do buffer copying in a void vertex function

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

@@ -24,6 +24,7 @@
 #include "HW/Latte/Renderer/Renderer.h"
 #include "Metal/MTLCommandBuffer.hpp"
 #include "Metal/MTLDevice.hpp"
+#include "Metal/MTLRenderCommandEncoder.hpp"
 #include "Metal/MTLRenderPass.hpp"
 #include "Metal/MTLRenderPipeline.hpp"
 #include "imgui.h"
@@ -95,7 +96,8 @@ MetalRenderer::MetalRenderer()
 #endif
 
     // Transform feedback
-    m_xfbRingBuffer = m_device->newBuffer(LatteStreamout_GetRingBufferSize() * 32, MTL::ResourceStorageModePrivate);
+    // HACK: using just LatteStreamout_GetRingBufferSize will cause page faults
+    m_xfbRingBuffer = m_device->newBuffer(LatteStreamout_GetRingBufferSize() * 4, MTL::ResourceStorageModePrivate);
 #ifdef CEMU_DEBUG_ASSERT
     m_xfbRingBuffer->setLabel(GetLabel("Transform feedback buffer", m_xfbRingBuffer));
 #endif
@@ -159,6 +161,7 @@ MetalRenderer::MetalRenderer()
     auto copyTextureToColorPipelineDescriptor = MTL::RenderPipelineDescriptor::alloc()->init();
 
     // Hybrid pipelines
+    m_copyBufferToBufferPipeline = new MetalHybridComputePipeline(this, utilityLibrary, "vertexCopyBufferToBuffer");
     //m_copyTextureToTexturePipeline = new MetalHybridComputePipeline(this, utilityLibrary, "vertexCopyTextureToTexture");
     m_restrideBufferPipeline = new MetalHybridComputePipeline(this, utilityLibrary, "vertexRestrideBuffer");
     utilityLibrary->release();
@@ -168,6 +171,7 @@ MetalRenderer::MetalRenderer()
 
 MetalRenderer::~MetalRenderer()
 {
+    delete m_copyBufferToBufferPipeline;
     //delete m_copyTextureToTexturePipeline;
     delete m_restrideBufferPipeline;
 
@@ -773,11 +777,32 @@ void MetalRenderer::bufferCache_copy(uint32 srcOffset, uint32 dstOffset, uint32
 }
 
 void MetalRenderer::bufferCache_copyStreamoutToMainBuffer(uint32 srcOffset, uint32 dstOffset, uint32 size)
+{
+    if (m_encoderType == MetalEncoderType::Render)
+    {
+        auto renderCommandEncoder = static_cast<MTL::RenderCommandEncoder*>(m_commandEncoder);
+
+		MTL::Resource* barrierBuffers[] = {m_xfbRingBuffer};
+        renderCommandEncoder->memoryBarrier(barrierBuffers, 1, MTL::RenderStageVertex | MTL::RenderStageFragment | MTL::RenderStageObject | MTL::RenderStageMesh, MTL::RenderStageVertex);
+
+		renderCommandEncoder->setRenderPipelineState(m_copyBufferToBufferPipeline->GetRenderPipelineState());
+		m_state.m_encoderState.m_renderPipelineState = m_copyBufferToBufferPipeline->GetRenderPipelineState();
+
+		SetBuffer(renderCommandEncoder, METAL_SHADER_TYPE_VERTEX, m_xfbRingBuffer, srcOffset, GET_HELPER_BUFFER_BINDING(0));
+		SetBuffer(renderCommandEncoder, METAL_SHADER_TYPE_VERTEX, m_memoryManager->GetBufferCache(), dstOffset, GET_HELPER_BUFFER_BINDING(1));
+
+		renderCommandEncoder->drawPrimitives(MTL::PrimitiveTypePoint, NS::UInteger(0), NS::UInteger(size));
+
+		barrierBuffers[0] = m_memoryManager->GetBufferCache();
+        renderCommandEncoder->memoryBarrier(barrierBuffers, 1, MTL::RenderStageVertex, MTL::RenderStageVertex | MTL::RenderStageFragment | MTL::RenderStageObject | MTL::RenderStageMesh);
+    }
+    else
     {
         auto blitCommandEncoder = GetBlitCommandEncoder();
 
         blitCommandEncoder->copyFromBuffer(m_xfbRingBuffer, srcOffset, m_memoryManager->GetBufferCache(), dstOffset, size);
     }
+}
 
 void MetalRenderer::buffer_bindVertexBuffer(uint32 bufferIndex, uint32 offset, uint32 size)
 {

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

@@ -440,6 +440,7 @@ private:
 	MTL::RenderPipelineState* m_presentPipelineSRGB;
 
 	// Hybrid pipelines
+	class MetalHybridComputePipeline* m_copyBufferToBufferPipeline;
 	//class MetalHybridComputePipeline* m_copyTextureToTexturePipeline;
 	class MetalHybridComputePipeline* m_restrideBufferPipeline;
 

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

@@ -31,6 +31,10 @@ fragment float4 fragmentPresent(VertexOut in [[stage_in]], texture2d<float> tex
     return tex.sample(samplr, in.texCoord);
 }
 
+vertex void vertexCopyBufferToBuffer(uint vid [[vertex_id]], device uint8_t* src [[buffer(GET_BUFFER_BINDING(0))]], device uint8_t* dst [[buffer(GET_BUFFER_BINDING(1))]]) {
+    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))]]) {
     uint2 coord = uint2(vid % width, vid / width);