Downloafy/suyu-forked
1
0
Fork 1
forked from suyu/suyu
Code Pull requests 1 Activity

Merge pull request #9588 from liamwhite/bylaws-reverts

Browse source 
Revert "shader_recompiler: Align SSBO offsets to meet host requirements"
This commit is contained in:
liamwhite 2023-02-19 13:11:57 -05:00 committed by GitHub
commit a4d0663158
No known key found for this signature in database
GPG key ID: 4AEE18F83AFDEB23
10 changed files with 9 additions and 34 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

2
src/shader_recompiler/frontend/maxwell/translate_program.cpp
View file

@ -292,7 +292,7 @@ IR::Program TranslateProgram(ObjectPool<IR::Inst>& inst_pool, ObjectPool<IR::Blo
Optimization::PositionPass(env, program);
Optimization::GlobalMemoryToStorageBufferPass(program, host_info);
Optimization::GlobalMemoryToStorageBufferPass(program);
Optimization::TexturePass(env, program, host_info);
if (Settings::values.resolution_info.active) {

1
src/shader_recompiler/host_translate_info.h
View file

@ -15,7 +15,6 @@ struct HostTranslateInfo {
bool needs_demote_reorder{}; ///< True when the device needs DemoteToHelperInvocation reordered
bool support_snorm_render_buffer{}; ///< True when the device supports SNORM render buffers
bool support_viewport_index_layer{}; ///< True when the device supports gl_Layer in VS
u32 min_ssbo_alignment{}; ///< Minimum alignment supported by the device for SSBOs
bool support_geometry_shader_passthrough{}; ///< True when the device supports geometry
///< passthrough shaders
};

13
src/shader_recompiler/ir_opt/global_memory_to_storage_buffer_pass.cpp
View file

@ -11,7 +11,6 @@
#include "shader_recompiler/frontend/ir/breadth_first_search.h"
#include "shader_recompiler/frontend/ir/ir_emitter.h"
#include "shader_recompiler/frontend/ir/value.h"
#include "shader_recompiler/host_translate_info.h"
#include "shader_recompiler/ir_opt/passes.h"
namespace Shader::Optimization {
@ -403,7 +402,7 @@ void CollectStorageBuffers(IR::Block& block, IR::Inst& inst, StorageInfo& info)
}
/// Returns the offset in indices (not bytes) for an equivalent storage instruction
IR::U32 StorageOffset(IR::Block& block, IR::Inst& inst, StorageBufferAddr buffer, u32 alignment) {
IR::U32 StorageOffset(IR::Block& block, IR::Inst& inst, StorageBufferAddr buffer) {
IR::IREmitter ir{block, IR::Block::InstructionList::s_iterator_to(inst)};
IR::U32 offset;
if (const std::optional<LowAddrInfo> low_addr{TrackLowAddress(&inst)}) {
@ -416,10 +415,7 @@ IR::U32 StorageOffset(IR::Block& block, IR::Inst& inst, StorageBufferAddr buffer
}
// Subtract the least significant 32 bits from the guest offset. The result is the storage
// buffer offset in bytes.
IR::U32 low_cbuf{ir.GetCbuf(ir.Imm32(buffer.index), ir.Imm32(buffer.offset))};
// Align the offset base to match the host alignment requirements
low_cbuf = ir.BitwiseAnd(low_cbuf, ir.Imm32(~(alignment - 1U)));
const IR::U32 low_cbuf{ir.GetCbuf(ir.Imm32(buffer.index), ir.Imm32(buffer.offset))};
return ir.ISub(offset, low_cbuf);
}
@ -514,7 +510,7 @@ void Replace(IR::Block& block, IR::Inst& inst, const IR::U32& storage_index,
}
} // Anonymous namespace
void GlobalMemoryToStorageBufferPass(IR::Program& program, const HostTranslateInfo& host_info) {
void GlobalMemoryToStorageBufferPass(IR::Program& program) {
StorageInfo info;
for (IR::Block* const block : program.post_order_blocks) {
for (IR::Inst& inst : block->Instructions()) {
@ -538,8 +534,7 @@ void GlobalMemoryToStorageBufferPass(IR::Program& program, const HostTranslateIn
const IR::U32 index{IR::Value{static_cast<u32>(info.set.index_of(it))}};
IR::Block* const block{storage_inst.block};
IR::Inst* const inst{storage_inst.inst};
const IR::U32 offset{
StorageOffset(*block, *inst, storage_buffer, host_info.min_ssbo_alignment)};
const IR::U32 offset{StorageOffset(*block, *inst, storage_buffer)};
Replace(*block, *inst, index, offset);
}
}

2
src/shader_recompiler/ir_opt/passes.h
View file

@ -15,7 +15,7 @@ namespace Shader::Optimization {
void CollectShaderInfoPass(Environment& env, IR::Program& program);
void ConstantPropagationPass(Environment& env, IR::Program& program);
void DeadCodeEliminationPass(IR::Program& program);
void GlobalMemoryToStorageBufferPass(IR::Program& program, const HostTranslateInfo& host_info);
void GlobalMemoryToStorageBufferPass(IR::Program& program);
void IdentityRemovalPass(IR::Program& program);
void LowerFp16ToFp32(IR::Program& program);
void LowerInt64ToInt32(IR::Program& program);

13
src/video_core/buffer_cache/buffer_cache.h
View file

@ -1938,21 +1938,14 @@ typename BufferCache<P>::Binding BufferCache<P>::StorageBufferBinding(GPUVAddr s
bool is_written) const {
const GPUVAddr gpu_addr = gpu_memory->Read<u64>(ssbo_addr);
const u32 size = gpu_memory->Read<u32>(ssbo_addr + 8);
const u32 alignment = runtime.GetStorageBufferAlignment();
const GPUVAddr aligned_gpu_addr = Common::AlignDown(gpu_addr, alignment);
const u32 aligned_size =
Common::AlignUp(static_cast<u32>(gpu_addr - aligned_gpu_addr) + size, alignment);
const std::optional<VAddr> cpu_addr = gpu_memory->GpuToCpuAddress(aligned_gpu_addr);
const std::optional<VAddr> cpu_addr = gpu_memory->GpuToCpuAddress(gpu_addr);
if (!cpu_addr || size == 0) {
return NULL_BINDING;
}
const VAddr cpu_end = Common::AlignUp(*cpu_addr + aligned_size, Core::Memory::YUZU_PAGESIZE);
const VAddr cpu_end = Common::AlignUp(*cpu_addr + size, Core::Memory::YUZU_PAGESIZE);
const Binding binding{
.cpu_addr = *cpu_addr,
.size = is_written ? aligned_size : static_cast<u32>(cpu_end - *cpu_addr),
.size = is_written ? size : static_cast<u32>(cpu_end - *cpu_addr),
.buffer_id = BufferId{},
};
return binding;

4
src/video_core/renderer_opengl/gl_buffer_cache.h
View file

@ -160,10 +160,6 @@ public:
return device.CanReportMemoryUsage();
}
u32 GetStorageBufferAlignment() const {
return static_cast<u32>(device.GetShaderStorageBufferAlignment());
}
private:
static constexpr std::array PABO_LUT{
GL_VERTEX_PROGRAM_PARAMETER_BUFFER_NV, GL_TESS_CONTROL_PROGRAM_PARAMETER_BUFFER_NV,

1
src/video_core/renderer_opengl/gl_shader_cache.cpp
View file

@ -236,7 +236,6 @@ ShaderCache::ShaderCache(RasterizerOpenGL& rasterizer_, Core::Frontend::EmuWindo
.needs_demote_reorder = device.IsAmd(),
.support_snorm_render_buffer = false,
.support_viewport_index_layer = device.HasVertexViewportLayer(),
.min_ssbo_alignment = static_cast<u32>(device.GetShaderStorageBufferAlignment()),
.support_geometry_shader_passthrough = device.HasGeometryShaderPassthrough(),
} {
if (use_asynchronous_shaders) {

4
src/video_core/renderer_vulkan/vk_buffer_cache.cpp
View file

@ -330,10 +330,6 @@ bool BufferCacheRuntime::CanReportMemoryUsage() const {
return device.CanReportMemoryUsage();
}
u32 BufferCacheRuntime::GetStorageBufferAlignment() const {
return static_cast<u32>(device.GetStorageBufferAlignment());
}
void BufferCacheRuntime::Finish() {
scheduler.Finish();
}

2
src/video_core/renderer_vulkan/vk_buffer_cache.h
View file

@ -73,8 +73,6 @@ public:
bool CanReportMemoryUsage() const;
u32 GetStorageBufferAlignment() const;
[[nodiscard]] StagingBufferRef UploadStagingBuffer(size_t size);
[[nodiscard]] StagingBufferRef DownloadStagingBuffer(size_t size);

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

@ -344,7 +344,6 @@ PipelineCache::PipelineCache(RasterizerVulkan& rasterizer_, const Device& device
driver_id == VK_DRIVER_ID_AMD_PROPRIETARY || driver_id == VK_DRIVER_ID_AMD_OPEN_SOURCE,
.support_snorm_render_buffer = true,
.support_viewport_index_layer = device.IsExtShaderViewportIndexLayerSupported(),
.min_ssbo_alignment = static_cast<u32>(device.GetStorageBufferAlignment()),
.support_geometry_shader_passthrough = device.IsNvGeometryShaderPassthroughSupported(),
};