Merge pull request #1966 from magumagu/unify-efb-encode

Unify EFB encoding shader generation

Source/Core/VideoBackends/D3D/PSTextureEncoder.h

@@ -24,9 +24,7 @@ namespace DX11
 
 class PSTextureEncoder : public TextureEncoder
 {
-
 public:
-
 	PSTextureEncoder();
 
 	void Init();
@@ -36,25 +34,14 @@ public:
 	              bool isIntensity, bool scaleByHalf);
 
 private:
-
 	bool m_ready;
 
 	ID3D11Texture2D* m_out;
 	ID3D11RenderTargetView* m_outRTV;
 	ID3D11Texture2D* m_outStage;
 	ID3D11Buffer* m_encodeParams;
-	ID3D11Buffer* m_quad;
-	ID3D11VertexShader* m_vShader;
-	ID3D11InputLayout* m_quadLayout;
-	ID3D11BlendState* m_efbEncodeBlendState;
-	ID3D11DepthStencilState* m_efbEncodeDepthState;
-	ID3D11RasterizerState* m_efbEncodeRastState;
-	ID3D11SamplerState* m_efbSampler;
 
-	// Stuff only used in static-linking mode (SM4.0-compatible)
+	ID3D11PixelShader* SetStaticShader(unsigned int dstFormat,
-
-	bool InitStaticMode();
-	bool SetStaticShader(unsigned int dstFormat,
 		PEControl::PixelFormat srcFormat, bool isIntensity, bool scaleByHalf);
 
 	typedef unsigned int ComboKey; // Key for a shader combination
@@ -69,35 +56,6 @@ private:
 	typedef std::map<ComboKey, ID3D11PixelShader*> ComboMap;
 
 	ComboMap m_staticShaders;
-
-	// Stuff only used for dynamic-linking mode (SM5.0+, available as soon as
-	// Microsoft fixes their bloody HLSL compiler)
-
-	bool InitDynamicMode();
-	bool SetDynamicShader(unsigned int dstFormat,
-		PEControl::PixelFormat srcFormat, bool isIntensity, bool scaleByHalf);
-
-	ID3D11PixelShader* m_dynamicShader;
-	ID3D11ClassLinkage* m_classLinkage;
-
-	// Interface slots
-	UINT m_fetchSlot;
-	UINT m_scaledFetchSlot;
-	UINT m_intensitySlot;
-	UINT m_generatorSlot;
-
-	// Class instances
-	// Fetch: 0 is RGB, 1 is RGBA, 2 is RGB565, 3 is Z
-	ID3D11ClassInstance* m_fetchClass[4];
-	// ScaledFetch: 0 is off, 1 is on
-	ID3D11ClassInstance* m_scaledFetchClass[2];
-	// Intensity: 0 is off, 1 is on
-	ID3D11ClassInstance* m_intensityClass[2];
-	// Generator: one for each dst format, 16 total
-	ID3D11ClassInstance* m_generatorClass[16];
-
-	std::vector<ID3D11ClassInstance*> m_linkageArray;
-
 };
 
 }

Source/Core/VideoBackends/OGL/TextureConverter.cpp

@@ -35,7 +35,7 @@ static GLuint s_texConvFrameBuffer[2] = {0,0};
 static GLuint s_srcTexture = 0; // for decoding from RAM
 static GLuint s_dstTexture = 0; // for encoding to RAM
 
-const int renderBufferWidth = 1024;
+const int renderBufferWidth = EFB_WIDTH * 4;
 const int renderBufferHeight = 1024;
 
 static SHADER s_rgbToYuyvProgram;
@@ -320,14 +320,16 @@ int EncodeToRamFromTexture(u32 address,GLuint source_texture, bool bFromZBuffer,
 		source.left, source.top,
 		expandedWidth, bScaleByHalf ? 2 : 1);
 
-	int cacheBytes = 32;
+	unsigned int numBlocksX = expandedWidth / TexDecoder_GetBlockWidthInTexels(format);
+	unsigned int numBlocksY = expandedHeight / TexDecoder_GetBlockHeightInTexels(format);
+	unsigned int cacheLinesPerRow;
 	if ((format & 0x0f) == 6)
-		cacheBytes = 64;
+		cacheLinesPerRow = numBlocksX * 2;
+	else
+		cacheLinesPerRow = numBlocksX;
 
-	int readStride = (expandedWidth * cacheBytes) /
-		TexDecoder_GetBlockWidthInTexels(format);
 	EncodeToRamUsingShader(source_texture,
-		dest_ptr, expandedWidth / samples, expandedHeight, readStride,
+		dest_ptr, cacheLinesPerRow * 8, numBlocksY, cacheLinesPerRow * 32,
 		bScaleByHalf > 0 && !bFromZBuffer);
 	return size_in_bytes; // TODO: D3D11 is calculating this value differently!
 

Source/Core/VideoCommon/TextureConversionShader.cpp

@@ -70,36 +70,38 @@ static void WriteSwizzler(char*& p, u32 format, API_TYPE ApiType)
 
 		WRITE(p, "  out vec4 ocol0;\n");
 		WRITE(p, "void main()\n");
-	}
-	else // D3D
-	{
-		WRITE(p,"sampler samp0 : register(s0);\n");
-		WRITE(p, "Texture2D Tex0 : register(t0);\n");
-
-		WRITE(p,"void main(\n");
-		WRITE(p,"  out float4 ocol0 : SV_Target)\n");
-	}
-
 		WRITE(p, "{\n"
 			"  int2 sampleUv;\n"
 			"  int2 uv1 = int2(gl_FragCoord.xy);\n"
 			);
+	}
+	else // D3D
+	{
+		WRITE(p, "sampler samp0 : register(s0);\n");
+		WRITE(p, "Texture2D Tex0 : register(t0);\n");
 
-	WRITE(p, "  int y_block_position = uv1.y & %d;\n", ~(blkH - 1));
+		WRITE(p, "void main(\n");
-	WRITE(p, "  int y_offset_in_block = uv1.y & %d;\n", blkH - 1);
+		WRITE(p, "  out float4 ocol0 : SV_Target, in float4 rawpos : SV_Position)\n");
-	WRITE(p, "  int x_virtual_position = (uv1.x << %d) + y_offset_in_block * position.z;\n", IntLog2(samples));
+		WRITE(p, "{\n"
-	WRITE(p, "  int x_block_position = (x_virtual_position >> %d) & %d;\n", IntLog2(blkH), ~(blkW - 1));
+			"  int2 sampleUv;\n"
+			"  int2 uv1 = int2(rawpos.xy);\n"
+			);
+	}
+
+	WRITE(p, "  int x_block_position = (uv1.x >> %d) << %d;\n", IntLog2(blkH * blkW / samples), IntLog2(blkW));
+	WRITE(p, "  int y_block_position = uv1.y << %d;\n", IntLog2(blkH));
 	if (samples == 1)
 	{
-		// 32 bit textures (RGBA8 and Z24) are stored in 2 cache line increments
+		// With samples == 1, we write out pairs of blocks; one A8R8, one G8B8.
-		WRITE(p, "  bool first = 0 == (x_virtual_position & %d);\n", 8 * samples); // first cache line, used in the encoders
+		WRITE(p, "  bool first = (uv1.x & %d) == 0;\n", blkH * blkW / 2);
-		WRITE(p, "  x_virtual_position = x_virtual_position << 1;\n");
+		samples = 2;
 	}
-	WRITE(p, "  int x_offset_in_block = x_virtual_position & %d;\n", blkW - 1);
+	WRITE(p, "  int offset_in_block = uv1.x & %d;\n", (blkH * blkW / samples) - 1);
-	WRITE(p, "  int y_offset = (x_virtual_position >> %d) & %d;\n", IntLog2(blkW), blkH - 1);
+	WRITE(p, "  int y_offset_in_block = offset_in_block >> %d;\n", IntLog2(blkW / samples));
+	WRITE(p, "  int x_offset_in_block = (offset_in_block & %d) << %d;\n", (blkW / samples) - 1, IntLog2(samples));
 
-	WRITE(p, "  sampleUv.x = x_offset_in_block + x_block_position;\n");
+	WRITE(p, "  sampleUv.x = x_block_position + x_offset_in_block;\n");
-	WRITE(p, "  sampleUv.y = y_block_position + y_offset;\n");
+	WRITE(p, "  sampleUv.y = y_block_position + y_offset_in_block;\n");
 
 	WRITE(p, "  float2 uv0 = float2(sampleUv);\n");                // sampleUv is the sample position in (int)gx_coords
 	WRITE(p, "  uv0 += float2(0.5, 0.5);\n");                      // move to center of pixel
@@ -116,9 +118,18 @@ static void WriteSwizzler(char*& p, u32 format, API_TYPE ApiType)
 
 static void WriteSampleColor(char*& p, const char* colorComp, const char* dest, int xoffset, API_TYPE ApiType)
 {
+	if (ApiType == API_OPENGL)
+	{
 		WRITE(p, "  %s = texture(samp0, float3(uv0 + float2(%d, 0) * sample_offset, 0.0)).%s;\n",
 			dest, xoffset, colorComp
 			);
+	}
+	else
+	{
+		WRITE(p, "  %s = Tex0.Sample(samp0, uv0 + float2(%d, 0) * sample_offset).%s;\n",
+			dest, xoffset, colorComp
+			);
+	}
 }
 
 static void WriteColorToIntensity(char*& p, const char* src, const char* dest)