early-access version 4142

README.md

@@ -1,7 +1,7 @@
 yuzu emulator early access
 =============
 
-This is the source code for early-access 4141.
+This is the source code for early-access 4142.
 
 ## Legal Notice
 

src/shader_recompiler/backend/spirv/emit_spirv_image.cpp

@@ -60,11 +60,10 @@ public:
         Add(spv::ImageOperandsMask::ConstOffsets, offsets);
     }
 
-    explicit ImageOperands(EmitContext& ctx, const IR::Value& offset, Id lod, Id ms) {
+    explicit ImageOperands(Id lod, Id ms) {
         if (Sirit::ValidId(lod)) {
             Add(spv::ImageOperandsMask::Lod, lod);
         }
-        AddOffset(ctx, offset, ImageFetchOffsetAllowed);
         if (Sirit::ValidId(ms)) {
             Add(spv::ImageOperandsMask::Sample, ms);
         }
@@ -312,6 +311,43 @@ Id ImageGatherSubpixelOffset(EmitContext& ctx, const IR::TextureInstInfo& info,
         return coords;
     }
 }
+
+void AddOffsetToCoordinates(EmitContext& ctx, const IR::TextureInstInfo& info, Id& coords,
+                            Id offset) {
+    if (!Sirit::ValidId(offset)) {
+        return;
+    }
+
+    Id result_type{};
+    switch (info.type) {
+    case TextureType::Buffer:
+    case TextureType::Color1D: {
+        result_type = ctx.U32[1];
+        break;
+    }
+    case TextureType::ColorArray1D:
+        offset = ctx.OpCompositeConstruct(ctx.U32[2], offset, ctx.u32_zero_value);
+        [[fallthrough]];
+    case TextureType::Color2D:
+    case TextureType::Color2DRect: {
+        result_type = ctx.U32[2];
+        break;
+    }
+    case TextureType::ColorArray2D:
+        offset = ctx.OpCompositeConstruct(ctx.U32[3], ctx.OpCompositeExtract(ctx.U32[1], coords, 0),
+                                          ctx.OpCompositeExtract(ctx.U32[1], coords, 1),
+                                          ctx.u32_zero_value);
+        [[fallthrough]];
+    case TextureType::Color3D: {
+        result_type = ctx.U32[3];
+        break;
+    }
+    case TextureType::ColorCube:
+    case TextureType::ColorArrayCube:
+        return;
+    }
+    coords = ctx.OpIAdd(result_type, coords, offset);
+}
 } // Anonymous namespace
 
 Id EmitBindlessImageSampleImplicitLod(EmitContext&) {
@@ -494,9 +530,10 @@ Id EmitImageGatherDref(EmitContext& ctx, IR::Inst* inst, const IR::Value& index,
                 operands.Span());
 }
 
-Id EmitImageFetch(EmitContext& ctx, IR::Inst* inst, const IR::Value& index, Id coords,
+Id EmitImageFetch(EmitContext& ctx, IR::Inst* inst, const IR::Value& index, Id coords, Id offset,
-                  const IR::Value& offset, Id lod, Id ms) {
+                  Id lod, Id ms) {
     const auto info{inst->Flags<IR::TextureInstInfo>()};
+    AddOffsetToCoordinates(ctx, info, coords, offset);
     if (info.type == TextureType::Buffer) {
         lod = Id{};
     }
@@ -504,7 +541,7 @@ Id EmitImageFetch(EmitContext& ctx, IR::Inst* inst, const IR::Value& index, Id c
         // This image is multisampled, lod must be implicit
         lod = Id{};
     }
-    const ImageOperands operands(ctx, offset, lod, ms);
+    const ImageOperands operands(lod, ms);
     return Emit(&EmitContext::OpImageSparseFetch, &EmitContext::OpImageFetch, ctx, inst, ctx.F32[4],
                 TextureImage(ctx, info, index), coords, operands.MaskOptional(), operands.Span());
 }

src/shader_recompiler/backend/spirv/emit_spirv_instructions.h

@@ -537,8 +537,8 @@ Id EmitImageGather(EmitContext& ctx, IR::Inst* inst, const IR::Value& index, Id
                    const IR::Value& offset, const IR::Value& offset2);
 Id EmitImageGatherDref(EmitContext& ctx, IR::Inst* inst, const IR::Value& index, Id coords,
                        const IR::Value& offset, const IR::Value& offset2, Id dref);
-Id EmitImageFetch(EmitContext& ctx, IR::Inst* inst, const IR::Value& index, Id coords,
+Id EmitImageFetch(EmitContext& ctx, IR::Inst* inst, const IR::Value& index, Id coords, Id offset,
-                  const IR::Value& offset, Id lod, Id ms);
+                  Id lod, Id ms);
 Id EmitImageQueryDimensions(EmitContext& ctx, IR::Inst* inst, const IR::Value& index, Id lod,
                             const IR::Value& skip_mips);
 Id EmitImageQueryLod(EmitContext& ctx, IR::Inst* inst, const IR::Value& index, Id coords);

src/video_core/buffer_cache/buffer_cache.h

@@ -35,7 +35,7 @@ BufferCache<P>::BufferCache(Tegra::MaxwellDeviceMemoryManager& device_memory_, R
     const s64 min_spacing_critical = device_local_memory - 512_MiB;
     const s64 mem_threshold = std::min(device_local_memory, TARGET_THRESHOLD);
     const s64 min_vacancy_expected = (6 * mem_threshold) / 10;
-    const s64 min_vacancy_critical = (3 * mem_threshold) / 10;
+    const s64 min_vacancy_critical = (2 * mem_threshold) / 10;
     minimum_memory = static_cast<u64>(
         std::max(std::min(device_local_memory - min_vacancy_expected, min_spacing_expected),
                  DEFAULT_EXPECTED_MEMORY));

src/video_core/surface.cpp

@@ -3,6 +3,7 @@
 
 #include "common/common_types.h"
 #include "common/math_util.h"
+#include "common/settings.h"
 #include "video_core/surface.h"
 
 namespace VideoCore::Surface {
@@ -400,11 +401,20 @@ std::pair<u32, u32> GetASTCBlockSize(PixelFormat format) {
     return {DefaultBlockWidth(format), DefaultBlockHeight(format)};
 }
 
-u64 EstimatedDecompressedSize(u64 base_size, PixelFormat format) {
+u64 TranscodedAstcSize(u64 base_size, PixelFormat format) {
     constexpr u64 RGBA8_PIXEL_SIZE = 4;
     const u64 base_block_size = static_cast<u64>(DefaultBlockWidth(format)) *
                                 static_cast<u64>(DefaultBlockHeight(format)) * RGBA8_PIXEL_SIZE;
-    return (base_size * base_block_size) / BytesPerBlock(format);
+    const u64 uncompressed_size = (base_size * base_block_size) / BytesPerBlock(format);
+
+    switch (Settings::values.astc_recompression.GetValue()) {
+    case Settings::AstcRecompression::Bc1:
+        return uncompressed_size / 8;
+    case Settings::AstcRecompression::Bc3:
+        return uncompressed_size / 4;
+    default:
+        return uncompressed_size;
+    }
 }
 
 } // namespace VideoCore::Surface

src/video_core/surface.h

@@ -517,6 +517,6 @@ size_t PixelComponentSizeBitsInteger(PixelFormat format);
 
 std::pair<u32, u32> GetASTCBlockSize(PixelFormat format);
 
-u64 EstimatedDecompressedSize(u64 base_size, PixelFormat format);
+u64 TranscodedAstcSize(u64 base_size, PixelFormat format);
 
 } // namespace VideoCore::Surface

src/video_core/texture_cache/texture_cache.h

@@ -55,7 +55,7 @@ TextureCache<P>::TextureCache(Runtime& runtime_, Tegra::MaxwellDeviceMemoryManag
         const s64 min_spacing_critical = device_local_memory - 512_MiB;
         const s64 mem_threshold = std::min(device_local_memory, TARGET_THRESHOLD);
         const s64 min_vacancy_expected = (6 * mem_threshold) / 10;
-        const s64 min_vacancy_critical = (3 * mem_threshold) / 10;
+        const s64 min_vacancy_critical = (2 * mem_threshold) / 10;
         expected_memory = static_cast<u64>(
             std::max(std::min(device_local_memory - min_vacancy_expected, min_spacing_expected),
                      DEFAULT_EXPECTED_MEMORY));
@@ -1979,7 +1979,7 @@ void TextureCache<P>::RegisterImage(ImageId image_id) {
     if ((IsPixelFormatASTC(image.info.format) &&
          True(image.flags & ImageFlagBits::AcceleratedUpload)) ||
         True(image.flags & ImageFlagBits::Converted)) {
-        tentative_size = EstimatedDecompressedSize(tentative_size, image.info.format);
+        tentative_size = TranscodedAstcSize(tentative_size, image.info.format);
     }
     total_used_memory += Common::AlignUp(tentative_size, 1024);
     image.lru_index = lru_cache.Insert(image_id, frame_tick);
@@ -2149,7 +2149,7 @@ void TextureCache<P>::DeleteImage(ImageId image_id, bool immediate_delete) {
     if ((IsPixelFormatASTC(image.info.format) &&
          True(image.flags & ImageFlagBits::AcceleratedUpload)) ||
         True(image.flags & ImageFlagBits::Converted)) {
-        tentative_size = EstimatedDecompressedSize(tentative_size, image.info.format);
+        tentative_size = TranscodedAstcSize(tentative_size, image.info.format);
     }
     total_used_memory -= Common::AlignUp(tentative_size, 1024);
     const GPUVAddr gpu_addr = image.gpu_addr;

src/video_core/vulkan_common/vulkan_device.cpp

@@ -1290,6 +1290,10 @@ u64 Device::GetDeviceMemoryUsage() const {
 }
 
 void Device::CollectPhysicalMemoryInfo() {
+    // Account for resolution scaling in memory limits
+    const size_t normal_memory = 6_GiB;
+    const size_t scaler_memory = 1_GiB * Settings::values.resolution_info.ScaleUp(1);
+
     // Calculate limits using memory budget
     VkPhysicalDeviceMemoryBudgetPropertiesEXT budget{};
     budget.sType = VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_MEMORY_BUDGET_PROPERTIES_EXT;
@@ -1320,6 +1324,7 @@ void Device::CollectPhysicalMemoryInfo() {
     if (!is_integrated) {
         const u64 reserve_memory = std::min<u64>(device_access_memory / 8, 1_GiB);
         device_access_memory -= reserve_memory;
+        device_access_memory = std::min<u64>(device_access_memory, normal_memory + scaler_memory);
         return;
     }
     const s64 available_memory = static_cast<s64>(device_access_memory - device_initial_usage);