diff options
5 files changed, 61 insertions, 57 deletions
diff --git a/src/shader_recompiler/backend/spirv/emit_spirv_image_atomic.cpp b/src/shader_recompiler/backend/spirv/emit_spirv_image_atomic.cpp index 2a12feddc..dde0f6e9c 100644 --- a/src/shader_recompiler/backend/spirv/emit_spirv_image_atomic.cpp +++ b/src/shader_recompiler/backend/spirv/emit_spirv_image_atomic.cpp @@ -7,15 +7,12 @@ namespace Shader::Backend::SPIRV { namespace { -Id Image(EmitContext& ctx, const IR::Value& index, IR::TextureInstInfo info) { - if (!index.IsImmediate()) { - throw NotImplementedException("Indirect image indexing"); - } +Id Image(EmitContext& ctx, IR::TextureInstInfo info) { if (info.type == TextureType::Buffer) { - const ImageBufferDefinition def{ctx.image_buffers.at(index.U32())}; + const ImageBufferDefinition def{ctx.image_buffers.at(info.descriptor_index)}; return def.id; } else { - const ImageDefinition def{ctx.images.at(index.U32())}; + const ImageDefinition def{ctx.images.at(info.descriptor_index)}; return def.id; } } @@ -28,8 +25,12 @@ std::pair<Id, Id> AtomicArgs(EmitContext& ctx) { Id ImageAtomicU32(EmitContext& ctx, IR::Inst* inst, const IR::Value& index, Id coords, Id value, Id (Sirit::Module::*atomic_func)(Id, Id, Id, Id, Id)) { + if (!index.IsImmediate() || index.U32() != 0) { + // TODO: handle layers + throw NotImplementedException("Image indexing"); + } const auto info{inst->Flags<IR::TextureInstInfo>()}; - const Id image{Image(ctx, index, info)}; + const Id image{Image(ctx, info)}; const Id pointer{ctx.OpImageTexelPointer(ctx.image_u32, image, coords, ctx.Const(0U))}; const auto [scope, semantics]{AtomicArgs(ctx)}; return (ctx.*atomic_func)(ctx.U32[1], pointer, scope, semantics, value); diff --git a/src/shader_recompiler/backend/spirv/spirv_emit_context.cpp b/src/shader_recompiler/backend/spirv/spirv_emit_context.cpp index 72f69b7aa..57df6fc34 100644 --- a/src/shader_recompiler/backend/spirv/spirv_emit_context.cpp +++ b/src/shader_recompiler/backend/spirv/spirv_emit_context.cpp @@ -74,11 +74,6 @@ spv::ImageFormat GetImageFormat(ImageFormat format) { throw InvalidArgument("Invalid image format {}", format); } -spv::ImageFormat GetImageFormatForBuffer(ImageFormat format) { - const auto spv_format = GetImageFormat(format); - return spv_format == spv::ImageFormat::Unknown ? spv::ImageFormat::R32ui : spv_format; -} - Id ImageType(EmitContext& ctx, const ImageDescriptor& desc) { const spv::ImageFormat format{GetImageFormat(desc.format)}; const Id type{ctx.U32[1]}; @@ -1275,7 +1270,7 @@ void EmitContext::DefineImageBuffers(const Info& info, u32& binding) { if (desc.count != 1) { throw NotImplementedException("Array of image buffers"); } - const spv::ImageFormat format{GetImageFormatForBuffer(desc.format)}; + const spv::ImageFormat format{GetImageFormat(desc.format)}; const Id image_type{TypeImage(U32[1], 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)}; diff --git a/src/video_core/renderer_vulkan/vk_texture_cache.cpp b/src/video_core/renderer_vulkan/vk_texture_cache.cpp index 285a50ea4..1f9e7acaa 100644 --- a/src/video_core/renderer_vulkan/vk_texture_cache.cpp +++ b/src/video_core/renderer_vulkan/vk_texture_cache.cpp @@ -120,19 +120,9 @@ constexpr VkBorderColor ConvertBorderColor(const std::array<float, 4>& color) { return usage; } -/// Returns the preferred format for a VkImage -[[nodiscard]] PixelFormat StorageFormat(PixelFormat format) { - switch (format) { - case PixelFormat::A8B8G8R8_SRGB: - return PixelFormat::A8B8G8R8_UNORM; - default: - return format; - } -} - [[nodiscard]] VkImageCreateInfo MakeImageCreateInfo(const Device& device, const ImageInfo& info) { - const PixelFormat format = StorageFormat(info.format); - const auto format_info = MaxwellToVK::SurfaceFormat(device, FormatType::Optimal, false, format); + const auto format_info = + MaxwellToVK::SurfaceFormat(device, FormatType::Optimal, false, info.format); VkImageCreateFlags flags{}; if (info.type == ImageType::e2D && info.resources.layers >= 6 && info.size.width == info.size.height && !device.HasBrokenCubeImageCompability()) { @@ -157,7 +147,7 @@ constexpr VkBorderColor ConvertBorderColor(const std::array<float, 4>& color) { .arrayLayers = static_cast<u32>(info.resources.layers), .samples = ConvertSampleCount(info.num_samples), .tiling = VK_IMAGE_TILING_OPTIMAL, - .usage = ImageUsageFlags(format_info, format), + .usage = ImageUsageFlags(format_info, info.format), .sharingMode = VK_SHARING_MODE_EXCLUSIVE, .queueFamilyIndexCount = 0, .pQueueFamilyIndices = nullptr, @@ -1049,15 +1039,27 @@ void TextureCacheRuntime::BlitImage(Framebuffer* dst_framebuffer, ImageView& dst dst_region, src_region, filter, operation); return; } + ASSERT(src.format == dst.format); if (aspect_mask == (VK_IMAGE_ASPECT_DEPTH_BIT | VK_IMAGE_ASPECT_STENCIL_BIT)) { - if (!device.IsBlitDepthStencilSupported()) { + const auto format = src.format; + const auto can_blit_depth_stencil = [this, format] { + switch (format) { + case VideoCore::Surface::PixelFormat::D24_UNORM_S8_UINT: + case VideoCore::Surface::PixelFormat::S8_UINT_D24_UNORM: + return device.IsBlitDepth24Stencil8Supported(); + case VideoCore::Surface::PixelFormat::D32_FLOAT_S8_UINT: + return device.IsBlitDepth32Stencil8Supported(); + default: + UNREACHABLE(); + } + }(); + if (!can_blit_depth_stencil) { UNIMPLEMENTED_IF(is_src_msaa || is_dst_msaa); blit_image_helper.BlitDepthStencil(dst_framebuffer, src.DepthView(), src.StencilView(), dst_region, src_region, filter, operation); return; } } - ASSERT(src.format == dst.format); ASSERT(!(is_dst_msaa && !is_src_msaa)); ASSERT(operation == Fermi2D::Operation::SrcCopy); @@ -1631,8 +1633,8 @@ bool Image::NeedsScaleHelper() const { return true; } static constexpr auto OPTIMAL_FORMAT = FormatType::Optimal; - const PixelFormat format = StorageFormat(info.format); - const auto vk_format = MaxwellToVK::SurfaceFormat(device, OPTIMAL_FORMAT, false, format).format; + const auto vk_format = + MaxwellToVK::SurfaceFormat(device, OPTIMAL_FORMAT, false, info.format).format; const auto blit_usage = VK_FORMAT_FEATURE_BLIT_SRC_BIT | VK_FORMAT_FEATURE_BLIT_DST_BIT; const bool needs_blit_helper = !device.IsFormatSupported(vk_format, blit_usage, OPTIMAL_FORMAT); return needs_blit_helper; diff --git a/src/video_core/vulkan_common/vulkan_device.cpp b/src/video_core/vulkan_common/vulkan_device.cpp index a88ff5ca5..18185610f 100644 --- a/src/video_core/vulkan_common/vulkan_device.cpp +++ b/src/video_core/vulkan_common/vulkan_device.cpp @@ -428,7 +428,8 @@ Device::Device(VkInstance instance_, vk::PhysicalDevice physical_, VkSurfaceKHR first_next = &diagnostics_nv; } - is_blit_depth_stencil_supported = TestDepthStencilBlits(); + is_blit_depth24_stencil8_supported = TestDepthStencilBlits(VK_FORMAT_D24_UNORM_S8_UINT); + is_blit_depth32_stencil8_supported = TestDepthStencilBlits(VK_FORMAT_D32_SFLOAT_S8_UINT); is_optimal_astc_supported = ComputeIsOptimalAstcSupported(); is_warp_potentially_bigger = !extensions.subgroup_size_control || properties.subgroup_size_control.maxSubgroupSize > GuestWarpSize; @@ -782,14 +783,13 @@ bool Device::ComputeIsOptimalAstcSupported() const { return true; } -bool Device::TestDepthStencilBlits() const { +bool Device::TestDepthStencilBlits(VkFormat format) const { static constexpr VkFormatFeatureFlags required_features = VK_FORMAT_FEATURE_BLIT_SRC_BIT | VK_FORMAT_FEATURE_BLIT_DST_BIT; const auto test_features = [](VkFormatProperties props) { return (props.optimalTilingFeatures & required_features) == required_features; }; - return test_features(format_properties.at(VK_FORMAT_D32_SFLOAT_S8_UINT)) && - test_features(format_properties.at(VK_FORMAT_D24_UNORM_S8_UINT)); + return test_features(format_properties.at(format)); } bool Device::IsFormatSupported(VkFormat wanted_format, VkFormatFeatureFlags wanted_usage, diff --git a/src/video_core/vulkan_common/vulkan_device.h b/src/video_core/vulkan_common/vulkan_device.h index 6c7fa34e5..8c5355a28 100644 --- a/src/video_core/vulkan_common/vulkan_device.h +++ b/src/video_core/vulkan_common/vulkan_device.h @@ -362,9 +362,14 @@ public: return features.features.depthBounds; } - /// Returns true when blitting from and to depth stencil images is supported. - bool IsBlitDepthStencilSupported() const { - return is_blit_depth_stencil_supported; + /// Returns true when blitting from and to D24S8 images is supported. + bool IsBlitDepth24Stencil8Supported() const { + return is_blit_depth24_stencil8_supported; + } + + /// Returns true when blitting from and to D32S8 images is supported. + bool IsBlitDepth32Stencil8Supported() const { + return is_blit_depth32_stencil8_supported; } /// Returns true if the device supports VK_NV_viewport_swizzle. @@ -674,7 +679,7 @@ private: bool ComputeIsOptimalAstcSupported() const; /// Returns true if the device natively supports blitting depth stencil images. - bool TestDepthStencilBlits() const; + bool TestDepthStencilBlits(VkFormat format) const; private: VkInstance instance; ///< Vulkan instance. @@ -738,25 +743,26 @@ private: VkPhysicalDeviceProperties2 properties2{}; // Misc features - bool is_optimal_astc_supported{}; ///< Support for all guest ASTC formats. - bool is_blit_depth_stencil_supported{}; ///< Support for blitting from and to depth stencil. - bool is_warp_potentially_bigger{}; ///< Host warp size can be bigger than guest. - bool is_integrated{}; ///< Is GPU an iGPU. - bool is_virtual{}; ///< Is GPU a virtual GPU. - bool is_non_gpu{}; ///< Is SoftwareRasterizer, FPGA, non-GPU device. - bool has_broken_compute{}; ///< Compute shaders can cause crashes - bool has_broken_cube_compatibility{}; ///< Has broken cube compatibility bit - bool has_renderdoc{}; ///< Has RenderDoc attached - bool has_nsight_graphics{}; ///< Has Nsight Graphics attached - bool supports_d24_depth{}; ///< Supports D24 depth buffers. - bool cant_blit_msaa{}; ///< Does not support MSAA<->MSAA blitting. - bool must_emulate_scaled_formats{}; ///< Requires scaled vertex format emulation - bool must_emulate_bgr565{}; ///< Emulates BGR565 by swizzling RGB565 format. - bool dynamic_state3_blending{}; ///< Has all blending features of dynamic_state3. - bool dynamic_state3_enables{}; ///< Has all enables features of dynamic_state3. - bool supports_conditional_barriers{}; ///< Allows barriers in conditional control flow. - u64 device_access_memory{}; ///< Total size of device local memory in bytes. - u32 sets_per_pool{}; ///< Sets per Description Pool + bool is_optimal_astc_supported{}; ///< Support for all guest ASTC formats. + bool is_blit_depth24_stencil8_supported{}; ///< Support for blitting from and to D24S8. + bool is_blit_depth32_stencil8_supported{}; ///< Support for blitting from and to D32S8. + bool is_warp_potentially_bigger{}; ///< Host warp size can be bigger than guest. + bool is_integrated{}; ///< Is GPU an iGPU. + bool is_virtual{}; ///< Is GPU a virtual GPU. + bool is_non_gpu{}; ///< Is SoftwareRasterizer, FPGA, non-GPU device. + bool has_broken_compute{}; ///< Compute shaders can cause crashes + bool has_broken_cube_compatibility{}; ///< Has broken cube compatibility bit + bool has_renderdoc{}; ///< Has RenderDoc attached + bool has_nsight_graphics{}; ///< Has Nsight Graphics attached + bool supports_d24_depth{}; ///< Supports D24 depth buffers. + bool cant_blit_msaa{}; ///< Does not support MSAA<->MSAA blitting. + bool must_emulate_scaled_formats{}; ///< Requires scaled vertex format emulation + bool must_emulate_bgr565{}; ///< Emulates BGR565 by swizzling RGB565 format. + bool dynamic_state3_blending{}; ///< Has all blending features of dynamic_state3. + bool dynamic_state3_enables{}; ///< Has all enables features of dynamic_state3. + bool supports_conditional_barriers{}; ///< Allows barriers in conditional control flow. + u64 device_access_memory{}; ///< Total size of device local memory in bytes. + u32 sets_per_pool{}; ///< Sets per Description Pool // Telemetry parameters std::set<std::string, std::less<>> supported_extensions; ///< Reported Vulkan extensions. |