The-Homebrew-Cloud/Cemu
2
0
Fork 0
mirror of https://github.com/cemu-project/Cemu.git synced 2024-12-01 21:44:17 +01:00
Code Issues Packages Projects Releases Wiki Activity

use texture views for surface copies

Browse Source
This commit is contained in:
Samuliak 2024-08-28 13:19:51 +02:00
parent 366be049a4
commit 163eeea102
4 changed files with 54 additions and 42 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

4
src/Cafe/HW/Latte/Renderer/Metal/MetalHybridComputePipeline.cpp
View File

@ -1,6 +1,6 @@
#include "Cafe/HW/Latte/Renderer/Metal/MetalHybridComputePipeline.h" #include "Cafe/HW/Latte/Renderer/Metal/MetalHybridComputePipeline.h"
MetalHybridComputePipeline::MetalHybridComputePipeline(class MetalRenderer* mtlRenderer, MTL::Library* library, const char* vertexFunctionName, const char* kernelFunctionName) MetalHybridComputePipeline::MetalHybridComputePipeline(class MetalRenderer* mtlRenderer, MTL::Library* library, const std::string& vertexFunctionName/*, const std::string& kernelFunctionName*/)
{ {
// Render pipeline state // Render pipeline state
MTL::Function* vertexFunction = library->newFunction(ToNSString(vertexFunctionName)); MTL::Function* vertexFunction = library->newFunction(ToNSString(vertexFunctionName));
@ -20,12 +20,10 @@ MetalHybridComputePipeline::MetalHybridComputePipeline(class MetalRenderer* mtlR
} }
// Compute pipeline state // Compute pipeline state
// TODO
} }
MetalHybridComputePipeline::~MetalHybridComputePipeline() MetalHybridComputePipeline::~MetalHybridComputePipeline()
{ {
m_renderPipelineState->release(); m_renderPipelineState->release();
// TODO: uncomment
//m_computePipelineState->release(); //m_computePipelineState->release();
} }

7
src/Cafe/HW/Latte/Renderer/Metal/MetalHybridComputePipeline.h
View File

@ -3,17 +3,18 @@
#include "Metal/MTLLibrary.hpp" #include "Metal/MTLLibrary.hpp"
#include "Metal/MTLRenderPipeline.hpp" #include "Metal/MTLRenderPipeline.hpp"
// TODO: rename to MetalVoidVertexPipeline
class MetalHybridComputePipeline class MetalHybridComputePipeline
{ {
public: public:
MetalHybridComputePipeline(class MetalRenderer* mtlRenderer, MTL::Library* library, const char* vertexFunctionName, const char* kernelFunctionName); MetalHybridComputePipeline(class MetalRenderer* mtlRenderer, MTL::Library* library, const std::string& vertexFunctionName/*, const std::string& kernelFunctionName*/);
~MetalHybridComputePipeline(); ~MetalHybridComputePipeline();
MTL::RenderPipelineState* GetRenderPipelineState() const { return m_renderPipelineState; } MTL::RenderPipelineState* GetRenderPipelineState() const { return m_renderPipelineState; }
MTL::RenderPipelineState* GetComputePipelineState() const { return m_computePipelineState; } //MTL::RenderPipelineState* GetComputePipelineState() const { return m_computePipelineState; }
private: private:
MTL::RenderPipelineState* m_renderPipelineState; MTL::RenderPipelineState* m_renderPipelineState;
MTL::RenderPipelineState* m_computePipelineState; //MTL::RenderPipelineState* m_computePipelineState;
}; };

71
src/Cafe/HW/Latte/Renderer/Metal/MetalRenderer.cpp
View File

@ -16,6 +16,7 @@
#include "Cafe/HW/Latte/Core/LatteShader.h" #include "Cafe/HW/Latte/Core/LatteShader.h"
#include "Cafe/HW/Latte/Core/LatteIndices.h" #include "Cafe/HW/Latte/Core/LatteIndices.h"
#include "Cemu/Logging/CemuDebugLogging.h" #include "Cemu/Logging/CemuDebugLogging.h"
#include "Foundation/NSTypes.hpp"
#include "HW/Latte/Core/LatteConst.h" #include "HW/Latte/Core/LatteConst.h"
#include "HW/Latte/Renderer/Metal/MetalCommon.h" #include "HW/Latte/Renderer/Metal/MetalCommon.h"
#include "HW/Latte/Renderer/Metal/MetalLayerHandle.h" #include "HW/Latte/Renderer/Metal/MetalLayerHandle.h"
@ -23,7 +24,9 @@
#include "Metal/MTLCommandBuffer.hpp" #include "Metal/MTLCommandBuffer.hpp"
#include "Metal/MTLDevice.hpp" #include "Metal/MTLDevice.hpp"
#include "Metal/MTLRenderPass.hpp" #include "Metal/MTLRenderPass.hpp"
#include "Metal/MTLRenderPipeline.hpp"
#include "imgui.h" #include "imgui.h"
#include <cstddef>
#define IMGUI_IMPL_METAL_CPP #define IMGUI_IMPL_METAL_CPP
#include "imgui/imgui_extension.h" #include "imgui/imgui_extension.h"
@ -117,13 +120,13 @@ MetalRenderer::MetalRenderer()
} }
// Present pipeline // Present pipeline
MTL::Function* presentVertexFunction = utilityLibrary->newFunction(ToNSString("vertexFullscreen")); MTL::Function* fullscreenVertexFunction = utilityLibrary->newFunction(ToNSString("vertexFullscreen"));
MTL::Function* presentFragmentFunction = utilityLibrary->newFunction(ToNSString("fragmentPresent")); MTL::Function* presentFragmentFunction = utilityLibrary->newFunction(ToNSString("fragmentPresent"));
MTL::RenderPipelineDescriptor* renderPipelineDescriptor = MTL::RenderPipelineDescriptor::alloc()->init(); MTL::RenderPipelineDescriptor* renderPipelineDescriptor = MTL::RenderPipelineDescriptor::alloc()->init();
renderPipelineDescriptor->setVertexFunction(presentVertexFunction); renderPipelineDescriptor->setVertexFunction(fullscreenVertexFunction);
renderPipelineDescriptor->setFragmentFunction(presentFragmentFunction); renderPipelineDescriptor->setFragmentFunction(presentFragmentFunction);
presentVertexFunction->release(); fullscreenVertexFunction->release();
presentFragmentFunction->release(); presentFragmentFunction->release();
error = nullptr; error = nullptr;
@ -151,9 +154,12 @@ MetalRenderer::MetalRenderer()
error->release(); error->release();
} }
// Copy texture pipelines
auto copyTextureToColorPipelineDescriptor = MTL::RenderPipelineDescriptor::alloc()->init();
// Hybrid pipelines // Hybrid pipelines
m_copyTextureToTexturePipeline = new MetalHybridComputePipeline(this, utilityLibrary, "vertexCopyTextureToTexture", "kernelCopyTextureToTexture"); m_copyTextureToTexturePipeline = new MetalHybridComputePipeline(this, utilityLibrary, "vertexCopyTextureToTexture");
m_restrideBufferPipeline = new MetalHybridComputePipeline(this, utilityLibrary, "vertexRestrideBuffer", "kernelRestrideBuffer"); m_restrideBufferPipeline = new MetalHybridComputePipeline(this, utilityLibrary, "vertexRestrideBuffer");
utilityLibrary->release(); utilityLibrary->release();
m_memoryManager->SetRestrideBufferPipeline(m_restrideBufferPipeline); m_memoryManager->SetRestrideBufferPipeline(m_restrideBufferPipeline);
@ -674,8 +680,6 @@ LatteTextureReadbackInfo* MetalRenderer::texture_createReadback(LatteTextureView
void MetalRenderer::surfaceCopy_copySurfaceWithFormatConversion(LatteTexture* sourceTexture, sint32 srcMip, sint32 srcSlice, LatteTexture* destinationTexture, sint32 dstMip, sint32 dstSlice, sint32 width, sint32 height) void MetalRenderer::surfaceCopy_copySurfaceWithFormatConversion(LatteTexture* sourceTexture, sint32 srcMip, sint32 srcSlice, LatteTexture* destinationTexture, sint32 dstMip, sint32 dstSlice, sint32 width, sint32 height)
{ {
return;
// scale copy size to effective size // scale copy size to effective size
sint32 effectiveCopyWidth = width; sint32 effectiveCopyWidth = width;
sint32 effectiveCopyHeight = height; sint32 effectiveCopyHeight = height;
@ -688,33 +692,25 @@ void MetalRenderer::surfaceCopy_copySurfaceWithFormatConversion(LatteTexture* so
sint32 texDstMip = dstMip; sint32 texDstMip = dstMip;
sint32 texDstSlice = dstSlice; sint32 texDstSlice = dstSlice;
LatteTextureMtl* srcTextureMtl = static_cast<LatteTextureMtl*>(sourceTexture); // Create texture views
LatteTextureMtl* dstTextureMtl = static_cast<LatteTextureMtl*>(destinationTexture); LatteTextureViewMtl* srcTextureMtl = static_cast<LatteTextureViewMtl*>(sourceTexture->GetOrCreateView(srcMip, 1, srcSlice, 1));
LatteTextureViewMtl* dstTextureMtl = static_cast<LatteTextureViewMtl*>(destinationTexture->GetOrCreateView(dstMip, 1, dstSlice, 1));
// check if texture rescale ratios match // check if texture rescale ratios match
// todo - if not, we have to use drawcall based copying // todo - if not, we have to use drawcall based copying
if (!LatteTexture_doesEffectiveRescaleRatioMatch(srcTextureMtl, texSrcMip, dstTextureMtl, texDstMip)) if (!LatteTexture_doesEffectiveRescaleRatioMatch(sourceTexture, texSrcMip, destinationTexture, texDstMip))
{ {
cemuLog_logDebug(LogType::Force, "surfaceCopy_copySurfaceWithFormatConversion(): Mismatching dimensions"); cemuLog_logDebug(LogType::Force, "surfaceCopy_copySurfaceWithFormatConversion(): Mismatching dimensions");
return; return;
} }
// check if bpp size matches // check if bpp size matches
if (srcTextureMtl->GetBPP() != dstTextureMtl->GetBPP()) if (sourceTexture->GetBPP() != destinationTexture->GetBPP())
{ {
cemuLog_logDebug(LogType::Force, "surfaceCopy_copySurfaceWithFormatConversion(): Mismatching BPP"); cemuLog_logDebug(LogType::Force, "surfaceCopy_copySurfaceWithFormatConversion(): Mismatching BPP");
return; return;
} }
struct CopyParams
{
uint32 width;
uint32 srcMip;
uint32 srcSlice;
uint32 dstMip;
uint32 dstSlice;
} params{(uint32)effectiveCopyWidth, (uint32)texSrcMip, (uint32)texSrcSlice, (uint32)texDstMip, (uint32)texDstSlice};
if (m_encoderType == MetalEncoderType::Render) if (m_encoderType == MetalEncoderType::Render)
{ {
auto renderCommandEncoder = static_cast<MTL::RenderCommandEncoder*>(m_commandEncoder); auto renderCommandEncoder = static_cast<MTL::RenderCommandEncoder*>(m_commandEncoder);
@ -722,17 +718,42 @@ void MetalRenderer::surfaceCopy_copySurfaceWithFormatConversion(LatteTexture* so
renderCommandEncoder->setRenderPipelineState(m_copyTextureToTexturePipeline->GetRenderPipelineState()); renderCommandEncoder->setRenderPipelineState(m_copyTextureToTexturePipeline->GetRenderPipelineState());
m_state.m_encoderState.m_renderPipelineState = m_copyTextureToTexturePipeline->GetRenderPipelineState(); m_state.m_encoderState.m_renderPipelineState = m_copyTextureToTexturePipeline->GetRenderPipelineState();
SetTexture(renderCommandEncoder, METAL_SHADER_TYPE_VERTEX, srcTextureMtl->GetTexture(), GET_HELPER_TEXTURE_BINDING(0)); SetTexture(renderCommandEncoder, METAL_SHADER_TYPE_VERTEX, srcTextureMtl->GetRGBAView(), GET_HELPER_TEXTURE_BINDING(0));
SetTexture(renderCommandEncoder, METAL_SHADER_TYPE_VERTEX, dstTextureMtl->GetTexture(), GET_HELPER_TEXTURE_BINDING(1)); SetTexture(renderCommandEncoder, METAL_SHADER_TYPE_VERTEX, dstTextureMtl->GetRGBAView(), GET_HELPER_TEXTURE_BINDING(1));
renderCommandEncoder->setVertexBytes(&params, sizeof(params), GET_HELPER_BUFFER_BINDING(0)); renderCommandEncoder->setVertexBytes(&effectiveCopyWidth, sizeof(effectiveCopyWidth), GET_HELPER_BUFFER_BINDING(0));
m_state.m_encoderState.m_buffers[METAL_SHADER_TYPE_VERTEX][GET_HELPER_BUFFER_BINDING(0)] = {nullptr}; m_state.m_encoderState.m_buffers[METAL_SHADER_TYPE_VERTEX][GET_HELPER_BUFFER_BINDING(0)] = {nullptr};
renderCommandEncoder->drawPrimitives(MTL::PrimitiveTypeTriangle, NS::UInteger(0), NS::UInteger(3)); renderCommandEncoder->drawPrimitives(MTL::PrimitiveTypeTriangle, NS::UInteger(0), NS::UInteger(3));
} }
else else
{ {
// TODO: do the copy in a compute shader // TODO: uncomment
debug_printf("surfaceCopy_copySurfaceWithFormatConversion: no active render command encoder, skipping copy\n"); /*
bool copyingToWholeRegion = ((effectiveCopyWidth == dstTextureMtl->GetMipWidth(dstMip) && effectiveCopyHeight == dstTextureMtl->GetMipHeight(dstMip)));
auto renderPassDescriptor = MTL::RenderPassDescriptor::alloc()->init();
auto colorAttachment = renderPassDescriptor->colorAttachments()->object(0);
colorAttachment->setTexture(dstTextureMtl->GetTexture());
// We don't care about previous contents if we are about to overwrite the whole region
colorAttachment->setLoadAction(copyingToWholeRegion ? MTL::LoadActionDontCare : MTL::LoadActionLoad);
colorAttachment->setStoreAction(MTL::StoreActionStore);
colorAttachment->setSlice(dstSlice);
colorAttachment->setLevel(dstMip);
auto renderCommandEncoder = GetTemporaryRenderCommandEncoder(renderPassDescriptor);
auto pipeline = (srcTextureMtl->IsDepth() ? m_copyTextureToColorPipeline : m_copyTextureToDepthPipeline);
renderCommandEncoder->setRenderPipelineState(pipeline);
renderCommandEncoder->setFragmentTexture(srcTextureMtl->GetTexture(), GET_HELPER_TEXTURE_BINDING(0));
renderCommandEncoder->setFragmentBytes(&effectiveCopyWidth, offsetof(effectiveCopyWidth), GET_HELPER_BUFFER_BINDING(0));
renderCommandEncoder->drawPrimitives(MTL::PrimitiveTypeTriangle, NS::UInteger(0), NS::UInteger(3));
EndEncoding();
*/
debug_printf("surface copy with no render command encoder, skipping copy\n");
} }
} }

14
src/Cafe/HW/Latte/Renderer/Metal/UtilityShaderSource.h
View File

@ -31,17 +31,9 @@ fragment float4 fragmentPresent(VertexOut in [[stage_in]], texture2d<float> tex
return tex.sample(samplr, in.texCoord); return tex.sample(samplr, in.texCoord);
} }
struct CopyParams { vertex void vertexCopyTextureToTexture(uint vid [[vertex_id]], texture2d<float, access::read> src [[texture(GET_TEXTURE_BINDING(0))]], texture2d<float, access::write> dst [[texture(GET_TEXTURE_BINDING(1))]], constant uint32_t& width [[buffer(GET_BUFFER_BINDING(0))]]) {
uint width; uint2 coord = uint2(vid % width, vid / width);
uint srcMip; return dst.write(float4(src.read(coord).r, 0.0, 0.0, 0.0), coord);
uint srcSlice;
uint dstMip;
uint dstSlice;
};
vertex void vertexCopyTextureToTexture(uint vid [[vertex_id]], texture2d_array<float, access::read> src [[texture(GET_TEXTURE_BINDING(0))]], texture2d_array<float, access::write> dst [[texture(GET_TEXTURE_BINDING(1))]], constant CopyParams& params [[buffer(GET_BUFFER_BINDING(0))]]) {
uint2 coord = uint2(vid % params.width, vid / params.width);
return dst.write(float4(src.read(coord, params.srcSlice, params.srcMip).r, 0.0, 0.0, 0.0), coord, params.dstSlice, params.dstMip);
} }
struct RestrideParams { struct RestrideParams {