
 }
 
 } // namespace Vulkan

src/video_core/renderer_vulkan/vk_buffer_cache.h

@@ -63,6 +63,7 @@ private:
     vk::Buffer buffer;
     std::vector<BufferView> views;
     VideoCommon::UsageTracker tracker;
+    bool is_null{};
 };
 
 class QuadArrayIndexBuffer;
@@ -151,6 +152,7 @@ private:
     }
 
     void ReserveNullBuffer();
+    vk::Buffer CreateNullBuffer();
 
     const Device& device;
     MemoryAllocator& memory_allocator;

src/video_core/renderer_vulkan/vk_pipeline_cache.cpp

@@ -54,7 +54,7 @@ using VideoCommon::FileEnvironment;
 using VideoCommon::GenericEnvironment;
 using VideoCommon::GraphicsEnvironment;
 
-constexpr u32 CACHE_VERSION = 10;
+constexpr u32 CACHE_VERSION = 11;
 constexpr std::array<char, 8> VULKAN_CACHE_MAGIC_NUMBER{'y', 'u', 'z', 'u', 'v', 'k', 'c', 'h'};
 
 template <typename Container>
@@ -374,6 +374,7 @@ PipelineCache::PipelineCache(RasterizerVulkan& rasterizer_, const Device& device
         .has_broken_robust =
             device.IsNvidia() && device.GetNvidiaArch() <= NvidiaArchitecture::Arch_Pascal,
         .min_ssbo_alignment = device.GetStorageBufferAlignment(),
+        .max_user_clip_distances = device.GetMaxUserClipDistances(),
     };
 
     host_info = Shader::HostTranslateInfo{

src/video_core/renderer_vulkan/vk_query_cache.cpp

@@ -289,12 +289,15 @@ public:
         }
 
         if (has_multi_queries) {
-            size_t intermediary_buffer_index = ObtainBuffer<false>(num_slots_used);
+            const size_t min_accumulation_limit =
+                std::min(first_accumulation_checkpoint, num_slots_used);
+            const size_t max_accumulation_limit =
+                std::max(last_accumulation_checkpoint, num_slots_used);
+            const size_t intermediary_buffer_index = ObtainBuffer<false>(num_slots_used);
             resolve_buffers.push_back(intermediary_buffer_index);
             queries_prefix_scan_pass->Run(*accumulation_buffer, *buffers[intermediary_buffer_index],
                                           *buffers[resolve_buffer_index], num_slots_used,
-                                          std::min(first_accumulation_checkpoint, num_slots_used),
-                                          last_accumulation_checkpoint);
+                                          min_accumulation_limit, max_accumulation_limit);
 
         } else {
             scheduler.RequestOutsideRenderPassOperationContext();

src/video_core/renderer_vulkan/vk_rasterizer.cpp

@@ -485,6 +485,10 @@ void RasterizerVulkan::DispatchCompute() {
 }
 
 void RasterizerVulkan::ResetCounter(VideoCommon::QueryType type) {
+    if (type != VideoCommon::QueryType::ZPassPixelCount64) {
+        LOG_DEBUG(Render_Vulkan, "Unimplemented counter reset={}", type);
+        return;
+    }
     query_cache.CounterReset(type);
 }
 

src/video_core/shader_environment.cpp

@@ -62,23 +62,9 @@ static Shader::TextureType ConvertTextureType(const Tegra::Texture::TICEntry& en
 }
 
 static Shader::TexturePixelFormat ConvertTexturePixelFormat(const Tegra::Texture::TICEntry& entry) {
-    switch (PixelFormatFromTextureInfo(entry.format, entry.r_type, entry.g_type, entry.b_type,
-                                       entry.a_type, entry.srgb_conversion)) {
-    case VideoCore::Surface::PixelFormat::A8B8G8R8_SNORM:
-        return Shader::TexturePixelFormat::A8B8G8R8_SNORM;
-    case VideoCore::Surface::PixelFormat::R8_SNORM:
-        return Shader::TexturePixelFormat::R8_SNORM;
-    case VideoCore::Surface::PixelFormat::R8G8_SNORM:
-        return Shader::TexturePixelFormat::R8G8_SNORM;
-    case VideoCore::Surface::PixelFormat::R16G16B16A16_SNORM:
-        return Shader::TexturePixelFormat::R16G16B16A16_SNORM;
-    case VideoCore::Surface::PixelFormat::R16G16_SNORM:
-        return Shader::TexturePixelFormat::R16G16_SNORM;
-    case VideoCore::Surface::PixelFormat::R16_SNORM:
-        return Shader::TexturePixelFormat::R16_SNORM;
-    default:
-        return Shader::TexturePixelFormat::OTHER;
-    }
+    return static_cast<Shader::TexturePixelFormat>(
+        PixelFormatFromTextureInfo(entry.format, entry.r_type, entry.g_type, entry.b_type,
+                                   entry.a_type, entry.srgb_conversion));
 }
 
 static std::string_view StageToPrefix(Shader::Stage stage) {
@@ -398,6 +384,11 @@ Shader::TexturePixelFormat GraphicsEnvironment::ReadTexturePixelFormat(u32 handl
     return result;
 }
 
+bool GraphicsEnvironment::IsTexturePixelFormatInteger(u32 handle) {
+    return VideoCore::Surface::IsPixelFormatInteger(
+        static_cast<VideoCore::Surface::PixelFormat>(ReadTexturePixelFormat(handle)));
+}
+
 u32 GraphicsEnvironment::ReadViewportTransformState() {
     const auto& regs{maxwell3d->regs};
     viewport_transform_state = regs.viewport_scale_offset_enabled;
@@ -448,6 +439,11 @@ Shader::TexturePixelFormat ComputeEnvironment::ReadTexturePixelFormat(u32 handle
     return result;
 }
 
+bool ComputeEnvironment::IsTexturePixelFormatInteger(u32 handle) {
+    return VideoCore::Surface::IsPixelFormatInteger(
+        static_cast<VideoCore::Surface::PixelFormat>(ReadTexturePixelFormat(handle)));
+}
+
 u32 ComputeEnvironment::ReadViewportTransformState() {
     return viewport_transform_state;
 }
@@ -551,6 +547,11 @@ Shader::TexturePixelFormat FileEnvironment::ReadTexturePixelFormat(u32 handle) {
     return it->second;
 }
 
+bool FileEnvironment::IsTexturePixelFormatInteger(u32 handle) {
+    return VideoCore::Surface::IsPixelFormatInteger(
+        static_cast<VideoCore::Surface::PixelFormat>(ReadTexturePixelFormat(handle)));
+}
+
 u32 FileEnvironment::ReadViewportTransformState() {
     return viewport_transform_state;
 }

src/video_core/shader_environment.h

@@ -115,6 +115,8 @@ public:
 
     Shader::TexturePixelFormat ReadTexturePixelFormat(u32 handle) override;
 
+    bool IsTexturePixelFormatInteger(u32 handle) override;
+
     u32 ReadViewportTransformState() override;
 
     std::optional<Shader::ReplaceConstant> GetReplaceConstBuffer(u32 bank, u32 offset) override;
@@ -139,6 +141,8 @@ public:
 
     Shader::TexturePixelFormat ReadTexturePixelFormat(u32 handle) override;
 
+    bool IsTexturePixelFormatInteger(u32 handle) override;
+
     u32 ReadViewportTransformState() override;
 
     std::optional<Shader::ReplaceConstant> GetReplaceConstBuffer(
@@ -171,6 +175,8 @@ public:
 
     [[nodiscard]] Shader::TexturePixelFormat ReadTexturePixelFormat(u32 handle) override;
 
+    [[nodiscard]] bool IsTexturePixelFormatInteger(u32 handle) override;
+
     [[nodiscard]] u32 ReadViewportTransformState() override;
 
     [[nodiscard]] u32 LocalMemorySize() const override;

src/video_core/texture_cache/texture_cache.h

@@ -1084,6 +1084,7 @@ ImageViewId TextureCache<P>::FindImageView(const TICEntry& config) {
     ImageViewId& image_view_id = pair->second;
     if (is_new) {
         image_view_id = CreateImageView(config);
+        channel_state->image_views_inv[image_view_id] = config;
     }
     return image_view_id;
 }
@@ -2218,14 +2219,17 @@ template <class P>
 void TextureCache<P>::RemoveImageViewReferences(std::span<const ImageViewId> removed_views) {
     for (size_t c : active_channel_ids) {
         auto& channel_info = channel_storage[c];
-        auto it = channel_info.image_views.begin();
-        while (it != channel_info.image_views.end()) {
-            const auto found = std::ranges::find(removed_views, it->second);
-            if (found != removed_views.end()) {
-                it = channel_info.image_views.erase(it);
-            } else {
-                ++it;
+        for (auto image_view_id : removed_views) {
+            auto it_v = channel_info.image_views_inv.find(image_view_id);
+            if (it_v == channel_info.image_views_inv.end()) {
+                continue;
             }
+            auto it = channel_info.image_views.find(it_v->second);
+            channel_info.image_views_inv.erase(it_v);
+            if (it == channel_info.image_views.end()) {
+                continue;
+            }
+            channel_info.image_views.erase(it);
         }
     }
 }

src/video_core/texture_cache/texture_cache_base.h

@@ -81,6 +81,7 @@ public:
     std::vector<ImageViewId> compute_image_view_ids;
 
     std::unordered_map<TICEntry, ImageViewId> image_views;
+    std::unordered_map<ImageViewId, TICEntry> image_views_inv;
     std::unordered_map<TSCEntry, SamplerId> samplers;
 
     TextureCacheGPUMap* gpu_page_table;

src/video_core/vulkan_common/vulkan_device.cpp

@@ -695,6 +695,11 @@ Device::Device(VkInstance instance_, vk::PhysicalDevice physical_, VkSurfaceKHR
             std::min(properties.properties.limits.maxVertexInputBindings, 16U);
     }
 
+    if (is_turnip) {
+        LOG_WARNING(Render_Vulkan, "Turnip requires higher-than-reported binding limits");
+        properties.properties.limits.maxVertexInputBindings = 32;
+    }
+
     if (!extensions.extended_dynamic_state && extensions.extended_dynamic_state2) {
         LOG_INFO(Render_Vulkan,
                  "Removing extendedDynamicState2 due to missing extendedDynamicState");

src/video_core/vulkan_common/vulkan_device.h

@@ -665,6 +665,10 @@ public:
         return properties.properties.limits.maxViewports;
     }
 
+    u32 GetMaxUserClipDistances() const {
+        return properties.properties.limits.maxClipDistances;
+    }
+
     bool SupportsConditionalBarriers() const {
         return supports_conditional_barriers;
     }

src/video_core/vulkan_common/vulkan_wrapper.cpp

@@ -377,6 +377,8 @@ const char* ToString(VkResult result) noexcept {
         return "VK_OPERATION_DEFERRED_KHR";
     case VkResult::VK_OPERATION_NOT_DEFERRED_KHR:
         return "VK_OPERATION_NOT_DEFERRED_KHR";
+    case VkResult::VK_ERROR_INVALID_VIDEO_STD_PARAMETERS_KHR:
+        return "VK_ERROR_INVALID_VIDEO_STD_PARAMETERS_KHR";
     case VkResult::VK_PIPELINE_COMPILE_REQUIRED_EXT:
         return "VK_PIPELINE_COMPILE_REQUIRED_EXT";
     case VkResult::VK_RESULT_MAX_ENUM:

src/yuzu/configuration/configure_graphics.cpp

@@ -106,32 +106,30 @@ ConfigureGraphics::ConfigureGraphics(
                                                 Settings::values.bg_green.GetValue(),
                                                 Settings::values.bg_blue.GetValue()));
     UpdateAPILayout();
-    PopulateVSyncModeSelection(); //< must happen after UpdateAPILayout
+    PopulateVSyncModeSelection(false); //< must happen after UpdateAPILayout
 
     // VSync setting needs to be determined after populating the VSync combobox
-    if (Settings::IsConfiguringGlobal()) {
-        const auto vsync_mode_setting = Settings::values.vsync_mode.GetValue();
-        const auto vsync_mode = VSyncSettingToMode(vsync_mode_setting);
-        int index{};
-        for (const auto mode : vsync_mode_combobox_enum_map) {
-            if (mode == vsync_mode) {
-                break;
-            }
-            index++;
-        }
-        if (static_cast<unsigned long>(index) < vsync_mode_combobox_enum_map.size()) {
-            vsync_mode_combobox->setCurrentIndex(index);
+    const auto vsync_mode_setting = Settings::values.vsync_mode.GetValue();
+    const auto vsync_mode = VSyncSettingToMode(vsync_mode_setting);
+    int index{};
+    for (const auto mode : vsync_mode_combobox_enum_map) {
+        if (mode == vsync_mode) {
+            break;
         }
+        index++;
+    }
+    if (static_cast<unsigned long>(index) < vsync_mode_combobox_enum_map.size()) {
+        vsync_mode_combobox->setCurrentIndex(index);
     }
 
     connect(api_combobox, qOverload<int>(&QComboBox::activated), this, [this] {
         UpdateAPILayout();
-        PopulateVSyncModeSelection();
+        PopulateVSyncModeSelection(false);
     });
     connect(vulkan_device_combobox, qOverload<int>(&QComboBox::activated), this,
             [this](int device) {
                 UpdateDeviceSelection(device);
-                PopulateVSyncModeSelection();
+                PopulateVSyncModeSelection(false);
             });
     connect(shader_backend_combobox, qOverload<int>(&QComboBox::activated), this,
             [this](int backend) { UpdateShaderBackendSelection(backend); });
@@ -147,8 +145,9 @@ ConfigureGraphics::ConfigureGraphics(
     const auto& update_screenshot_info = [this, &builder]() {
         const auto& combobox_enumerations = builder.ComboboxTranslations().at(
             Settings::EnumMetadata<Settings::AspectRatio>::Index());
-        const auto index = aspect_ratio_combobox->currentIndex();
-        const auto ratio = static_cast<Settings::AspectRatio>(combobox_enumerations[index].first);
+        const auto ratio_index = aspect_ratio_combobox->currentIndex();
+        const auto ratio =
+            static_cast<Settings::AspectRatio>(combobox_enumerations[ratio_index].first);
 
         const auto& combobox_enumerations_resolution = builder.ComboboxTranslations().at(
             Settings::EnumMetadata<Settings::ResolutionSetup>::Index());
@@ -174,11 +173,7 @@ ConfigureGraphics::ConfigureGraphics(
     }
 }
 
-void ConfigureGraphics::PopulateVSyncModeSelection() {
-    if (!Settings::IsConfiguringGlobal()) {
-        return;
-    }
-
+void ConfigureGraphics::PopulateVSyncModeSelection(bool use_setting) {
     const Settings::RendererBackend backend{GetCurrentGraphicsBackend()};
     if (backend == Settings::RendererBackend::Null) {
         vsync_mode_combobox->setEnabled(false);
@@ -189,8 +184,9 @@ void ConfigureGraphics::PopulateVSyncModeSelection() {
     const int current_index = //< current selected vsync mode from combobox
         vsync_mode_combobox->currentIndex();
     const auto current_mode = //< current selected vsync mode as a VkPresentModeKHR
-        current_index == -1 ? VSyncSettingToMode(Settings::values.vsync_mode.GetValue())
-                            : vsync_mode_combobox_enum_map[current_index];
+        current_index == -1 || use_setting
+            ? VSyncSettingToMode(Settings::values.vsync_mode.GetValue())
+            : vsync_mode_combobox_enum_map[current_index];
     int index{};
     const int device{vulkan_device_combobox->currentIndex()}; //< current selected Vulkan device
 
@@ -214,6 +210,23 @@ void ConfigureGraphics::PopulateVSyncModeSelection() {
         }
         index++;
     }
+
+    if (!Settings::IsConfiguringGlobal()) {
+        vsync_restore_global_button->setVisible(!Settings::values.vsync_mode.UsingGlobal());
+
+        const Settings::VSyncMode global_vsync_mode = Settings::values.vsync_mode.GetValue(true);
+        vsync_restore_global_button->setEnabled(
+            (backend == Settings::RendererBackend::OpenGL &&
+             (global_vsync_mode == Settings::VSyncMode::Immediate ||
+              global_vsync_mode == Settings::VSyncMode::Fifo)) ||
+            backend == Settings::RendererBackend::Vulkan);
+    }
+}
+
+void ConfigureGraphics::UpdateVsyncSetting() const {
+    const auto mode = vsync_mode_combobox_enum_map[vsync_mode_combobox->currentIndex()];
+    const auto vsync_mode = PresentModeToSetting(mode);
+    Settings::values.vsync_mode.SetValue(vsync_mode);
 }
 
 void ConfigureGraphics::UpdateDeviceSelection(int device) {
@@ -299,6 +312,33 @@ void ConfigureGraphics::Setup(const ConfigurationShared::Builder& builder) {
         } else if (setting->Id() == Settings::values.vsync_mode.Id()) {
             // Keep track of vsync_mode's combobox so we can populate it
             vsync_mode_combobox = widget->combobox;
+
+            // Since vsync is populated at runtime, we have to manually set up the button for
+            // restoring the global setting.
+            if (!Settings::IsConfiguringGlobal()) {
+                QPushButton* restore_button =
+                    ConfigurationShared::Widget::CreateRestoreGlobalButton(
+                        Settings::values.vsync_mode.UsingGlobal(), widget);
+                restore_button->setEnabled(true);
+                widget->layout()->addWidget(restore_button);
+
+                QObject::connect(restore_button, &QAbstractButton::clicked,
+                                 [restore_button, this](bool) {
+                                     Settings::values.vsync_mode.SetGlobal(true);
+                                     PopulateVSyncModeSelection(true);
+
+                                     restore_button->setVisible(false);
+                                 });
+
+                std::function<void()> set_non_global = [restore_button, this]() {
+                    Settings::values.vsync_mode.SetGlobal(false);
+                    UpdateVsyncSetting();
+                    restore_button->setVisible(true);
+                };
+                QObject::connect(widget->combobox, QOverload<int>::of(&QComboBox::activated),
+                                 [set_non_global]() { set_non_global(); });
+                vsync_restore_global_button = restore_button;
+            }
             hold_graphics.emplace(setting->Id(), widget);
         } else if (setting->Id() == Settings::values.aspect_ratio.Id()) {
             // Keep track of the aspect ratio combobox to update other UI tabs that need it
@@ -400,11 +440,7 @@ void ConfigureGraphics::ApplyConfiguration() {
         func(powered_on);
     }
 
-    if (Settings::IsConfiguringGlobal()) {
-        const auto mode = vsync_mode_combobox_enum_map[vsync_mode_combobox->currentIndex()];
-        const auto vsync_mode = PresentModeToSetting(mode);
-        Settings::values.vsync_mode.SetValue(vsync_mode);
-    }
+    UpdateVsyncSetting();
 
     Settings::values.vulkan_device.SetGlobal(true);
     Settings::values.shader_backend.SetGlobal(true);

src/yuzu/configuration/configure_graphics.h

@@ -62,7 +62,8 @@ private:
 
     void Setup(const ConfigurationShared::Builder& builder);
 
-    void PopulateVSyncModeSelection();
+    void PopulateVSyncModeSelection(bool use_setting);
+    void UpdateVsyncSetting() const;
     void UpdateBackgroundColorButton(QColor color);
     void UpdateAPILayout();
     void UpdateDeviceSelection(int device);
@@ -104,6 +105,7 @@ private:
     QComboBox* api_combobox;
     QComboBox* shader_backend_combobox;
     QComboBox* vsync_mode_combobox;
+    QPushButton* vsync_restore_global_button;
     QWidget* vulkan_device_widget;
     QWidget* api_widget;
     QWidget* shader_backend_widget;

src/yuzu/configuration/qt_config.cpp

@@ -225,6 +225,8 @@ void QtConfig::ReadPathValues() {
         QString::fromStdString(ReadStringSetting(std::string("recentFiles")))
             .split(QStringLiteral(", "), Qt::SkipEmptyParts, Qt::CaseSensitive);
 
+    ReadCategory(Settings::Category::Paths);
+
     EndGroup();
 }
 
@@ -405,6 +407,8 @@ void QtConfig::SaveQtControlValues() {
 void QtConfig::SavePathValues() {
     BeginGroup(Settings::TranslateCategory(Settings::Category::Paths));
 
+    WriteCategory(Settings::Category::Paths);
+
     WriteSetting(std::string("romsPath"), UISettings::values.roms_path);
     BeginArray(std::string("gamedirs"));
     for (int i = 0; i < UISettings::values.game_dirs.size(); ++i) {