Shader Disk Cache implementation for D3D. Saves generated shaders on disk. Eliminates "freeze jerks" in D3D plugin the _second_ and later times you play something.... not much to do about the first time. The D3D shader compiler is just slow.

Also assorted cleanup around the shader code.

git-svn-id: https://dolphin-emu.googlecode.com/svn/trunk@4869 8ced0084-cf51-0410-be5f-012b33b47a6e

5 files changed, 33 insertions, 30 deletions

diff --git a/Source/Core/VideoCommon/Src/PixelShaderGen.cpp b/Source/Core/VideoCommon/Src/PixelShaderGen.cpp
index faf0c97eff..54a433d71c 100644
--- a/Source/Core/VideoCommon/Src/PixelShaderGen.cpp
+++ b/Source/Core/VideoCommon/Src/PixelShaderGen.cpp
@@ -31,7 +31,7 @@ PIXELSHADERUID last_pixel_shader_uid;
 // a unique identifier, basically containing all the bits. Yup, it's a lot ....
 // It would likely be a lot more efficient to build this incrementally as the attributes
 // are set...
-void GetPixelShaderId(PIXELSHADERUID &uid, u32 texturemask, u32 dstAlphaEnable)
+void GetPixelShaderId(PIXELSHADERUID *uid, u32 texturemask, u32 dstAlphaEnable)
 {
 	u32 projtexcoords = 0;
 	for (u32 i = 0; i < (u32)bpmem.genMode.numtevstages + 1; i++) 
@@ -43,7 +43,7 @@ void GetPixelShaderId(PIXELSHADERUID &uid, u32 texturemask, u32 dstAlphaEnable)
 				projtexcoords |= 1 << texcoord;
 		}
 	}
-	uid.values[0] = (u32)bpmem.genMode.numtevstages |
+	uid->values[0] = (u32)bpmem.genMode.numtevstages |
 				   ((u32)bpmem.genMode.numindstages << 4) |
 				   ((u32)bpmem.genMode.numtexgens << 7) |
 				   ((u32)dstAlphaEnable << 11) |
@@ -51,21 +51,21 @@ void GetPixelShaderId(PIXELSHADERUID &uid, u32 texturemask, u32 dstAlphaEnable)
 				   (projtexcoords << 20) |
 				   ((u32)bpmem.ztex2.op << 28);
 
-	uid.values[0] = (uid.values[0] & ~0x0ff00000) | (projtexcoords << 20);
+	uid->values[0] = (uid->values[0] & ~0x0ff00000) | (projtexcoords << 20);
 	// swap table
 	for (int i = 0; i < 8; i += 2)
-		((u8*)&uid.values[1])[i / 2] = (bpmem.tevksel[i].hex & 0xf) | ((bpmem.tevksel[i + 1].hex & 0xf) << 4);
+		((u8*)&uid->values[1])[i / 2] = (bpmem.tevksel[i].hex & 0xf) | ((bpmem.tevksel[i + 1].hex & 0xf) << 4);
 
-	uid.values[2] = texturemask;
+	uid->values[2] = texturemask;
 
     u32 enableZTexture = (!bpmem.zcontrol.zcomploc && bpmem.zmode.testenable && bpmem.zmode.updateenable)?1:0;
 
-    uid.values[3] = (u32)bpmem.fog.c_proj_fsel.fsel |
+    uid->values[3] = (u32)bpmem.fog.c_proj_fsel.fsel |
                    ((u32)bpmem.fog.c_proj_fsel.proj << 3) |
                    ((u32)enableZTexture << 4);
 
 	int hdr = 4;
-	u32* pcurvalue = &uid.values[hdr];
+	u32 *pcurvalue = &uid->values[hdr];
 	for (u32 i = 0; i < (u32)bpmem.genMode.numtevstages + 1; ++i) 
 	{
 		TevStageCombiner::ColorCombiner &cc = bpmem.combiners[i].colorC;
@@ -119,7 +119,7 @@ void GetPixelShaderId(PIXELSHADERUID &uid, u32 texturemask, u32 dstAlphaEnable)
 	if ((bpmem.genMode.numtevstages % 3) != 2)
 		++pcurvalue;
 
-	uid.tevstages = (u32)(pcurvalue - &uid.values[0] - hdr);
+	uid->tevstages = (u32)(pcurvalue - &uid->values[0] - hdr);
 
 	for (u32 i = 0; i < bpmem.genMode.numindstages; ++i) 
 	{
@@ -134,7 +134,7 @@ void GetPixelShaderId(PIXELSHADERUID &uid, u32 texturemask, u32 dstAlphaEnable)
 	}
 
 	// yeah, well ....
-	uid.indstages = (u32)(pcurvalue - &uid.values[0] - (hdr - 1) - uid.tevstages);
+	uid->indstages = (u32)(pcurvalue - &uid->values[0] - (hdr - 1) - uid->tevstages);
 }
 
 //   old tev->pixelshader notes
@@ -385,7 +385,7 @@ static void BuildSwapModeTable()
     }
 }
 
-const char *GeneratePixelShader(u32 texture_mask, bool dstAlphaEnable, u32 HLSL)
+const char *GeneratePixelShaderCode(u32 texture_mask, bool dstAlphaEnable, u32 HLSL)
 {
 	setlocale(LC_NUMERIC, "C"); // Reset locale for compilation
 	text[sizeof(text) - 1] = 0x7C;  // canary
diff --git a/Source/Core/VideoCommon/Src/PixelShaderGen.h b/Source/Core/VideoCommon/Src/PixelShaderGen.h
index 995ea55b00..37bcdc0d9e 100644
--- a/Source/Core/VideoCommon/Src/PixelShaderGen.h
+++ b/Source/Core/VideoCommon/Src/PixelShaderGen.h
@@ -42,6 +42,7 @@
 #define C_COLORMATRIX  (C_FOG + 2)
 #define PIXELSHADERUID_MAX_VALUES (5 + 32 + 6 + 11)
 
+// DO NOT make anything in this class virtual.
 class PIXELSHADERUID
 {
 public:
@@ -100,8 +101,9 @@ public:
 	}
 };
 
-const char *GeneratePixelShader(u32 texture_mask, bool dstAlphaEnable, u32 HLSL = 0);
-void GetPixelShaderId(PIXELSHADERUID &, u32 texturemask, u32 dstAlphaEnable);
+const char *GeneratePixelShaderCode(u32 texture_mask, bool dstAlphaEnable, u32 HLSL = 0);
+void GetPixelShaderId(PIXELSHADERUID *uid, u32 texturemask, u32 dstAlphaEnable);
+
 extern PIXELSHADERUID last_pixel_shader_uid;
 
 #endif // GCOGL_PIXELSHADER_H
diff --git a/Source/Core/VideoCommon/Src/TextureConversionShader.h b/Source/Core/VideoCommon/Src/TextureConversionShader.h
index 7e81c4cc4f..e49f2a24ef 100644
--- a/Source/Core/VideoCommon/Src/TextureConversionShader.h
+++ b/Source/Core/VideoCommon/Src/TextureConversionShader.h
@@ -25,7 +25,7 @@ namespace TextureConversionShader
 {
 u16 GetEncodedSampleCount(u32 format);
 
-const char *GenerateEncodingShader(u32 format,bool HLSL = false);
+const char *GenerateEncodingShader(u32 format, bool HLSL = false);
 
 void SetShaderParameters(float width, float height, float offsetX, float offsetY, float widthStride, float heightStride,float buffW = 0.0f,float buffH = 0.0f);
 
diff --git a/Source/Core/VideoCommon/Src/VertexShaderGen.cpp b/Source/Core/VideoCommon/Src/VertexShaderGen.cpp
index ad8d227813..934990a9f6 100644
--- a/Source/Core/VideoCommon/Src/VertexShaderGen.cpp
+++ b/Source/Core/VideoCommon/Src/VertexShaderGen.cpp
@@ -29,27 +29,27 @@ VERTEXSHADERUID  last_vertex_shader_uid;
 
 // Mash together all the inputs that contribute to the code of a generated vertex shader into
 // a unique identifier, basically containing all the bits. Yup, it's a lot ....
-void GetVertexShaderId(VERTEXSHADERUID& vid, u32 components)
+void GetVertexShaderId(VERTEXSHADERUID *uid, u32 components)
 {
-	vid.values[0] = components |
+	uid->values[0] = components |
 		(xfregs.numTexGens << 23) |
 		(xfregs.nNumChans << 27) |
 		((u32)xfregs.bEnableDualTexTransform << 29);
 
 	for (int i = 0; i < 2; ++i) {
-		vid.values[1+i] = xfregs.colChans[i].color.enablelighting ?
+		uid->values[1+i] = xfregs.colChans[i].color.enablelighting ?
 			(u32)xfregs.colChans[i].color.hex :
 			(u32)xfregs.colChans[i].color.matsource;
-		vid.values[1+i] |= (xfregs.colChans[i].alpha.enablelighting ?
+		uid->values[1+i] |= (xfregs.colChans[i].alpha.enablelighting ?
 			(u32)xfregs.colChans[i].alpha.hex :
 			(u32)xfregs.colChans[i].alpha.matsource) << 15;
 	}
 
 	// fog
-	vid.values[1] |= (((u32)bpmem.fog.c_proj_fsel.fsel & 3) << 30);
-	vid.values[2] |= (((u32)bpmem.fog.c_proj_fsel.fsel >> 2) << 30);
+	uid->values[1] |= (((u32)bpmem.fog.c_proj_fsel.fsel & 3) << 30);
+	uid->values[2] |= (((u32)bpmem.fog.c_proj_fsel.fsel >> 2) << 30);
 
-	u32* pcurvalue = &vid.values[3];
+	u32 *pcurvalue = &uid->values[3];
 	for (int i = 0; i < xfregs.numTexGens; ++i) {
 		TexMtxInfo tinfo = xfregs.texcoords[i].texmtxinfo;
 		if (tinfo.texgentype != XF_TEXGEN_EMBOSS_MAP)
@@ -78,16 +78,16 @@ static char text[16384];
 
 #define LIGHTS_POS ""
 
-char *GenerateLightShader(char* p, int index, const LitChannel& chan, const char* dest, int coloralpha);
+char *GenerateLightShader(char *p, int index, const LitChannel& chan, const char *dest, int coloralpha);
 
-const char *GenerateVertexShader(u32 components, bool D3D)
+const char *GenerateVertexShaderCode(u32 components, bool D3D)
 {
 	setlocale(LC_NUMERIC, "C"); // Reset locale for compilation
 	text[sizeof(text) - 1] = 0x7C;  // canary
     DVSTARTPROFILE();
 
-    _assert_( bpmem.genMode.numtexgens == xfregs.numTexGens);
-    _assert_( bpmem.genMode.numcolchans == xfregs.nNumChans);
+    _assert_(bpmem.genMode.numtexgens == xfregs.numTexGens);
+    _assert_(bpmem.genMode.numcolchans == xfregs.nNumChans);
     
     u32 lightMask = 0;
     if (xfregs.nNumChans > 0)
@@ -125,9 +125,8 @@ const char *GenerateVertexShader(u32 components, bool D3D)
     WRITE(p, "};\n");
 
     // uniforms
-    // bool bTexMtx = ((components & VB_HAS_TEXMTXIDXALL)<<VB_HAS_UVTEXMTXSHIFT)!=0; unused TODO: keep?
 
-    WRITE(p, "uniform s_"I_TRANSFORMMATRICES" "I_TRANSFORMMATRICES" : register(c%d);\n", C_TRANSFORMMATRICES);
+	WRITE(p, "uniform s_"I_TRANSFORMMATRICES" "I_TRANSFORMMATRICES" : register(c%d);\n", C_TRANSFORMMATRICES);
     WRITE(p, "uniform s_"I_TEXMATRICES" "I_TEXMATRICES" : register(c%d);\n", C_TEXMATRICES); // also using tex matrices
     WRITE(p, "uniform s_"I_NORMALMATRICES" "I_NORMALMATRICES" : register(c%d);\n", C_NORMALMATRICES);
     WRITE(p, "uniform s_"I_POSNORMALMATRIX" "I_POSNORMALMATRIX" : register(c%d);\n", C_POSNORMALMATRIX);
@@ -406,7 +405,7 @@ const char *GenerateVertexShader(u32 components, bool D3D)
                 break;
         }
 
-        if(xfregs.bEnableDualTexTransform && texinfo.texgentype == XF_TEXGEN_REGULAR) { // only works for regular tex gen types?
+        if (xfregs.bEnableDualTexTransform && texinfo.texgentype == XF_TEXGEN_REGULAR) { // only works for regular tex gen types?
 			int postidx = xfregs.texcoords[i].postmtxinfo.index;
 			WRITE(p, "float4 P0 = "I_POSTTRANSFORMMATRICES".T[%d].t;\n"
 				"float4 P1 = "I_POSTTRANSFORMMATRICES".T[%d].t;\n"
@@ -461,7 +460,7 @@ const char *GenerateVertexShader(u32 components, bool D3D)
 }
 
 // coloralpha - 1 if color, 2 if alpha
-char* GenerateLightShader(char* p, int index, const LitChannel& chan, const char* dest, int coloralpha)
+char *GenerateLightShader(char *p, int index, const LitChannel& chan, const char *dest, int coloralpha)
 {
     const char* swizzle = "xyzw";
     if (coloralpha == 1 ) swizzle = "xyz";
diff --git a/Source/Core/VideoCommon/Src/VertexShaderGen.h b/Source/Core/VideoCommon/Src/VertexShaderGen.h
index f3a3046efa..631787425c 100644
--- a/Source/Core/VideoCommon/Src/VertexShaderGen.h
+++ b/Source/Core/VideoCommon/Src/VertexShaderGen.h
@@ -99,8 +99,10 @@ public:
 	}
 };
 
-const char *GenerateVertexShader(u32 components, bool D3D);
-void GetVertexShaderId(VERTEXSHADERUID& vid, u32 components);
+// components is included in the uid.
+const char *GenerateVertexShaderCode(u32 components, bool D3D);
+void GetVertexShaderId(VERTEXSHADERUID *uid, u32 components);
+
 extern VERTEXSHADERUID  last_vertex_shader_uid;
 
 #endif // GCOGL_VERTEXSHADER_H