raphiaura/KytyPS5
1
0
Fork 0
mirror of https://github.com/KytyPS5/KytyPS5.git synced 2026-10-20 02:47:02 +02:00
Code Issues Packages Projects Releases Wiki Activity

Fix wave64 LDS synchronization on subgroup32 hosts

Browse Source
This commit is contained in:
nmzik
2026-08-23 00:46:37 +02:00
parent 0d92ce49da
commit 3e25eb0998
10 changed files with 348 additions and 15 deletions
Download Patch File Download Diff File Expand all files Collapse all files
src/graphics/host_gpu/renderer/pipeline/shaders.cpp
+1 -2
View File
@@ -1019,8 +1019,7 @@ void CreatePipelineInternal(GraphicContext& graphics, PipelineCache::ComputePipe
comp_shader_stage_info.pSpecializationInfo = nullptr;
EXIT_IF(!input_info.stage);
const auto wave_size = input_info.stage.program->wave_size;
if (graphics.subgroup_size_control_enabled &&
(graphics.required_subgroup_size_stages & vk::ShaderStageFlagBits::eCompute) &&
if (graphics.compute_subgroup_size_control_enabled &&
wave_size >= graphics.min_subgroup_size && wave_size <= graphics.max_subgroup_size) {
comp_subgroup_size.sType =
vk::StructureType::ePipelineShaderStageRequiredSubgroupSizeCreateInfo;
src/graphics/host_gpu/renderer/renderCompute.cpp
+3 -1
View File
@@ -210,7 +210,9 @@ void RenderExecutor::DispatchDirect(uint64_t submit_id, RenderCommandBuffer& buf
ShaderComputeInputInfo input_info {};
std::span<const uint32_t> cs_shader;
if (!ShaderCompileInfoCS(cs_regs, sh_regs, input_info, cs_shader)) {
if (!ShaderCompileInfoCS(cs_regs, sh_regs,
!m_context.GetGraphics().compute_wave64_supported, input_info,
cs_shader)) {
EXIT("ShaderCompileInfoCS failed for dispatch with CS shader 0x%016" PRIx64 "\n",
cs_regs.cs_regs.data_addr);
}
src/graphics/host_gpu/vulkanInstance.h
+2 -1
View File
@@ -25,7 +25,8 @@ struct VulkanInstance {
VmaAllocator allocator = nullptr;
bool memory_budget_ext_enabled = false;
bool rt_extensions_enabled = false;
bool subgroup_size_control_enabled = false;
bool compute_subgroup_size_control_enabled = false;
bool compute_wave64_supported = false;
bool sample_rate_shading_enabled = false;
uint32_t subgroup_size = 0;
uint32_t min_subgroup_size = 0;