mirrors/shadPS4
1
0
Fork 0
mirror of https://github.com/shadps4-emu/shadPS4.git synced 2025-07-30 22:14:56 +00:00
Code Issues Packages Projects Releases Wiki Activity

Handle missing VK_EXT_shader_atomic_float2

Browse Source
This commit is contained in:
Marcin Mikołajczyk 2025-04-22 20:07:03 +01:00
parent b327f4a4bc
commit 03b19926ae
7 changed files with 57 additions and 4 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

2
externals/sirit vendored

@ -1 +1 @@
Subproject commit 4744731a1c3d7868d43036c507909dae8321ad99
Subproject commit 09a1416ab1b59ddfebd2618412f118f2004f3b2c

30
src/shader_recompiler/backend/spirv/emit_spirv_atomic.cpp
View File

@ -196,11 +196,37 @@ Id EmitImageAtomicUMax32(EmitContext& ctx, IR::Inst* inst, u32 handle, Id coords
}
Id EmitImageAtomicFMax32(EmitContext& ctx, IR::Inst* inst, u32 handle, Id coords, Id value) {
return ImageAtomicF32(ctx, inst, handle, coords, value, &Sirit::Module::OpAtomicFMax);
if (ctx.profile.supports_image_fp32_atomic_min_max) {
return ImageAtomicF32(ctx, inst, handle, coords, value, &Sirit::Module::OpAtomicFMax);
}
const auto u32_value = ctx.OpBitcast(ctx.U32[1], value);
const auto sign_bit_set =
ctx.OpBitFieldUExtract(ctx.U32[1], u32_value, ctx.ConstU32(31u), ctx.ConstU32(1u));
const auto result = ctx.OpSelect(
ctx.F32[1], sign_bit_set,
EmitBitCastF32U32(ctx, EmitImageAtomicUMin32(ctx, inst, handle, coords, u32_value)),
EmitBitCastF32U32(ctx, EmitImageAtomicSMax32(ctx, inst, handle, coords, u32_value)));
return result;
}
Id EmitImageAtomicFMin32(EmitContext& ctx, IR::Inst* inst, u32 handle, Id coords, Id value) {
return ImageAtomicF32(ctx, inst, handle, coords, value, &Sirit::Module::OpAtomicFMin);
if (ctx.profile.supports_image_fp32_atomic_min_max) {
return ImageAtomicF32(ctx, inst, handle, coords, value, &Sirit::Module::OpAtomicFMin);
}
const auto u32_value = ctx.OpBitcast(ctx.U32[1], value);
const auto sign_bit_set =
ctx.OpBitFieldUExtract(ctx.U32[1], u32_value, ctx.ConstU32(31u), ctx.ConstU32(1u));
const auto result = ctx.OpSelect(
ctx.F32[1], sign_bit_set,
EmitBitCastF32U32(ctx, EmitImageAtomicUMax32(ctx, inst, handle, coords, u32_value)),
EmitBitCastF32U32(ctx, EmitImageAtomicSMin32(ctx, inst, handle, coords, u32_value)));
return result;
}
Id EmitImageAtomicInc32(EmitContext&, IR::Inst*, u32, Id, Id) {

2
src/shader_recompiler/frontend/translate/vector_memory.cpp
View File

@ -115,6 +115,8 @@ void Translator::EmitVectorMemory(const GcnInst& inst) {
return IMAGE_ATOMIC(AtomicOp::Smin, inst);
case Opcode::IMAGE_ATOMIC_UMIN:
return IMAGE_ATOMIC(AtomicOp::Umin, inst);
case Opcode::IMAGE_ATOMIC_FMIN:
return IMAGE_ATOMIC(AtomicOp::Fmin, inst);
case Opcode::IMAGE_ATOMIC_SMAX:
return IMAGE_ATOMIC(AtomicOp::Smax, inst);
case Opcode::IMAGE_ATOMIC_FMAX:

1
src/shader_recompiler/profile.h
View File

@ -29,6 +29,7 @@ struct Profile {
bool supports_native_cube_calc{};
bool supports_trinary_minmax{};
bool supports_robust_buffer_access{};
bool supports_image_fp32_atomic_min_max{};
bool has_broken_spirv_clamp{};
bool lower_left_origin_mode{};
bool needs_manual_interpolation{};

17
src/video_core/renderer_vulkan/vk_instance.cpp
View File

@ -210,7 +210,8 @@ bool Instance::CreateDevice() {
vk::PhysicalDeviceRobustness2FeaturesEXT,
vk::PhysicalDeviceExtendedDynamicState3FeaturesEXT,
vk::PhysicalDevicePrimitiveTopologyListRestartFeaturesEXT,
vk::PhysicalDevicePortabilitySubsetFeaturesKHR>();
vk::PhysicalDevicePortabilitySubsetFeaturesKHR,
vk::PhysicalDeviceShaderAtomicFloat2FeaturesEXT>();
features = feature_chain.get().features;
const vk::StructureChain properties_chain = physical_device.getProperties2<
@ -272,6 +273,13 @@ bool Instance::CreateDevice() {
image_load_store_lod = add_extension(VK_AMD_SHADER_IMAGE_LOAD_STORE_LOD_EXTENSION_NAME);
amd_gcn_shader = add_extension(VK_AMD_GCN_SHADER_EXTENSION_NAME);
amd_shader_trinary_minmax = add_extension(VK_AMD_SHADER_TRINARY_MINMAX_EXTENSION_NAME);
shader_atomic_float2 = add_extension(VK_EXT_SHADER_ATOMIC_FLOAT_2_EXTENSION_NAME);
if (shader_atomic_float2) {
shader_atomic_float2_features =
feature_chain.get<vk::PhysicalDeviceShaderAtomicFloat2FeaturesEXT>();
LOG_INFO(Render_Vulkan, "- shaderImageFloat32AtomicMinMax: {}",
shader_atomic_float2_features.shaderImageFloat32AtomicMinMax);
}
const bool calibrated_timestamps =
TRACY_GPU_ENABLED ? add_extension(VK_EXT_CALIBRATED_TIMESTAMPS_EXTENSION_NAME) : false;
@ -401,6 +409,10 @@ bool Instance::CreateDevice() {
vk::PhysicalDeviceLegacyVertexAttributesFeaturesEXT{
.legacyVertexAttributes = true,
},
vk::PhysicalDeviceShaderAtomicFloat2FeaturesEXT{
.shaderImageFloat32AtomicMinMax =
shader_atomic_float2_features.shaderImageFloat32AtomicMinMax,
},
#ifdef __APPLE__
portability_features,
#endif
@ -430,6 +442,9 @@ bool Instance::CreateDevice() {
if (!legacy_vertex_attributes) {
device_chain.unlink<vk::PhysicalDeviceLegacyVertexAttributesFeaturesEXT>();
}
if (!shader_atomic_float2) {
device_chain.unlink<vk::PhysicalDeviceShaderAtomicFloat2FeaturesEXT>();
}
auto [device_result, dev] = physical_device.createDeviceUnique(device_chain.get());
if (device_result != vk::Result::eSuccess) {

8
src/video_core/renderer_vulkan/vk_instance.h
View File

@ -165,6 +165,12 @@ public:
return amd_shader_trinary_minmax;
}
/// Returns true when the shaderImageFloat32AtomicMinMax feature of
/// VK_EXT_shader_atomic_float2 is supported.
bool IsShaderAtomicFloatImage32MinMaxSupported() const {
return shader_atomic_float2 && shader_atomic_float2_features.shaderImageFloat32AtomicMinMax;
}
/// Returns true when geometry shaders are supported by the device
bool IsGeometryStageSupported() const {
return features.geometryShader;
@ -336,6 +342,7 @@ private:
vk::PhysicalDevicePortabilitySubsetFeaturesKHR portability_features;
vk::PhysicalDeviceExtendedDynamicState3FeaturesEXT dynamic_state_3_features;
vk::PhysicalDeviceRobustness2FeaturesEXT robustness2_features;
vk::PhysicalDeviceShaderAtomicFloat2FeaturesEXT shader_atomic_float2_features;
vk::DriverIdKHR driver_id;
vk::UniqueDebugUtilsMessengerEXT debug_callback{};
std::string vendor_name;
@ -360,6 +367,7 @@ private:
bool image_load_store_lod{};
bool amd_gcn_shader{};
bool amd_shader_trinary_minmax{};
bool shader_atomic_float2{};
bool portability_subset{};
};

1
src/video_core/renderer_vulkan/vk_pipeline_cache.cpp
View File

@ -206,6 +206,7 @@ PipelineCache::PipelineCache(const Instance& instance_, Scheduler& scheduler_,
.supports_native_cube_calc = instance_.IsAmdGcnShaderSupported(),
.supports_trinary_minmax = instance_.IsAmdShaderTrinaryMinMaxSupported(),
.supports_robust_buffer_access = instance_.IsRobustBufferAccess2Supported(),
.supports_image_fp32_atomic_min_max = instance_.IsShaderAtomicFloatImage32MinMaxSupported(),
.needs_manual_interpolation = instance.IsFragmentShaderBarycentricSupported() &&
instance.GetDriverID() == vk::DriverId::eNvidiaProprietary,
.needs_lds_barriers = instance.GetDriverID() == vk::DriverId::eNvidiaProprietary ||