Merge pull request #12 from SamoZ256/metal-state5

Accurate special state 5

src/Cafe/HW/Latte/Renderer/Metal/LatteTextureViewMtl.cpp

@@ -2,7 +2,6 @@
 #include "Cafe/HW/Latte/Renderer/Metal/LatteTextureMtl.h"
 #include "Cafe/HW/Latte/Renderer/Metal/MetalRenderer.h"
 #include "Cafe/HW/Latte/Renderer/Metal/LatteToMtl.h"
-#include "Metal/MTLTexture.hpp"
 
 uint32 LatteTextureMtl_AdjustTextureCompSel(Latte::E_GX2SURFFMT format, uint32 compSel)
 {
@@ -159,17 +158,17 @@ MTL::Texture* LatteTextureViewMtl::CreateSwizzledView(uint32 gpuSamplerSwizzle)
 
     uint32 baseLevel = firstMip;
     uint32 levelCount = this->numMip;
-    uint32 baseLayer = 0;
+    uint32 baseLayer;
-    uint32 layerCount = 1;
+    uint32 layerCount;
-
+    // TODO: check if base texture is 3D texture as well
-    // TODO: check if base texture is 3D texture as well?
     if (textureType == MTL::TextureType3D)
     {
         cemu_assert_debug(firstMip == 0);
         cemu_assert_debug(this->numSlice == baseTexture->depth);
+        baseLayer = 0;
+        layerCount = 1;
     }
-    // Cube array needs to have layer count multiple of 6 as opposed to when creating a texture
+    else
-    else if (textureType == MTL::TextureTypeCubeArray || textureType == MTL::TextureType2DArray)
     {
         baseLayer = firstSlice;
         layerCount = this->numSlice;

src/Cafe/HW/Latte/Renderer/Metal/MetalRenderer.cpp

@@ -125,9 +125,18 @@ MetalRenderer::MetalRenderer()
 	if (error)
     {
         cemuLog_log(LogType::Force, "failed to create utility library (error: {})", error->localizedDescription()->utf8String());
-        return;
     }
 
+    // Pipelines
+    MTL::Function* vertexFullscreenFunction = utilityLibrary->newFunction(ToNSString("vertexFullscreen"));
+    MTL::Function* fragmentCopyDepthToColorFunction = utilityLibrary->newFunction(ToNSString("fragmentCopyDepthToColor"));
+
+    m_copyDepthToColorDesc = MTL::RenderPipelineDescriptor::alloc()->init();
+    m_copyDepthToColorDesc->setVertexFunction(vertexFullscreenFunction);
+    m_copyDepthToColorDesc->setFragmentFunction(fragmentCopyDepthToColorFunction);
+    vertexFullscreenFunction->release();
+    fragmentCopyDepthToColorFunction->release();
+
     // Void vertex pipelines
     if (m_isAppleGPU)
         m_copyBufferToBufferPipeline = new MetalVoidVertexPipeline(this, utilityLibrary, "vertexCopyBufferToBuffer");
@@ -142,8 +151,9 @@ MetalRenderer::~MetalRenderer()
     //delete m_copyTextureToTexturePipeline;
     //delete m_restrideBufferPipeline;
 
-    //m_presentPipelineLinear->release();
+    m_copyDepthToColorDesc->release();
-    //m_presentPipelineSRGB->release();
+    for (const auto [pixelFormat, pipeline] : m_copyDepthToColorPipelines)
+        pipeline->release();
 
     delete m_outputShaderCache;
     delete m_pipelineCache;
@@ -1348,14 +1358,44 @@ void MetalRenderer::draw_handleSpecialState5()
 
 	LatteTextureView* colorBuffer = LatteMRT::GetColorAttachment(0);
 	LatteTextureView* depthBuffer = LatteMRT::GetDepthAttachment();
+	auto colorTextureMtl = static_cast<LatteTextureViewMtl*>(colorBuffer);
+	auto depthTextureMtl = static_cast<LatteTextureViewMtl*>(depthBuffer);
 
 	sint32 vpWidth, vpHeight;
 	LatteMRT::GetVirtualViewportDimensions(vpWidth, vpHeight);
 
-	surfaceCopy_copySurfaceWithFormatConversion(
+	// Get the pipeline
-		depthBuffer->baseTexture, depthBuffer->firstMip, depthBuffer->firstSlice,
+	MTL::PixelFormat colorPixelFormat = colorTextureMtl->GetRGBAView()->pixelFormat();
-		colorBuffer->baseTexture, colorBuffer->firstMip, colorBuffer->firstSlice,
+	auto& pipeline = m_copyDepthToColorPipelines[colorPixelFormat];
-		vpWidth, vpHeight);
+	if (!pipeline)
+	{
+	    m_copyDepthToColorDesc->colorAttachments()->object(0)->setPixelFormat(colorPixelFormat);
+
+        NS::Error* error = nullptr;
+        pipeline = m_device->newRenderPipelineState(m_copyDepthToColorDesc, &error);
+        if (error)
+        {
+            cemuLog_log(LogType::Force, "failed to create copy depth to color pipeline (error: {})", error->localizedDescription()->utf8String());
+        }
+	}
+
+	// Sadly, we need to end encoding to ensure that the depth data is up-to-date
+	EndEncoding();
+
+	// Copy depth to color
+	auto renderCommandEncoder = GetRenderCommandEncoder();
+
+	auto& encoderState = m_state.m_encoderState;
+
+	renderCommandEncoder->setRenderPipelineState(pipeline);
+	// TODO: make a helper function for this
+	encoderState.m_renderPipelineState = pipeline;
+	SetTexture(renderCommandEncoder, METAL_SHADER_TYPE_FRAGMENT, depthTextureMtl->GetRGBAView(), GET_HELPER_TEXTURE_BINDING(0));
+	// TODO: make a helper function for this
+	renderCommandEncoder->setFragmentBytes(&vpWidth, sizeof(sint32), GET_HELPER_BUFFER_BINDING(0));
+	encoderState.m_buffers[METAL_SHADER_TYPE_FRAGMENT][GET_HELPER_BUFFER_BINDING(0)] = {nullptr};
+
+	renderCommandEncoder->drawPrimitives(MTL::PrimitiveTypeTriangle,  NS::UInteger(0),  NS::UInteger(3));
 }
 
 void* MetalRenderer::indexData_reserveIndexMemory(uint32 size, uint32& offset, uint32& bufferIndex)

src/Cafe/HW/Latte/Renderer/Metal/MetalRenderer.h

@@ -471,6 +471,10 @@ private:
 	class MetalDepthStencilCache* m_depthStencilCache;
 	class MetalSamplerCache* m_samplerCache;
 
+	// Pipelines
+	MTL::RenderPipelineDescriptor* m_copyDepthToColorDesc;
+	std::map<MTL::PixelFormat, MTL::RenderPipelineState*> m_copyDepthToColorPipelines;
+
 	// Void vertex pipelines
 	class MetalVoidVertexPipeline* m_copyBufferToBufferPipeline;
 

src/Cafe/HW/Latte/Renderer/Metal/UtilityShaderSource.h

@@ -8,24 +8,24 @@ using namespace metal;
 
 #define GET_BUFFER_BINDING(index) (28 + index)
 #define GET_TEXTURE_BINDING(index) (29 + index)
-#define GET_SAMPLER_BINDING(index) (14 + index)\n
+#define GET_SAMPLER_BINDING(index) (14 + index)
+
+constant float2 positions[] = {float2(-1.0, -3.0), float2(-1.0, 1.0), float2(3.0, 1.0)};
+
+struct VertexOut {
+    float4 position [[position]];
+    float2 texCoord;
+};
+
+vertex VertexOut vertexFullscreen(ushort vid [[vertex_id]]) {
+    VertexOut out;
+    out.position = float4(positions[vid], 0.0, 1.0);
+    out.texCoord = positions[vid] * 0.5 + 0.5;
+    out.texCoord.y = 1.0 - out.texCoord.y;
+
+    return out;
+}
 
-//constant float2 positions[] = {float2(-1.0, -3.0), float2(-1.0, 1.0), float2(3.0, 1.0)};
-//
-//struct VertexOut {
-//    float4 position [[position]];
-//    float2 texCoord;
-//};
-//
-//vertex VertexOut vertexFullscreen(ushort vid [[vertex_id]]) {
-//    VertexOut out;
-//    out.position = float4(positions[vid], 0.0, 1.0);
-//    out.texCoord = positions[vid] * 0.5 + 0.5;
-//    out.texCoord.y = 1.0 - out.texCoord.y;
-//
-//    return out;
-//}
-//
 //fragment float4 fragmentPresent(VertexOut in [[stage_in]], texture2d<float> tex [[texture(0)]], //sampler samplr [[sampler(0)]]) {
 //    return tex.sample(samplr, in.texCoord);
 //}
@@ -34,19 +34,18 @@ vertex void vertexCopyBufferToBuffer(uint vid [[vertex_id]], device uint8_t* src
     dst[vid] = src[vid];
 }
 
-//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))]]) {
+fragment float4 fragmentCopyDepthToColor(VertexOut in [[stage_in]], texture2d<float, access::read> src [[texture(GET_TEXTURE_BINDING(0))]]) {
-//    uint2 coord = uint2(vid % width, vid / width);
+    return float4(src.read(uint2(in.position.xy)).r, 0.0, 0.0, 0.0);
-//    return dst.write(float4(src.read(coord).r, 0.0, 0.0, 0.0), coord);
+}
+
+//struct RestrideParams {
+//    uint oldStride;
+//    uint newStride;
+//};
+
+//vertex void vertexRestrideBuffer(uint vid [[vertex_id]], device uint8_t* src [[buffer//(GET_BUFFER_BINDING(0))]], device uint8_t* dst [[buffer(GET_BUFFER_BINDING(1))]], constant //RestrideParams& params [[buffer(GET_BUFFER_BINDING(2))]]) {
+//    for (uint32_t i = 0; i < params.oldStride; i++) {
+//        dst[vid * params.newStride + i] = src[vid * params.oldStride + i];
+//    }
 //}
-
-struct RestrideParams {
-    uint oldStride;
-    uint newStride;
-};
-
-vertex void vertexRestrideBuffer(uint vid [[vertex_id]], device uint8_t* src [[buffer(GET_BUFFER_BINDING(0))]], device uint8_t* dst [[buffer(GET_BUFFER_BINDING(1))]], constant RestrideParams& params [[buffer(GET_BUFFER_BINDING(2))]]) {
-    for (uint32_t i = 0; i < params.oldStride; i++) {
-        dst[vid * params.newStride + i] = src[vid * params.oldStride + i];
-    }
-}
 )";