The-Homebrew-Cloud/Cemu
2
0
Fork 0
mirror of https://github.com/cemu-project/Cemu.git synced 2024-12-02 05:54:18 +01:00
Code Issues Packages Projects Releases Wiki Activity

remove more unnecessary rebinds

Browse Source
This commit is contained in:
Samuliak 2024-08-15 10:58:33 +02:00
parent 9a215e064f
commit 9982ac7acb
4 changed files with 103 additions and 31 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

5
src/Cafe/HW/Latte/Renderer/Metal/MetalMemoryManager.cpp
View File

@ -104,9 +104,13 @@ MetalRestridedBufferRange MetalVertexBufferCache::RestrideBufferIfNeeded(MTL::Bu
auto renderCommandEncoder = static_cast<MTL::RenderCommandEncoder*>(m_mtlr->GetCommandEncoder()); auto renderCommandEncoder = static_cast<MTL::RenderCommandEncoder*>(m_mtlr->GetCommandEncoder());
renderCommandEncoder->setRenderPipelineState(m_restrideBufferPipeline->GetRenderPipelineState()); renderCommandEncoder->setRenderPipelineState(m_restrideBufferPipeline->GetRenderPipelineState());
m_mtlr->GetEncoderState().m_renderPipelineState = m_restrideBufferPipeline->GetRenderPipelineState();
MTL::Buffer* buffers[] = {bufferCache, buffer}; MTL::Buffer* buffers[] = {bufferCache, buffer};
size_t offsets[] = {vertexBufferRange.offset, restrideInfo.allocation.bufferOffset}; size_t offsets[] = {vertexBufferRange.offset, restrideInfo.allocation.bufferOffset};
renderCommandEncoder->setVertexBuffers(buffers, offsets, NS::Range(GET_HELPER_BUFFER_BINDING(0), 2)); renderCommandEncoder->setVertexBuffers(buffers, offsets, NS::Range(GET_HELPER_BUFFER_BINDING(0), 2));
m_mtlr->GetEncoderState().m_uniformBufferOffsets[METAL_SHADER_TYPE_VERTEX][GET_HELPER_BUFFER_BINDING(0)] = INVALID_OFFSET;
m_mtlr->GetEncoderState().m_uniformBufferOffsets[METAL_SHADER_TYPE_VERTEX][GET_HELPER_BUFFER_BINDING(1)] = INVALID_OFFSET;
struct struct
{ {
@ -114,6 +118,7 @@ MetalRestridedBufferRange MetalVertexBufferCache::RestrideBufferIfNeeded(MTL::Bu
uint32 newStride; uint32 newStride;
} strideData = {static_cast<uint32>(stride), static_cast<uint32>(newStride)}; } strideData = {static_cast<uint32>(stride), static_cast<uint32>(newStride)};
renderCommandEncoder->setVertexBytes(&strideData, sizeof(strideData), GET_HELPER_BUFFER_BINDING(2)); renderCommandEncoder->setVertexBytes(&strideData, sizeof(strideData), GET_HELPER_BUFFER_BINDING(2));
m_mtlr->GetEncoderState().m_uniformBufferOffsets[METAL_SHADER_TYPE_VERTEX][GET_HELPER_BUFFER_BINDING(2)] = INVALID_OFFSET;
renderCommandEncoder->drawPrimitives(MTL::PrimitiveTypePoint, NS::UInteger(0), vertexBufferRange.size / stride); renderCommandEncoder->drawPrimitives(MTL::PrimitiveTypePoint, NS::UInteger(0), vertexBufferRange.size / stride);

111
src/Cafe/HW/Latte/Renderer/Metal/MetalRenderer.cpp
View File

@ -17,6 +17,7 @@
#include "Cemu/Logging/CemuDebugLogging.h" #include "Cemu/Logging/CemuDebugLogging.h"
#include "Common/precompiled.h" #include "Common/precompiled.h"
#include "HW/Latte/Renderer/Metal/MetalCommon.h" #include "HW/Latte/Renderer/Metal/MetalCommon.h"
#include "Metal/MTLRenderCommandEncoder.hpp"
#include "Metal/MTLRenderPass.hpp" #include "Metal/MTLRenderPass.hpp"
#include "gui/guiWrapper.h" #include "gui/guiWrapper.h"
@ -232,8 +233,8 @@ void MetalRenderer::DrawBackbufferQuad(LatteTextureView* texView, RendererOutput
// Draw to Metal layer // Draw to Metal layer
renderCommandEncoder->setRenderPipelineState(m_state.m_usesSRGB ? m_presentPipelineSRGB : m_presentPipelineLinear); renderCommandEncoder->setRenderPipelineState(m_state.m_usesSRGB ? m_presentPipelineSRGB : m_presentPipelineLinear);
renderCommandEncoder->setFragmentTexture(presentTexture, GET_HELPER_TEXTURE_BINDING(0)); renderCommandEncoder->setFragmentTexture(presentTexture, 0);
renderCommandEncoder->setFragmentSamplerState((useLinearTexFilter ? m_linearSampler : m_nearestSampler), GET_HELPER_SAMPLER_BINDING(0)); renderCommandEncoder->setFragmentSamplerState((useLinearTexFilter ? m_linearSampler : m_nearestSampler), 0);
renderCommandEncoder->drawPrimitives(MTL::PrimitiveTypeTriangle, NS::UInteger(0), NS::UInteger(3)); renderCommandEncoder->drawPrimitives(MTL::PrimitiveTypeTriangle, NS::UInteger(0), NS::UInteger(3));
@ -535,12 +536,13 @@ void MetalRenderer::surfaceCopy_copySurfaceWithFormatConversion(LatteTexture* so
auto renderCommandEncoder = static_cast<MTL::RenderCommandEncoder*>(m_commandEncoder); auto renderCommandEncoder = static_cast<MTL::RenderCommandEncoder*>(m_commandEncoder);
renderCommandEncoder->setRenderPipelineState(m_copyTextureToTexturePipeline->GetRenderPipelineState()); renderCommandEncoder->setRenderPipelineState(m_copyTextureToTexturePipeline->GetRenderPipelineState());
m_state.m_encoderState.m_renderPipelineState = m_copyTextureToTexturePipeline->GetRenderPipelineState();
renderCommandEncoder->setViewport(MTL::Viewport{0.0, 0.0, (double)effectiveCopyWidth, (double)effectiveCopyHeight, 0.0, 1.0}); renderCommandEncoder->setVertexTextures(textures, NS::Range(GET_HELPER_TEXTURE_BINDING(0), 2));
renderCommandEncoder->setScissorRect(MTL::ScissorRect{0, 0, (uint32)effectiveCopyWidth, (uint32)effectiveCopyHeight}); m_state.m_encoderState.m_textures[METAL_SHADER_TYPE_VERTEX][GET_HELPER_TEXTURE_BINDING(0)] = {(LatteTextureViewMtl*)textures[0]};
m_state.m_encoderState.m_textures[METAL_SHADER_TYPE_VERTEX][GET_HELPER_TEXTURE_BINDING(1)] = {(LatteTextureViewMtl*)textures[1]};
renderCommandEncoder->setVertexTextures(textures, NS::Range(GET_HELPER_BUFFER_BINDING(0), 2));
renderCommandEncoder->setVertexBytes(&params, sizeof(params), GET_HELPER_BUFFER_BINDING(0)); renderCommandEncoder->setVertexBytes(&params, sizeof(params), GET_HELPER_BUFFER_BINDING(0));
m_state.m_encoderState.m_uniformBufferOffsets[METAL_SHADER_TYPE_VERTEX][GET_HELPER_BUFFER_BINDING(0)] = INVALID_OFFSET;
renderCommandEncoder->drawPrimitives(MTL::PrimitiveTypeTriangle, NS::UInteger(0), NS::UInteger(3)); renderCommandEncoder->drawPrimitives(MTL::PrimitiveTypeTriangle, NS::UInteger(0), NS::UInteger(3));
} }
@ -687,6 +689,8 @@ void MetalRenderer::draw_execute(uint32 baseVertex, uint32 baseInstance, uint32
// return; // return;
//} //}
auto& encoderState = m_state.m_encoderState;
// Render pass // Render pass
auto renderCommandEncoder = GetRenderCommandEncoder(); auto renderCommandEncoder = GetRenderCommandEncoder();
@ -702,7 +706,11 @@ void MetalRenderer::draw_execute(uint32 baseVertex, uint32 baseInstance, uint32
// Depth stencil state // Depth stencil state
MTL::DepthStencilState* depthStencilState = m_depthStencilCache->GetDepthStencilState(LatteGPUState.contextNew); MTL::DepthStencilState* depthStencilState = m_depthStencilCache->GetDepthStencilState(LatteGPUState.contextNew);
renderCommandEncoder->setDepthStencilState(depthStencilState); if (depthStencilState != encoderState.m_depthStencilState)
{
renderCommandEncoder->setDepthStencilState(depthStencilState);
encoderState.m_depthStencilState = depthStencilState;
}
// Stencil reference // Stencil reference
bool stencilEnable = LatteGPUState.contextNew.DB_DEPTH_CONTROL.get_STENCIL_ENABLE(); bool stencilEnable = LatteGPUState.contextNew.DB_DEPTH_CONTROL.get_STENCIL_ENABLE();
@ -710,11 +718,19 @@ void MetalRenderer::draw_execute(uint32 baseVertex, uint32 baseInstance, uint32
{ {
bool backStencilEnable = LatteGPUState.contextNew.DB_DEPTH_CONTROL.get_BACK_STENCIL_ENABLE(); bool backStencilEnable = LatteGPUState.contextNew.DB_DEPTH_CONTROL.get_BACK_STENCIL_ENABLE();
uint32 stencilRefFront = LatteGPUState.contextNew.DB_STENCILREFMASK.get_STENCILREF_F(); uint32 stencilRefFront = LatteGPUState.contextNew.DB_STENCILREFMASK.get_STENCILREF_F();
uint32 stencilRefBack = LatteGPUState.contextNew.DB_STENCILREFMASK_BF.get_STENCILREF_B(); uint32 stencilRefBack;
if (backStencilEnable) if (backStencilEnable)
renderCommandEncoder->setStencilReferenceValues(stencilRefFront, stencilRefBack); stencilRefBack = LatteGPUState.contextNew.DB_STENCILREFMASK_BF.get_STENCILREF_B();
else else
renderCommandEncoder->setStencilReferenceValue(stencilRefFront); stencilRefBack = stencilRefFront;
if (stencilRefFront != encoderState.m_stencilRefFront || stencilRefBack != encoderState.m_stencilRefBack)
{
renderCommandEncoder->setStencilReferenceValues(stencilRefFront, stencilRefBack);
encoderState.m_stencilRefFront = stencilRefFront;
encoderState.m_stencilRefBack = stencilRefBack;
}
} }
// Primitive type // Primitive type
@ -739,21 +755,35 @@ void MetalRenderer::draw_execute(uint32 baseVertex, uint32 baseInstance, uint32
if (polyOffsetFrontEnable) if (polyOffsetFrontEnable)
{ {
//uint32 frontScaleU32 = LatteGPUState.contextNew.PA_SU_POLY_OFFSET_FRONT_SCALE.getRawValue(); uint32 frontScaleU32 = LatteGPUState.contextNew.PA_SU_POLY_OFFSET_FRONT_SCALE.getRawValue();
//uint32 frontOffsetU32 = LatteGPUState.contextNew.PA_SU_POLY_OFFSET_FRONT_OFFSET.getRawValue(); uint32 frontOffsetU32 = LatteGPUState.contextNew.PA_SU_POLY_OFFSET_FRONT_OFFSET.getRawValue();
//uint32 offsetClampU32 = LatteGPUState.contextNew.PA_SU_POLY_OFFSET_CLAMP.getRawValue(); uint32 offsetClampU32 = LatteGPUState.contextNew.PA_SU_POLY_OFFSET_CLAMP.getRawValue();
float frontScale = LatteGPUState.contextNew.PA_SU_POLY_OFFSET_FRONT_SCALE.get_SCALE(); if (frontOffsetU32 != encoderState.m_depthBias || frontScaleU32 != encoderState.m_depthSlope || offsetClampU32 != encoderState.m_depthClamp)
float frontOffset = LatteGPUState.contextNew.PA_SU_POLY_OFFSET_FRONT_OFFSET.get_OFFSET(); {
float offsetClamp = LatteGPUState.contextNew.PA_SU_POLY_OFFSET_CLAMP.get_CLAMP(); float frontScale = LatteGPUState.contextNew.PA_SU_POLY_OFFSET_FRONT_SCALE.get_SCALE();
float frontOffset = LatteGPUState.contextNew.PA_SU_POLY_OFFSET_FRONT_OFFSET.get_OFFSET();
float offsetClamp = LatteGPUState.contextNew.PA_SU_POLY_OFFSET_CLAMP.get_CLAMP();
frontScale /= 16.0f; frontScale /= 16.0f;
renderCommandEncoder->setDepthBias(frontOffset, frontScale, offsetClamp); renderCommandEncoder->setDepthBias(frontOffset, frontScale, offsetClamp);
encoderState.m_depthBias = frontOffsetU32;
encoderState.m_depthSlope = frontScaleU32;
encoderState.m_depthClamp = offsetClampU32;
}
} }
else else
{ {
renderCommandEncoder->setDepthBias(0.0f, 0.0f, 0.0f); if (0 != encoderState.m_depthBias || 0 != encoderState.m_depthSlope || 0 != encoderState.m_depthClamp)
{
renderCommandEncoder->setDepthBias(0.0f, 0.0f, 0.0f);
encoderState.m_depthBias = 0;
encoderState.m_depthSlope = 0;
encoderState.m_depthClamp = 0;
}
} }
// todo - how does culling behave with rects? // todo - how does culling behave with rects?
@ -766,19 +796,36 @@ void MetalRenderer::draw_execute(uint32 baseVertex, uint32 baseInstance, uint32
cullBack = cullFront; cullBack = cullFront;
} }
// Cull mode
if (cullFront && cullBack) if (cullFront && cullBack)
return; // We can just skip the draw (TODO: can we?) return; // We can just skip the draw (TODO: can we?)
else if (cullFront)
renderCommandEncoder->setCullMode(MTL::CullModeFront);
else if (cullBack)
renderCommandEncoder->setCullMode(MTL::CullModeBack);
else
renderCommandEncoder->setCullMode(MTL::CullModeNone);
MTL::CullMode cullMode;
if (cullFront)
cullMode = MTL::CullModeFront;
else if (cullBack)
cullMode = MTL::CullModeBack;
else
cullMode = MTL::CullModeNone;
if (cullMode != encoderState.m_cullMode)
{
renderCommandEncoder->setCullMode(cullMode);
encoderState.m_cullMode = cullMode;
}
// Front face
MTL::Winding frontFaceWinding;
if (frontFace == Latte::LATTE_PA_SU_SC_MODE_CNTL::E_FRONTFACE::CCW) if (frontFace == Latte::LATTE_PA_SU_SC_MODE_CNTL::E_FRONTFACE::CCW)
renderCommandEncoder->setFrontFacingWinding(MTL::WindingCounterClockwise); frontFaceWinding = MTL::WindingCounterClockwise;
else else
renderCommandEncoder->setFrontFacingWinding(MTL::WindingClockwise); frontFaceWinding = MTL::WindingClockwise;
if (frontFaceWinding != encoderState.m_frontFaceWinding)
{
renderCommandEncoder->setFrontFacingWinding(frontFaceWinding);
encoderState.m_frontFaceWinding = frontFaceWinding;
}
// Resources // Resources
@ -817,7 +864,11 @@ void MetalRenderer::draw_execute(uint32 baseVertex, uint32 baseInstance, uint32
// Render pipeline state // Render pipeline state
MTL::RenderPipelineState* renderPipelineState = m_pipelineCache->GetPipelineState(fetchShader, vertexShader, pixelShader, m_state.m_lastUsedFBO, LatteGPUState.contextNew); MTL::RenderPipelineState* renderPipelineState = m_pipelineCache->GetPipelineState(fetchShader, vertexShader, pixelShader, m_state.m_lastUsedFBO, LatteGPUState.contextNew);
renderCommandEncoder->setRenderPipelineState(renderPipelineState); if (renderPipelineState != encoderState.m_renderPipelineState)
{
renderCommandEncoder->setRenderPipelineState(renderPipelineState);
encoderState.m_renderPipelineState = renderPipelineState;
}
// Uniform buffers, textures and samplers // Uniform buffers, textures and samplers
BindStageResources(renderCommandEncoder, vertexShader); BindStageResources(renderCommandEncoder, vertexShader);

16
src/Cafe/HW/Latte/Renderer/Metal/MetalRenderer.h
View File

@ -7,6 +7,8 @@
#include "Cafe/HW/Latte/Renderer/Renderer.h" #include "Cafe/HW/Latte/Renderer/Renderer.h"
#include "Cafe/HW/Latte/Renderer/Metal/MetalMemoryManager.h" #include "Cafe/HW/Latte/Renderer/Metal/MetalMemoryManager.h"
#include "Metal/MTLDepthStencil.hpp"
#include "Metal/MTLRenderCommandEncoder.hpp"
struct MetalBoundBuffer struct MetalBoundBuffer
{ {
@ -40,6 +42,15 @@ inline MetalShaderType GetMtlShaderType(LatteConst::ShaderType shaderType)
struct MetalEncoderState struct MetalEncoderState
{ {
MTL::RenderPipelineState* m_renderPipelineState = nullptr;
MTL::DepthStencilState* m_depthStencilState = nullptr;
MTL::CullMode m_cullMode = MTL::CullModeNone;
MTL::Winding m_frontFaceWinding = MTL::WindingClockwise;
uint32 m_stencilRefFront = 0;
uint32 m_stencilRefBack = 0;
uint32 m_depthBias = 0;
uint32 m_depthSlope = 0;
uint32 m_depthClamp = 0;
struct { struct {
class LatteTextureViewMtl* m_textureView = nullptr; class LatteTextureViewMtl* m_textureView = nullptr;
uint32 m_word4 = INVALID_UINT32; uint32 m_word4 = INVALID_UINT32;
@ -288,6 +299,11 @@ public:
} }
} }
MetalEncoderState& GetEncoderState()
{
return m_state.m_encoderState;
}
MTL::CommandBuffer* GetCommandBuffer(); MTL::CommandBuffer* GetCommandBuffer();
bool CommandBufferCompleted(MTL::CommandBuffer* commandBuffer); bool CommandBufferCompleted(MTL::CommandBuffer* commandBuffer);
void WaitForCommandBufferCompletion(MTL::CommandBuffer* commandBuffer); void WaitForCommandBufferCompletion(MTL::CommandBuffer* commandBuffer);

2
src/Cafe/HW/Latte/Renderer/Metal/UtilityShaderSource.h
View File

@ -20,7 +20,7 @@ vertex VertexOut vertexFullscreen(ushort vid [[vertex_id]]) {
return out; return out;
} }
fragment float4 fragmentPresent(VertexOut in [[stage_in]], texture2d<float> tex [[texture(GET_TEXTURE_BINDING(0))]], sampler samplr [[sampler(GET_SAMPLER_BINDING(0))]]) { fragment float4 fragmentPresent(VertexOut in [[stage_in]], texture2d<float> tex [[texture(0)]], sampler samplr [[sampler(0)]]) {
return tex.sample(samplr, in.texCoord); return tex.sample(samplr, in.texCoord);
} }