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

do texture load on GPU if no unified memory

Browse Source
This commit is contained in:
Samuliak 2024-08-18 14:49:25 +02:00
parent 485a652c85
commit be0a69a5e2
2 changed files with 22 additions and 3 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

3
src/Cafe/HW/Latte/Renderer/Metal/LatteTextureMtl.cpp
View File

@ -3,13 +3,14 @@
#include "Cafe/HW/Latte/Renderer/Metal/MetalRenderer.h" #include "Cafe/HW/Latte/Renderer/Metal/MetalRenderer.h"
#include "Cafe/HW/Latte/Renderer/Metal/LatteToMtl.h" #include "Cafe/HW/Latte/Renderer/Metal/LatteToMtl.h"
#include "Common/precompiled.h" #include "Common/precompiled.h"
#include "Metal/MTLResource.hpp"
LatteTextureMtl::LatteTextureMtl(class MetalRenderer* mtlRenderer, Latte::E_DIM dim, MPTR physAddress, MPTR physMipAddress, Latte::E_GX2SURFFMT format, uint32 width, uint32 height, uint32 depth, uint32 pitch, uint32 mipLevels, uint32 swizzle, LatteTextureMtl::LatteTextureMtl(class MetalRenderer* mtlRenderer, Latte::E_DIM dim, MPTR physAddress, MPTR physMipAddress, Latte::E_GX2SURFFMT format, uint32 width, uint32 height, uint32 depth, uint32 pitch, uint32 mipLevels, uint32 swizzle,
Latte::E_HWTILEMODE tileMode, bool isDepth) Latte::E_HWTILEMODE tileMode, bool isDepth)
: LatteTexture(dim, physAddress, physMipAddress, format, width, height, depth, pitch, mipLevels, swizzle, tileMode, isDepth), m_mtlr(mtlRenderer), m_format(format), m_isDepth(isDepth) : LatteTexture(dim, physAddress, physMipAddress, format, width, height, depth, pitch, mipLevels, swizzle, tileMode, isDepth), m_mtlr(mtlRenderer), m_format(format), m_isDepth(isDepth)
{ {
MTL::TextureDescriptor* desc = MTL::TextureDescriptor::alloc()->init(); MTL::TextureDescriptor* desc = MTL::TextureDescriptor::alloc()->init();
desc->setStorageMode(m_mtlr->GetOptimalStorageMode()); desc->setStorageMode(m_mtlr->HasUnifiedMemory() ? MTL::StorageModeShared : MTL::StorageModePrivate);
sint32 effectiveBaseWidth = width; sint32 effectiveBaseWidth = width;
sint32 effectiveBaseHeight = height; sint32 effectiveBaseHeight = height;

20
src/Cafe/HW/Latte/Renderer/Metal/MetalRenderer.cpp
View File

@ -378,6 +378,7 @@ void MetalRenderer::texture_clearSlice(LatteTexture* hostTexture, sint32 sliceIn
} }
} }
// TODO: do a GPU blit even on unified memory? That would mean we could use private storage mode for all textures
void MetalRenderer::texture_loadSlice(LatteTexture* hostTexture, sint32 width, sint32 height, sint32 depth, void* pixelData, sint32 sliceIndex, sint32 mipIndex, uint32 compressedImageSize) void MetalRenderer::texture_loadSlice(LatteTexture* hostTexture, sint32 width, sint32 height, sint32 depth, void* pixelData, sint32 sliceIndex, sint32 mipIndex, uint32 compressedImageSize)
{ {
auto textureMtl = (LatteTextureMtl*)hostTexture; auto textureMtl = (LatteTextureMtl*)hostTexture;
@ -390,10 +391,27 @@ void MetalRenderer::texture_loadSlice(LatteTexture* hostTexture, sint32 width, s
} }
size_t bytesPerRow = GetMtlTextureBytesPerRow(textureMtl->GetFormat(), textureMtl->IsDepth(), width); size_t bytesPerRow = GetMtlTextureBytesPerRow(textureMtl->GetFormat(), textureMtl->IsDepth(), width);
// No need to calculate bytesPerImage for 3D textures, since we always load just one slice // No need to set bytesPerImage for 3D textures, since we always load just one slice
//size_t bytesPerImage = GetMtlTextureBytesPerImage(textureMtl->GetFormat(), textureMtl->IsDepth(), height, bytesPerRow); //size_t bytesPerImage = GetMtlTextureBytesPerImage(textureMtl->GetFormat(), textureMtl->IsDepth(), height, bytesPerRow);
if (HasUnifiedMemory())
{
textureMtl->GetTexture()->replaceRegion(MTL::Region(0, 0, offsetZ, width, height, 1), mipIndex, sliceIndex, pixelData, bytesPerRow, 0); textureMtl->GetTexture()->replaceRegion(MTL::Region(0, 0, offsetZ, width, height, 1), mipIndex, sliceIndex, pixelData, bytesPerRow, 0);
} }
else
{
auto blitCommandEncoder = GetBlitCommandEncoder();
// Allocate a temporary buffer
auto allocation = m_memoryManager->GetTemporaryBufferAllocator().GetBufferAllocation(compressedImageSize);
auto buffer = m_memoryManager->GetTemporaryBufferAllocator().GetBuffer(allocation.bufferIndex);
// Copy the data to the temporary buffer
memcpy(allocation.data, pixelData, compressedImageSize);
// Copy the data from the temporary buffer to the texture
blitCommandEncoder->copyFromBuffer(buffer, allocation.offset, bytesPerRow, 0, MTL::Size(width, height, 1), textureMtl->GetTexture(), sliceIndex, mipIndex, MTL::Origin(0, 0, offsetZ));
}
}
void MetalRenderer::texture_clearColorSlice(LatteTexture* hostTexture, sint32 sliceIndex, sint32 mipIndex, float r, float g, float b, float a) void MetalRenderer::texture_clearColorSlice(LatteTexture* hostTexture, sint32 sliceIndex, sint32 mipIndex, float r, float g, float b, float a)
{ {