From 8a17e159435f8c10e930c4bcf3f6f2ed9fd303de Mon Sep 17 00:00:00 2001 From: NeoBrainX Date: Tue, 30 Aug 2011 20:58:43 +0200 Subject: Various cleanups and TODOs. --- Source/Core/VideoCommon/Src/PixelShaderGen.cpp | 110 ++++++++++--------------- 1 file changed, 44 insertions(+), 66 deletions(-) (limited to 'Source/Core/VideoCommon/Src/PixelShaderGen.cpp') diff --git a/Source/Core/VideoCommon/Src/PixelShaderGen.cpp b/Source/Core/VideoCommon/Src/PixelShaderGen.cpp index 530ecd54bb..d05997c0c4 100644 --- a/Source/Core/VideoCommon/Src/PixelShaderGen.cpp +++ b/Source/Core/VideoCommon/Src/PixelShaderGen.cpp @@ -333,12 +333,6 @@ static const char *tevRasTable[] = "float4(0.0f, 0.0f, 0.0f, 0.0f)", // zero }; -static const char *alphaRef[2] = -{ - I_ALPHA"[0].r", - I_ALPHA"[0].g" -}; - //static const char *tevTexFunc[] = { "tex2D", "texRECT" }; static const char *tevCOutputTable[] = { "prev.rgb", "c0.rgb", "c1.rgb", "c2.rgb" }; @@ -385,7 +379,7 @@ const char *GeneratePixelShaderCode(DSTALPHA_MODE dstAlphaMode, API_TYPE ApiType setlocale(LC_NUMERIC, "C"); // Reset locale for compilation text[sizeof(text) - 1] = 0x7C; // canary - BuildSwapModeTable(); + BuildSwapModeTable(); // Needed for WriteStage int numStages = bpmem.genMode.numtevstages + 1; int numTexgen = bpmem.genMode.numtexgens; @@ -505,10 +499,6 @@ const char *GeneratePixelShaderCode(DSTALPHA_MODE dstAlphaMode, API_TYPE ApiType { // alpha test will always fail, so restart the shader and just make it an empty function WRITE(p, "ocol0 = 0;\n"); - if(DepthTextureEnable) - WRITE(p, "depth = 1.f;\n"); - if(dstAlphaMode == DSTALPHA_DUAL_SOURCE_BLEND) - WRITE(p, "ocol1 = 0;\n"); WRITE(p, "discard;\n"); if(ApiType != API_D3D11) WRITE(p, "return;\n"); @@ -626,6 +616,7 @@ const char *GeneratePixelShaderCode(DSTALPHA_MODE dstAlphaMode, API_TYPE ApiType if(RegisterStates[0].AlphaNeedOverflowControl || RegisterStates[0].ColorNeedOverflowControl) WRITE(p, "prev = frac(prev * (255.0f/256.0f)) * (256.0f/255.0f);\n"); + // TODO: Why are we doing a second alpha pretest here? if (!WriteAlphaTest(p, ApiType, dstAlphaMode)) { // alpha test will always fail, so restart the shader and just make it an empty function @@ -740,10 +731,6 @@ static const char *TEVCMPAlphaOPTable[16] = static void WriteStage(char *&p, int n, API_TYPE ApiType) { - char *rasswap = swapModeTable[bpmem.combiners[n].alphaC.rswap]; - char *texswap = swapModeTable[bpmem.combiners[n].alphaC.tswap]; - - int texcoord = bpmem.tevorders[n/2].getTexCoord(n&1); bool bHasTexCoord = (u32)texcoord < bpmem.genMode.numtexgens; bool bHasIndStage = bpmem.tevind[n].IsActive() && bpmem.tevind[n].bt < bpmem.genMode.numindstages; @@ -780,11 +767,13 @@ static void WriteStage(char *&p, int n, API_TYPE ApiType) } else if (bpmem.tevind[n].mid <= 7 && bHasTexCoord) { // s matrix + // TODO: Might become negative? int mtxidx = 2*(bpmem.tevind[n].mid-5); WRITE(p, "float2 indtevtrans%d = "I_INDTEXMTX"[%d].ww * uv%d.xy * indtevcrd%d.xx;\n", n, mtxidx, texcoord, n); } else if (bpmem.tevind[n].mid <= 11 && bHasTexCoord) { // t matrix + // TODO: Might become negative? int mtxidx = 2*(bpmem.tevind[n].mid-9); WRITE(p, "float2 indtevtrans%d = "I_INDTEXMTX"[%d].ww * uv%d.xy * indtevcrd%d.yy;\n", n, mtxidx, texcoord, n); } @@ -823,11 +812,14 @@ static void WriteStage(char *&p, int n, API_TYPE ApiType) TevStageCombiner::ColorCombiner &cc = bpmem.combiners[n].colorC; TevStageCombiner::AlphaCombiner &ac = bpmem.combiners[n].alphaC; - bool bCRas = cc.a == TEVCOLORARG_RASA || cc.a == TEVCOLORARG_RASC || cc.b == TEVCOLORARG_RASA || cc.b == TEVCOLORARG_RASC || cc.c == TEVCOLORARG_RASA || cc.c == TEVCOLORARG_RASC || cc.d == TEVCOLORARG_RASA || cc.d == TEVCOLORARG_RASC; - bool bARas = ac.a == TEVALPHAARG_RASA || ac.b == TEVALPHAARG_RASA || ac.c == TEVALPHAARG_RASA || ac.d == TEVALPHAARG_RASA; - - if(bCRas || bARas) + if(cc.a == TEVCOLORARG_RASA || cc.a == TEVCOLORARG_RASC + || cc.b == TEVCOLORARG_RASA || cc.b == TEVCOLORARG_RASC + || cc.c == TEVCOLORARG_RASA || cc.c == TEVCOLORARG_RASC + || cc.d == TEVCOLORARG_RASA || cc.d == TEVCOLORARG_RASC + || ac.a == TEVALPHAARG_RASA || ac.b == TEVALPHAARG_RASA + || ac.c == TEVALPHAARG_RASA || ac.d == TEVALPHAARG_RASA) { + char *rasswap = swapModeTable[bpmem.combiners[n].alphaC.rswap]; WRITE(p, "rastemp = %s.%s;\n", tevRasTable[bpmem.tevorders[n / 2].getColorChan(n & 1)], rasswap); WRITE(p, "crastemp = frac(rastemp * (255.0f/256.0f)) * (256.0f/255.0f);\n"); } @@ -835,7 +827,6 @@ static void WriteStage(char *&p, int n, API_TYPE ApiType) if (bpmem.tevorders[n/2].getEnable(n&1)) { - int texmap = bpmem.tevorders[n/2].getTexMap(n&1); if(!bHasIndStage) { // calc tevcord @@ -845,20 +836,19 @@ static void WriteStage(char *&p, int n, API_TYPE ApiType) WRITE(p, "tevcoord.xy = float2(0.0f, 0.0f);\n"); } + char *texswap = swapModeTable[bpmem.combiners[n].alphaC.tswap]; + int texmap = bpmem.tevorders[n/2].getTexMap(n&1); SampleTexture(p, "textemp", "tevcoord", texswap, texmap, ApiType); } else WRITE(p, "textemp = float4(1.0f, 1.0f, 1.0f, 1.0f);\n"); - int kc = bpmem.tevksel[n / 2].getKC(n & 1); - int ka = bpmem.tevksel[n / 2].getKA(n & 1); - - - bool bCKonst = cc.a == TEVCOLORARG_KONST || cc.b == TEVCOLORARG_KONST || cc.c == TEVCOLORARG_KONST || cc.d == TEVCOLORARG_KONST; - bool bAKonst = ac.a == TEVALPHAARG_KONST || ac.b == TEVALPHAARG_KONST || ac.c == TEVALPHAARG_KONST || ac.d == TEVALPHAARG_KONST; - if (bCKonst || bAKonst ) + if (cc.a == TEVCOLORARG_KONST || cc.b == TEVCOLORARG_KONST || cc.c == TEVCOLORARG_KONST || cc.d == TEVCOLORARG_KONST + || ac.a == TEVALPHAARG_KONST || ac.b == TEVALPHAARG_KONST || ac.c == TEVALPHAARG_KONST || ac.d == TEVALPHAARG_KONST) { + int kc = bpmem.tevksel[n / 2].getKC(n & 1); + int ka = bpmem.tevksel[n / 2].getKA(n & 1); WRITE(p, "konsttemp = float4(%s, %s);\n", tevKSelTableC[kc], tevKSelTableA[ka]); if(kc > 7 || ka > 7) { @@ -870,15 +860,10 @@ static void WriteStage(char *&p, int n, API_TYPE ApiType) } } - if(cc.a == TEVCOLORARG_CPREV - || cc.a == TEVCOLORARG_APREV - || cc.b == TEVCOLORARG_CPREV - || cc.b == TEVCOLORARG_APREV - || cc.c == TEVCOLORARG_CPREV - || cc.c == TEVCOLORARG_APREV - || ac.a == TEVALPHAARG_APREV - || ac.b == TEVALPHAARG_APREV - || ac.c == TEVALPHAARG_APREV) + if(cc.a == TEVCOLORARG_CPREV || cc.a == TEVCOLORARG_APREV + || cc.b == TEVCOLORARG_CPREV || cc.b == TEVCOLORARG_APREV + || cc.c == TEVCOLORARG_CPREV || cc.c == TEVCOLORARG_APREV + || ac.a == TEVALPHAARG_APREV || ac.b == TEVALPHAARG_APREV || ac.c == TEVALPHAARG_APREV) { if(RegisterStates[0].AlphaNeedOverflowControl || RegisterStates[0].ColorNeedOverflowControl) { @@ -893,15 +878,10 @@ static void WriteStage(char *&p, int n, API_TYPE ApiType) RegisterStates[0].AuxStored = true; } - if(cc.a == TEVCOLORARG_C0 - || cc.a == TEVCOLORARG_A0 - || cc.b == TEVCOLORARG_C0 - || cc.b == TEVCOLORARG_A0 - || cc.c == TEVCOLORARG_C0 - || cc.c == TEVCOLORARG_A0 - || ac.a == TEVALPHAARG_A0 - || ac.b == TEVALPHAARG_A0 - || ac.c == TEVALPHAARG_A0) + if(cc.a == TEVCOLORARG_C0 || cc.a == TEVCOLORARG_A0 + || cc.b == TEVCOLORARG_C0 || cc.b == TEVCOLORARG_A0 + || cc.c == TEVCOLORARG_C0 || cc.c == TEVCOLORARG_A0 + || ac.a == TEVALPHAARG_A0 || ac.b == TEVALPHAARG_A0 || ac.c == TEVALPHAARG_A0) { if(RegisterStates[1].AlphaNeedOverflowControl || RegisterStates[1].ColorNeedOverflowControl) { @@ -916,15 +896,10 @@ static void WriteStage(char *&p, int n, API_TYPE ApiType) RegisterStates[1].AuxStored = true; } - if(cc.a == TEVCOLORARG_C1 - || cc.a == TEVCOLORARG_A1 - || cc.b == TEVCOLORARG_C1 - || cc.b == TEVCOLORARG_A1 - || cc.c == TEVCOLORARG_C1 - || cc.c == TEVCOLORARG_A1 - || ac.a == TEVALPHAARG_A1 - || ac.b == TEVALPHAARG_A1 - || ac.c == TEVALPHAARG_A1) + if(cc.a == TEVCOLORARG_C1 || cc.a == TEVCOLORARG_A1 + || cc.b == TEVCOLORARG_C1 || cc.b == TEVCOLORARG_A1 + || cc.c == TEVCOLORARG_C1 || cc.c == TEVCOLORARG_A1 + || ac.a == TEVALPHAARG_A1 || ac.b == TEVALPHAARG_A1 || ac.c == TEVALPHAARG_A1) { if(RegisterStates[2].AlphaNeedOverflowControl || RegisterStates[2].ColorNeedOverflowControl) { @@ -939,15 +914,10 @@ static void WriteStage(char *&p, int n, API_TYPE ApiType) RegisterStates[2].AuxStored = true; } - if(cc.a == TEVCOLORARG_C2 - || cc.a == TEVCOLORARG_A2 - || cc.b == TEVCOLORARG_C2 - || cc.b == TEVCOLORARG_A2 - || cc.c == TEVCOLORARG_C2 - || cc.c == TEVCOLORARG_A2 - || ac.a == TEVALPHAARG_A2 - || ac.b == TEVALPHAARG_A2 - || ac.c == TEVALPHAARG_A2) + if(cc.a == TEVCOLORARG_C2 || cc.a == TEVCOLORARG_A2 + || cc.b == TEVCOLORARG_C2 || cc.b == TEVCOLORARG_A2 + || cc.c == TEVCOLORARG_C2 || cc.c == TEVCOLORARG_A2 + || ac.a == TEVALPHAARG_A2 || ac.b == TEVALPHAARG_A2 || ac.c == TEVALPHAARG_A2) { if(RegisterStates[3].AlphaNeedOverflowControl || RegisterStates[3].ColorNeedOverflowControl) { @@ -994,7 +964,7 @@ static void WriteStage(char *&p, int n, API_TYPE ApiType) WRITE(p, "%s", tevBiasTable[cc.bias]); - if (cc.shift > 0) + if (cc.shift > TEVSCALE_1) WRITE(p, ")"); } else @@ -1122,6 +1092,11 @@ static int AlphaPreTest() static bool WriteAlphaTest(char *&p, API_TYPE ApiType,DSTALPHA_MODE dstAlphaMode) { + static const char *alphaRef[2] = + { + I_ALPHA"[0].r", + I_ALPHA"[0].g" + }; int Pretest = AlphaPreTest(); if(Pretest >= 0) @@ -1139,7 +1114,10 @@ static bool WriteAlphaTest(char *&p, API_TYPE ApiType,DSTALPHA_MODE dstAlphaMode compindex = bpmem.alphaFunc.comp1 % 8; WRITE(p, tevAlphaFuncsTable[compindex],alphaRef[1]);//lookup the second component from the alpha function table - WRITE(p, ")){ocol0 = 0;%s%sdiscard;%s}\n",dstAlphaMode == DSTALPHA_DUAL_SOURCE_BLEND ? "ocol1 = 0;" : "",DepthTextureEnable ? "depth = 1.f;" : "",(ApiType != API_D3D11)? "return;" : ""); + WRITE(p, ")){ocol0 = 0;%s%s discard;%s}\n", + dstAlphaMode == DSTALPHA_DUAL_SOURCE_BLEND ? "ocol1 = 0;" : "", + DepthTextureEnable ? "depth = 1.f;" : "", + (ApiType != API_D3D11) ? "return;" : ""); return true; } @@ -1197,4 +1175,4 @@ static void WriteFog(char *&p) WRITE(p, " prev.rgb = lerp(prev.rgb,"I_FOG"[0].rgb,fog);\n"); -} +} \ No newline at end of file -- cgit v1.2.3 From 08e06b22937fa79db165e5a31dfd4ccf67ef57db Mon Sep 17 00:00:00 2001 From: NeoBrainX Date: Wed, 31 Aug 2011 16:09:54 +0200 Subject: Partially revert revision d511b506120c. Slightly slows down emulation, but deobfuscates the pixel shader gen greatly... --- Source/Core/VideoCommon/Src/PixelShaderGen.cpp | 102 +++---------------------- 1 file changed, 12 insertions(+), 90 deletions(-) (limited to 'Source/Core/VideoCommon/Src/PixelShaderGen.cpp') diff --git a/Source/Core/VideoCommon/Src/PixelShaderGen.cpp b/Source/Core/VideoCommon/Src/PixelShaderGen.cpp index d05997c0c4..ade8a0865e 100644 --- a/Source/Core/VideoCommon/Src/PixelShaderGen.cpp +++ b/Source/Core/VideoCommon/Src/PixelShaderGen.cpp @@ -353,15 +353,6 @@ static char swapModeTable[4][5]; static char text[16384]; static bool DepthTextureEnable; -struct RegisterState -{ - bool ColorNeedOverflowControl; - bool AlphaNeedOverflowControl; - bool AuxStored; -}; - -static RegisterState RegisterStates[4]; - static void BuildSwapModeTable() { for (int i = 0; i < 4; i++) @@ -582,16 +573,6 @@ const char *GeneratePixelShaderCode(DSTALPHA_MODE dstAlphaMode, API_TYPE ApiType } } - RegisterStates[0].AlphaNeedOverflowControl = false; - RegisterStates[0].ColorNeedOverflowControl = false; - RegisterStates[0].AuxStored = false; - for(int i = 1; i < 4; i++) - { - RegisterStates[i].AlphaNeedOverflowControl = true; - RegisterStates[i].ColorNeedOverflowControl = true; - RegisterStates[i].AuxStored = false; - } - for (int i = 0; i < numStages; i++) WriteStage(p, i, ApiType); //build the equation for this stage @@ -599,22 +580,11 @@ const char *GeneratePixelShaderCode(DSTALPHA_MODE dstAlphaMode, API_TYPE ApiType { // The results of the last texenv stage are put onto the screen, // regardless of the used destination register - if(bpmem.combiners[numStages - 1].colorC.dest != 0) - { - bool retrieveFromAuxRegister = !RegisterStates[bpmem.combiners[numStages - 1].colorC.dest].ColorNeedOverflowControl && RegisterStates[bpmem.combiners[numStages - 1].colorC.dest].AuxStored; - WRITE(p, "prev.rgb = %s%s;\n", retrieveFromAuxRegister ? "c" : "" , tevCOutputTable[bpmem.combiners[numStages - 1].colorC.dest]); - RegisterStates[0].ColorNeedOverflowControl = RegisterStates[bpmem.combiners[numStages - 1].colorC.dest].ColorNeedOverflowControl; - } - if(bpmem.combiners[numStages - 1].alphaC.dest != 0) - { - bool retrieveFromAuxRegister = !RegisterStates[bpmem.combiners[numStages - 1].alphaC.dest].AlphaNeedOverflowControl && RegisterStates[bpmem.combiners[numStages - 1].alphaC.dest].AuxStored; - WRITE(p, "prev.a = %s%s;\n", retrieveFromAuxRegister ? "c" : "" , tevAOutputTable[bpmem.combiners[numStages - 1].alphaC.dest]); - RegisterStates[0].AlphaNeedOverflowControl = RegisterStates[bpmem.combiners[numStages - 1].alphaC.dest].AlphaNeedOverflowControl; - } - } - // emulation of unisgned 8 overflow when casting if needed - if(RegisterStates[0].AlphaNeedOverflowControl || RegisterStates[0].ColorNeedOverflowControl) - WRITE(p, "prev = frac(prev * (255.0f/256.0f)) * (256.0f/255.0f);\n"); + WRITE(p, "prev.rgb = %s;\n",tevCOutputTable[bpmem.combiners[numStages-1].colorC.dest]); + WRITE(p, "prev.a = %s;\n",tevAOutputTable[bpmem.combiners[numStages-1].alphaC.dest]); + } + // emulation of unsigned 8 overflow when casting + WRITE(p, "prev = frac(4.0f + prev * (255.0f/256.0f)) * (256.0f/255.0f);\n"); // TODO: Why are we doing a second alpha pretest here? if (!WriteAlphaTest(p, ApiType, dstAlphaMode)) @@ -864,76 +834,30 @@ static void WriteStage(char *&p, int n, API_TYPE ApiType) || cc.b == TEVCOLORARG_CPREV || cc.b == TEVCOLORARG_APREV || cc.c == TEVCOLORARG_CPREV || cc.c == TEVCOLORARG_APREV || ac.a == TEVALPHAARG_APREV || ac.b == TEVALPHAARG_APREV || ac.c == TEVALPHAARG_APREV) - { - if(RegisterStates[0].AlphaNeedOverflowControl || RegisterStates[0].ColorNeedOverflowControl) - { - WRITE(p, "cprev = frac(prev * (255.0f/256.0f)) * (256.0f/255.0f);\n"); - RegisterStates[0].AlphaNeedOverflowControl = false; - RegisterStates[0].ColorNeedOverflowControl = false; - } - else - { - WRITE(p, "cprev = prev;\n"); - } - RegisterStates[0].AuxStored = true; - } + WRITE(p, "cprev = frac(prev * (255.0f/256.0f)) * (256.0f/255.0f);\n"); + if(cc.a == TEVCOLORARG_C0 || cc.a == TEVCOLORARG_A0 || cc.b == TEVCOLORARG_C0 || cc.b == TEVCOLORARG_A0 || cc.c == TEVCOLORARG_C0 || cc.c == TEVCOLORARG_A0 || ac.a == TEVALPHAARG_A0 || ac.b == TEVALPHAARG_A0 || ac.c == TEVALPHAARG_A0) - { - if(RegisterStates[1].AlphaNeedOverflowControl || RegisterStates[1].ColorNeedOverflowControl) - { - WRITE(p, "cc0 = frac(c0 * (255.0f/256.0f)) * (256.0f/255.0f);\n"); - RegisterStates[1].AlphaNeedOverflowControl = false; - RegisterStates[1].ColorNeedOverflowControl = false; - } - else - { - WRITE(p, "cc0 = c0;\n"); - } - RegisterStates[1].AuxStored = true; - } + WRITE(p, "cc0 = frac(c0 * (255.0f/256.0f)) * (256.0f/255.0f);\n"); + if(cc.a == TEVCOLORARG_C1 || cc.a == TEVCOLORARG_A1 || cc.b == TEVCOLORARG_C1 || cc.b == TEVCOLORARG_A1 || cc.c == TEVCOLORARG_C1 || cc.c == TEVCOLORARG_A1 || ac.a == TEVALPHAARG_A1 || ac.b == TEVALPHAARG_A1 || ac.c == TEVALPHAARG_A1) - { - if(RegisterStates[2].AlphaNeedOverflowControl || RegisterStates[2].ColorNeedOverflowControl) - { - WRITE(p, "cc1 = frac(c1 * (255.0f/256.0f)) * (256.0f/255.0f);\n"); - RegisterStates[2].AlphaNeedOverflowControl = false; - RegisterStates[2].ColorNeedOverflowControl = false; - } - else - { - WRITE(p, "cc1 = c1;\n"); - } - RegisterStates[2].AuxStored = true; - } + WRITE(p, "cc1 = frac(c1 * (255.0f/256.0f)) * (256.0f/255.0f);\n"); + if(cc.a == TEVCOLORARG_C2 || cc.a == TEVCOLORARG_A2 || cc.b == TEVCOLORARG_C2 || cc.b == TEVCOLORARG_A2 || cc.c == TEVCOLORARG_C2 || cc.c == TEVCOLORARG_A2 || ac.a == TEVALPHAARG_A2 || ac.b == TEVALPHAARG_A2 || ac.c == TEVALPHAARG_A2) - { - if(RegisterStates[3].AlphaNeedOverflowControl || RegisterStates[3].ColorNeedOverflowControl) - { WRITE(p, "cc2 = frac(c2 * (255.0f/256.0f)) * (256.0f/255.0f);\n"); - RegisterStates[3].AlphaNeedOverflowControl = false; - RegisterStates[3].ColorNeedOverflowControl = false; - } - else - { - WRITE(p, "cc2 = c2;\n"); - } - RegisterStates[3].AuxStored = true; - } - RegisterStates[cc.dest].ColorNeedOverflowControl = (cc.clamp == 0); - RegisterStates[cc.dest].AuxStored = false; + if (cc.clamp) WRITE(p, "%s = saturate(", tevCOutputTable[cc.dest]); else @@ -980,8 +904,6 @@ static void WriteStage(char *&p, int n, API_TYPE ApiType) WRITE(p, ")"); WRITE(p,";\n"); - RegisterStates[ac.dest].AlphaNeedOverflowControl = (ac.clamp == 0); - RegisterStates[ac.dest].AuxStored = false; // combine the alpha channel if (ac.clamp) WRITE(p, "%s = saturate(", tevAOutputTable[ac.dest]); -- cgit v1.2.3 From 7f01139d13ad917d333548918492e92e2bcb436a Mon Sep 17 00:00:00 2001 From: NeoBrainX Date: Wed, 31 Aug 2011 03:18:19 +0200 Subject: Replace the pixel shader UID generation algorithm with a better one which reduces the number of redundant shader compilations by around 30% (can be optimized even further though). This should help some games which suffer from heavy stuttering like e.g. F-Zero GX or Red Steel 2. --- Source/Core/VideoCommon/Src/PixelShaderGen.cpp | 2360 +++++++++++++----------- 1 file changed, 1265 insertions(+), 1095 deletions(-) (limited to 'Source/Core/VideoCommon/Src/PixelShaderGen.cpp') diff --git a/Source/Core/VideoCommon/Src/PixelShaderGen.cpp b/Source/Core/VideoCommon/Src/PixelShaderGen.cpp index ade8a0865e..6d0cc64964 100644 --- a/Source/Core/VideoCommon/Src/PixelShaderGen.cpp +++ b/Source/Core/VideoCommon/Src/PixelShaderGen.cpp @@ -1,1100 +1,1270 @@ -// Copyright (C) 2003 Dolphin Project. - -// This program is free software: you can redistribute it and/or modify -// it under the terms of the GNU General Public License as published by -// the Free Software Foundation, version 2.0. - -// This program is distributed in the hope that it will be useful, -// but WITHOUT ANY WARRANTY; without even the implied warranty of -// MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the -// GNU General Public License 2.0 for more details. - -// A copy of the GPL 2.0 should have been included with the program. -// If not, see http://www.gnu.org/licenses/ - -// Official SVN repository and contact information can be found at -// http://code.google.com/p/dolphin-emu/ - -#include -#include -#include -#include - -#include "LightingShaderGen.h" -#include "PixelShaderGen.h" -#include "XFMemory.h" // for texture projection mode -#include "BPMemory.h" -#include "VideoConfig.h" -#include "NativeVertexFormat.h" - -PIXELSHADERUID last_pixel_shader_uid; - -// Mash together all the inputs that contribute to the code of a generated pixel shader into -// 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, DSTALPHA_MODE dstAlphaMode) -{ - u32 numstages = bpmem.genMode.numtevstages + 1; - u32 projtexcoords = 0; - for (u32 i = 0; i < numstages; i++) - { - if (bpmem.tevorders[i/2].getEnable(i & 1)) - { - int texcoord = bpmem.tevorders[i / 2].getTexCoord(i & 1); - if (xfregs.texMtxInfo[i].projection) - projtexcoords |= 1 << texcoord; - } - } - uid->values[0] = (u32)bpmem.genMode.numtevstages | - ((u32)bpmem.genMode.numindstages << 4) | - ((u32)bpmem.genMode.numtexgens << 7) | - ((u32)dstAlphaMode << 11) | - ((u32)((bpmem.alphaFunc.hex >> 16) & 0xff) << 13) | - (projtexcoords << 21) | - ((u32)bpmem.ztex2.op << 29); - - // 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); - - u32 enableZTexture = (bpmem.ztex2.op != ZTEXTURE_DISABLE && !bpmem.zcontrol.zcomploc && bpmem.zmode.testenable && bpmem.zmode.updateenable) || g_ActiveConfig.bEnablePerPixelDepth ? 1 : 0; - - uid->values[2] = (u32)bpmem.fog.c_proj_fsel.fsel | - ((u32)bpmem.fog.c_proj_fsel.proj << 3) | - ((u32)enableZTexture << 4) | ((u32)bpmem.fogRange.Base.Enabled << 5); - - if(g_ActiveConfig.bEnablePixelLighting && g_ActiveConfig.backend_info.bSupportsPixelLighting) - { - for (int i = 0; i < 2; ++i) { - uid->values[3 + i] = xfregs.color[i].enablelighting ? - (u32)xfregs.color[i].hex : - (u32)xfregs.color[i].matsource; - uid->values[3 + i] |= (xfregs.alpha[i].enablelighting ? - (u32)xfregs.alpha[i].hex : - (u32)xfregs.alpha[i].matsource) << 15; - } - } - uid->values[4] |= (g_ActiveConfig.bEnablePixelLighting && g_ActiveConfig.backend_info.bSupportsPixelLighting) << 31; - - int hdr = 5; - u32 *pcurvalue = &uid->values[hdr]; - for (u32 i = 0; i < numstages; ++i) - { - TevStageCombiner::ColorCombiner &cc = bpmem.combiners[i].colorC; - TevStageCombiner::AlphaCombiner &ac = bpmem.combiners[i].alphaC; - - u32 val0 = cc.hex & 0xffffff; - u32 val1 = ac.hex & 0xffffff; - val0 |= bpmem.tevksel[i / 2].getKC(i & 1) << 24; - val1 |= bpmem.tevksel[i / 2].getKA(i & 1) << 24; - pcurvalue[0] = val0; - pcurvalue[1] = val1; - pcurvalue += 2; - } - - for (u32 i = 0; i < numstages / 2; ++i) - { - u32 val0, val1; - if (bpmem.tevorders[i].hex & 0x40) - val0 = bpmem.tevorders[i].hex & 0x3ff; - else - val0 = bpmem.tevorders[i].hex & 0x380; - if (bpmem.tevorders[i].hex & 0x40000) - val1 = (bpmem.tevorders[i].hex & 0x3ff000) >> 12; - else - val1 = (bpmem.tevorders[i].hex & 0x380000) >> 12; - - switch (i % 3) { - case 0: pcurvalue[0] = val0|(val1<<10); break; - case 1: pcurvalue[0] |= val0<<20; pcurvalue[1] = val1; pcurvalue++; break; - case 2: pcurvalue[1] |= (val0<<10)|(val1<<20); pcurvalue++; break; - default: PanicAlert("Unknown case for Tev Stages / 2: %08x", (i % 3)); - } - } - - if (numstages & 1) { // odd - u32 val0; - if (bpmem.tevorders[bpmem.genMode.numtevstages/2].hex & 0x40) - val0 = bpmem.tevorders[bpmem.genMode.numtevstages/2].hex & 0x3ff; - else - val0 = bpmem.tevorders[bpmem.genMode.numtevstages/2].hex & 0x380; - - switch (bpmem.genMode.numtevstages % 3) - { - case 0: pcurvalue[0] = val0; break; - case 1: pcurvalue[0] |= val0 << 20; break; - case 2: pcurvalue[1] |= val0 << 10; pcurvalue++; break; - default: PanicAlert("Unknown case for Tev Stages: %08x", bpmem.genMode.numtevstages % 3); - } - } - - if ((bpmem.genMode.numtevstages % 3) != 2) - ++pcurvalue; - - uid->tevstages = (u32)(pcurvalue - &uid->values[0] - hdr); - - for (u32 i = 0; i < bpmem.genMode.numindstages; ++i) - { - u32 val = bpmem.tevind[i].hex & 0x1fffff; // 21 bits - switch (i % 3) - { - case 0: pcurvalue[0] = val; break; - case 1: pcurvalue[0] |= val << 21; pcurvalue[1] = val >> 11; ++pcurvalue; break; - case 2: pcurvalue[0] |= val << 10; ++pcurvalue; break; - default: PanicAlert("Unknown case for Ind Stages: %08x", (i % 3)); - } - } - - // yeah, well .... - uid->indstages = (u32)(pcurvalue - &uid->values[0] - (hdr - 1) - uid->tevstages); - -} - -// old tev->pixelshader notes -// -// color for this stage (alpha, color) is given by bpmem.tevorders[0].colorchan0 -// konstant for this stage (alpha, color) is given by bpmem.tevksel -// inputs are given by bpmem.combiners[0].colorC.a/b/c/d << could be current chan color -// according to GXTevColorArg table above -// output is given by .outreg -// tevtemp is set according to swapmodetables and - -static void WriteStage(char *&p, int n, API_TYPE ApiType); -static void SampleTexture(char *&p, const char *destination, const char *texcoords, const char *texswap, int texmap, API_TYPE ApiType); -// static void WriteAlphaCompare(char *&p, int num, int comp); -static bool WriteAlphaTest(char *&p, API_TYPE ApiType,DSTALPHA_MODE dstAlphaMode); -static void WriteFog(char *&p); -static int AlphaPreTest(); - -static const char *tevKSelTableC[] = // KCSEL -{ - "1.0f,1.0f,1.0f", // 1 = 0x00 - "0.875f,0.875f,0.875f", // 7_8 = 0x01 - "0.75f,0.75f,0.75f", // 3_4 = 0x02 - "0.625f,0.625f,0.625f", // 5_8 = 0x03 - "0.5f,0.5f,0.5f", // 1_2 = 0x04 - "0.375f,0.375f,0.375f", // 3_8 = 0x05 - "0.25f,0.25f,0.25f", // 1_4 = 0x06 - "0.125f,0.125f,0.125f", // 1_8 = 0x07 - "ERROR", // 0x08 - "ERROR", // 0x09 - "ERROR", // 0x0a - "ERROR", // 0x0b - I_KCOLORS"[0].rgb", // K0 = 0x0C - I_KCOLORS"[1].rgb", // K1 = 0x0D - I_KCOLORS"[2].rgb", // K2 = 0x0E - I_KCOLORS"[3].rgb", // K3 = 0x0F - I_KCOLORS"[0].rrr", // K0_R = 0x10 - I_KCOLORS"[1].rrr", // K1_R = 0x11 - I_KCOLORS"[2].rrr", // K2_R = 0x12 - I_KCOLORS"[3].rrr", // K3_R = 0x13 - I_KCOLORS"[0].ggg", // K0_G = 0x14 - I_KCOLORS"[1].ggg", // K1_G = 0x15 - I_KCOLORS"[2].ggg", // K2_G = 0x16 - I_KCOLORS"[3].ggg", // K3_G = 0x17 - I_KCOLORS"[0].bbb", // K0_B = 0x18 - I_KCOLORS"[1].bbb", // K1_B = 0x19 - I_KCOLORS"[2].bbb", // K2_B = 0x1A - I_KCOLORS"[3].bbb", // K3_B = 0x1B - I_KCOLORS"[0].aaa", // K0_A = 0x1C - I_KCOLORS"[1].aaa", // K1_A = 0x1D - I_KCOLORS"[2].aaa", // K2_A = 0x1E - I_KCOLORS"[3].aaa", // K3_A = 0x1F -}; - -static const char *tevKSelTableA[] = // KASEL -{ - "1.0f", // 1 = 0x00 - "0.875f",// 7_8 = 0x01 - "0.75f", // 3_4 = 0x02 - "0.625f",// 5_8 = 0x03 - "0.5f", // 1_2 = 0x04 - "0.375f",// 3_8 = 0x05 - "0.25f", // 1_4 = 0x06 - "0.125f",// 1_8 = 0x07 - "ERROR", // 0x08 - "ERROR", // 0x09 - "ERROR", // 0x0a - "ERROR", // 0x0b - "ERROR", // 0x0c - "ERROR", // 0x0d - "ERROR", // 0x0e - "ERROR", // 0x0f - I_KCOLORS"[0].r", // K0_R = 0x10 - I_KCOLORS"[1].r", // K1_R = 0x11 - I_KCOLORS"[2].r", // K2_R = 0x12 - I_KCOLORS"[3].r", // K3_R = 0x13 - I_KCOLORS"[0].g", // K0_G = 0x14 - I_KCOLORS"[1].g", // K1_G = 0x15 - I_KCOLORS"[2].g", // K2_G = 0x16 - I_KCOLORS"[3].g", // K3_G = 0x17 - I_KCOLORS"[0].b", // K0_B = 0x18 - I_KCOLORS"[1].b", // K1_B = 0x19 - I_KCOLORS"[2].b", // K2_B = 0x1A - I_KCOLORS"[3].b", // K3_B = 0x1B - I_KCOLORS"[0].a", // K0_A = 0x1C - I_KCOLORS"[1].a", // K1_A = 0x1D - I_KCOLORS"[2].a", // K2_A = 0x1E - I_KCOLORS"[3].a", // K3_A = 0x1F -}; - -static const char *tevScaleTable[] = // CS -{ - "1.0f", // SCALE_1 - "2.0f", // SCALE_2 - "4.0f", // SCALE_4 - "0.5f", // DIVIDE_2 -}; - -static const char *tevBiasTable[] = // TB -{ - "", // ZERO, - "+0.5f", // ADDHALF, - "-0.5f", // SUBHALF, - "", -}; - -static const char *tevOpTable[] = { // TEV - "+", // TEVOP_ADD = 0, - "-", // TEVOP_SUB = 1, -}; - -static const char *tevCInputTable[] = // CC -{ - "(prev.rgb)", // CPREV, - "(prev.aaa)", // APREV, - "(c0.rgb)", // C0, - "(c0.aaa)", // A0, - "(c1.rgb)", // C1, - "(c1.aaa)", // A1, - "(c2.rgb)", // C2, - "(c2.aaa)", // A2, - "(textemp.rgb)", // TEXC, - "(textemp.aaa)", // TEXA, - "(rastemp.rgb)", // RASC, - "(rastemp.aaa)", // RASA, - "float3(1.0f, 1.0f, 1.0f)", // ONE - "float3(0.5f, 0.5f, 0.5f)", // HALF - "(konsttemp.rgb)", //"konsttemp.rgb", // KONST - "float3(0.0f, 0.0f, 0.0f)", // ZERO - ///aded extra values to map clamped values - "(cprev.rgb)", // CPREV, - "(cprev.aaa)", // APREV, - "(cc0.rgb)", // C0, - "(cc0.aaa)", // A0, - "(cc1.rgb)", // C1, - "(cc1.aaa)", // A1, - "(cc2.rgb)", // C2, - "(cc2.aaa)", // A2, - "(textemp.rgb)", // TEXC, - "(textemp.aaa)", // TEXA, - "(crastemp.rgb)", // RASC, - "(crastemp.aaa)", // RASA, - "float3(1.0f, 1.0f, 1.0f)", // ONE - "float3(0.5f, 0.5f, 0.5f)", // HALF - "(ckonsttemp.rgb)", //"konsttemp.rgb", // KONST - "float3(0.0f, 0.0f, 0.0f)", // ZERO - "PADERROR", "PADERROR", "PADERROR", "PADERROR" -}; - -static const char *tevAInputTable[] = // CA -{ - "prev", // APREV, - "c0", // A0, - "c1", // A1, - "c2", // A2, - "textemp", // TEXA, - "rastemp", // RASA, - "konsttemp", // KONST, (hw1 had quarter) - "float4(0.0f, 0.0f, 0.0f, 0.0f)", // ZERO - ///aded extra values to map clamped values - "cprev", // APREV, - "cc0", // A0, - "cc1", // A1, - "cc2", // A2, - "textemp", // TEXA, - "crastemp", // RASA, - "ckonsttemp", // KONST, (hw1 had quarter) - "float4(0.0f, 0.0f, 0.0f, 0.0f)", // ZERO - "PADERROR", "PADERROR", "PADERROR", "PADERROR", - "PADERROR", "PADERROR", "PADERROR", "PADERROR", -}; - -static const char *tevRasTable[] = -{ - "colors_0", - "colors_1", - "ERROR", //2 - "ERROR", //3 - "ERROR", //4 - "alphabump", // use bump alpha - "(alphabump*(255.0f/248.0f))", //normalized - "float4(0.0f, 0.0f, 0.0f, 0.0f)", // zero -}; - -//static const char *tevTexFunc[] = { "tex2D", "texRECT" }; - -static const char *tevCOutputTable[] = { "prev.rgb", "c0.rgb", "c1.rgb", "c2.rgb" }; -static const char *tevAOutputTable[] = { "prev.a", "c0.a", "c1.a", "c2.a" }; -static const char *tevIndAlphaSel[] = {"", "x", "y", "z"}; -//static const char *tevIndAlphaScale[] = {"", "*32", "*16", "*8"}; -static const char *tevIndAlphaScale[] = {"*(248.0f/255.0f)", "*(224.0f/255.0f)", "*(240.0f/255.0f)", "*(248.0f/255.0f)"}; -static const char *tevIndBiasField[] = {"", "x", "y", "xy", "z", "xz", "yz", "xyz"}; // indexed by bias -static const char *tevIndBiasAdd[] = {"-128.0f", "1.0f", "1.0f", "1.0f" }; // indexed by fmt -static const char *tevIndWrapStart[] = {"0.0f", "256.0f", "128.0f", "64.0f", "32.0f", "16.0f", "0.001f" }; -static const char *tevIndFmtScale[] = {"255.0f", "31.0f", "15.0f", "7.0f" }; - -#define WRITE p+=sprintf - -static const char *swapColors = "rgba"; -static char swapModeTable[4][5]; - -static char text[16384]; -static bool DepthTextureEnable; - -static void BuildSwapModeTable() -{ - for (int i = 0; i < 4; i++) - { - swapModeTable[i][0] = swapColors[bpmem.tevksel[i*2].swap1]; - swapModeTable[i][1] = swapColors[bpmem.tevksel[i*2].swap2]; - swapModeTable[i][2] = swapColors[bpmem.tevksel[i*2+1].swap1]; - swapModeTable[i][3] = swapColors[bpmem.tevksel[i*2+1].swap2]; - swapModeTable[i][4] = 0; - } -} - -const char *GeneratePixelShaderCode(DSTALPHA_MODE dstAlphaMode, API_TYPE ApiType, u32 components) -{ - setlocale(LC_NUMERIC, "C"); // Reset locale for compilation - text[sizeof(text) - 1] = 0x7C; // canary - - BuildSwapModeTable(); // Needed for WriteStage - int numStages = bpmem.genMode.numtevstages + 1; - int numTexgen = bpmem.genMode.numtexgens; - - char *p = text; - WRITE(p, "//Pixel Shader for TEV stages\n"); - WRITE(p, "//%i TEV stages, %i texgens, %i IND stages\n", - numStages, numTexgen, bpmem.genMode.numindstages); - - int nIndirectStagesUsed = 0; - if (bpmem.genMode.numindstages > 0) - { - for (int i = 0; i < numStages; ++i) - { - if (bpmem.tevind[i].IsActive() && bpmem.tevind[i].bt < bpmem.genMode.numindstages) - nIndirectStagesUsed |= 1 << bpmem.tevind[i].bt; - } - } - DepthTextureEnable = (bpmem.ztex2.op != ZTEXTURE_DISABLE && !bpmem.zcontrol.zcomploc && bpmem.zmode.testenable && bpmem.zmode.updateenable) || g_ActiveConfig.bEnablePerPixelDepth ; - // Declare samplers - - if(ApiType != API_D3D11) - { - WRITE(p, "uniform sampler2D "); - } - else - { - WRITE(p, "sampler "); - } - - bool bfirst = true; - for (int i = 0; i < 8; ++i) - { - WRITE(p, "%s samp%d : register(s%d)", bfirst?"":",", i, i); - bfirst = false; - } - WRITE(p, ";\n"); - if(ApiType == API_D3D11) - { - WRITE(p, "Texture2D "); - bfirst = true; - for (int i = 0; i < 8; ++i) - { - WRITE(p, "%s Tex%d : register(t%d)", bfirst?"":",", i, i); - bfirst = false; - } - WRITE(p, ";\n"); - } - - WRITE(p, "\n"); - - WRITE(p, "uniform float4 "I_COLORS"[4] : register(c%d);\n", C_COLORS); - WRITE(p, "uniform float4 "I_KCOLORS"[4] : register(c%d);\n", C_KCOLORS); - WRITE(p, "uniform float4 "I_ALPHA"[1] : register(c%d);\n", C_ALPHA); - WRITE(p, "uniform float4 "I_TEXDIMS"[8] : register(c%d);\n", C_TEXDIMS); - WRITE(p, "uniform float4 "I_ZBIAS"[2] : register(c%d);\n", C_ZBIAS); - WRITE(p, "uniform float4 "I_INDTEXSCALE"[2] : register(c%d);\n", C_INDTEXSCALE); - WRITE(p, "uniform float4 "I_INDTEXMTX"[6] : register(c%d);\n", C_INDTEXMTX); - WRITE(p, "uniform float4 "I_FOG"[3] : register(c%d);\n", C_FOG); - - if(g_ActiveConfig.bEnablePixelLighting && g_ActiveConfig.backend_info.bSupportsPixelLighting) - { - WRITE(p,"typedef struct { float4 col; float4 cosatt; float4 distatt; float4 pos; float4 dir; } Light;\n"); - WRITE(p,"typedef struct { Light lights[8]; } s_"I_PLIGHTS";\n"); - WRITE(p, "uniform s_"I_PLIGHTS" "I_PLIGHTS" : register(c%d);\n", C_PLIGHTS); - WRITE(p, "typedef struct { float4 C0, C1, C2, C3; } s_"I_PMATERIALS";\n"); - WRITE(p, "uniform s_"I_PMATERIALS" "I_PMATERIALS" : register(c%d);\n", C_PMATERIALS); - } - - WRITE(p, "void main(\n"); - if(ApiType != API_D3D11) - { - WRITE(p, " out float4 ocol0 : COLOR0,%s%s\n in float4 rawpos : %s,\n", - dstAlphaMode == DSTALPHA_DUAL_SOURCE_BLEND ? "\n out float4 ocol1 : COLOR1," : "", - DepthTextureEnable ? "\n out float depth : DEPTH," : "", - ApiType & API_OPENGL ? "WPOS" : ApiType & API_D3D9_SM20 ? "POSITION" : "VPOS"); - } - else - { - WRITE(p, " out float4 ocol0 : SV_Target0,%s%s\n in float4 rawpos : SV_Position,\n", - dstAlphaMode == DSTALPHA_DUAL_SOURCE_BLEND ? "\n out float4 ocol1 : SV_Target1," : "", - DepthTextureEnable ? "\n out float depth : SV_Depth," : ""); - } - - WRITE(p, " in float4 colors_0 : COLOR0,\n"); - WRITE(p, " in float4 colors_1 : COLOR1"); - - // compute window position if needed because binding semantic WPOS is not widely supported - if (numTexgen < 7) - { - for (int i = 0; i < numTexgen; ++i) - WRITE(p, ",\n in float3 uv%d : TEXCOORD%d", i, i); - WRITE(p, ",\n in float4 clipPos : TEXCOORD%d", numTexgen); - if(g_ActiveConfig.bEnablePixelLighting && g_ActiveConfig.backend_info.bSupportsPixelLighting) - WRITE(p, ",\n in float4 Normal : TEXCOORD%d", numTexgen + 1); - } - else - { - // wpos is in w of first 4 texcoords - if(g_ActiveConfig.bEnablePixelLighting && g_ActiveConfig.backend_info.bSupportsPixelLighting) - { - for (int i = 0; i < 8; ++i) - WRITE(p, ",\n in float4 uv%d : TEXCOORD%d", i, i); - } - else - { - for (unsigned int i = 0; i < xfregs.numTexGen.numTexGens; ++i) - WRITE(p, ",\n in float%d uv%d : TEXCOORD%d", i < 4 ? 4 : 3 , i, i); - } - } - WRITE(p, " ) {\n"); - - char* pmainstart = p; - int Pretest = AlphaPreTest(); - if (dstAlphaMode == DSTALPHA_ALPHA_PASS && !DepthTextureEnable && Pretest >= 0) - { - if (!Pretest) - { - // alpha test will always fail, so restart the shader and just make it an empty function - WRITE(p, "ocol0 = 0;\n"); - WRITE(p, "discard;\n"); - if(ApiType != API_D3D11) - WRITE(p, "return;\n"); - } - else - { - WRITE(p, " ocol0 = "I_ALPHA"[0].aaaa;\n"); - } - WRITE(p, "}\n"); - return text; - } - - WRITE(p, " float4 c0 = "I_COLORS"[1], c1 = "I_COLORS"[2], c2 = "I_COLORS"[3], prev = float4(0.0f, 0.0f, 0.0f, 0.0f), textemp = float4(0.0f, 0.0f, 0.0f, 0.0f), rastemp = float4(0.0f, 0.0f, 0.0f, 0.0f), konsttemp = float4(0.0f, 0.0f, 0.0f, 0.0f);\n" - " float3 comp16 = float3(1.0f, 255.0f, 0.0f), comp24 = float3(1.0f, 255.0f, 255.0f*255.0f);\n" - " float4 alphabump=0;\n" - " float3 tevcoord;\n" - " float2 wrappedcoord, tempcoord;\n" - " float4 cc0, cc1, cc2, cprev,crastemp,ckonsttemp;\n\n"); - - if(g_ActiveConfig.bEnablePixelLighting && g_ActiveConfig.backend_info.bSupportsPixelLighting) - { - if (xfregs.numTexGen.numTexGens < 7) - { - WRITE(p,"float3 _norm0 = normalize(Normal.xyz);\n\n"); - WRITE(p,"float3 pos = float3(clipPos.x,clipPos.y,Normal.w);\n"); - } - else - { - WRITE(p," float3 _norm0 = normalize(float3(uv4.w,uv5.w,uv6.w));\n\n"); - WRITE(p,"float3 pos = float3(uv0.w,uv1.w,uv7.w);\n"); - } - - - WRITE(p, "float4 mat, lacc;\n" - "float3 ldir, h;\n" - "float dist, dist2, attn;\n"); - - p = GenerateLightingShader(p, components, I_PMATERIALS, I_PLIGHTS, "colors_", "colors_"); - } - - if (numTexgen < 7) - WRITE(p, "clipPos = float4(rawpos.x, rawpos.y, clipPos.z, clipPos.w);\n"); - else - WRITE(p, "float4 clipPos = float4(rawpos.x, rawpos.y, uv2.w, uv3.w);\n"); - - // HACK to handle cases where the tex gen is not enabled - if (numTexgen == 0) - { - WRITE(p, "float3 uv0 = float3(0.0f, 0.0f, 0.0f);\n"); - } - else - { - for (int i = 0; i < numTexgen; ++i) - { - // optional perspective divides - if (xfregs.texMtxInfo[i].projection == XF_TEXPROJ_STQ) - { - WRITE(p, "if (uv%d.z)", i); - WRITE(p, " uv%d.xy = uv%d.xy / uv%d.z;\n", i, i, i); - } - - WRITE(p, "uv%d.xy = uv%d.xy * "I_TEXDIMS"[%d].zw;\n", i, i, i); - } - } - - // indirect texture map lookup - for(u32 i = 0; i < bpmem.genMode.numindstages; ++i) - { - if (nIndirectStagesUsed & (1< +#include +#include +#include + +#include "LightingShaderGen.h" +#include "PixelShaderGen.h" +#include "XFMemory.h" // for texture projection mode +#include "BPMemory.h" +#include "VideoConfig.h" +#include "NativeVertexFormat.h" + +PIXELSHADERUID last_pixel_shader_uid; + +static int AlphaPreTest(); + +static void StageHash(int stage, u32* out) +{ + out[0] |= bpmem.combiners[stage].colorC.hex & 0xFFFFFF; // 24 + u32 alphaC = bpmem.combiners[stage].alphaC.hex & 0xFFFFF0; // 24, strip out tswap and rswap for now + out[0] |= (alphaC&0xF0) << 24; // 8 + out[1] |= alphaC >> 8; // 16 + + // reserve 3 bits for bpmem.tevorders[stage/2].getTexMap + out[1] |= bpmem.tevorders[stage/2].getTexCoord(stage&1) << 19; // 3 + out[1] |= bpmem.tevorders[stage/2].getEnable(stage&1) << 22; // 1 + // reserve 3 bits for bpmem.tevorders[stage/2].getColorChan + + bool bHasIndStage = bpmem.tevind[stage].IsActive() && bpmem.tevind[stage].bt < bpmem.genMode.numindstages; + out[2] |= bHasIndStage << 2; // 1 + + bool needstexcoord = false; + + if (bHasIndStage) + { + out[2] |= (bpmem.tevind[stage].hex & 0x17FFFF) << 3; // 21, TODO: needs an explanation + needstexcoord = true; + } + + + TevStageCombiner::ColorCombiner& cc = bpmem.combiners[stage].colorC; + TevStageCombiner::AlphaCombiner& ac = bpmem.combiners[stage].alphaC; + + if(cc.a == TEVCOLORARG_RASA || cc.a == TEVCOLORARG_RASC + || cc.b == TEVCOLORARG_RASA || cc.b == TEVCOLORARG_RASC + || cc.c == TEVCOLORARG_RASA || cc.c == TEVCOLORARG_RASC + || cc.d == TEVCOLORARG_RASA || cc.d == TEVCOLORARG_RASC + || ac.a == TEVALPHAARG_RASA || ac.b == TEVALPHAARG_RASA + || ac.c == TEVALPHAARG_RASA || ac.d == TEVALPHAARG_RASA) + { + out[0] |= bpmem.combiners[stage].alphaC.rswap; + out[2] |= bpmem.tevksel[bpmem.combiners[stage].alphaC.rswap*2].swap1 << 24; // 2 + out[2] |= bpmem.tevksel[bpmem.combiners[stage].alphaC.rswap*2].swap2 << 26; // 2 + out[2] |= bpmem.tevksel[bpmem.combiners[stage].alphaC.rswap*2+1].swap1 << 28; // 2 + out[2] |= bpmem.tevksel[bpmem.combiners[stage].alphaC.rswap*2+1].swap2 << 30; // 2 + out[1] |= (bpmem.tevorders[stage/2].getColorChan(stage&1)&1) << 23; + out[2] |= (bpmem.tevorders[stage/2].getColorChan(stage&1)&0x6) >> 1; + } + + out[3] |= bpmem.tevorders[stage/2].getEnable(stage&1); + if (bpmem.tevorders[stage/2].getEnable(stage&1)) + { + if (bHasIndStage) needstexcoord = true; + + out[0] |= bpmem.combiners[stage].alphaC.tswap; + out[3] |= bpmem.tevksel[bpmem.combiners[stage].alphaC.tswap*2].swap1 << 1; // 2 + out[3] |= bpmem.tevksel[bpmem.combiners[stage].alphaC.tswap*2].swap2 << 3; // 2 + out[3] |= bpmem.tevksel[bpmem.combiners[stage].alphaC.tswap*2+1].swap1 << 5; // 2 + out[3] |= bpmem.tevksel[bpmem.combiners[stage].alphaC.tswap*2+1].swap2 << 7; // 2 + out[1] |= bpmem.tevorders[stage/2].getTexMap(stage&1) << 16; + } + + if (cc.a == TEVCOLORARG_KONST || cc.b == TEVCOLORARG_KONST || cc.c == TEVCOLORARG_KONST || cc.d == TEVCOLORARG_KONST + || ac.a == TEVALPHAARG_KONST || ac.b == TEVALPHAARG_KONST || ac.c == TEVALPHAARG_KONST || ac.d == TEVALPHAARG_KONST) + { + out[3] |= bpmem.tevksel[stage/2].getKC(stage&1) << 9; // 5 + out[3] |= bpmem.tevksel[stage/2].getKA(stage&1) << 14; // 5 + } + + if (needstexcoord) + { + out[1] |= bpmem.tevorders[stage/2].getTexCoord(stage&1) << 16; + } +} + +void GetPixelShaderId(PIXELSHADERUID *uid, DSTALPHA_MODE dstAlphaMode) +{ + uid->values[0] |= bpmem.genMode.numtevstages; // 4 + uid->values[0] |= bpmem.genMode.numtexgens << 4; // 4 + uid->values[0] |= dstAlphaMode << 8; // 2 + + uid->tevstages = bpmem.genMode.numtevstages; + + bool DepthTextureEnable = (bpmem.ztex2.op != ZTEXTURE_DISABLE && !bpmem.zcontrol.zcomploc && bpmem.zmode.testenable && bpmem.zmode.updateenable) || g_ActiveConfig.bEnablePerPixelDepth; + + uid->values[0] |= DepthTextureEnable << 10; // 1 + + bool enablePL = g_ActiveConfig.bEnablePixelLighting && g_ActiveConfig.backend_info.bSupportsPixelLighting; + uid->values[0] |= enablePL << 11; // 1 + + if (!enablePL) uid->values[0] |= xfregs.numTexGen.numTexGens << 12; // 4 + u32 alphaPreTest = AlphaPreTest()+1; + + uid->values[0] |= alphaPreTest << 16; // 2 + + if (alphaPreTest == 1 || (alphaPreTest && !DepthTextureEnable && dstAlphaMode == DSTALPHA_ALPHA_PASS)) + { + // Courtesy of PreAlphaTest, we're done already ;) + // TODO: There's a comment including bpmem.genmode.numindstages.. shouldnt really bother about that though. + return; + } + + if (enablePL) + { + // TODO: Include register states for lighting shader + } + + for (unsigned int i = 0; i < bpmem.genMode.numtexgens; ++i) + { + if (18+i < 32) + uid->values[0] |= xfregs.texMtxInfo[i].projection << (18+i); // 1 + else + uid->values[1] |= xfregs.texMtxInfo[i].projection << (i - 14); // 1 + } + + uid->indstages = bpmem.genMode.numindstages; + + uid->values[1] = bpmem.genMode.numindstages << 2; // 3 + u32 indirectStagesUsed = 0; + for (unsigned int i = 0; i < bpmem.genMode.numindstages; ++i) + if (bpmem.tevind[i].IsActive() && bpmem.tevind[i].bt < bpmem.genMode.numindstages) + indirectStagesUsed |= (1 << bpmem.tevind[i].bt); + + assert(indirectStagesUsed == (indirectStagesUsed & 0xF)); + + uid->values[1] |= indirectStagesUsed << 5; // 4; + + for (unsigned int i = 0; i < bpmem.genMode.numindstages; ++i) + { + if (indirectStagesUsed & (1 << i)) + { + uid->values[1] |= (bpmem.tevindref.getTexCoord(i) < bpmem.genMode.numtexgens) << (9 + 3*i); // 1 + if (bpmem.tevindref.getTexCoord(i) < bpmem.genMode.numtexgens) + uid->values[1] |= bpmem.tevindref.getTexCoord(i) << (10 + 3*i); // 2 + } + } + + u32* ptr = &uid->values[2]; + for (unsigned int i = 0; i < bpmem.genMode.numtevstages+1; ++i) + { + StageHash(i, ptr); + ptr += 4; // max: ptr = &uid->values[66] + } + + if (alphaPreTest == 0 || alphaPreTest == 2) + { + ptr[0] |= bpmem.fog.c_proj_fsel.fsel; // 3 + if (DepthTextureEnable) + { + ptr[0] |= bpmem.ztex2.op << 3; // 2 + ptr[0] |= bpmem.zcontrol.zcomploc << 5; // 1 + ptr[0] |= bpmem.zmode.testenable << 6; // 1 + ptr[0] |= bpmem.zmode.updateenable << 7; // 1 + } + } + + if (dstAlphaMode != DSTALPHA_ALPHA_PASS) + { + if (bpmem.fog.c_proj_fsel.fsel != 0) + { + ptr[0] |= bpmem.fog.c_proj_fsel.proj << 8; // 1 + ptr[0] |= bpmem.fogRange.Base.Enabled << 9; // 1 + } + } + uid->tevstages = (ptr+1) - uid->values; +} + +// Mash together all the inputs that contribute to the code of a generated pixel shader into +// 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, DSTALPHA_MODE dstAlphaMode) +{ + u32 numstages = bpmem.genMode.numtevstages + 1; + u32 projtexcoords = 0; + for (u32 i = 0; i < numstages; i++) + { + if (bpmem.tevorders[i/2].getEnable(i & 1)) + { + int texcoord = bpmem.tevorders[i / 2].getTexCoord(i & 1); + if (xfregs.texMtxInfo[i].projection) + projtexcoords |= 1 << texcoord; + } + } + uid->values[0] = (u32)bpmem.genMode.numtevstages | + ((u32)bpmem.genMode.numindstages << 4) | + ((u32)bpmem.genMode.numtexgens << 7) | + ((u32)dstAlphaMode << 11) | + ((u32)((bpmem.alphaFunc.hex >> 16) & 0xff) << 13) | + (projtexcoords << 21) | + ((u32)bpmem.ztex2.op << 29); + + // 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); + + u32 enableZTexture = (bpmem.ztex2.op != ZTEXTURE_DISABLE && !bpmem.zcontrol.zcomploc && bpmem.zmode.testenable && bpmem.zmode.updateenable) || g_ActiveConfig.bEnablePerPixelDepth ? 1 : 0; + + uid->values[2] = (u32)bpmem.fog.c_proj_fsel.fsel | + ((u32)bpmem.fog.c_proj_fsel.proj << 3) | + ((u32)enableZTexture << 4) | ((u32)bpmem.fogRange.Base.Enabled << 5); + + if(g_ActiveConfig.bEnablePixelLighting && g_ActiveConfig.backend_info.bSupportsPixelLighting) + { + for (int i = 0; i < 2; ++i) { + uid->values[3 + i] = xfregs.color[i].enablelighting ? + (u32)xfregs.color[i].hex : + (u32)xfregs.color[i].matsource; + uid->values[3 + i] |= (xfregs.alpha[i].enablelighting ? + (u32)xfregs.alpha[i].hex : + (u32)xfregs.alpha[i].matsource) << 15; + } + } + uid->values[4] |= (g_ActiveConfig.bEnablePixelLighting && g_ActiveConfig.backend_info.bSupportsPixelLighting) << 31; + + int hdr = 5; + u32 *pcurvalue = &uid->values[hdr]; + for (u32 i = 0; i < numstages; ++i) + { + TevStageCombiner::ColorCombiner &cc = bpmem.combiners[i].colorC; + TevStageCombiner::AlphaCombiner &ac = bpmem.combiners[i].alphaC; + + u32 val0 = cc.hex & 0xffffff; + u32 val1 = ac.hex & 0xffffff; + val0 |= bpmem.tevksel[i / 2].getKC(i & 1) << 24; + val1 |= bpmem.tevksel[i / 2].getKA(i & 1) << 24; + pcurvalue[0] = val0; + pcurvalue[1] = val1; + pcurvalue += 2; + } + + for (u32 i = 0; i < numstages / 2; ++i) + { + u32 val0, val1; + if (bpmem.tevorders[i].hex & 0x40) + val0 = bpmem.tevorders[i].hex & 0x3ff; + else + val0 = bpmem.tevorders[i].hex & 0x380; + if (bpmem.tevorders[i].hex & 0x40000) + val1 = (bpmem.tevorders[i].hex & 0x3ff000) >> 12; + else + val1 = (bpmem.tevorders[i].hex & 0x380000) >> 12; + + switch (i % 3) { + case 0: pcurvalue[0] = val0|(val1<<10); break; + case 1: pcurvalue[0] |= val0<<20; pcurvalue[1] = val1; pcurvalue++; break; + case 2: pcurvalue[1] |= (val0<<10)|(val1<<20); pcurvalue++; break; + default: PanicAlert("Unknown case for Tev Stages / 2: %08x", (i % 3)); + } + } + + if (numstages & 1) { // odd + u32 val0; + if (bpmem.tevorders[bpmem.genMode.numtevstages/2].hex & 0x40) + val0 = bpmem.tevorders[bpmem.genMode.numtevstages/2].hex & 0x3ff; + else + val0 = bpmem.tevorders[bpmem.genMode.numtevstages/2].hex & 0x380; + + switch (bpmem.genMode.numtevstages % 3) + { + case 0: pcurvalue[0] = val0; break; + case 1: pcurvalue[0] |= val0 << 20; break; + case 2: pcurvalue[1] |= val0 << 10; pcurvalue++; break; + default: PanicAlert("Unknown case for Tev Stages: %08x", bpmem.genMode.numtevstages % 3); + } + } + + if ((bpmem.genMode.numtevstages % 3) != 2) + ++pcurvalue; + + uid->tevstages = (u32)(pcurvalue - &uid->values[0] - hdr); + + for (u32 i = 0; i < bpmem.genMode.numindstages; ++i) + { + u32 val = bpmem.tevind[i].hex & 0x1fffff; // 21 bits + switch (i % 3) + { + case 0: pcurvalue[0] = val; break; + case 1: pcurvalue[0] |= val << 21; pcurvalue[1] = val >> 11; ++pcurvalue; break; + case 2: pcurvalue[0] |= val << 10; ++pcurvalue; break; + default: PanicAlert("Unknown case for Ind Stages: %08x", (i % 3)); + } + } + + // yeah, well .... + uid->indstages = (u32)(pcurvalue - &uid->values[0] - (hdr - 1) - uid->tevstages); + +} + +// old tev->pixelshader notes +// +// color for this stage (alpha, color) is given by bpmem.tevorders[0].colorchan0 +// konstant for this stage (alpha, color) is given by bpmem.tevksel +// inputs are given by bpmem.combiners[0].colorC.a/b/c/d << could be current chan color +// according to GXTevColorArg table above +// output is given by .outreg +// tevtemp is set according to swapmodetables and + +static void WriteStage(char *&p, int n, API_TYPE ApiType); +static void SampleTexture(char *&p, const char *destination, const char *texcoords, const char *texswap, int texmap, API_TYPE ApiType); +// static void WriteAlphaCompare(char *&p, int num, int comp); +static bool WriteAlphaTest(char *&p, API_TYPE ApiType,DSTALPHA_MODE dstAlphaMode); +static void WriteFog(char *&p); + +static const char *tevKSelTableC[] = // KCSEL +{ + "1.0f,1.0f,1.0f", // 1 = 0x00 + "0.875f,0.875f,0.875f", // 7_8 = 0x01 + "0.75f,0.75f,0.75f", // 3_4 = 0x02 + "0.625f,0.625f,0.625f", // 5_8 = 0x03 + "0.5f,0.5f,0.5f", // 1_2 = 0x04 + "0.375f,0.375f,0.375f", // 3_8 = 0x05 + "0.25f,0.25f,0.25f", // 1_4 = 0x06 + "0.125f,0.125f,0.125f", // 1_8 = 0x07 + "ERROR", // 0x08 + "ERROR", // 0x09 + "ERROR", // 0x0a + "ERROR", // 0x0b + I_KCOLORS"[0].rgb", // K0 = 0x0C + I_KCOLORS"[1].rgb", // K1 = 0x0D + I_KCOLORS"[2].rgb", // K2 = 0x0E + I_KCOLORS"[3].rgb", // K3 = 0x0F + I_KCOLORS"[0].rrr", // K0_R = 0x10 + I_KCOLORS"[1].rrr", // K1_R = 0x11 + I_KCOLORS"[2].rrr", // K2_R = 0x12 + I_KCOLORS"[3].rrr", // K3_R = 0x13 + I_KCOLORS"[0].ggg", // K0_G = 0x14 + I_KCOLORS"[1].ggg", // K1_G = 0x15 + I_KCOLORS"[2].ggg", // K2_G = 0x16 + I_KCOLORS"[3].ggg", // K3_G = 0x17 + I_KCOLORS"[0].bbb", // K0_B = 0x18 + I_KCOLORS"[1].bbb", // K1_B = 0x19 + I_KCOLORS"[2].bbb", // K2_B = 0x1A + I_KCOLORS"[3].bbb", // K3_B = 0x1B + I_KCOLORS"[0].aaa", // K0_A = 0x1C + I_KCOLORS"[1].aaa", // K1_A = 0x1D + I_KCOLORS"[2].aaa", // K2_A = 0x1E + I_KCOLORS"[3].aaa", // K3_A = 0x1F +}; + +static const char *tevKSelTableA[] = // KASEL +{ + "1.0f", // 1 = 0x00 + "0.875f",// 7_8 = 0x01 + "0.75f", // 3_4 = 0x02 + "0.625f",// 5_8 = 0x03 + "0.5f", // 1_2 = 0x04 + "0.375f",// 3_8 = 0x05 + "0.25f", // 1_4 = 0x06 + "0.125f",// 1_8 = 0x07 + "ERROR", // 0x08 + "ERROR", // 0x09 + "ERROR", // 0x0a + "ERROR", // 0x0b + "ERROR", // 0x0c + "ERROR", // 0x0d + "ERROR", // 0x0e + "ERROR", // 0x0f + I_KCOLORS"[0].r", // K0_R = 0x10 + I_KCOLORS"[1].r", // K1_R = 0x11 + I_KCOLORS"[2].r", // K2_R = 0x12 + I_KCOLORS"[3].r", // K3_R = 0x13 + I_KCOLORS"[0].g", // K0_G = 0x14 + I_KCOLORS"[1].g", // K1_G = 0x15 + I_KCOLORS"[2].g", // K2_G = 0x16 + I_KCOLORS"[3].g", // K3_G = 0x17 + I_KCOLORS"[0].b", // K0_B = 0x18 + I_KCOLORS"[1].b", // K1_B = 0x19 + I_KCOLORS"[2].b", // K2_B = 0x1A + I_KCOLORS"[3].b", // K3_B = 0x1B + I_KCOLORS"[0].a", // K0_A = 0x1C + I_KCOLORS"[1].a", // K1_A = 0x1D + I_KCOLORS"[2].a", // K2_A = 0x1E + I_KCOLORS"[3].a", // K3_A = 0x1F +}; + +static const char *tevScaleTable[] = // CS +{ + "1.0f", // SCALE_1 + "2.0f", // SCALE_2 + "4.0f", // SCALE_4 + "0.5f", // DIVIDE_2 +}; + +static const char *tevBiasTable[] = // TB +{ + "", // ZERO, + "+0.5f", // ADDHALF, + "-0.5f", // SUBHALF, + "", +}; + +static const char *tevOpTable[] = { // TEV + "+", // TEVOP_ADD = 0, + "-", // TEVOP_SUB = 1, +}; + +static const char *tevCInputTable[] = // CC +{ + "(prev.rgb)", // CPREV, + "(prev.aaa)", // APREV, + "(c0.rgb)", // C0, + "(c0.aaa)", // A0, + "(c1.rgb)", // C1, + "(c1.aaa)", // A1, + "(c2.rgb)", // C2, + "(c2.aaa)", // A2, + "(textemp.rgb)", // TEXC, + "(textemp.aaa)", // TEXA, + "(rastemp.rgb)", // RASC, + "(rastemp.aaa)", // RASA, + "float3(1.0f, 1.0f, 1.0f)", // ONE + "float3(0.5f, 0.5f, 0.5f)", // HALF + "(konsttemp.rgb)", //"konsttemp.rgb", // KONST + "float3(0.0f, 0.0f, 0.0f)", // ZERO + ///aded extra values to map clamped values + "(cprev.rgb)", // CPREV, + "(cprev.aaa)", // APREV, + "(cc0.rgb)", // C0, + "(cc0.aaa)", // A0, + "(cc1.rgb)", // C1, + "(cc1.aaa)", // A1, + "(cc2.rgb)", // C2, + "(cc2.aaa)", // A2, + "(textemp.rgb)", // TEXC, + "(textemp.aaa)", // TEXA, + "(crastemp.rgb)", // RASC, + "(crastemp.aaa)", // RASA, + "float3(1.0f, 1.0f, 1.0f)", // ONE + "float3(0.5f, 0.5f, 0.5f)", // HALF + "(ckonsttemp.rgb)", //"konsttemp.rgb", // KONST + "float3(0.0f, 0.0f, 0.0f)", // ZERO + "PADERROR", "PADERROR", "PADERROR", "PADERROR" +}; + +static const char *tevAInputTable[] = // CA +{ + "prev", // APREV, + "c0", // A0, + "c1", // A1, + "c2", // A2, + "textemp", // TEXA, + "rastemp", // RASA, + "konsttemp", // KONST, (hw1 had quarter) + "float4(0.0f, 0.0f, 0.0f, 0.0f)", // ZERO + ///aded extra values to map clamped values + "cprev", // APREV, + "cc0", // A0, + "cc1", // A1, + "cc2", // A2, + "textemp", // TEXA, + "crastemp", // RASA, + "ckonsttemp", // KONST, (hw1 had quarter) + "float4(0.0f, 0.0f, 0.0f, 0.0f)", // ZERO + "PADERROR", "PADERROR", "PADERROR", "PADERROR", + "PADERROR", "PADERROR", "PADERROR", "PADERROR", +}; + +static const char *tevRasTable[] = +{ + "colors_0", + "colors_1", + "ERROR", //2 + "ERROR", //3 + "ERROR", //4 + "alphabump", // use bump alpha + "(alphabump*(255.0f/248.0f))", //normalized + "float4(0.0f, 0.0f, 0.0f, 0.0f)", // zero +}; + +//static const char *tevTexFunc[] = { "tex2D", "texRECT" }; + +static const char *tevCOutputTable[] = { "prev.rgb", "c0.rgb", "c1.rgb", "c2.rgb" }; +static const char *tevAOutputTable[] = { "prev.a", "c0.a", "c1.a", "c2.a" }; +static const char *tevIndAlphaSel[] = {"", "x", "y", "z"}; +//static const char *tevIndAlphaScale[] = {"", "*32", "*16", "*8"}; +static const char *tevIndAlphaScale[] = {"*(248.0f/255.0f)", "*(224.0f/255.0f)", "*(240.0f/255.0f)", "*(248.0f/255.0f)"}; +static const char *tevIndBiasField[] = {"", "x", "y", "xy", "z", "xz", "yz", "xyz"}; // indexed by bias +static const char *tevIndBiasAdd[] = {"-128.0f", "1.0f", "1.0f", "1.0f" }; // indexed by fmt +static const char *tevIndWrapStart[] = {"0.0f", "256.0f", "128.0f", "64.0f", "32.0f", "16.0f", "0.001f" }; +static const char *tevIndFmtScale[] = {"255.0f", "31.0f", "15.0f", "7.0f" }; + +#define WRITE p+=sprintf + +static char swapModeTable[4][5]; + +static char text[16384]; +static bool DepthTextureEnable; + +static void BuildSwapModeTable() +{ + static const char *swapColors = "rgba"; + for (int i = 0; i < 4; i++) + { + swapModeTable[i][0] = swapColors[bpmem.tevksel[i*2].swap1]; + swapModeTable[i][1] = swapColors[bpmem.tevksel[i*2].swap2]; + swapModeTable[i][2] = swapColors[bpmem.tevksel[i*2+1].swap1]; + swapModeTable[i][3] = swapColors[bpmem.tevksel[i*2+1].swap2]; + swapModeTable[i][4] = 0; + } +} + +const char *GeneratePixelShaderCode(DSTALPHA_MODE dstAlphaMode, API_TYPE ApiType, u32 components) +{ + setlocale(LC_NUMERIC, "C"); // Reset locale for compilation + text[sizeof(text) - 1] = 0x7C; // canary + + BuildSwapModeTable(); // Needed for WriteStage + int numStages = bpmem.genMode.numtevstages + 1; + int numTexgen = bpmem.genMode.numtexgens; + + char *p = text; + WRITE(p, "//Pixel Shader for TEV stages\n"); + WRITE(p, "//%i TEV stages, %i texgens, %i IND stages\n", + numStages, numTexgen, bpmem.genMode.numindstages); + + int nIndirectStagesUsed = 0; + if (bpmem.genMode.numindstages > 0) + { + for (int i = 0; i < numStages; ++i) + { + if (bpmem.tevind[i].IsActive() && bpmem.tevind[i].bt < bpmem.genMode.numindstages) + nIndirectStagesUsed |= 1 << bpmem.tevind[i].bt; + } + } + DepthTextureEnable = (bpmem.ztex2.op != ZTEXTURE_DISABLE && !bpmem.zcontrol.zcomploc && bpmem.zmode.testenable && bpmem.zmode.updateenable) || g_ActiveConfig.bEnablePerPixelDepth ; + // Declare samplers + + if(ApiType != API_D3D11) + { + WRITE(p, "uniform sampler2D "); + } + else + { + WRITE(p, "sampler "); + } + + bool bfirst = true; + for (int i = 0; i < 8; ++i) + { + WRITE(p, "%s samp%d : register(s%d)", bfirst?"":",", i, i); + bfirst = false; + } + WRITE(p, ";\n"); + if(ApiType == API_D3D11) + { + WRITE(p, "Texture2D "); + bfirst = true; + for (int i = 0; i < 8; ++i) + { + WRITE(p, "%s Tex%d : register(t%d)", bfirst?"":",", i, i); + bfirst = false; + } + WRITE(p, ";\n"); + } + + WRITE(p, "\n"); + + WRITE(p, "uniform float4 "I_COLORS"[4] : register(c%d);\n", C_COLORS); + WRITE(p, "uniform float4 "I_KCOLORS"[4] : register(c%d);\n", C_KCOLORS); + WRITE(p, "uniform float4 "I_ALPHA"[1] : register(c%d);\n", C_ALPHA); + WRITE(p, "uniform float4 "I_TEXDIMS"[8] : register(c%d);\n", C_TEXDIMS); + WRITE(p, "uniform float4 "I_ZBIAS"[2] : register(c%d);\n", C_ZBIAS); + WRITE(p, "uniform float4 "I_INDTEXSCALE"[2] : register(c%d);\n", C_INDTEXSCALE); + WRITE(p, "uniform float4 "I_INDTEXMTX"[6] : register(c%d);\n", C_INDTEXMTX); + WRITE(p, "uniform float4 "I_FOG"[3] : register(c%d);\n", C_FOG); + + if(g_ActiveConfig.bEnablePixelLighting && g_ActiveConfig.backend_info.bSupportsPixelLighting) + { + WRITE(p,"typedef struct { float4 col; float4 cosatt; float4 distatt; float4 pos; float4 dir; } Light;\n"); + WRITE(p,"typedef struct { Light lights[8]; } s_"I_PLIGHTS";\n"); + WRITE(p, "uniform s_"I_PLIGHTS" "I_PLIGHTS" : register(c%d);\n", C_PLIGHTS); + WRITE(p, "typedef struct { float4 C0, C1, C2, C3; } s_"I_PMATERIALS";\n"); + WRITE(p, "uniform s_"I_PMATERIALS" "I_PMATERIALS" : register(c%d);\n", C_PMATERIALS); + } + + WRITE(p, "void main(\n"); + if(ApiType != API_D3D11) + { + WRITE(p, " out float4 ocol0 : COLOR0,%s%s\n in float4 rawpos : %s,\n", + dstAlphaMode == DSTALPHA_DUAL_SOURCE_BLEND ? "\n out float4 ocol1 : COLOR1," : "", + DepthTextureEnable ? "\n out float depth : DEPTH," : "", + ApiType & API_OPENGL ? "WPOS" : ApiType & API_D3D9_SM20 ? "POSITION" : "VPOS"); + } + else + { + WRITE(p, " out float4 ocol0 : SV_Target0,%s%s\n in float4 rawpos : SV_Position,\n", + dstAlphaMode == DSTALPHA_DUAL_SOURCE_BLEND ? "\n out float4 ocol1 : SV_Target1," : "", + DepthTextureEnable ? "\n out float depth : SV_Depth," : ""); + } + + WRITE(p, " in float4 colors_0 : COLOR0,\n"); + WRITE(p, " in float4 colors_1 : COLOR1"); + + // compute window position if needed because binding semantic WPOS is not widely supported + if (numTexgen < 7) + { + for (int i = 0; i < numTexgen; ++i) + WRITE(p, ",\n in float3 uv%d : TEXCOORD%d", i, i); + WRITE(p, ",\n in float4 clipPos : TEXCOORD%d", numTexgen); + if(g_ActiveConfig.bEnablePixelLighting && g_ActiveConfig.backend_info.bSupportsPixelLighting) + WRITE(p, ",\n in float4 Normal : TEXCOORD%d", numTexgen + 1); + } + else + { + // wpos is in w of first 4 texcoords + if(g_ActiveConfig.bEnablePixelLighting && g_ActiveConfig.backend_info.bSupportsPixelLighting) + { + for (int i = 0; i < 8; ++i) + WRITE(p, ",\n in float4 uv%d : TEXCOORD%d", i, i); + } + else + { + for (unsigned int i = 0; i < xfregs.numTexGen.numTexGens; ++i) + WRITE(p, ",\n in float%d uv%d : TEXCOORD%d", i < 4 ? 4 : 3 , i, i); + } + } + WRITE(p, " ) {\n"); + + char* pmainstart = p; + int Pretest = AlphaPreTest(); + if (dstAlphaMode == DSTALPHA_ALPHA_PASS && !DepthTextureEnable && Pretest >= 0) + { + if (!Pretest) + { + // alpha test will always fail, so restart the shader and just make it an empty function + WRITE(p, "ocol0 = 0;\n"); + WRITE(p, "discard;\n"); + if(ApiType != API_D3D11) + WRITE(p, "return;\n"); + } + else + { + WRITE(p, " ocol0 = "I_ALPHA"[0].aaaa;\n"); + } + WRITE(p, "}\n"); + return text; + } + + WRITE(p, " float4 c0 = "I_COLORS"[1], c1 = "I_COLORS"[2], c2 = "I_COLORS"[3], prev = float4(0.0f, 0.0f, 0.0f, 0.0f), textemp = float4(0.0f, 0.0f, 0.0f, 0.0f), rastemp = float4(0.0f, 0.0f, 0.0f, 0.0f), konsttemp = float4(0.0f, 0.0f, 0.0f, 0.0f);\n" + " float3 comp16 = float3(1.0f, 255.0f, 0.0f), comp24 = float3(1.0f, 255.0f, 255.0f*255.0f);\n" + " float4 alphabump=0;\n" + " float3 tevcoord;\n" + " float2 wrappedcoord, tempcoord;\n" + " float4 cc0, cc1, cc2, cprev,crastemp,ckonsttemp;\n\n"); + + if(g_ActiveConfig.bEnablePixelLighting && g_ActiveConfig.backend_info.bSupportsPixelLighting) + { + if (xfregs.numTexGen.numTexGens < 7) + { + WRITE(p,"float3 _norm0 = normalize(Normal.xyz);\n\n"); + WRITE(p,"float3 pos = float3(clipPos.x,clipPos.y,Normal.w);\n"); + } + else + { + WRITE(p," float3 _norm0 = normalize(float3(uv4.w,uv5.w,uv6.w));\n\n"); + WRITE(p,"float3 pos = float3(uv0.w,uv1.w,uv7.w);\n"); + } + + + WRITE(p, "float4 mat, lacc;\n" + "float3 ldir, h;\n" + "float dist, dist2, attn;\n"); + + p = GenerateLightingShader(p, components, I_PMATERIALS, I_PLIGHTS, "colors_", "colors_"); + } + + if (numTexgen < 7) + WRITE(p, "clipPos = float4(rawpos.x, rawpos.y, clipPos.z, clipPos.w);\n"); + else + WRITE(p, "float4 clipPos = float4(rawpos.x, rawpos.y, uv2.w, uv3.w);\n"); + + // HACK to handle cases where the tex gen is not enabled + if (numTexgen == 0) + { + WRITE(p, "float3 uv0 = float3(0.0f, 0.0f, 0.0f);\n"); + } + else + { + for (int i = 0; i < numTexgen; ++i) + { + // optional perspective divides + if (xfregs.texMtxInfo[i].projection == XF_TEXPROJ_STQ) + { + WRITE(p, "if (uv%d.z)", i); + WRITE(p, " uv%d.xy = uv%d.xy / uv%d.z;\n", i, i, i); + } + + WRITE(p, "uv%d.xy = uv%d.xy * "I_TEXDIMS"[%d].zw;\n", i, i, i); + } + } + + // indirect texture map lookup + for(u32 i = 0; i < bpmem.genMode.numindstages; ++i) + { + if (nIndirectStagesUsed & (1<= %s.r + (0.25f/255.0f)) ? %s : float3(0.0f, 0.0f, 0.0f))",//#define TEVCMP_R8_GT 8 - " %s + ((abs(%s.r - %s.r) < (0.5f/255.0f)) ? %s : float3(0.0f, 0.0f, 0.0f))",//#define TEVCMP_R8_EQ 9 - " %s + (( dot(%s.rgb, comp16) >= (dot(%s.rgb, comp16) + (0.25f/255.0f))) ? %s : float3(0.0f, 0.0f, 0.0f))",//#define TEVCMP_GR16_GT 10 - " %s + (abs(dot(%s.rgb, comp16) - dot(%s.rgb, comp16)) < (0.5f/255.0f) ? %s : float3(0.0f, 0.0f, 0.0f))",//#define TEVCMP_GR16_EQ 11 - " %s + (( dot(%s.rgb, comp24) >= (dot(%s.rgb, comp24) + (0.25f/255.0f))) ? %s : float3(0.0f, 0.0f, 0.0f))",//#define TEVCMP_BGR24_GT 12 - " %s + (abs(dot(%s.rgb, comp24) - dot(%s.rgb, comp24)) < (0.5f/255.0f) ? %s : float3(0.0f, 0.0f, 0.0f))",//#define TEVCMP_BGR24_EQ 13 - " %s + (max(sign(%s.rgb - %s.rgb - (0.25f/255.0f)), float3(0.0f, 0.0f, 0.0f)) * %s)",//#define TEVCMP_RGB8_GT 14 - " %s + ((float3(1.0f, 1.0f, 1.0f) - max(sign(abs(%s.rgb - %s.rgb) - (0.5f/255.0f)), float3(0.0f, 0.0f, 0.0f))) * %s)"//#define TEVCMP_RGB8_EQ 15 -}; - -//table with the alpha compare operations -static const char *TEVCMPAlphaOPTable[16] = -{ - "0.0f",//0 - "0.0f",//1 - "0.0f",//2 - "0.0f",//3 - "0.0f",//4 - "0.0f",//5 - "0.0f",//6 - "0.0f",//7 - " %s.a + ((%s.r >= (%s.r + (0.25f/255.0f))) ? %s.a : 0.0f)",//#define TEVCMP_R8_GT 8 - " %s.a + (abs(%s.r - %s.r) < (0.5f/255.0f) ? %s.a : 0.0f)",//#define TEVCMP_R8_EQ 9 - " %s.a + ((dot(%s.rgb, comp16) >= (dot(%s.rgb, comp16) + (0.25f/255.0f))) ? %s.a : 0.0f)",//#define TEVCMP_GR16_GT 10 - " %s.a + (abs(dot(%s.rgb, comp16) - dot(%s.rgb, comp16)) < (0.5f/255.0f) ? %s.a : 0.0f)",//#define TEVCMP_GR16_EQ 11 - " %s.a + ((dot(%s.rgb, comp24) >= (dot(%s.rgb, comp24) + (0.25f/255.0f))) ? %s.a : 0.0f)",//#define TEVCMP_BGR24_GT 12 - " %s.a + (abs(dot(%s.rgb, comp24) - dot(%s.rgb, comp24)) < (0.5f/255.0f) ? %s.a : 0.0f)",//#define TEVCMP_BGR24_EQ 13 - " %s.a + ((%s.a >= (%s.a + (0.25f/255.0f))) ? %s.a : 0.0f)",//#define TEVCMP_A8_GT 14 - " %s.a + (abs(%s.a - %s.a) < (0.5f/255.0f) ? %s.a : 0.0f)"//#define TEVCMP_A8_EQ 15 - -}; - - -static void WriteStage(char *&p, int n, API_TYPE ApiType) -{ - int texcoord = bpmem.tevorders[n/2].getTexCoord(n&1); - bool bHasTexCoord = (u32)texcoord < bpmem.genMode.numtexgens; - bool bHasIndStage = bpmem.tevind[n].IsActive() && bpmem.tevind[n].bt < bpmem.genMode.numindstages; - - // HACK to handle cases where the tex gen is not enabled - if (!bHasTexCoord) - texcoord = 0; - - if (bHasIndStage) - { - // perform the indirect op on the incoming regular coordinates using indtex%d as the offset coords - if (bpmem.tevind[n].bs != ITBA_OFF) - { - WRITE(p, "alphabump = indtex%d.%s %s;\n", - bpmem.tevind[n].bt, - tevIndAlphaSel[bpmem.tevind[n].bs], - tevIndAlphaScale[bpmem.tevind[n].fmt]); - } - // format - WRITE(p, "float3 indtevcrd%d = indtex%d * %s;\n", n, bpmem.tevind[n].bt, tevIndFmtScale[bpmem.tevind[n].fmt]); - - // bias - if (bpmem.tevind[n].bias != ITB_NONE ) - WRITE(p, "indtevcrd%d.%s += %s;\n", n, tevIndBiasField[bpmem.tevind[n].bias], tevIndBiasAdd[bpmem.tevind[n].fmt]); - - // multiply by offset matrix and scale - if (bpmem.tevind[n].mid != 0) - { - if (bpmem.tevind[n].mid <= 3) - { - int mtxidx = 2*(bpmem.tevind[n].mid-1); - WRITE(p, "float2 indtevtrans%d = float2(dot("I_INDTEXMTX"[%d].xyz, indtevcrd%d), dot("I_INDTEXMTX"[%d].xyz, indtevcrd%d));\n", - n, mtxidx, n, mtxidx+1, n); - } - else if (bpmem.tevind[n].mid <= 7 && bHasTexCoord) - { // s matrix - // TODO: Might become negative? - int mtxidx = 2*(bpmem.tevind[n].mid-5); - WRITE(p, "float2 indtevtrans%d = "I_INDTEXMTX"[%d].ww * uv%d.xy * indtevcrd%d.xx;\n", n, mtxidx, texcoord, n); - } - else if (bpmem.tevind[n].mid <= 11 && bHasTexCoord) - { // t matrix - // TODO: Might become negative? - int mtxidx = 2*(bpmem.tevind[n].mid-9); - WRITE(p, "float2 indtevtrans%d = "I_INDTEXMTX"[%d].ww * uv%d.xy * indtevcrd%d.yy;\n", n, mtxidx, texcoord, n); - } - else - WRITE(p, "float2 indtevtrans%d = 0;\n", n); - } - else - WRITE(p, "float2 indtevtrans%d = 0;\n", n); - - // --------- - // Wrapping - // --------- - - // wrap S - if (bpmem.tevind[n].sw == ITW_OFF) - WRITE(p, "wrappedcoord.x = uv%d.x;\n", texcoord); - else if (bpmem.tevind[n].sw == ITW_0) - WRITE(p, "wrappedcoord.x = 0.0f;\n"); - else - WRITE(p, "wrappedcoord.x = fmod( uv%d.x, %s );\n", texcoord, tevIndWrapStart[bpmem.tevind[n].sw]); - - // wrap T - if (bpmem.tevind[n].tw == ITW_OFF) - WRITE(p, "wrappedcoord.y = uv%d.y;\n", texcoord); - else if (bpmem.tevind[n].tw == ITW_0) - WRITE(p, "wrappedcoord.y = 0.0f;\n"); - else - WRITE(p, "wrappedcoord.y = fmod( uv%d.y, %s );\n", texcoord, tevIndWrapStart[bpmem.tevind[n].tw]); - - if (bpmem.tevind[n].fb_addprev) // add previous tevcoord - WRITE(p, "tevcoord.xy += wrappedcoord + indtevtrans%d;\n", n); - else - WRITE(p, "tevcoord.xy = wrappedcoord + indtevtrans%d;\n", n); - } - - TevStageCombiner::ColorCombiner &cc = bpmem.combiners[n].colorC; - TevStageCombiner::AlphaCombiner &ac = bpmem.combiners[n].alphaC; - - if(cc.a == TEVCOLORARG_RASA || cc.a == TEVCOLORARG_RASC - || cc.b == TEVCOLORARG_RASA || cc.b == TEVCOLORARG_RASC - || cc.c == TEVCOLORARG_RASA || cc.c == TEVCOLORARG_RASC - || cc.d == TEVCOLORARG_RASA || cc.d == TEVCOLORARG_RASC - || ac.a == TEVALPHAARG_RASA || ac.b == TEVALPHAARG_RASA - || ac.c == TEVALPHAARG_RASA || ac.d == TEVALPHAARG_RASA) - { - char *rasswap = swapModeTable[bpmem.combiners[n].alphaC.rswap]; - WRITE(p, "rastemp = %s.%s;\n", tevRasTable[bpmem.tevorders[n / 2].getColorChan(n & 1)], rasswap); - WRITE(p, "crastemp = frac(rastemp * (255.0f/256.0f)) * (256.0f/255.0f);\n"); - } - - - if (bpmem.tevorders[n/2].getEnable(n&1)) - { - if(!bHasIndStage) - { - // calc tevcord - if(bHasTexCoord) - WRITE(p, "tevcoord.xy = uv%d.xy;\n", texcoord); - else - WRITE(p, "tevcoord.xy = float2(0.0f, 0.0f);\n"); - } - - char *texswap = swapModeTable[bpmem.combiners[n].alphaC.tswap]; - int texmap = bpmem.tevorders[n/2].getTexMap(n&1); - SampleTexture(p, "textemp", "tevcoord", texswap, texmap, ApiType); - } - else - WRITE(p, "textemp = float4(1.0f, 1.0f, 1.0f, 1.0f);\n"); - - - if (cc.a == TEVCOLORARG_KONST || cc.b == TEVCOLORARG_KONST || cc.c == TEVCOLORARG_KONST || cc.d == TEVCOLORARG_KONST - || ac.a == TEVALPHAARG_KONST || ac.b == TEVALPHAARG_KONST || ac.c == TEVALPHAARG_KONST || ac.d == TEVALPHAARG_KONST) - { - int kc = bpmem.tevksel[n / 2].getKC(n & 1); - int ka = bpmem.tevksel[n / 2].getKA(n & 1); - WRITE(p, "konsttemp = float4(%s, %s);\n", tevKSelTableC[kc], tevKSelTableA[ka]); - if(kc > 7 || ka > 7) - { - WRITE(p, "ckonsttemp = frac(konsttemp * (255.0f/256.0f)) * (256.0f/255.0f);\n"); - } - else - { - WRITE(p, "ckonsttemp = konsttemp;\n"); - } - } - - if(cc.a == TEVCOLORARG_CPREV || cc.a == TEVCOLORARG_APREV - || cc.b == TEVCOLORARG_CPREV || cc.b == TEVCOLORARG_APREV - || cc.c == TEVCOLORARG_CPREV || cc.c == TEVCOLORARG_APREV - || ac.a == TEVALPHAARG_APREV || ac.b == TEVALPHAARG_APREV || ac.c == TEVALPHAARG_APREV) - WRITE(p, "cprev = frac(prev * (255.0f/256.0f)) * (256.0f/255.0f);\n"); - - - if(cc.a == TEVCOLORARG_C0 || cc.a == TEVCOLORARG_A0 - || cc.b == TEVCOLORARG_C0 || cc.b == TEVCOLORARG_A0 - || cc.c == TEVCOLORARG_C0 || cc.c == TEVCOLORARG_A0 - || ac.a == TEVALPHAARG_A0 || ac.b == TEVALPHAARG_A0 || ac.c == TEVALPHAARG_A0) - WRITE(p, "cc0 = frac(c0 * (255.0f/256.0f)) * (256.0f/255.0f);\n"); - - - if(cc.a == TEVCOLORARG_C1 || cc.a == TEVCOLORARG_A1 - || cc.b == TEVCOLORARG_C1 || cc.b == TEVCOLORARG_A1 - || cc.c == TEVCOLORARG_C1 || cc.c == TEVCOLORARG_A1 - || ac.a == TEVALPHAARG_A1 || ac.b == TEVALPHAARG_A1 || ac.c == TEVALPHAARG_A1) - WRITE(p, "cc1 = frac(c1 * (255.0f/256.0f)) * (256.0f/255.0f);\n"); - - - if(cc.a == TEVCOLORARG_C2 || cc.a == TEVCOLORARG_A2 - || cc.b == TEVCOLORARG_C2 || cc.b == TEVCOLORARG_A2 - || cc.c == TEVCOLORARG_C2 || cc.c == TEVCOLORARG_A2 - || ac.a == TEVALPHAARG_A2 || ac.b == TEVALPHAARG_A2 || ac.c == TEVALPHAARG_A2) - WRITE(p, "cc2 = frac(c2 * (255.0f/256.0f)) * (256.0f/255.0f);\n"); - - - if (cc.clamp) - WRITE(p, "%s = saturate(", tevCOutputTable[cc.dest]); - else - WRITE(p, "%s = ", tevCOutputTable[cc.dest]); - - // combine the color channel - if (cc.bias != TevBias_COMPARE) // if not compare - { - //normal color combiner goes here - if (cc.shift > TEVSCALE_1) - WRITE(p, "%s*(", tevScaleTable[cc.shift]); - - if(!(cc.d == TEVCOLORARG_ZERO && cc.op == TEVOP_ADD)) - WRITE(p, "%s%s", tevCInputTable[cc.d], tevOpTable[cc.op]); - - if (cc.a == cc.b) - WRITE(p, "%s", tevCInputTable[cc.a + 16]); - else if (cc.c == TEVCOLORARG_ZERO) - WRITE(p, "%s", tevCInputTable[cc.a + 16]); - else if (cc.c == TEVCOLORARG_ONE) - WRITE(p, "%s", tevCInputTable[cc.b + 16]); - else if (cc.a == TEVCOLORARG_ZERO) - WRITE(p, "%s*%s", tevCInputTable[cc.b + 16], tevCInputTable[cc.c + 16]); - else if (cc.b == TEVCOLORARG_ZERO) - WRITE(p, "%s*(float3(1.0f, 1.0f, 1.0f)-%s)", tevCInputTable[cc.a + 16], tevCInputTable[cc.c + 16]); - else - WRITE(p, "lerp(%s, %s, %s)", tevCInputTable[cc.a + 16], tevCInputTable[cc.b + 16], tevCInputTable[cc.c + 16]); - - WRITE(p, "%s", tevBiasTable[cc.bias]); - - if (cc.shift > TEVSCALE_1) - WRITE(p, ")"); - } - else - { - int cmp = (cc.shift<<1)|cc.op|8; // comparemode stored here - WRITE(p, TEVCMPColorOPTable[cmp],//lookup the function from the op table - tevCInputTable[cc.d], - tevCInputTable[cc.a + 16], - tevCInputTable[cc.b + 16], - tevCInputTable[cc.c + 16]); - } - if (cc.clamp) - WRITE(p, ")"); - WRITE(p,";\n"); - - // combine the alpha channel - if (ac.clamp) - WRITE(p, "%s = saturate(", tevAOutputTable[ac.dest]); - else - WRITE(p, "%s = ", tevAOutputTable[ac.dest]); - - if (ac.bias != TevBias_COMPARE) // if not compare - { - //normal alpha combiner goes here - if (ac.shift > TEVSCALE_1) - WRITE(p, "%s*(", tevScaleTable[ac.shift]); - - if(!(ac.d == TEVALPHAARG_ZERO && ac.op == TEVOP_ADD)) - WRITE(p, "%s.a%s", tevAInputTable[ac.d], tevOpTable[ac.op]); - - if (ac.a == ac.b) - WRITE(p, "%s.a", tevAInputTable[ac.a + 8]); - else if (ac.c == TEVALPHAARG_ZERO) - WRITE(p, "%s.a", tevAInputTable[ac.a + 8]); - else if (ac.a == TEVALPHAARG_ZERO) - WRITE(p, "%s.a*%s.a", tevAInputTable[ac.b + 8], tevAInputTable[ac.c + 8]); - else if (ac.b == TEVALPHAARG_ZERO) - WRITE(p, "%s.a*(1.0f-%s.a)", tevAInputTable[ac.a + 8], tevAInputTable[ac.c + 8]); - else - WRITE(p, "lerp(%s.a, %s.a, %s.a)", tevAInputTable[ac.a + 8], tevAInputTable[ac.b + 8], tevAInputTable[ac.c + 8]); - - WRITE(p, "%s",tevBiasTable[ac.bias]); - - if (ac.shift>0) - WRITE(p, ")"); - - } - else - { - //compare alpha combiner goes here - int cmp = (ac.shift<<1)|ac.op|8; // comparemode stored here - WRITE(p, TEVCMPAlphaOPTable[cmp], - tevAInputTable[ac.d], - tevAInputTable[ac.a + 8], - tevAInputTable[ac.b + 8], - tevAInputTable[ac.c + 8]); - } - if (ac.clamp) - WRITE(p, ")"); - WRITE(p, ";\n\n"); -} - -void SampleTexture(char *&p, const char *destination, const char *texcoords, const char *texswap, int texmap, API_TYPE ApiType) -{ - if (ApiType == API_D3D11) - WRITE(p, "%s=Tex%d.Sample(samp%d,%s.xy * "I_TEXDIMS"[%d].xy).%s;\n", destination, texmap,texmap, texcoords, texmap, texswap); - else - WRITE(p, "%s=tex2D(samp%d,%s.xy * "I_TEXDIMS"[%d].xy).%s;\n", destination, texmap, texcoords, texmap, texswap); -} - -static const char *tevAlphaFuncsTable[] = -{ - "(false)", //ALPHACMP_NEVER 0 - "(prev.a <= %s - (0.25f/255.0f))", //ALPHACMP_LESS 1 - "(abs( prev.a - %s ) < (0.5f/255.0f))", //ALPHACMP_EQUAL 2 - "(prev.a < %s + (0.25f/255.0f))", //ALPHACMP_LEQUAL 3 - "(prev.a >= %s + (0.25f/255.0f))", //ALPHACMP_GREATER 4 - "(abs( prev.a - %s ) >= (0.5f/255.0f))", //ALPHACMP_NEQUAL 5 - "(prev.a > %s - (0.25f/255.0f))", //ALPHACMP_GEQUAL 6 - "(true)" //ALPHACMP_ALWAYS 7 -}; - -static const char *tevAlphaFunclogicTable[] = -{ - " && ", // and - " || ", // or - " != ", // xor - " == " // xnor -}; -static int AlphaPreTest() -{ - u32 op = bpmem.alphaFunc.logic; - u32 comp[2] = {bpmem.alphaFunc.comp0, bpmem.alphaFunc.comp1}; - - // First kill all the simple cases - switch(op) - { - case 0: // AND - if (comp[0] == ALPHACMP_ALWAYS && comp[1] == ALPHACMP_ALWAYS) return true; - if (comp[0] == ALPHACMP_NEVER || comp[1] == ALPHACMP_NEVER) return false; - break; - case 1: // OR - if (comp[0] == ALPHACMP_ALWAYS || comp[1] == ALPHACMP_ALWAYS) return true; - if (comp[0] == ALPHACMP_NEVER && comp[1] == ALPHACMP_NEVER)return false; - break; - case 2: // XOR - if ((comp[0] == ALPHACMP_ALWAYS && comp[1] == ALPHACMP_NEVER) || (comp[0] == ALPHACMP_NEVER && comp[1] == ALPHACMP_ALWAYS)) - return true; - if ((comp[0] == ALPHACMP_ALWAYS && comp[1] == ALPHACMP_ALWAYS) || (comp[0] == ALPHACMP_NEVER && comp[1] == ALPHACMP_NEVER)) - return false; - break; - case 3: // XNOR - if ((comp[0] == ALPHACMP_ALWAYS && comp[1] == ALPHACMP_NEVER) || (comp[0] == ALPHACMP_NEVER && comp[1] == ALPHACMP_ALWAYS)) - return false; - if ((comp[0] == ALPHACMP_ALWAYS && comp[1] == ALPHACMP_ALWAYS) || (comp[0] == ALPHACMP_NEVER && comp[1] == ALPHACMP_NEVER)) - return true; - break; - default: PanicAlert("bad logic for alpha test? %08x", op); - } - return -1; -} - - -static bool WriteAlphaTest(char *&p, API_TYPE ApiType,DSTALPHA_MODE dstAlphaMode) -{ - static const char *alphaRef[2] = - { - I_ALPHA"[0].r", - I_ALPHA"[0].g" - }; - - int Pretest = AlphaPreTest(); - if(Pretest >= 0) - { - return Pretest != 0; - } - - // using discard then return works the same in cg and dx9 but not in dx11 - WRITE(p, "if(!( "); - - int compindex = bpmem.alphaFunc.comp0 % 8; - WRITE(p, tevAlphaFuncsTable[compindex],alphaRef[0]);//lookup the first component from the alpha function table - - WRITE(p, "%s", tevAlphaFunclogicTable[bpmem.alphaFunc.logic % 4]);//lookup the logic op - - compindex = bpmem.alphaFunc.comp1 % 8; - WRITE(p, tevAlphaFuncsTable[compindex],alphaRef[1]);//lookup the second component from the alpha function table - WRITE(p, ")){ocol0 = 0;%s%s discard;%s}\n", - dstAlphaMode == DSTALPHA_DUAL_SOURCE_BLEND ? "ocol1 = 0;" : "", - DepthTextureEnable ? "depth = 1.f;" : "", - (ApiType != API_D3D11) ? "return;" : ""); - return true; -} - -static const char *tevFogFuncsTable[] = -{ - "", //No Fog - "", //? - "", //Linear - "", //? - " fog = 1.0f - pow(2.0f, -8.0f * fog);\n", //exp - " fog = 1.0f - pow(2.0f, -8.0f * fog * fog);\n", //exp2 - " fog = pow(2.0f, -8.0f * (1.0f - fog));\n", //backward exp - " fog = 1.0f - fog;\n fog = pow(2.0f, -8.0f * fog * fog);\n" //backward exp2 -}; - -static void WriteFog(char *&p) -{ - if(bpmem.fog.c_proj_fsel.fsel == 0)return;//no Fog - - if (bpmem.fog.c_proj_fsel.proj == 0) - { - // perspective - // ze = A/(B - (Zs >> B_SHF) - WRITE (p, " float ze = "I_FOG"[1].x / ("I_FOG"[1].y - (zCoord / "I_FOG"[1].w));\n"); - } - else - { - // orthographic - // ze = a*Zs (here, no B_SHF) - WRITE (p, " float ze = "I_FOG"[1].x * zCoord;\n"); - } - - // x_adjust = sqrt((x-center)^2 + k^2)/k - // ze *= x_adjust - //this is complitly teorical as the real hard seems to use a table intead of calculate the values. - if(bpmem.fogRange.Base.Enabled) - { - WRITE (p, " float x_adjust = (2.0f * (clipPos.x / "I_FOG"[2].y)) - 1.0f - "I_FOG"[2].x;\n"); - WRITE (p, " x_adjust = sqrt(x_adjust * x_adjust + "I_FOG"[2].z * "I_FOG"[2].z) / "I_FOG"[2].z;\n"); - WRITE (p, " ze *= x_adjust;\n"); - } - - WRITE (p, " float fog = saturate(ze - "I_FOG"[1].z);\n"); - - if(bpmem.fog.c_proj_fsel.fsel > 3) - { - WRITE(p, "%s", tevFogFuncsTable[bpmem.fog.c_proj_fsel.fsel]); - } - else - { - if(bpmem.fog.c_proj_fsel.fsel != 2) - WARN_LOG(VIDEO, "Unknown Fog Type! %08x", bpmem.fog.c_proj_fsel.fsel); - } - - WRITE(p, " prev.rgb = lerp(prev.rgb,"I_FOG"[0].rgb,fog);\n"); - - + WRITE(p, "prev = frac(4.0f + prev * (255.0f/256.0f)) * (256.0f/255.0f);\n"); + + // TODO: Why are we doing a second alpha pretest here? + if (!WriteAlphaTest(p, ApiType, dstAlphaMode)) + { + // alpha test will always fail, so restart the shader and just make it an empty function + p = pmainstart; + WRITE(p, "ocol0 = 0;\n"); + if(DepthTextureEnable) + WRITE(p, "depth = 1.f;\n"); + if(dstAlphaMode == DSTALPHA_DUAL_SOURCE_BLEND) + WRITE(p, "ocol1 = 0;\n"); + WRITE(p, "discard;\n"); + if(ApiType != API_D3D11) + WRITE(p, "return;\n"); + } + else + { + if((bpmem.fog.c_proj_fsel.fsel != 0) || DepthTextureEnable) + { + + // the screen space depth value = far z + (clip z / clip w) * z range + WRITE(p, "float zCoord = "I_ZBIAS"[1].x + (clipPos.z / clipPos.w) * "I_ZBIAS"[1].y;\n"); + } + + if (DepthTextureEnable) + { + // use the texture input of the last texture stage (textemp), hopefully this has been read and is in correct format... + if (bpmem.ztex2.op != ZTEXTURE_DISABLE && !bpmem.zcontrol.zcomploc && bpmem.zmode.testenable && bpmem.zmode.updateenable) + { + if (bpmem.ztex2.op == ZTEXTURE_ADD) + WRITE(p, "zCoord = dot("I_ZBIAS"[0].xyzw, textemp.xyzw) + "I_ZBIAS"[1].w + zCoord;\n"); + else + WRITE(p, "zCoord = dot("I_ZBIAS"[0].xyzw, textemp.xyzw) + "I_ZBIAS"[1].w;\n"); + + // scale to make result from frac correct + WRITE(p, "zCoord = zCoord * (16777215.0f/16777216.0f);\n"); + WRITE(p, "zCoord = frac(zCoord);\n"); + WRITE(p, "zCoord = zCoord * (16777216.0f/16777215.0f);\n"); + } + WRITE(p, "depth = zCoord;\n"); + } + + if (dstAlphaMode == DSTALPHA_ALPHA_PASS) + WRITE(p, " ocol0 = float4(prev.rgb, "I_ALPHA"[0].a);\n"); + else + { + WriteFog(p); + WRITE(p, " ocol0 = prev;\n"); + } + + // On D3D11, use dual-source color blending to perform dst alpha in a + // single pass + if (dstAlphaMode == DSTALPHA_DUAL_SOURCE_BLEND) + { + // Colors will be blended against the alpha from ocol1... + WRITE(p, " ocol1 = ocol0;\n"); + // ...and the alpha from ocol0 will be written to the framebuffer. + WRITE(p, " ocol0.a = "I_ALPHA"[0].a;\n"); + } + } + WRITE(p, "}\n"); + if (text[sizeof(text) - 1] != 0x7C) + PanicAlert("PixelShader generator - buffer too small, canary has been eaten!"); + + setlocale(LC_NUMERIC, ""); // restore locale + return text; +} + + + +//table with the color compare operations +static const char *TEVCMPColorOPTable[16] = +{ + "float3(0.0f, 0.0f, 0.0f)",//0 + "float3(0.0f, 0.0f, 0.0f)",//1 + "float3(0.0f, 0.0f, 0.0f)",//2 + "float3(0.0f, 0.0f, 0.0f)",//3 + "float3(0.0f, 0.0f, 0.0f)",//4 + "float3(0.0f, 0.0f, 0.0f)",//5 + "float3(0.0f, 0.0f, 0.0f)",//6 + "float3(0.0f, 0.0f, 0.0f)",//7 + " %s + ((%s.r >= %s.r + (0.25f/255.0f)) ? %s : float3(0.0f, 0.0f, 0.0f))",//#define TEVCMP_R8_GT 8 + " %s + ((abs(%s.r - %s.r) < (0.5f/255.0f)) ? %s : float3(0.0f, 0.0f, 0.0f))",//#define TEVCMP_R8_EQ 9 + " %s + (( dot(%s.rgb, comp16) >= (dot(%s.rgb, comp16) + (0.25f/255.0f))) ? %s : float3(0.0f, 0.0f, 0.0f))",//#define TEVCMP_GR16_GT 10 + " %s + (abs(dot(%s.rgb, comp16) - dot(%s.rgb, comp16)) < (0.5f/255.0f) ? %s : float3(0.0f, 0.0f, 0.0f))",//#define TEVCMP_GR16_EQ 11 + " %s + (( dot(%s.rgb, comp24) >= (dot(%s.rgb, comp24) + (0.25f/255.0f))) ? %s : float3(0.0f, 0.0f, 0.0f))",//#define TEVCMP_BGR24_GT 12 + " %s + (abs(dot(%s.rgb, comp24) - dot(%s.rgb, comp24)) < (0.5f/255.0f) ? %s : float3(0.0f, 0.0f, 0.0f))",//#define TEVCMP_BGR24_EQ 13 + " %s + (max(sign(%s.rgb - %s.rgb - (0.25f/255.0f)), float3(0.0f, 0.0f, 0.0f)) * %s)",//#define TEVCMP_RGB8_GT 14 + " %s + ((float3(1.0f, 1.0f, 1.0f) - max(sign(abs(%s.rgb - %s.rgb) - (0.5f/255.0f)), float3(0.0f, 0.0f, 0.0f))) * %s)"//#define TEVCMP_RGB8_EQ 15 +}; + +//table with the alpha compare operations +static const char *TEVCMPAlphaOPTable[16] = +{ + "0.0f",//0 + "0.0f",//1 + "0.0f",//2 + "0.0f",//3 + "0.0f",//4 + "0.0f",//5 + "0.0f",//6 + "0.0f",//7 + " %s.a + ((%s.r >= (%s.r + (0.25f/255.0f))) ? %s.a : 0.0f)",//#define TEVCMP_R8_GT 8 + " %s.a + (abs(%s.r - %s.r) < (0.5f/255.0f) ? %s.a : 0.0f)",//#define TEVCMP_R8_EQ 9 + " %s.a + ((dot(%s.rgb, comp16) >= (dot(%s.rgb, comp16) + (0.25f/255.0f))) ? %s.a : 0.0f)",//#define TEVCMP_GR16_GT 10 + " %s.a + (abs(dot(%s.rgb, comp16) - dot(%s.rgb, comp16)) < (0.5f/255.0f) ? %s.a : 0.0f)",//#define TEVCMP_GR16_EQ 11 + " %s.a + ((dot(%s.rgb, comp24) >= (dot(%s.rgb, comp24) + (0.25f/255.0f))) ? %s.a : 0.0f)",//#define TEVCMP_BGR24_GT 12 + " %s.a + (abs(dot(%s.rgb, comp24) - dot(%s.rgb, comp24)) < (0.5f/255.0f) ? %s.a : 0.0f)",//#define TEVCMP_BGR24_EQ 13 + " %s.a + ((%s.a >= (%s.a + (0.25f/255.0f))) ? %s.a : 0.0f)",//#define TEVCMP_A8_GT 14 + " %s.a + (abs(%s.a - %s.a) < (0.5f/255.0f) ? %s.a : 0.0f)"//#define TEVCMP_A8_EQ 15 + +}; + + +static void WriteStage(char *&p, int n, API_TYPE ApiType) +{ + int texcoord = bpmem.tevorders[n/2].getTexCoord(n&1); + bool bHasTexCoord = (u32)texcoord < bpmem.genMode.numtexgens; + bool bHasIndStage = bpmem.tevind[n].IsActive() && bpmem.tevind[n].bt < bpmem.genMode.numindstages; + + // HACK to handle cases where the tex gen is not enabled + if (!bHasTexCoord) + texcoord = 0; + + WRITE(p, "// TEV stage %d\n", n); + + if (bHasIndStage) + { + WRITE(p, "// indirect op\n", n); + // perform the indirect op on the incoming regular coordinates using indtex%d as the offset coords + if (bpmem.tevind[n].bs != ITBA_OFF) + { + WRITE(p, "alphabump = indtex%d.%s %s;\n", + bpmem.tevind[n].bt, + tevIndAlphaSel[bpmem.tevind[n].bs], + tevIndAlphaScale[bpmem.tevind[n].fmt]); + } + // format + WRITE(p, "float3 indtevcrd%d = indtex%d * %s;\n", n, bpmem.tevind[n].bt, tevIndFmtScale[bpmem.tevind[n].fmt]); + + // bias + if (bpmem.tevind[n].bias != ITB_NONE ) + WRITE(p, "indtevcrd%d.%s += %s;\n", n, tevIndBiasField[bpmem.tevind[n].bias], tevIndBiasAdd[bpmem.tevind[n].fmt]); + + // multiply by offset matrix and scale + if (bpmem.tevind[n].mid != 0) + { + if (bpmem.tevind[n].mid <= 3) + { + int mtxidx = 2*(bpmem.tevind[n].mid-1); + WRITE(p, "float2 indtevtrans%d = float2(dot("I_INDTEXMTX"[%d].xyz, indtevcrd%d), dot("I_INDTEXMTX"[%d].xyz, indtevcrd%d));\n", + n, mtxidx, n, mtxidx+1, n); + } + else if (bpmem.tevind[n].mid <= 7 && bHasTexCoord) + { // s matrix + // TODO: Might become negative? + int mtxidx = 2*(bpmem.tevind[n].mid-5); + WRITE(p, "float2 indtevtrans%d = "I_INDTEXMTX"[%d].ww * uv%d.xy * indtevcrd%d.xx;\n", n, mtxidx, texcoord, n); + } + else if (bpmem.tevind[n].mid <= 11 && bHasTexCoord) + { // t matrix + // TODO: Might become negative? + int mtxidx = 2*(bpmem.tevind[n].mid-9); + WRITE(p, "float2 indtevtrans%d = "I_INDTEXMTX"[%d].ww * uv%d.xy * indtevcrd%d.yy;\n", n, mtxidx, texcoord, n); + } + else + WRITE(p, "float2 indtevtrans%d = 0;\n", n); + } + else + WRITE(p, "float2 indtevtrans%d = 0;\n", n); + + // --------- + // Wrapping + // --------- + + // wrap S + if (bpmem.tevind[n].sw == ITW_OFF) + WRITE(p, "wrappedcoord.x = uv%d.x;\n", texcoord); + else if (bpmem.tevind[n].sw == ITW_0) + WRITE(p, "wrappedcoord.x = 0.0f;\n"); + else + WRITE(p, "wrappedcoord.x = fmod( uv%d.x, %s );\n", texcoord, tevIndWrapStart[bpmem.tevind[n].sw]); + + // wrap T + if (bpmem.tevind[n].tw == ITW_OFF) + WRITE(p, "wrappedcoord.y = uv%d.y;\n", texcoord); + else if (bpmem.tevind[n].tw == ITW_0) + WRITE(p, "wrappedcoord.y = 0.0f;\n"); + else + WRITE(p, "wrappedcoord.y = fmod( uv%d.y, %s );\n", texcoord, tevIndWrapStart[bpmem.tevind[n].tw]); + + if (bpmem.tevind[n].fb_addprev) // add previous tevcoord + WRITE(p, "tevcoord.xy += wrappedcoord + indtevtrans%d;\n", n); + else + WRITE(p, "tevcoord.xy = wrappedcoord + indtevtrans%d;\n", n); + } + + TevStageCombiner::ColorCombiner &cc = bpmem.combiners[n].colorC; + TevStageCombiner::AlphaCombiner &ac = bpmem.combiners[n].alphaC; + + // blah1 + if(cc.a == TEVCOLORARG_RASA || cc.a == TEVCOLORARG_RASC + || cc.b == TEVCOLORARG_RASA || cc.b == TEVCOLORARG_RASC + || cc.c == TEVCOLORARG_RASA || cc.c == TEVCOLORARG_RASC + || cc.d == TEVCOLORARG_RASA || cc.d == TEVCOLORARG_RASC + || ac.a == TEVALPHAARG_RASA || ac.b == TEVALPHAARG_RASA + || ac.c == TEVALPHAARG_RASA || ac.d == TEVALPHAARG_RASA) + { + char *rasswap = swapModeTable[bpmem.combiners[n].alphaC.rswap]; + WRITE(p, "rastemp = %s.%s;\n", tevRasTable[bpmem.tevorders[n / 2].getColorChan(n & 1)], rasswap); + WRITE(p, "crastemp = frac(rastemp * (255.0f/256.0f)) * (256.0f/255.0f);\n"); + } + + + if (bpmem.tevorders[n/2].getEnable(n&1)) + { + if(!bHasIndStage) + { + // calc tevcord + if(bHasTexCoord) + WRITE(p, "tevcoord.xy = uv%d.xy;\n", texcoord); + else + WRITE(p, "tevcoord.xy = float2(0.0f, 0.0f);\n"); + } + + char *texswap = swapModeTable[bpmem.combiners[n].alphaC.tswap]; + int texmap = bpmem.tevorders[n/2].getTexMap(n&1); + SampleTexture(p, "textemp", "tevcoord", texswap, texmap, ApiType); + } + else + WRITE(p, "textemp = float4(1.0f, 1.0f, 1.0f, 1.0f);\n"); + + + // blah2 + if (cc.a == TEVCOLORARG_KONST || cc.b == TEVCOLORARG_KONST || cc.c == TEVCOLORARG_KONST || cc.d == TEVCOLORARG_KONST + || ac.a == TEVALPHAARG_KONST || ac.b == TEVALPHAARG_KONST || ac.c == TEVALPHAARG_KONST || ac.d == TEVALPHAARG_KONST) + { + int kc = bpmem.tevksel[n / 2].getKC(n & 1); + int ka = bpmem.tevksel[n / 2].getKA(n & 1); + WRITE(p, "konsttemp = float4(%s, %s);\n", tevKSelTableC[kc], tevKSelTableA[ka]); + if(kc > 7 || ka > 7) + { + WRITE(p, "ckonsttemp = frac(konsttemp * (255.0f/256.0f)) * (256.0f/255.0f);\n"); + } + else + { + WRITE(p, "ckonsttemp = konsttemp;\n"); + } + } + + if(cc.a == TEVCOLORARG_CPREV || cc.a == TEVCOLORARG_APREV + || cc.b == TEVCOLORARG_CPREV || cc.b == TEVCOLORARG_APREV + || cc.c == TEVCOLORARG_CPREV || cc.c == TEVCOLORARG_APREV + || ac.a == TEVALPHAARG_APREV || ac.b == TEVALPHAARG_APREV || ac.c == TEVALPHAARG_APREV) + WRITE(p, "cprev = frac(prev * (255.0f/256.0f)) * (256.0f/255.0f);\n"); + + + if(cc.a == TEVCOLORARG_C0 || cc.a == TEVCOLORARG_A0 + || cc.b == TEVCOLORARG_C0 || cc.b == TEVCOLORARG_A0 + || cc.c == TEVCOLORARG_C0 || cc.c == TEVCOLORARG_A0 + || ac.a == TEVALPHAARG_A0 || ac.b == TEVALPHAARG_A0 || ac.c == TEVALPHAARG_A0) + WRITE(p, "cc0 = frac(c0 * (255.0f/256.0f)) * (256.0f/255.0f);\n"); + + + if(cc.a == TEVCOLORARG_C1 || cc.a == TEVCOLORARG_A1 + || cc.b == TEVCOLORARG_C1 || cc.b == TEVCOLORARG_A1 + || cc.c == TEVCOLORARG_C1 || cc.c == TEVCOLORARG_A1 + || ac.a == TEVALPHAARG_A1 || ac.b == TEVALPHAARG_A1 || ac.c == TEVALPHAARG_A1) + WRITE(p, "cc1 = frac(c1 * (255.0f/256.0f)) * (256.0f/255.0f);\n"); + + + if(cc.a == TEVCOLORARG_C2 || cc.a == TEVCOLORARG_A2 + || cc.b == TEVCOLORARG_C2 || cc.b == TEVCOLORARG_A2 + || cc.c == TEVCOLORARG_C2 || cc.c == TEVCOLORARG_A2 + || ac.a == TEVALPHAARG_A2 || ac.b == TEVALPHAARG_A2 || ac.c == TEVALPHAARG_A2) + WRITE(p, "cc2 = frac(c2 * (255.0f/256.0f)) * (256.0f/255.0f);\n"); + + + WRITE(p, "// color combine\n", n); + if (cc.clamp) + WRITE(p, "%s = saturate(", tevCOutputTable[cc.dest]); + else + WRITE(p, "%s = ", tevCOutputTable[cc.dest]); + + // combine the color channel + if (cc.bias != TevBias_COMPARE) // if not compare + { + //normal color combiner goes here + if (cc.shift > TEVSCALE_1) + WRITE(p, "%s*(", tevScaleTable[cc.shift]); + + if(!(cc.d == TEVCOLORARG_ZERO && cc.op == TEVOP_ADD)) + WRITE(p, "%s%s", tevCInputTable[cc.d], tevOpTable[cc.op]); + + if (cc.a == cc.b) + WRITE(p, "%s", tevCInputTable[cc.a + 16]); + else if (cc.c == TEVCOLORARG_ZERO) + WRITE(p, "%s", tevCInputTable[cc.a + 16]); + else if (cc.c == TEVCOLORARG_ONE) + WRITE(p, "%s", tevCInputTable[cc.b + 16]); + else if (cc.a == TEVCOLORARG_ZERO) + WRITE(p, "%s*%s", tevCInputTable[cc.b + 16], tevCInputTable[cc.c + 16]); + else if (cc.b == TEVCOLORARG_ZERO) + WRITE(p, "%s*(float3(1.0f, 1.0f, 1.0f)-%s)", tevCInputTable[cc.a + 16], tevCInputTable[cc.c + 16]); + else + WRITE(p, "lerp(%s, %s, %s)", tevCInputTable[cc.a + 16], tevCInputTable[cc.b + 16], tevCInputTable[cc.c + 16]); + + WRITE(p, "%s", tevBiasTable[cc.bias]); + + if (cc.shift > TEVSCALE_1) + WRITE(p, ")"); + } + else + { + int cmp = (cc.shift<<1)|cc.op|8; // comparemode stored here + WRITE(p, TEVCMPColorOPTable[cmp],//lookup the function from the op table + tevCInputTable[cc.d], + tevCInputTable[cc.a + 16], + tevCInputTable[cc.b + 16], + tevCInputTable[cc.c + 16]); + } + if (cc.clamp) + WRITE(p, ")"); + WRITE(p,";\n"); + + WRITE(p, "// alpha combine\n", n); + // combine the alpha channel + if (ac.clamp) + WRITE(p, "%s = saturate(", tevAOutputTable[ac.dest]); + else + WRITE(p, "%s = ", tevAOutputTable[ac.dest]); + + if (ac.bias != TevBias_COMPARE) // if not compare + { + //normal alpha combiner goes here + if (ac.shift > TEVSCALE_1) + WRITE(p, "%s*(", tevScaleTable[ac.shift]); + + if(!(ac.d == TEVALPHAARG_ZERO && ac.op == TEVOP_ADD)) + WRITE(p, "%s.a%s", tevAInputTable[ac.d], tevOpTable[ac.op]); + + if (ac.a == ac.b) + WRITE(p, "%s.a", tevAInputTable[ac.a + 8]); + else if (ac.c == TEVALPHAARG_ZERO) + WRITE(p, "%s.a", tevAInputTable[ac.a + 8]); + else if (ac.a == TEVALPHAARG_ZERO) + WRITE(p, "%s.a*%s.a", tevAInputTable[ac.b + 8], tevAInputTable[ac.c + 8]); + else if (ac.b == TEVALPHAARG_ZERO) + WRITE(p, "%s.a*(1.0f-%s.a)", tevAInputTable[ac.a + 8], tevAInputTable[ac.c + 8]); + else + WRITE(p, "lerp(%s.a, %s.a, %s.a)", tevAInputTable[ac.a + 8], tevAInputTable[ac.b + 8], tevAInputTable[ac.c + 8]); + + WRITE(p, "%s",tevBiasTable[ac.bias]); + + if (ac.shift>0) + WRITE(p, ")"); + + } + else + { + //compare alpha combiner goes here + int cmp = (ac.shift<<1)|ac.op|8; // comparemode stored here + WRITE(p, TEVCMPAlphaOPTable[cmp], + tevAInputTable[ac.d], + tevAInputTable[ac.a + 8], + tevAInputTable[ac.b + 8], + tevAInputTable[ac.c + 8]); + } + if (ac.clamp) + WRITE(p, ")"); + WRITE(p, ";\n\n"); + WRITE(p, "// TEV done\n", n); +} + +void SampleTexture(char *&p, const char *destination, const char *texcoords, const char *texswap, int texmap, API_TYPE ApiType) +{ + if (ApiType == API_D3D11) + WRITE(p, "%s=Tex%d.Sample(samp%d,%s.xy * "I_TEXDIMS"[%d].xy).%s;\n", destination, texmap,texmap, texcoords, texmap, texswap); + else + WRITE(p, "%s=tex2D(samp%d,%s.xy * "I_TEXDIMS"[%d].xy).%s;\n", destination, texmap, texcoords, texmap, texswap); +} + +static const char *tevAlphaFuncsTable[] = +{ + "(false)", //ALPHACMP_NEVER 0 + "(prev.a <= %s - (0.25f/255.0f))", //ALPHACMP_LESS 1 + "(abs( prev.a - %s ) < (0.5f/255.0f))", //ALPHACMP_EQUAL 2 + "(prev.a < %s + (0.25f/255.0f))", //ALPHACMP_LEQUAL 3 + "(prev.a >= %s + (0.25f/255.0f))", //ALPHACMP_GREATER 4 + "(abs( prev.a - %s ) >= (0.5f/255.0f))", //ALPHACMP_NEQUAL 5 + "(prev.a > %s - (0.25f/255.0f))", //ALPHACMP_GEQUAL 6 + "(true)" //ALPHACMP_ALWAYS 7 +}; + +static const char *tevAlphaFunclogicTable[] = +{ + " && ", // and + " || ", // or + " != ", // xor + " == " // xnor +}; +static int AlphaPreTest() +{ + u32 op = bpmem.alphaFunc.logic; + u32 comp[2] = {bpmem.alphaFunc.comp0, bpmem.alphaFunc.comp1}; + + // First kill all the simple cases + switch(op) + { + case 0: // AND + if (comp[0] == ALPHACMP_ALWAYS && comp[1] == ALPHACMP_ALWAYS) return true; + if (comp[0] == ALPHACMP_NEVER || comp[1] == ALPHACMP_NEVER) return false; + break; + case 1: // OR + if (comp[0] == ALPHACMP_ALWAYS || comp[1] == ALPHACMP_ALWAYS) return true; + if (comp[0] == ALPHACMP_NEVER && comp[1] == ALPHACMP_NEVER)return false; + break; + case 2: // XOR + if ((comp[0] == ALPHACMP_ALWAYS && comp[1] == ALPHACMP_NEVER) || (comp[0] == ALPHACMP_NEVER && comp[1] == ALPHACMP_ALWAYS)) + return true; + if ((comp[0] == ALPHACMP_ALWAYS && comp[1] == ALPHACMP_ALWAYS) || (comp[0] == ALPHACMP_NEVER && comp[1] == ALPHACMP_NEVER)) + return false; + break; + case 3: // XNOR + if ((comp[0] == ALPHACMP_ALWAYS && comp[1] == ALPHACMP_NEVER) || (comp[0] == ALPHACMP_NEVER && comp[1] == ALPHACMP_ALWAYS)) + return false; + if ((comp[0] == ALPHACMP_ALWAYS && comp[1] == ALPHACMP_ALWAYS) || (comp[0] == ALPHACMP_NEVER && comp[1] == ALPHACMP_NEVER)) + return true; + break; + default: PanicAlert("bad logic for alpha test? %08x", op); + } + return -1; +} + + +static bool WriteAlphaTest(char *&p, API_TYPE ApiType,DSTALPHA_MODE dstAlphaMode) +{ + static const char *alphaRef[2] = + { + I_ALPHA"[0].r", + I_ALPHA"[0].g" + }; + + int Pretest = AlphaPreTest(); + if(Pretest >= 0) + { + return Pretest != 0; + } + + // using discard then return works the same in cg and dx9 but not in dx11 + WRITE(p, "if(!( "); + + int compindex = bpmem.alphaFunc.comp0 % 8; + WRITE(p, tevAlphaFuncsTable[compindex],alphaRef[0]);//lookup the first component from the alpha function table + + WRITE(p, "%s", tevAlphaFunclogicTable[bpmem.alphaFunc.logic % 4]);//lookup the logic op + + compindex = bpmem.alphaFunc.comp1 % 8; + WRITE(p, tevAlphaFuncsTable[compindex],alphaRef[1]);//lookup the second component from the alpha function table + WRITE(p, ")){ocol0 = 0;%s%s discard;%s}\n", + dstAlphaMode == DSTALPHA_DUAL_SOURCE_BLEND ? "ocol1 = 0;" : "", + DepthTextureEnable ? "depth = 1.f;" : "", + (ApiType != API_D3D11) ? "return;" : ""); + return true; +} + +static const char *tevFogFuncsTable[] = +{ + "", //No Fog + "", //? + "", //Linear + "", //? + " fog = 1.0f - pow(2.0f, -8.0f * fog);\n", //exp + " fog = 1.0f - pow(2.0f, -8.0f * fog * fog);\n", //exp2 + " fog = pow(2.0f, -8.0f * (1.0f - fog));\n", //backward exp + " fog = 1.0f - fog;\n fog = pow(2.0f, -8.0f * fog * fog);\n" //backward exp2 +}; + +static void WriteFog(char *&p) +{ + if(bpmem.fog.c_proj_fsel.fsel == 0)return;//no Fog + + if (bpmem.fog.c_proj_fsel.proj == 0) + { + // perspective + // ze = A/(B - (Zs >> B_SHF) + WRITE (p, " float ze = "I_FOG"[1].x / ("I_FOG"[1].y - (zCoord / "I_FOG"[1].w));\n"); + } + else + { + // orthographic + // ze = a*Zs (here, no B_SHF) + WRITE (p, " float ze = "I_FOG"[1].x * zCoord;\n"); + } + + // x_adjust = sqrt((x-center)^2 + k^2)/k + // ze *= x_adjust + //this is complitly teorical as the real hard seems to use a table intead of calculate the values. + if(bpmem.fogRange.Base.Enabled) + { + WRITE (p, " float x_adjust = (2.0f * (clipPos.x / "I_FOG"[2].y)) - 1.0f - "I_FOG"[2].x;\n"); + WRITE (p, " x_adjust = sqrt(x_adjust * x_adjust + "I_FOG"[2].z * "I_FOG"[2].z) / "I_FOG"[2].z;\n"); + WRITE (p, " ze *= x_adjust;\n"); + } + + WRITE (p, " float fog = saturate(ze - "I_FOG"[1].z);\n"); + + if(bpmem.fog.c_proj_fsel.fsel > 3) + { + WRITE(p, "%s", tevFogFuncsTable[bpmem.fog.c_proj_fsel.fsel]); + } + else + { + if(bpmem.fog.c_proj_fsel.fsel != 2) + WARN_LOG(VIDEO, "Unknown Fog Type! %08x", bpmem.fog.c_proj_fsel.fsel); + } + + WRITE(p, " prev.rgb = lerp(prev.rgb,"I_FOG"[0].rgb,fog);\n"); + + } \ No newline at end of file -- cgit v1.2.3 From 4137f287fd4d8778967bb8a82eafe5450687e2b7 Mon Sep 17 00:00:00 2001 From: NeoBrainX Date: Wed, 31 Aug 2011 19:45:28 +0200 Subject: Fix a critical bug which caused shaders to be redundantly recompiled when disabling per-pixel depth. As a nice side effect, the number of redundant shader compilations is now next to zero ;) --- Source/Core/VideoCommon/Src/PixelShaderGen.cpp | 1 + 1 file changed, 1 insertion(+) (limited to 'Source/Core/VideoCommon/Src/PixelShaderGen.cpp') diff --git a/Source/Core/VideoCommon/Src/PixelShaderGen.cpp b/Source/Core/VideoCommon/Src/PixelShaderGen.cpp index 6d0cc64964..ff82438305 100644 --- a/Source/Core/VideoCommon/Src/PixelShaderGen.cpp +++ b/Source/Core/VideoCommon/Src/PixelShaderGen.cpp @@ -124,6 +124,7 @@ void GetPixelShaderId(PIXELSHADERUID *uid, DSTALPHA_MODE dstAlphaMode) { // Courtesy of PreAlphaTest, we're done already ;) // TODO: There's a comment including bpmem.genmode.numindstages.. shouldnt really bother about that though. + uid->tevstages = 1; return; } -- cgit v1.2.3 From df4e337ac99b9fdb72bbc244bd437032c6109a3b Mon Sep 17 00:00:00 2001 From: NeoBrainX Date: Sat, 6 Aug 2011 21:35:16 +0000 Subject: Fix various pixel shader compilation errors caused by the Direct3D shader compiler going nuts due to uninitialized (and unused) shader variables. git-svn-id: https://dolphin-emu.googlecode.com/svn/trunk@7693 8ced0084-cf51-0410-be5f-012b33b47a6e --- Source/Core/VideoCommon/Src/PixelShaderGen.cpp | 10 ++++++---- 1 file changed, 6 insertions(+), 4 deletions(-) (limited to 'Source/Core/VideoCommon/Src/PixelShaderGen.cpp') diff --git a/Source/Core/VideoCommon/Src/PixelShaderGen.cpp b/Source/Core/VideoCommon/Src/PixelShaderGen.cpp index 530ecd54bb..6c709084e4 100644 --- a/Source/Core/VideoCommon/Src/PixelShaderGen.cpp +++ b/Source/Core/VideoCommon/Src/PixelShaderGen.cpp @@ -523,10 +523,12 @@ const char *GeneratePixelShaderCode(DSTALPHA_MODE dstAlphaMode, API_TYPE ApiType WRITE(p, " float4 c0 = "I_COLORS"[1], c1 = "I_COLORS"[2], c2 = "I_COLORS"[3], prev = float4(0.0f, 0.0f, 0.0f, 0.0f), textemp = float4(0.0f, 0.0f, 0.0f, 0.0f), rastemp = float4(0.0f, 0.0f, 0.0f, 0.0f), konsttemp = float4(0.0f, 0.0f, 0.0f, 0.0f);\n" " float3 comp16 = float3(1.0f, 255.0f, 0.0f), comp24 = float3(1.0f, 255.0f, 255.0f*255.0f);\n" - " float4 alphabump=0;\n" - " float3 tevcoord;\n" - " float2 wrappedcoord, tempcoord;\n" - " float4 cc0, cc1, cc2, cprev,crastemp,ckonsttemp;\n\n"); + " float4 alphabump=float4(0.0f,0.0f,0.0f,0.0f);\n" + " float3 tevcoord=float3(0.0f, 0.0f, 0.0f);\n" + " float2 wrappedcoord=float2(0.0f,0.0f), tempcoord=float2(0.0f,0.0f);\n" + " float4 cc0=float4(0.0f,0.0f,0.0f,0.0f), cc1=float4(0.0f,0.0f,0.0f,0.0f);\n" + " float4 cc2=float4(0.0f,0.0f,0.0f,0.0f), cprev=float4(0.0f,0.0f,0.0f,0.0f);\n" + " float4 crastemp=float4(0.0f,0.0f,0.0f,0.0f),ckonsttemp=float4(0.0f,0.0f,0.0f,0.0f);\n\n"); if(g_ActiveConfig.bEnablePixelLighting && g_ActiveConfig.backend_info.bSupportsPixelLighting) { -- cgit v1.2.3 From 4702de591eb69e815398b21c5b4321b1ac133552 Mon Sep 17 00:00:00 2001 From: NeoBrainX Date: Sun, 4 Sep 2011 02:10:07 +0200 Subject: Added safe pixel shader UIDs for debugging purposes. --- Source/Core/VideoCommon/Src/PixelShaderGen.cpp | 54 ++++++++++++++++++++++---- 1 file changed, 47 insertions(+), 7 deletions(-) (limited to 'Source/Core/VideoCommon/Src/PixelShaderGen.cpp') diff --git a/Source/Core/VideoCommon/Src/PixelShaderGen.cpp b/Source/Core/VideoCommon/Src/PixelShaderGen.cpp index ff82438305..9913ea7c5a 100644 --- a/Source/Core/VideoCommon/Src/PixelShaderGen.cpp +++ b/Source/Core/VideoCommon/Src/PixelShaderGen.cpp @@ -164,7 +164,7 @@ void GetPixelShaderId(PIXELSHADERUID *uid, DSTALPHA_MODE dstAlphaMode) } u32* ptr = &uid->values[2]; - for (unsigned int i = 0; i < bpmem.genMode.numtevstages+1; ++i) + for (int i = 0; i < bpmem.genMode.numtevstages+1; ++i) { StageHash(i, ptr); ptr += 4; // max: ptr = &uid->values[66] @@ -315,6 +315,46 @@ void _GetPixelShaderId(PIXELSHADERUID *uid, DSTALPHA_MODE dstAlphaMode) } +void GetSafePixelShaderId(PIXELSHADERUIDSAFE *uid, DSTALPHA_MODE dstAlphaMode) +{ + u32* ptr = uid->values; + *ptr++ = dstAlphaMode; + *ptr++ = bpmem.genMode.hex; + *ptr++ = bpmem.ztex2.hex; + *ptr++ = bpmem.zcontrol.hex; + *ptr++ = bpmem.zmode.hex; + *ptr++ = g_ActiveConfig.bEnablePerPixelDepth; + *ptr++ = g_ActiveConfig.bEnablePixelLighting && g_ActiveConfig.backend_info.bSupportsPixelLighting; + *ptr++ = xfregs.numTexGen.hex; + + if (g_ActiveConfig.bEnablePixelLighting && g_ActiveConfig.backend_info.bSupportsPixelLighting) + { + // TODO: Include register states for lighting shader + } + + for (unsigned int i = 0; i < 8; ++i) + *ptr++ = xfregs.texMtxInfo[i].hex; + + for (unsigned int i = 0; i < 16; ++i) + *ptr++ = bpmem.tevind[i].hex; + + *ptr++ = bpmem.tevindref.hex; + + for (int i = 0; i < bpmem.genMode.numtevstages+1; ++i) + { + // TODO ... + StageHash(i, ptr); + ptr += 4; // max: ptr = &uid->values[66] + } + + *ptr++ = bpmem.fog.c_proj_fsel.hex; + + *ptr++ = bpmem.fogRange.Base.hex; + + _assert_((ptr - uid->values) == uid->GetNumValues()); +} + + // old tev->pixelshader notes // // color for this stage (alpha, color) is given by bpmem.tevorders[0].colorchan0 @@ -539,8 +579,8 @@ const char *GeneratePixelShaderCode(DSTALPHA_MODE dstAlphaMode, API_TYPE ApiType char *p = text; WRITE(p, "//Pixel Shader for TEV stages\n"); - WRITE(p, "//%i TEV stages, %i texgens, %i IND stages\n", - numStages, numTexgen, bpmem.genMode.numindstages); + WRITE(p, "//%i TEV stages, %i texgens, XXX IND stages\n", + numStages, numTexgen/*, bpmem.genMode.numindstages*/); int nIndirectStagesUsed = 0; if (bpmem.genMode.numindstages > 0) @@ -876,7 +916,7 @@ static void WriteStage(char *&p, int n, API_TYPE ApiType) if (bHasIndStage) { - WRITE(p, "// indirect op\n", n); + WRITE(p, "// indirect op\n"); // perform the indirect op on the incoming regular coordinates using indtex%d as the offset coords if (bpmem.tevind[n].bs != ITBA_OFF) { @@ -1026,7 +1066,7 @@ static void WriteStage(char *&p, int n, API_TYPE ApiType) WRITE(p, "cc2 = frac(c2 * (255.0f/256.0f)) * (256.0f/255.0f);\n"); - WRITE(p, "// color combine\n", n); + WRITE(p, "// color combine\n"); if (cc.clamp) WRITE(p, "%s = saturate(", tevCOutputTable[cc.dest]); else @@ -1073,7 +1113,7 @@ static void WriteStage(char *&p, int n, API_TYPE ApiType) WRITE(p, ")"); WRITE(p,";\n"); - WRITE(p, "// alpha combine\n", n); + WRITE(p, "// alpha combine\n"); // combine the alpha channel if (ac.clamp) WRITE(p, "%s = saturate(", tevAOutputTable[ac.dest]); @@ -1119,7 +1159,7 @@ static void WriteStage(char *&p, int n, API_TYPE ApiType) if (ac.clamp) WRITE(p, ")"); WRITE(p, ";\n\n"); - WRITE(p, "// TEV done\n", n); + WRITE(p, "// TEV done\n"); } void SampleTexture(char *&p, const char *destination, const char *texcoords, const char *texswap, int texmap, API_TYPE ApiType) -- cgit v1.2.3 From 3939f9595a87393482df617465888f4ed46b6d94 Mon Sep 17 00:00:00 2001 From: NeoBrainX Date: Wed, 7 Sep 2011 20:20:29 +0200 Subject: Add runtime checks to make sure we aren't overoptimizing the pixel shader cache. --- Source/Core/VideoCommon/Src/PixelShaderGen.cpp | 49 ++++++++++++++------------ 1 file changed, 26 insertions(+), 23 deletions(-) (limited to 'Source/Core/VideoCommon/Src/PixelShaderGen.cpp') diff --git a/Source/Core/VideoCommon/Src/PixelShaderGen.cpp b/Source/Core/VideoCommon/Src/PixelShaderGen.cpp index 9913ea7c5a..8bdfc3db28 100644 --- a/Source/Core/VideoCommon/Src/PixelShaderGen.cpp +++ b/Source/Core/VideoCommon/Src/PixelShaderGen.cpp @@ -172,13 +172,13 @@ void GetPixelShaderId(PIXELSHADERUID *uid, DSTALPHA_MODE dstAlphaMode) if (alphaPreTest == 0 || alphaPreTest == 2) { - ptr[0] |= bpmem.fog.c_proj_fsel.fsel; // 3 + ptr[0] |= bpmem.fog.c_proj_fsel.fsel << 8; // 3 if (DepthTextureEnable) { - ptr[0] |= bpmem.ztex2.op << 3; // 2 - ptr[0] |= bpmem.zcontrol.zcomploc << 5; // 1 - ptr[0] |= bpmem.zmode.testenable << 6; // 1 - ptr[0] |= bpmem.zmode.updateenable << 7; // 1 + ptr[0] |= bpmem.ztex2.op << 11; // 2 + ptr[0] |= bpmem.zcontrol.zcomploc << 13; // 1 + ptr[0] |= bpmem.zmode.testenable << 14; // 1 + ptr[0] |= bpmem.zmode.updateenable << 15; // 1 } } @@ -186,8 +186,8 @@ void GetPixelShaderId(PIXELSHADERUID *uid, DSTALPHA_MODE dstAlphaMode) { if (bpmem.fog.c_proj_fsel.fsel != 0) { - ptr[0] |= bpmem.fog.c_proj_fsel.proj << 8; // 1 - ptr[0] |= bpmem.fogRange.Base.Enabled << 9; // 1 + ptr[0] |= bpmem.fog.c_proj_fsel.proj << 16; // 1 + ptr[0] |= bpmem.fogRange.Base.Enabled << 17; // 1 } } uid->tevstages = (ptr+1) - uid->values; @@ -318,14 +318,14 @@ void _GetPixelShaderId(PIXELSHADERUID *uid, DSTALPHA_MODE dstAlphaMode) void GetSafePixelShaderId(PIXELSHADERUIDSAFE *uid, DSTALPHA_MODE dstAlphaMode) { u32* ptr = uid->values; - *ptr++ = dstAlphaMode; - *ptr++ = bpmem.genMode.hex; - *ptr++ = bpmem.ztex2.hex; - *ptr++ = bpmem.zcontrol.hex; - *ptr++ = bpmem.zmode.hex; - *ptr++ = g_ActiveConfig.bEnablePerPixelDepth; - *ptr++ = g_ActiveConfig.bEnablePixelLighting && g_ActiveConfig.backend_info.bSupportsPixelLighting; - *ptr++ = xfregs.numTexGen.hex; + *ptr++ = dstAlphaMode; // 0 + *ptr++ = bpmem.genMode.hex; // 1 + *ptr++ = bpmem.ztex2.hex; // 2 + *ptr++ = bpmem.zcontrol.hex; // 3 + *ptr++ = bpmem.zmode.hex; // 4 + *ptr++ = g_ActiveConfig.bEnablePerPixelDepth; // 5 + *ptr++ = g_ActiveConfig.bEnablePixelLighting && g_ActiveConfig.backend_info.bSupportsPixelLighting; // 6 + *ptr++ = xfregs.numTexGen.hex; // 7 if (g_ActiveConfig.bEnablePixelLighting && g_ActiveConfig.backend_info.bSupportsPixelLighting) { @@ -333,25 +333,28 @@ void GetSafePixelShaderId(PIXELSHADERUIDSAFE *uid, DSTALPHA_MODE dstAlphaMode) } for (unsigned int i = 0; i < 8; ++i) - *ptr++ = xfregs.texMtxInfo[i].hex; + *ptr++ = xfregs.texMtxInfo[i].hex; // 8-15 for (unsigned int i = 0; i < 16; ++i) - *ptr++ = bpmem.tevind[i].hex; + *ptr++ = bpmem.tevind[i].hex; // 16-31 - *ptr++ = bpmem.tevindref.hex; + *ptr++ = bpmem.tevindref.hex; // 32 - for (int i = 0; i < bpmem.genMode.numtevstages+1; ++i) + for (int i = 0; i < bpmem.genMode.numtevstages+1; ++i) // up to 16 times { // TODO ... StageHash(i, ptr); - ptr += 4; // max: ptr = &uid->values[66] + ptr += 4; // max: ptr = &uid->values[33+63] } - *ptr++ = bpmem.fog.c_proj_fsel.hex; + ptr = &uid->values[97]; + + *ptr++ = bpmem.alphaFunc.hex; // 97 - *ptr++ = bpmem.fogRange.Base.hex; + *ptr++ = bpmem.fog.c_proj_fsel.hex; // 98 + *ptr++ = bpmem.fogRange.Base.hex; // 99 - _assert_((ptr - uid->values) == uid->GetNumValues()); + _assert_((ptr - uid->values) <= uid->GetNumValues()); } -- cgit v1.2.3 From 98b62d836290ff0ea46f73a6f6bdb0d3c91df923 Mon Sep 17 00:00:00 2001 From: NeoBrainX Date: Wed, 7 Sep 2011 21:12:17 +0200 Subject: Track alpha blending paramaters in the pixel shader UID. --- Source/Core/VideoCommon/Src/PixelShaderGen.cpp | 4 ++++ 1 file changed, 4 insertions(+) (limited to 'Source/Core/VideoCommon/Src/PixelShaderGen.cpp') diff --git a/Source/Core/VideoCommon/Src/PixelShaderGen.cpp b/Source/Core/VideoCommon/Src/PixelShaderGen.cpp index 8bdfc3db28..46f139f105 100644 --- a/Source/Core/VideoCommon/Src/PixelShaderGen.cpp +++ b/Source/Core/VideoCommon/Src/PixelShaderGen.cpp @@ -170,6 +170,10 @@ void GetPixelShaderId(PIXELSHADERUID *uid, DSTALPHA_MODE dstAlphaMode) ptr += 4; // max: ptr = &uid->values[66] } + ptr[0] |= bpmem.alphaFunc.comp0; // 3 + ptr[0] |= bpmem.alphaFunc.comp1 << 3; // 3 + ptr[0] |= bpmem.alphaFunc.logic << 6; // 2 + if (alphaPreTest == 0 || alphaPreTest == 2) { ptr[0] |= bpmem.fog.c_proj_fsel.fsel << 8; // 3 -- cgit v1.2.3 From 6c7bda6851b39d358f9b951739156e07345ccfa4 Mon Sep 17 00:00:00 2001 From: NeoBrainX Date: Thu, 8 Sep 2011 02:09:44 +0200 Subject: Various fixes and cleanups. --- Source/Core/VideoCommon/Src/PixelShaderGen.cpp | 152 +++---------------------- 1 file changed, 16 insertions(+), 136 deletions(-) (limited to 'Source/Core/VideoCommon/Src/PixelShaderGen.cpp') diff --git a/Source/Core/VideoCommon/Src/PixelShaderGen.cpp b/Source/Core/VideoCommon/Src/PixelShaderGen.cpp index 46f139f105..464c16fa58 100644 --- a/Source/Core/VideoCommon/Src/PixelShaderGen.cpp +++ b/Source/Core/VideoCommon/Src/PixelShaderGen.cpp @@ -100,14 +100,16 @@ static void StageHash(int stage, u32* out) } } +// Mash together all the inputs that contribute to the code of a generated pixel shader into +// 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, DSTALPHA_MODE dstAlphaMode) { uid->values[0] |= bpmem.genMode.numtevstages; // 4 uid->values[0] |= bpmem.genMode.numtexgens << 4; // 4 uid->values[0] |= dstAlphaMode << 8; // 2 - uid->tevstages = bpmem.genMode.numtevstages; - bool DepthTextureEnable = (bpmem.ztex2.op != ZTEXTURE_DISABLE && !bpmem.zcontrol.zcomploc && bpmem.zmode.testenable && bpmem.zmode.updateenable) || g_ActiveConfig.bEnablePerPixelDepth; uid->values[0] |= DepthTextureEnable << 10; // 1 @@ -124,7 +126,7 @@ void GetPixelShaderId(PIXELSHADERUID *uid, DSTALPHA_MODE dstAlphaMode) { // Courtesy of PreAlphaTest, we're done already ;) // TODO: There's a comment including bpmem.genmode.numindstages.. shouldnt really bother about that though. - uid->tevstages = 1; + uid->num_values = 1; return; } @@ -141,8 +143,6 @@ void GetPixelShaderId(PIXELSHADERUID *uid, DSTALPHA_MODE dstAlphaMode) uid->values[1] |= xfregs.texMtxInfo[i].projection << (i - 14); // 1 } - uid->indstages = bpmem.genMode.numindstages; - uid->values[1] = bpmem.genMode.numindstages << 2; // 3 u32 indirectStagesUsed = 0; for (unsigned int i = 0; i < bpmem.genMode.numindstages; ++i) @@ -194,129 +194,7 @@ void GetPixelShaderId(PIXELSHADERUID *uid, DSTALPHA_MODE dstAlphaMode) ptr[0] |= bpmem.fogRange.Base.Enabled << 17; // 1 } } - uid->tevstages = (ptr+1) - uid->values; -} - -// Mash together all the inputs that contribute to the code of a generated pixel shader into -// 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, DSTALPHA_MODE dstAlphaMode) -{ - u32 numstages = bpmem.genMode.numtevstages + 1; - u32 projtexcoords = 0; - for (u32 i = 0; i < numstages; i++) - { - if (bpmem.tevorders[i/2].getEnable(i & 1)) - { - int texcoord = bpmem.tevorders[i / 2].getTexCoord(i & 1); - if (xfregs.texMtxInfo[i].projection) - projtexcoords |= 1 << texcoord; - } - } - uid->values[0] = (u32)bpmem.genMode.numtevstages | - ((u32)bpmem.genMode.numindstages << 4) | - ((u32)bpmem.genMode.numtexgens << 7) | - ((u32)dstAlphaMode << 11) | - ((u32)((bpmem.alphaFunc.hex >> 16) & 0xff) << 13) | - (projtexcoords << 21) | - ((u32)bpmem.ztex2.op << 29); - - // 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); - - u32 enableZTexture = (bpmem.ztex2.op != ZTEXTURE_DISABLE && !bpmem.zcontrol.zcomploc && bpmem.zmode.testenable && bpmem.zmode.updateenable) || g_ActiveConfig.bEnablePerPixelDepth ? 1 : 0; - - uid->values[2] = (u32)bpmem.fog.c_proj_fsel.fsel | - ((u32)bpmem.fog.c_proj_fsel.proj << 3) | - ((u32)enableZTexture << 4) | ((u32)bpmem.fogRange.Base.Enabled << 5); - - if(g_ActiveConfig.bEnablePixelLighting && g_ActiveConfig.backend_info.bSupportsPixelLighting) - { - for (int i = 0; i < 2; ++i) { - uid->values[3 + i] = xfregs.color[i].enablelighting ? - (u32)xfregs.color[i].hex : - (u32)xfregs.color[i].matsource; - uid->values[3 + i] |= (xfregs.alpha[i].enablelighting ? - (u32)xfregs.alpha[i].hex : - (u32)xfregs.alpha[i].matsource) << 15; - } - } - uid->values[4] |= (g_ActiveConfig.bEnablePixelLighting && g_ActiveConfig.backend_info.bSupportsPixelLighting) << 31; - - int hdr = 5; - u32 *pcurvalue = &uid->values[hdr]; - for (u32 i = 0; i < numstages; ++i) - { - TevStageCombiner::ColorCombiner &cc = bpmem.combiners[i].colorC; - TevStageCombiner::AlphaCombiner &ac = bpmem.combiners[i].alphaC; - - u32 val0 = cc.hex & 0xffffff; - u32 val1 = ac.hex & 0xffffff; - val0 |= bpmem.tevksel[i / 2].getKC(i & 1) << 24; - val1 |= bpmem.tevksel[i / 2].getKA(i & 1) << 24; - pcurvalue[0] = val0; - pcurvalue[1] = val1; - pcurvalue += 2; - } - - for (u32 i = 0; i < numstages / 2; ++i) - { - u32 val0, val1; - if (bpmem.tevorders[i].hex & 0x40) - val0 = bpmem.tevorders[i].hex & 0x3ff; - else - val0 = bpmem.tevorders[i].hex & 0x380; - if (bpmem.tevorders[i].hex & 0x40000) - val1 = (bpmem.tevorders[i].hex & 0x3ff000) >> 12; - else - val1 = (bpmem.tevorders[i].hex & 0x380000) >> 12; - - switch (i % 3) { - case 0: pcurvalue[0] = val0|(val1<<10); break; - case 1: pcurvalue[0] |= val0<<20; pcurvalue[1] = val1; pcurvalue++; break; - case 2: pcurvalue[1] |= (val0<<10)|(val1<<20); pcurvalue++; break; - default: PanicAlert("Unknown case for Tev Stages / 2: %08x", (i % 3)); - } - } - - if (numstages & 1) { // odd - u32 val0; - if (bpmem.tevorders[bpmem.genMode.numtevstages/2].hex & 0x40) - val0 = bpmem.tevorders[bpmem.genMode.numtevstages/2].hex & 0x3ff; - else - val0 = bpmem.tevorders[bpmem.genMode.numtevstages/2].hex & 0x380; - - switch (bpmem.genMode.numtevstages % 3) - { - case 0: pcurvalue[0] = val0; break; - case 1: pcurvalue[0] |= val0 << 20; break; - case 2: pcurvalue[1] |= val0 << 10; pcurvalue++; break; - default: PanicAlert("Unknown case for Tev Stages: %08x", bpmem.genMode.numtevstages % 3); - } - } - - if ((bpmem.genMode.numtevstages % 3) != 2) - ++pcurvalue; - - uid->tevstages = (u32)(pcurvalue - &uid->values[0] - hdr); - - for (u32 i = 0; i < bpmem.genMode.numindstages; ++i) - { - u32 val = bpmem.tevind[i].hex & 0x1fffff; // 21 bits - switch (i % 3) - { - case 0: pcurvalue[0] = val; break; - case 1: pcurvalue[0] |= val << 21; pcurvalue[1] = val >> 11; ++pcurvalue; break; - case 2: pcurvalue[0] |= val << 10; ++pcurvalue; break; - default: PanicAlert("Unknown case for Ind Stages: %08x", (i % 3)); - } - } - - // yeah, well .... - uid->indstages = (u32)(pcurvalue - &uid->values[0] - (hdr - 1) - uid->tevstages); - + uid->num_values = (ptr+1) - uid->values; } void GetSafePixelShaderId(PIXELSHADERUIDSAFE *uid, DSTALPHA_MODE dstAlphaMode) @@ -346,19 +224,21 @@ void GetSafePixelShaderId(PIXELSHADERUIDSAFE *uid, DSTALPHA_MODE dstAlphaMode) for (int i = 0; i < bpmem.genMode.numtevstages+1; ++i) // up to 16 times { - // TODO ... - StageHash(i, ptr); - ptr += 4; // max: ptr = &uid->values[33+63] + *ptr++ = bpmem.combiners[i].colorC.hex; // 33+5*i + *ptr++ = bpmem.combiners[i].alphaC.hex; // 34+5*i + *ptr++ = bpmem.tevind[i].hex; // 35+5*i + *ptr++ = bpmem.tevksel[i/2].hex; // 36+5*i + *ptr++ = bpmem.tevorders[i/2].hex; // 37+5*i } - ptr = &uid->values[97]; + ptr = &uid->values[113]; - *ptr++ = bpmem.alphaFunc.hex; // 97 + *ptr++ = bpmem.alphaFunc.hex; // 113 - *ptr++ = bpmem.fog.c_proj_fsel.hex; // 98 - *ptr++ = bpmem.fogRange.Base.hex; // 99 + *ptr++ = bpmem.fog.c_proj_fsel.hex; // 114 + *ptr++ = bpmem.fogRange.Base.hex; // 115 - _assert_((ptr - uid->values) <= uid->GetNumValues()); + _assert_((ptr - uid->values) == uid->GetNumValues()); } -- cgit v1.2.3 From b28348066e083d532d5ab62c93ce0f1dc5c217ed Mon Sep 17 00:00:00 2001 From: NeoBrainX Date: Fri, 9 Sep 2011 00:32:04 +0200 Subject: Improve the shader UID debugging stuff and merge it to VideoCommon, effectively enabling it in D3D9 and D3D11 as well. --- Source/Core/VideoCommon/Src/PixelShaderGen.cpp | 45 ++++++++++++++++++++++---- 1 file changed, 39 insertions(+), 6 deletions(-) (limited to 'Source/Core/VideoCommon/Src/PixelShaderGen.cpp') diff --git a/Source/Core/VideoCommon/Src/PixelShaderGen.cpp b/Source/Core/VideoCommon/Src/PixelShaderGen.cpp index 8aa3a0a56e..25007be93b 100644 --- a/Source/Core/VideoCommon/Src/PixelShaderGen.cpp +++ b/Source/Core/VideoCommon/Src/PixelShaderGen.cpp @@ -241,6 +241,39 @@ void GetSafePixelShaderId(PIXELSHADERUIDSAFE *uid, DSTALPHA_MODE dstAlphaMode) _assert_((ptr - uid->values) == uid->GetNumValues()); } +void ValidatePixelShaderIDs(API_TYPE api, PIXELSHADERUIDSAFE old_id, const std::string& old_code, DSTALPHA_MODE dstAlphaMode, u32 components) +{ + PIXELSHADERUIDSAFE new_id; + GetSafePixelShaderId(&new_id, dstAlphaMode); + + if (!(old_id == new_id)) + { + std::string new_code(GeneratePixelShaderCode(dstAlphaMode, api, components)); + if (old_code != new_code) + { + _assert_(old_id.GetNumValues() == new_id.GetNumValues()); + + char msg[8192]; + char* ptr = msg; + ptr += sprintf(ptr, "Pixel shader IDs matched but unique IDs did not!\nUnique IDs (old <-> new):\n"); + const int N = new_id.GetNumValues(); + for (int i = 0; i < N/2; ++i) + ptr += sprintf(ptr, "%02d, %08X %08X | %08X %08X\n", 2*i, old_id.values[2*i], old_id.values[2*i+1], + new_id.values[2*i], new_id.values[2*i+1]); + if (N % 2) + ptr += sprintf(ptr, "%02d, %08X | %08X\n", N-1, old_id.values[N-1], new_id.values[N-1]); + + static int num_failures = 0; + char szTemp[MAX_PATH]; + sprintf(szTemp, "%spsuid_mismatch_%04i.txt", File::GetUserPath(D_DUMP_IDX).c_str(), num_failures++); + std::ofstream file(szTemp); + file << msg; + file.close(); + + PanicAlert("Unique pixel shader ID mismatch!\n\nReport this to the devs, along with the contents of %s.", szTemp); + } + } +} // old tev->pixelshader notes // @@ -594,12 +627,12 @@ const char *GeneratePixelShaderCode(DSTALPHA_MODE dstAlphaMode, API_TYPE ApiType WRITE(p, " float4 c0 = "I_COLORS"[1], c1 = "I_COLORS"[2], c2 = "I_COLORS"[3], prev = float4(0.0f, 0.0f, 0.0f, 0.0f), textemp = float4(0.0f, 0.0f, 0.0f, 0.0f), rastemp = float4(0.0f, 0.0f, 0.0f, 0.0f), konsttemp = float4(0.0f, 0.0f, 0.0f, 0.0f);\n" " float3 comp16 = float3(1.0f, 255.0f, 0.0f), comp24 = float3(1.0f, 255.0f, 255.0f*255.0f);\n" - " float4 alphabump=float4(0.0f,0.0f,0.0f,0.0f);\n" - " float3 tevcoord=float3(0.0f, 0.0f, 0.0f);\n" - " float2 wrappedcoord=float2(0.0f,0.0f), tempcoord=float2(0.0f,0.0f);\n" - " float4 cc0=float4(0.0f,0.0f,0.0f,0.0f), cc1=float4(0.0f,0.0f,0.0f,0.0f);\n" - " float4 cc2=float4(0.0f,0.0f,0.0f,0.0f), cprev=float4(0.0f,0.0f,0.0f,0.0f);\n" - " float4 crastemp=float4(0.0f,0.0f,0.0f,0.0f),ckonsttemp=float4(0.0f,0.0f,0.0f,0.0f);\n\n"); + " float4 alphabump=float4(0.0f,0.0f,0.0f,0.0f);\n" + " float3 tevcoord=float3(0.0f, 0.0f, 0.0f);\n" + " float2 wrappedcoord=float2(0.0f,0.0f), tempcoord=float2(0.0f,0.0f);\n" + " float4 cc0=float4(0.0f,0.0f,0.0f,0.0f), cc1=float4(0.0f,0.0f,0.0f,0.0f);\n" + " float4 cc2=float4(0.0f,0.0f,0.0f,0.0f), cprev=float4(0.0f,0.0f,0.0f,0.0f);\n" + " float4 crastemp=float4(0.0f,0.0f,0.0f,0.0f),ckonsttemp=float4(0.0f,0.0f,0.0f,0.0f);\n\n"); if(g_ActiveConfig.bEnablePixelLighting && g_ActiveConfig.backend_info.bSupportsPixelLighting) { -- cgit v1.2.3 From a021dd7b79062d20552852d32387cf6b36ab06e6 Mon Sep 17 00:00:00 2001 From: NeoBrainX Date: Fri, 9 Sep 2011 21:45:11 +0200 Subject: Small fix to the previous commit. --- Source/Core/VideoCommon/Src/PixelShaderGen.cpp | 3 +++ 1 file changed, 3 insertions(+) (limited to 'Source/Core/VideoCommon/Src/PixelShaderGen.cpp') diff --git a/Source/Core/VideoCommon/Src/PixelShaderGen.cpp b/Source/Core/VideoCommon/Src/PixelShaderGen.cpp index 25007be93b..1ea7479990 100644 --- a/Source/Core/VideoCommon/Src/PixelShaderGen.cpp +++ b/Source/Core/VideoCommon/Src/PixelShaderGen.cpp @@ -243,6 +243,9 @@ void GetSafePixelShaderId(PIXELSHADERUIDSAFE *uid, DSTALPHA_MODE dstAlphaMode) void ValidatePixelShaderIDs(API_TYPE api, PIXELSHADERUIDSAFE old_id, const std::string& old_code, DSTALPHA_MODE dstAlphaMode, u32 components) { + if (!g_ActiveConfig.bEnableShaderDebugging) + return; + PIXELSHADERUIDSAFE new_id; GetSafePixelShaderId(&new_id, dstAlphaMode); -- cgit v1.2.3 From 5d075ce507c180dc89b9731eaa2f52e729caf8f6 Mon Sep 17 00:00:00 2001 From: NeoBrainX Date: Sat, 10 Sep 2011 03:10:28 +0200 Subject: - D3D9: pass the correct API type to ValidatePixelShaderIDs - don't load shader cache from disk in d3d9/11 if shader debugging is enabled (we won't have any info about the source shader code otherwise, etc) - dump shader source codes on safe UIDs mismatch Thanks to LordMark and [SS] for reporting those to me ;) --- Source/Core/VideoCommon/Src/PixelShaderGen.cpp | 2 ++ 1 file changed, 2 insertions(+) (limited to 'Source/Core/VideoCommon/Src/PixelShaderGen.cpp') diff --git a/Source/Core/VideoCommon/Src/PixelShaderGen.cpp b/Source/Core/VideoCommon/Src/PixelShaderGen.cpp index 1ea7479990..7937c695ce 100644 --- a/Source/Core/VideoCommon/Src/PixelShaderGen.cpp +++ b/Source/Core/VideoCommon/Src/PixelShaderGen.cpp @@ -271,6 +271,8 @@ void ValidatePixelShaderIDs(API_TYPE api, PIXELSHADERUIDSAFE old_id, const std:: sprintf(szTemp, "%spsuid_mismatch_%04i.txt", File::GetUserPath(D_DUMP_IDX).c_str(), num_failures++); std::ofstream file(szTemp); file << msg; + file << "\n\nOld shader code:\n" << old_code; + file << "\n\nNew shader code:\n" << new_code; file.close(); PanicAlert("Unique pixel shader ID mismatch!\n\nReport this to the devs, along with the contents of %s.", szTemp); -- cgit v1.2.3 From ca7e8a9e88aeb2271ed04af4ff3994286c00a90c Mon Sep 17 00:00:00 2001 From: NeoBrainX Date: Thu, 29 Sep 2011 21:21:09 +0200 Subject: Fix pixel lighting. --- Source/Core/VideoCommon/Src/PixelShaderGen.cpp | 18 +++++++++++------- 1 file changed, 11 insertions(+), 7 deletions(-) (limited to 'Source/Core/VideoCommon/Src/PixelShaderGen.cpp') diff --git a/Source/Core/VideoCommon/Src/PixelShaderGen.cpp b/Source/Core/VideoCommon/Src/PixelShaderGen.cpp index 7937c695ce..52a4477390 100644 --- a/Source/Core/VideoCommon/Src/PixelShaderGen.cpp +++ b/Source/Core/VideoCommon/Src/PixelShaderGen.cpp @@ -125,16 +125,11 @@ void GetPixelShaderId(PIXELSHADERUID *uid, DSTALPHA_MODE dstAlphaMode) if (alphaPreTest == 1 || (alphaPreTest && !DepthTextureEnable && dstAlphaMode == DSTALPHA_ALPHA_PASS)) { // Courtesy of PreAlphaTest, we're done already ;) - // TODO: There's a comment including bpmem.genmode.numindstages.. shouldnt really bother about that though. + // NOTE: The comment header of generated shaders depends on the value of bpmem.genmode.numindstages.. shouldnt really bother about that though. uid->num_values = 1; return; } - if (enablePL) - { - // TODO: Include register states for lighting shader - } - for (unsigned int i = 0; i < bpmem.genMode.numtexgens; ++i) { if (18+i < 32) @@ -194,7 +189,12 @@ void GetPixelShaderId(PIXELSHADERUID *uid, DSTALPHA_MODE dstAlphaMode) ptr[0] |= bpmem.fogRange.Base.Enabled << 17; // 1 } } - uid->num_values = (ptr+1) - uid->values; + + ++ptr; + if (enablePL) + ptr += GetLightingShaderId(ptr); + + uid->num_values = ptr - uid->values; } void GetSafePixelShaderId(PIXELSHADERUIDSAFE *uid, DSTALPHA_MODE dstAlphaMode) @@ -212,6 +212,10 @@ void GetSafePixelShaderId(PIXELSHADERUIDSAFE *uid, DSTALPHA_MODE dstAlphaMode) if (g_ActiveConfig.bEnablePixelLighting && g_ActiveConfig.backend_info.bSupportsPixelLighting) { // TODO: Include register states for lighting shader + *ptr++ = xfregs.color[0].hex; + *ptr++ = xfregs.alpha[0].hex; + *ptr++ = xfregs.color[1].hex; + *ptr++ = xfregs.alpha[1].hex; } for (unsigned int i = 0; i < 8; ++i) -- cgit v1.2.3 From ddfe219293e3af9e04f7020f0557736e1762370b Mon Sep 17 00:00:00 2001 From: NeoBrainX Date: Thu, 29 Sep 2011 21:52:13 +0200 Subject: Fixup line endings. --- Source/Core/VideoCommon/Src/PixelShaderGen.cpp | 2482 ++++++++++++------------ 1 file changed, 1241 insertions(+), 1241 deletions(-) (limited to 'Source/Core/VideoCommon/Src/PixelShaderGen.cpp') diff --git a/Source/Core/VideoCommon/Src/PixelShaderGen.cpp b/Source/Core/VideoCommon/Src/PixelShaderGen.cpp index 52a4477390..f52c6b4190 100644 --- a/Source/Core/VideoCommon/Src/PixelShaderGen.cpp +++ b/Source/Core/VideoCommon/Src/PixelShaderGen.cpp @@ -1,1242 +1,1242 @@ -// Copyright (C) 2003 Dolphin Project. - -// This program is free software: you can redistribute it and/or modify -// it under the terms of the GNU General Public License as published by -// the Free Software Foundation, version 2.0. - -// This program is distributed in the hope that it will be useful, -// but WITHOUT ANY WARRANTY; without even the implied warranty of -// MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the -// GNU General Public License 2.0 for more details. - -// A copy of the GPL 2.0 should have been included with the program. -// If not, see http://www.gnu.org/licenses/ - -// Official SVN repository and contact information can be found at -// http://code.google.com/p/dolphin-emu/ - -#include -#include -#include -#include - -#include "LightingShaderGen.h" -#include "PixelShaderGen.h" -#include "XFMemory.h" // for texture projection mode -#include "BPMemory.h" -#include "VideoConfig.h" -#include "NativeVertexFormat.h" - -PIXELSHADERUID last_pixel_shader_uid; - -static int AlphaPreTest(); - -static void StageHash(int stage, u32* out) -{ - out[0] |= bpmem.combiners[stage].colorC.hex & 0xFFFFFF; // 24 - u32 alphaC = bpmem.combiners[stage].alphaC.hex & 0xFFFFF0; // 24, strip out tswap and rswap for now - out[0] |= (alphaC&0xF0) << 24; // 8 - out[1] |= alphaC >> 8; // 16 - - // reserve 3 bits for bpmem.tevorders[stage/2].getTexMap - out[1] |= bpmem.tevorders[stage/2].getTexCoord(stage&1) << 19; // 3 - out[1] |= bpmem.tevorders[stage/2].getEnable(stage&1) << 22; // 1 - // reserve 3 bits for bpmem.tevorders[stage/2].getColorChan - - bool bHasIndStage = bpmem.tevind[stage].IsActive() && bpmem.tevind[stage].bt < bpmem.genMode.numindstages; - out[2] |= bHasIndStage << 2; // 1 - - bool needstexcoord = false; - - if (bHasIndStage) - { - out[2] |= (bpmem.tevind[stage].hex & 0x17FFFF) << 3; // 21, TODO: needs an explanation - needstexcoord = true; - } - - - TevStageCombiner::ColorCombiner& cc = bpmem.combiners[stage].colorC; - TevStageCombiner::AlphaCombiner& ac = bpmem.combiners[stage].alphaC; - - if(cc.a == TEVCOLORARG_RASA || cc.a == TEVCOLORARG_RASC - || cc.b == TEVCOLORARG_RASA || cc.b == TEVCOLORARG_RASC - || cc.c == TEVCOLORARG_RASA || cc.c == TEVCOLORARG_RASC - || cc.d == TEVCOLORARG_RASA || cc.d == TEVCOLORARG_RASC - || ac.a == TEVALPHAARG_RASA || ac.b == TEVALPHAARG_RASA - || ac.c == TEVALPHAARG_RASA || ac.d == TEVALPHAARG_RASA) - { - out[0] |= bpmem.combiners[stage].alphaC.rswap; - out[2] |= bpmem.tevksel[bpmem.combiners[stage].alphaC.rswap*2].swap1 << 24; // 2 - out[2] |= bpmem.tevksel[bpmem.combiners[stage].alphaC.rswap*2].swap2 << 26; // 2 - out[2] |= bpmem.tevksel[bpmem.combiners[stage].alphaC.rswap*2+1].swap1 << 28; // 2 - out[2] |= bpmem.tevksel[bpmem.combiners[stage].alphaC.rswap*2+1].swap2 << 30; // 2 - out[1] |= (bpmem.tevorders[stage/2].getColorChan(stage&1)&1) << 23; - out[2] |= (bpmem.tevorders[stage/2].getColorChan(stage&1)&0x6) >> 1; - } - - out[3] |= bpmem.tevorders[stage/2].getEnable(stage&1); - if (bpmem.tevorders[stage/2].getEnable(stage&1)) - { - if (bHasIndStage) needstexcoord = true; - - out[0] |= bpmem.combiners[stage].alphaC.tswap; - out[3] |= bpmem.tevksel[bpmem.combiners[stage].alphaC.tswap*2].swap1 << 1; // 2 - out[3] |= bpmem.tevksel[bpmem.combiners[stage].alphaC.tswap*2].swap2 << 3; // 2 - out[3] |= bpmem.tevksel[bpmem.combiners[stage].alphaC.tswap*2+1].swap1 << 5; // 2 - out[3] |= bpmem.tevksel[bpmem.combiners[stage].alphaC.tswap*2+1].swap2 << 7; // 2 - out[1] |= bpmem.tevorders[stage/2].getTexMap(stage&1) << 16; - } - - if (cc.a == TEVCOLORARG_KONST || cc.b == TEVCOLORARG_KONST || cc.c == TEVCOLORARG_KONST || cc.d == TEVCOLORARG_KONST - || ac.a == TEVALPHAARG_KONST || ac.b == TEVALPHAARG_KONST || ac.c == TEVALPHAARG_KONST || ac.d == TEVALPHAARG_KONST) - { - out[3] |= bpmem.tevksel[stage/2].getKC(stage&1) << 9; // 5 - out[3] |= bpmem.tevksel[stage/2].getKA(stage&1) << 14; // 5 - } - - if (needstexcoord) - { - out[1] |= bpmem.tevorders[stage/2].getTexCoord(stage&1) << 16; - } -} - -// Mash together all the inputs that contribute to the code of a generated pixel shader into -// 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, DSTALPHA_MODE dstAlphaMode) -{ - uid->values[0] |= bpmem.genMode.numtevstages; // 4 - uid->values[0] |= bpmem.genMode.numtexgens << 4; // 4 - uid->values[0] |= dstAlphaMode << 8; // 2 - - bool DepthTextureEnable = (bpmem.ztex2.op != ZTEXTURE_DISABLE && !bpmem.zcontrol.zcomploc && bpmem.zmode.testenable && bpmem.zmode.updateenable) || g_ActiveConfig.bEnablePerPixelDepth; - - uid->values[0] |= DepthTextureEnable << 10; // 1 - - bool enablePL = g_ActiveConfig.bEnablePixelLighting && g_ActiveConfig.backend_info.bSupportsPixelLighting; - uid->values[0] |= enablePL << 11; // 1 - - if (!enablePL) uid->values[0] |= xfregs.numTexGen.numTexGens << 12; // 4 - u32 alphaPreTest = AlphaPreTest()+1; - - uid->values[0] |= alphaPreTest << 16; // 2 - - if (alphaPreTest == 1 || (alphaPreTest && !DepthTextureEnable && dstAlphaMode == DSTALPHA_ALPHA_PASS)) - { - // Courtesy of PreAlphaTest, we're done already ;) - // NOTE: The comment header of generated shaders depends on the value of bpmem.genmode.numindstages.. shouldnt really bother about that though. - uid->num_values = 1; - return; - } - - for (unsigned int i = 0; i < bpmem.genMode.numtexgens; ++i) - { - if (18+i < 32) - uid->values[0] |= xfregs.texMtxInfo[i].projection << (18+i); // 1 - else - uid->values[1] |= xfregs.texMtxInfo[i].projection << (i - 14); // 1 - } - - uid->values[1] = bpmem.genMode.numindstages << 2; // 3 - u32 indirectStagesUsed = 0; - for (unsigned int i = 0; i < bpmem.genMode.numindstages; ++i) - if (bpmem.tevind[i].IsActive() && bpmem.tevind[i].bt < bpmem.genMode.numindstages) - indirectStagesUsed |= (1 << bpmem.tevind[i].bt); - - assert(indirectStagesUsed == (indirectStagesUsed & 0xF)); - - uid->values[1] |= indirectStagesUsed << 5; // 4; - - for (unsigned int i = 0; i < bpmem.genMode.numindstages; ++i) - { - if (indirectStagesUsed & (1 << i)) - { - uid->values[1] |= (bpmem.tevindref.getTexCoord(i) < bpmem.genMode.numtexgens) << (9 + 3*i); // 1 - if (bpmem.tevindref.getTexCoord(i) < bpmem.genMode.numtexgens) - uid->values[1] |= bpmem.tevindref.getTexCoord(i) << (10 + 3*i); // 2 - } - } - - u32* ptr = &uid->values[2]; - for (int i = 0; i < bpmem.genMode.numtevstages+1; ++i) - { - StageHash(i, ptr); - ptr += 4; // max: ptr = &uid->values[66] - } - - ptr[0] |= bpmem.alphaFunc.comp0; // 3 - ptr[0] |= bpmem.alphaFunc.comp1 << 3; // 3 - ptr[0] |= bpmem.alphaFunc.logic << 6; // 2 - - if (alphaPreTest == 0 || alphaPreTest == 2) - { - ptr[0] |= bpmem.fog.c_proj_fsel.fsel << 8; // 3 - if (DepthTextureEnable) - { - ptr[0] |= bpmem.ztex2.op << 11; // 2 - ptr[0] |= bpmem.zcontrol.zcomploc << 13; // 1 - ptr[0] |= bpmem.zmode.testenable << 14; // 1 - ptr[0] |= bpmem.zmode.updateenable << 15; // 1 - } - } - - if (dstAlphaMode != DSTALPHA_ALPHA_PASS) - { - if (bpmem.fog.c_proj_fsel.fsel != 0) - { - ptr[0] |= bpmem.fog.c_proj_fsel.proj << 16; // 1 - ptr[0] |= bpmem.fogRange.Base.Enabled << 17; // 1 - } - } - - ++ptr; - if (enablePL) - ptr += GetLightingShaderId(ptr); - - uid->num_values = ptr - uid->values; -} - -void GetSafePixelShaderId(PIXELSHADERUIDSAFE *uid, DSTALPHA_MODE dstAlphaMode) -{ - u32* ptr = uid->values; - *ptr++ = dstAlphaMode; // 0 - *ptr++ = bpmem.genMode.hex; // 1 - *ptr++ = bpmem.ztex2.hex; // 2 - *ptr++ = bpmem.zcontrol.hex; // 3 - *ptr++ = bpmem.zmode.hex; // 4 - *ptr++ = g_ActiveConfig.bEnablePerPixelDepth; // 5 - *ptr++ = g_ActiveConfig.bEnablePixelLighting && g_ActiveConfig.backend_info.bSupportsPixelLighting; // 6 - *ptr++ = xfregs.numTexGen.hex; // 7 - - if (g_ActiveConfig.bEnablePixelLighting && g_ActiveConfig.backend_info.bSupportsPixelLighting) - { - // TODO: Include register states for lighting shader - *ptr++ = xfregs.color[0].hex; - *ptr++ = xfregs.alpha[0].hex; - *ptr++ = xfregs.color[1].hex; - *ptr++ = xfregs.alpha[1].hex; - } - - for (unsigned int i = 0; i < 8; ++i) - *ptr++ = xfregs.texMtxInfo[i].hex; // 8-15 - - for (unsigned int i = 0; i < 16; ++i) - *ptr++ = bpmem.tevind[i].hex; // 16-31 - - *ptr++ = bpmem.tevindref.hex; // 32 - - for (int i = 0; i < bpmem.genMode.numtevstages+1; ++i) // up to 16 times - { - *ptr++ = bpmem.combiners[i].colorC.hex; // 33+5*i - *ptr++ = bpmem.combiners[i].alphaC.hex; // 34+5*i - *ptr++ = bpmem.tevind[i].hex; // 35+5*i - *ptr++ = bpmem.tevksel[i/2].hex; // 36+5*i - *ptr++ = bpmem.tevorders[i/2].hex; // 37+5*i - } - - ptr = &uid->values[113]; - - *ptr++ = bpmem.alphaFunc.hex; // 113 - - *ptr++ = bpmem.fog.c_proj_fsel.hex; // 114 - *ptr++ = bpmem.fogRange.Base.hex; // 115 - - _assert_((ptr - uid->values) == uid->GetNumValues()); -} - -void ValidatePixelShaderIDs(API_TYPE api, PIXELSHADERUIDSAFE old_id, const std::string& old_code, DSTALPHA_MODE dstAlphaMode, u32 components) -{ - if (!g_ActiveConfig.bEnableShaderDebugging) - return; - - PIXELSHADERUIDSAFE new_id; - GetSafePixelShaderId(&new_id, dstAlphaMode); - - if (!(old_id == new_id)) - { - std::string new_code(GeneratePixelShaderCode(dstAlphaMode, api, components)); - if (old_code != new_code) - { - _assert_(old_id.GetNumValues() == new_id.GetNumValues()); - - char msg[8192]; - char* ptr = msg; - ptr += sprintf(ptr, "Pixel shader IDs matched but unique IDs did not!\nUnique IDs (old <-> new):\n"); - const int N = new_id.GetNumValues(); - for (int i = 0; i < N/2; ++i) - ptr += sprintf(ptr, "%02d, %08X %08X | %08X %08X\n", 2*i, old_id.values[2*i], old_id.values[2*i+1], - new_id.values[2*i], new_id.values[2*i+1]); - if (N % 2) - ptr += sprintf(ptr, "%02d, %08X | %08X\n", N-1, old_id.values[N-1], new_id.values[N-1]); - - static int num_failures = 0; - char szTemp[MAX_PATH]; - sprintf(szTemp, "%spsuid_mismatch_%04i.txt", File::GetUserPath(D_DUMP_IDX).c_str(), num_failures++); - std::ofstream file(szTemp); - file << msg; - file << "\n\nOld shader code:\n" << old_code; - file << "\n\nNew shader code:\n" << new_code; - file.close(); - - PanicAlert("Unique pixel shader ID mismatch!\n\nReport this to the devs, along with the contents of %s.", szTemp); - } - } -} - -// old tev->pixelshader notes -// -// color for this stage (alpha, color) is given by bpmem.tevorders[0].colorchan0 -// konstant for this stage (alpha, color) is given by bpmem.tevksel -// inputs are given by bpmem.combiners[0].colorC.a/b/c/d << could be current chan color -// according to GXTevColorArg table above -// output is given by .outreg -// tevtemp is set according to swapmodetables and - -static void WriteStage(char *&p, int n, API_TYPE ApiType); -static void SampleTexture(char *&p, const char *destination, const char *texcoords, const char *texswap, int texmap, API_TYPE ApiType); -// static void WriteAlphaCompare(char *&p, int num, int comp); -static bool WriteAlphaTest(char *&p, API_TYPE ApiType,DSTALPHA_MODE dstAlphaMode); -static void WriteFog(char *&p); - -static const char *tevKSelTableC[] = // KCSEL -{ - "1.0f,1.0f,1.0f", // 1 = 0x00 - "0.875f,0.875f,0.875f", // 7_8 = 0x01 - "0.75f,0.75f,0.75f", // 3_4 = 0x02 - "0.625f,0.625f,0.625f", // 5_8 = 0x03 - "0.5f,0.5f,0.5f", // 1_2 = 0x04 - "0.375f,0.375f,0.375f", // 3_8 = 0x05 - "0.25f,0.25f,0.25f", // 1_4 = 0x06 - "0.125f,0.125f,0.125f", // 1_8 = 0x07 - "ERROR", // 0x08 - "ERROR", // 0x09 - "ERROR", // 0x0a - "ERROR", // 0x0b - I_KCOLORS"[0].rgb", // K0 = 0x0C - I_KCOLORS"[1].rgb", // K1 = 0x0D - I_KCOLORS"[2].rgb", // K2 = 0x0E - I_KCOLORS"[3].rgb", // K3 = 0x0F - I_KCOLORS"[0].rrr", // K0_R = 0x10 - I_KCOLORS"[1].rrr", // K1_R = 0x11 - I_KCOLORS"[2].rrr", // K2_R = 0x12 - I_KCOLORS"[3].rrr", // K3_R = 0x13 - I_KCOLORS"[0].ggg", // K0_G = 0x14 - I_KCOLORS"[1].ggg", // K1_G = 0x15 - I_KCOLORS"[2].ggg", // K2_G = 0x16 - I_KCOLORS"[3].ggg", // K3_G = 0x17 - I_KCOLORS"[0].bbb", // K0_B = 0x18 - I_KCOLORS"[1].bbb", // K1_B = 0x19 - I_KCOLORS"[2].bbb", // K2_B = 0x1A - I_KCOLORS"[3].bbb", // K3_B = 0x1B - I_KCOLORS"[0].aaa", // K0_A = 0x1C - I_KCOLORS"[1].aaa", // K1_A = 0x1D - I_KCOLORS"[2].aaa", // K2_A = 0x1E - I_KCOLORS"[3].aaa", // K3_A = 0x1F -}; - -static const char *tevKSelTableA[] = // KASEL -{ - "1.0f", // 1 = 0x00 - "0.875f",// 7_8 = 0x01 - "0.75f", // 3_4 = 0x02 - "0.625f",// 5_8 = 0x03 - "0.5f", // 1_2 = 0x04 - "0.375f",// 3_8 = 0x05 - "0.25f", // 1_4 = 0x06 - "0.125f",// 1_8 = 0x07 - "ERROR", // 0x08 - "ERROR", // 0x09 - "ERROR", // 0x0a - "ERROR", // 0x0b - "ERROR", // 0x0c - "ERROR", // 0x0d - "ERROR", // 0x0e - "ERROR", // 0x0f - I_KCOLORS"[0].r", // K0_R = 0x10 - I_KCOLORS"[1].r", // K1_R = 0x11 - I_KCOLORS"[2].r", // K2_R = 0x12 - I_KCOLORS"[3].r", // K3_R = 0x13 - I_KCOLORS"[0].g", // K0_G = 0x14 - I_KCOLORS"[1].g", // K1_G = 0x15 - I_KCOLORS"[2].g", // K2_G = 0x16 - I_KCOLORS"[3].g", // K3_G = 0x17 - I_KCOLORS"[0].b", // K0_B = 0x18 - I_KCOLORS"[1].b", // K1_B = 0x19 - I_KCOLORS"[2].b", // K2_B = 0x1A - I_KCOLORS"[3].b", // K3_B = 0x1B - I_KCOLORS"[0].a", // K0_A = 0x1C - I_KCOLORS"[1].a", // K1_A = 0x1D - I_KCOLORS"[2].a", // K2_A = 0x1E - I_KCOLORS"[3].a", // K3_A = 0x1F -}; - -static const char *tevScaleTable[] = // CS -{ - "1.0f", // SCALE_1 - "2.0f", // SCALE_2 - "4.0f", // SCALE_4 - "0.5f", // DIVIDE_2 -}; - -static const char *tevBiasTable[] = // TB -{ - "", // ZERO, - "+0.5f", // ADDHALF, - "-0.5f", // SUBHALF, - "", -}; - -static const char *tevOpTable[] = { // TEV - "+", // TEVOP_ADD = 0, - "-", // TEVOP_SUB = 1, -}; - -static const char *tevCInputTable[] = // CC -{ - "(prev.rgb)", // CPREV, - "(prev.aaa)", // APREV, - "(c0.rgb)", // C0, - "(c0.aaa)", // A0, - "(c1.rgb)", // C1, - "(c1.aaa)", // A1, - "(c2.rgb)", // C2, - "(c2.aaa)", // A2, - "(textemp.rgb)", // TEXC, - "(textemp.aaa)", // TEXA, - "(rastemp.rgb)", // RASC, - "(rastemp.aaa)", // RASA, - "float3(1.0f, 1.0f, 1.0f)", // ONE - "float3(0.5f, 0.5f, 0.5f)", // HALF - "(konsttemp.rgb)", //"konsttemp.rgb", // KONST - "float3(0.0f, 0.0f, 0.0f)", // ZERO - ///aded extra values to map clamped values - "(cprev.rgb)", // CPREV, - "(cprev.aaa)", // APREV, - "(cc0.rgb)", // C0, - "(cc0.aaa)", // A0, - "(cc1.rgb)", // C1, - "(cc1.aaa)", // A1, - "(cc2.rgb)", // C2, - "(cc2.aaa)", // A2, - "(textemp.rgb)", // TEXC, - "(textemp.aaa)", // TEXA, - "(crastemp.rgb)", // RASC, - "(crastemp.aaa)", // RASA, - "float3(1.0f, 1.0f, 1.0f)", // ONE - "float3(0.5f, 0.5f, 0.5f)", // HALF - "(ckonsttemp.rgb)", //"konsttemp.rgb", // KONST - "float3(0.0f, 0.0f, 0.0f)", // ZERO - "PADERROR", "PADERROR", "PADERROR", "PADERROR" -}; - -static const char *tevAInputTable[] = // CA -{ - "prev", // APREV, - "c0", // A0, - "c1", // A1, - "c2", // A2, - "textemp", // TEXA, - "rastemp", // RASA, - "konsttemp", // KONST, (hw1 had quarter) - "float4(0.0f, 0.0f, 0.0f, 0.0f)", // ZERO - ///aded extra values to map clamped values - "cprev", // APREV, - "cc0", // A0, - "cc1", // A1, - "cc2", // A2, - "textemp", // TEXA, - "crastemp", // RASA, - "ckonsttemp", // KONST, (hw1 had quarter) - "float4(0.0f, 0.0f, 0.0f, 0.0f)", // ZERO - "PADERROR", "PADERROR", "PADERROR", "PADERROR", - "PADERROR", "PADERROR", "PADERROR", "PADERROR", -}; - -static const char *tevRasTable[] = -{ - "colors_0", - "colors_1", - "ERROR", //2 - "ERROR", //3 - "ERROR", //4 - "alphabump", // use bump alpha - "(alphabump*(255.0f/248.0f))", //normalized - "float4(0.0f, 0.0f, 0.0f, 0.0f)", // zero -}; - -//static const char *tevTexFunc[] = { "tex2D", "texRECT" }; - -static const char *tevCOutputTable[] = { "prev.rgb", "c0.rgb", "c1.rgb", "c2.rgb" }; -static const char *tevAOutputTable[] = { "prev.a", "c0.a", "c1.a", "c2.a" }; -static const char *tevIndAlphaSel[] = {"", "x", "y", "z"}; -//static const char *tevIndAlphaScale[] = {"", "*32", "*16", "*8"}; -static const char *tevIndAlphaScale[] = {"*(248.0f/255.0f)", "*(224.0f/255.0f)", "*(240.0f/255.0f)", "*(248.0f/255.0f)"}; -static const char *tevIndBiasField[] = {"", "x", "y", "xy", "z", "xz", "yz", "xyz"}; // indexed by bias -static const char *tevIndBiasAdd[] = {"-128.0f", "1.0f", "1.0f", "1.0f" }; // indexed by fmt -static const char *tevIndWrapStart[] = {"0.0f", "256.0f", "128.0f", "64.0f", "32.0f", "16.0f", "0.001f" }; -static const char *tevIndFmtScale[] = {"255.0f", "31.0f", "15.0f", "7.0f" }; - -#define WRITE p+=sprintf - -static char swapModeTable[4][5]; - -static char text[16384]; -static bool DepthTextureEnable; - -static void BuildSwapModeTable() -{ - static const char *swapColors = "rgba"; - for (int i = 0; i < 4; i++) - { - swapModeTable[i][0] = swapColors[bpmem.tevksel[i*2].swap1]; - swapModeTable[i][1] = swapColors[bpmem.tevksel[i*2].swap2]; - swapModeTable[i][2] = swapColors[bpmem.tevksel[i*2+1].swap1]; - swapModeTable[i][3] = swapColors[bpmem.tevksel[i*2+1].swap2]; - swapModeTable[i][4] = 0; - } -} - -const char *GeneratePixelShaderCode(DSTALPHA_MODE dstAlphaMode, API_TYPE ApiType, u32 components) -{ - setlocale(LC_NUMERIC, "C"); // Reset locale for compilation - text[sizeof(text) - 1] = 0x7C; // canary - - BuildSwapModeTable(); // Needed for WriteStage - int numStages = bpmem.genMode.numtevstages + 1; - int numTexgen = bpmem.genMode.numtexgens; - - char *p = text; - WRITE(p, "//Pixel Shader for TEV stages\n"); - WRITE(p, "//%i TEV stages, %i texgens, XXX IND stages\n", - numStages, numTexgen/*, bpmem.genMode.numindstages*/); - - int nIndirectStagesUsed = 0; - if (bpmem.genMode.numindstages > 0) - { - for (int i = 0; i < numStages; ++i) - { - if (bpmem.tevind[i].IsActive() && bpmem.tevind[i].bt < bpmem.genMode.numindstages) - nIndirectStagesUsed |= 1 << bpmem.tevind[i].bt; - } - } - DepthTextureEnable = (bpmem.ztex2.op != ZTEXTURE_DISABLE && !bpmem.zcontrol.zcomploc && bpmem.zmode.testenable && bpmem.zmode.updateenable) || g_ActiveConfig.bEnablePerPixelDepth ; - // Declare samplers - - if(ApiType != API_D3D11) - { - WRITE(p, "uniform sampler2D "); - } - else - { - WRITE(p, "sampler "); - } - - bool bfirst = true; - for (int i = 0; i < 8; ++i) - { - WRITE(p, "%s samp%d : register(s%d)", bfirst?"":",", i, i); - bfirst = false; - } - WRITE(p, ";\n"); - if(ApiType == API_D3D11) - { - WRITE(p, "Texture2D "); - bfirst = true; - for (int i = 0; i < 8; ++i) - { - WRITE(p, "%s Tex%d : register(t%d)", bfirst?"":",", i, i); - bfirst = false; - } - WRITE(p, ";\n"); - } - - WRITE(p, "\n"); - - WRITE(p, "uniform float4 "I_COLORS"[4] : register(c%d);\n", C_COLORS); - WRITE(p, "uniform float4 "I_KCOLORS"[4] : register(c%d);\n", C_KCOLORS); - WRITE(p, "uniform float4 "I_ALPHA"[1] : register(c%d);\n", C_ALPHA); - WRITE(p, "uniform float4 "I_TEXDIMS"[8] : register(c%d);\n", C_TEXDIMS); - WRITE(p, "uniform float4 "I_ZBIAS"[2] : register(c%d);\n", C_ZBIAS); - WRITE(p, "uniform float4 "I_INDTEXSCALE"[2] : register(c%d);\n", C_INDTEXSCALE); - WRITE(p, "uniform float4 "I_INDTEXMTX"[6] : register(c%d);\n", C_INDTEXMTX); - WRITE(p, "uniform float4 "I_FOG"[3] : register(c%d);\n", C_FOG); - - if(g_ActiveConfig.bEnablePixelLighting && g_ActiveConfig.backend_info.bSupportsPixelLighting) - { - WRITE(p,"typedef struct { float4 col; float4 cosatt; float4 distatt; float4 pos; float4 dir; } Light;\n"); - WRITE(p,"typedef struct { Light lights[8]; } s_"I_PLIGHTS";\n"); - WRITE(p, "uniform s_"I_PLIGHTS" "I_PLIGHTS" : register(c%d);\n", C_PLIGHTS); - WRITE(p, "typedef struct { float4 C0, C1, C2, C3; } s_"I_PMATERIALS";\n"); - WRITE(p, "uniform s_"I_PMATERIALS" "I_PMATERIALS" : register(c%d);\n", C_PMATERIALS); - } - - WRITE(p, "void main(\n"); - if(ApiType != API_D3D11) - { - WRITE(p, " out float4 ocol0 : COLOR0,%s%s\n in float4 rawpos : %s,\n", - dstAlphaMode == DSTALPHA_DUAL_SOURCE_BLEND ? "\n out float4 ocol1 : COLOR1," : "", - DepthTextureEnable ? "\n out float depth : DEPTH," : "", - ApiType & API_OPENGL ? "WPOS" : ApiType & API_D3D9_SM20 ? "POSITION" : "VPOS"); - } - else - { - WRITE(p, " out float4 ocol0 : SV_Target0,%s%s\n in float4 rawpos : SV_Position,\n", - dstAlphaMode == DSTALPHA_DUAL_SOURCE_BLEND ? "\n out float4 ocol1 : SV_Target1," : "", - DepthTextureEnable ? "\n out float depth : SV_Depth," : ""); - } - - WRITE(p, " in float4 colors_0 : COLOR0,\n"); - WRITE(p, " in float4 colors_1 : COLOR1"); - - // compute window position if needed because binding semantic WPOS is not widely supported - if (numTexgen < 7) - { - for (int i = 0; i < numTexgen; ++i) - WRITE(p, ",\n in float3 uv%d : TEXCOORD%d", i, i); - WRITE(p, ",\n in float4 clipPos : TEXCOORD%d", numTexgen); - if(g_ActiveConfig.bEnablePixelLighting && g_ActiveConfig.backend_info.bSupportsPixelLighting) - WRITE(p, ",\n in float4 Normal : TEXCOORD%d", numTexgen + 1); - } - else - { - // wpos is in w of first 4 texcoords - if(g_ActiveConfig.bEnablePixelLighting && g_ActiveConfig.backend_info.bSupportsPixelLighting) - { - for (int i = 0; i < 8; ++i) - WRITE(p, ",\n in float4 uv%d : TEXCOORD%d", i, i); - } - else - { - for (unsigned int i = 0; i < xfregs.numTexGen.numTexGens; ++i) - WRITE(p, ",\n in float%d uv%d : TEXCOORD%d", i < 4 ? 4 : 3 , i, i); - } - } - WRITE(p, " ) {\n"); - - char* pmainstart = p; - int Pretest = AlphaPreTest(); - if (dstAlphaMode == DSTALPHA_ALPHA_PASS && !DepthTextureEnable && Pretest >= 0) - { - if (!Pretest) - { - // alpha test will always fail, so restart the shader and just make it an empty function - WRITE(p, "ocol0 = 0;\n"); - WRITE(p, "discard;\n"); - if(ApiType != API_D3D11) - WRITE(p, "return;\n"); - } - else - { - WRITE(p, " ocol0 = "I_ALPHA"[0].aaaa;\n"); - } - WRITE(p, "}\n"); - return text; - } - - WRITE(p, " float4 c0 = "I_COLORS"[1], c1 = "I_COLORS"[2], c2 = "I_COLORS"[3], prev = float4(0.0f, 0.0f, 0.0f, 0.0f), textemp = float4(0.0f, 0.0f, 0.0f, 0.0f), rastemp = float4(0.0f, 0.0f, 0.0f, 0.0f), konsttemp = float4(0.0f, 0.0f, 0.0f, 0.0f);\n" - " float3 comp16 = float3(1.0f, 255.0f, 0.0f), comp24 = float3(1.0f, 255.0f, 255.0f*255.0f);\n" - " float4 alphabump=float4(0.0f,0.0f,0.0f,0.0f);\n" - " float3 tevcoord=float3(0.0f, 0.0f, 0.0f);\n" - " float2 wrappedcoord=float2(0.0f,0.0f), tempcoord=float2(0.0f,0.0f);\n" - " float4 cc0=float4(0.0f,0.0f,0.0f,0.0f), cc1=float4(0.0f,0.0f,0.0f,0.0f);\n" - " float4 cc2=float4(0.0f,0.0f,0.0f,0.0f), cprev=float4(0.0f,0.0f,0.0f,0.0f);\n" - " float4 crastemp=float4(0.0f,0.0f,0.0f,0.0f),ckonsttemp=float4(0.0f,0.0f,0.0f,0.0f);\n\n"); - - if(g_ActiveConfig.bEnablePixelLighting && g_ActiveConfig.backend_info.bSupportsPixelLighting) - { - if (xfregs.numTexGen.numTexGens < 7) - { - WRITE(p,"float3 _norm0 = normalize(Normal.xyz);\n\n"); - WRITE(p,"float3 pos = float3(clipPos.x,clipPos.y,Normal.w);\n"); - } - else - { - WRITE(p," float3 _norm0 = normalize(float3(uv4.w,uv5.w,uv6.w));\n\n"); - WRITE(p,"float3 pos = float3(uv0.w,uv1.w,uv7.w);\n"); - } - - - WRITE(p, "float4 mat, lacc;\n" - "float3 ldir, h;\n" - "float dist, dist2, attn;\n"); - - p = GenerateLightingShader(p, components, I_PMATERIALS, I_PLIGHTS, "colors_", "colors_"); - } - - if (numTexgen < 7) - WRITE(p, "clipPos = float4(rawpos.x, rawpos.y, clipPos.z, clipPos.w);\n"); - else - WRITE(p, "float4 clipPos = float4(rawpos.x, rawpos.y, uv2.w, uv3.w);\n"); - - // HACK to handle cases where the tex gen is not enabled - if (numTexgen == 0) - { - WRITE(p, "float3 uv0 = float3(0.0f, 0.0f, 0.0f);\n"); - } - else - { - for (int i = 0; i < numTexgen; ++i) - { - // optional perspective divides - if (xfregs.texMtxInfo[i].projection == XF_TEXPROJ_STQ) - { - WRITE(p, "if (uv%d.z)", i); - WRITE(p, " uv%d.xy = uv%d.xy / uv%d.z;\n", i, i, i); - } - - WRITE(p, "uv%d.xy = uv%d.xy * "I_TEXDIMS"[%d].zw;\n", i, i, i); - } - } - - // indirect texture map lookup - for(u32 i = 0; i < bpmem.genMode.numindstages; ++i) - { - if (nIndirectStagesUsed & (1<= %s.r + (0.25f/255.0f)) ? %s : float3(0.0f, 0.0f, 0.0f))",//#define TEVCMP_R8_GT 8 - " %s + ((abs(%s.r - %s.r) < (0.5f/255.0f)) ? %s : float3(0.0f, 0.0f, 0.0f))",//#define TEVCMP_R8_EQ 9 - " %s + (( dot(%s.rgb, comp16) >= (dot(%s.rgb, comp16) + (0.25f/255.0f))) ? %s : float3(0.0f, 0.0f, 0.0f))",//#define TEVCMP_GR16_GT 10 - " %s + (abs(dot(%s.rgb, comp16) - dot(%s.rgb, comp16)) < (0.5f/255.0f) ? %s : float3(0.0f, 0.0f, 0.0f))",//#define TEVCMP_GR16_EQ 11 - " %s + (( dot(%s.rgb, comp24) >= (dot(%s.rgb, comp24) + (0.25f/255.0f))) ? %s : float3(0.0f, 0.0f, 0.0f))",//#define TEVCMP_BGR24_GT 12 - " %s + (abs(dot(%s.rgb, comp24) - dot(%s.rgb, comp24)) < (0.5f/255.0f) ? %s : float3(0.0f, 0.0f, 0.0f))",//#define TEVCMP_BGR24_EQ 13 - " %s + (max(sign(%s.rgb - %s.rgb - (0.25f/255.0f)), float3(0.0f, 0.0f, 0.0f)) * %s)",//#define TEVCMP_RGB8_GT 14 - " %s + ((float3(1.0f, 1.0f, 1.0f) - max(sign(abs(%s.rgb - %s.rgb) - (0.5f/255.0f)), float3(0.0f, 0.0f, 0.0f))) * %s)"//#define TEVCMP_RGB8_EQ 15 -}; - -//table with the alpha compare operations -static const char *TEVCMPAlphaOPTable[16] = -{ - "0.0f",//0 - "0.0f",//1 - "0.0f",//2 - "0.0f",//3 - "0.0f",//4 - "0.0f",//5 - "0.0f",//6 - "0.0f",//7 - " %s.a + ((%s.r >= (%s.r + (0.25f/255.0f))) ? %s.a : 0.0f)",//#define TEVCMP_R8_GT 8 - " %s.a + (abs(%s.r - %s.r) < (0.5f/255.0f) ? %s.a : 0.0f)",//#define TEVCMP_R8_EQ 9 - " %s.a + ((dot(%s.rgb, comp16) >= (dot(%s.rgb, comp16) + (0.25f/255.0f))) ? %s.a : 0.0f)",//#define TEVCMP_GR16_GT 10 - " %s.a + (abs(dot(%s.rgb, comp16) - dot(%s.rgb, comp16)) < (0.5f/255.0f) ? %s.a : 0.0f)",//#define TEVCMP_GR16_EQ 11 - " %s.a + ((dot(%s.rgb, comp24) >= (dot(%s.rgb, comp24) + (0.25f/255.0f))) ? %s.a : 0.0f)",//#define TEVCMP_BGR24_GT 12 - " %s.a + (abs(dot(%s.rgb, comp24) - dot(%s.rgb, comp24)) < (0.5f/255.0f) ? %s.a : 0.0f)",//#define TEVCMP_BGR24_EQ 13 - " %s.a + ((%s.a >= (%s.a + (0.25f/255.0f))) ? %s.a : 0.0f)",//#define TEVCMP_A8_GT 14 - " %s.a + (abs(%s.a - %s.a) < (0.5f/255.0f) ? %s.a : 0.0f)"//#define TEVCMP_A8_EQ 15 - -}; - - -static void WriteStage(char *&p, int n, API_TYPE ApiType) -{ - int texcoord = bpmem.tevorders[n/2].getTexCoord(n&1); - bool bHasTexCoord = (u32)texcoord < bpmem.genMode.numtexgens; - bool bHasIndStage = bpmem.tevind[n].IsActive() && bpmem.tevind[n].bt < bpmem.genMode.numindstages; - - // HACK to handle cases where the tex gen is not enabled - if (!bHasTexCoord) - texcoord = 0; - - WRITE(p, "// TEV stage %d\n", n); - - if (bHasIndStage) - { - WRITE(p, "// indirect op\n"); - // perform the indirect op on the incoming regular coordinates using indtex%d as the offset coords - if (bpmem.tevind[n].bs != ITBA_OFF) - { - WRITE(p, "alphabump = indtex%d.%s %s;\n", - bpmem.tevind[n].bt, - tevIndAlphaSel[bpmem.tevind[n].bs], - tevIndAlphaScale[bpmem.tevind[n].fmt]); - } - // format - WRITE(p, "float3 indtevcrd%d = indtex%d * %s;\n", n, bpmem.tevind[n].bt, tevIndFmtScale[bpmem.tevind[n].fmt]); - - // bias - if (bpmem.tevind[n].bias != ITB_NONE ) - WRITE(p, "indtevcrd%d.%s += %s;\n", n, tevIndBiasField[bpmem.tevind[n].bias], tevIndBiasAdd[bpmem.tevind[n].fmt]); - - // multiply by offset matrix and scale - if (bpmem.tevind[n].mid != 0) - { - if (bpmem.tevind[n].mid <= 3) - { - int mtxidx = 2*(bpmem.tevind[n].mid-1); - WRITE(p, "float2 indtevtrans%d = float2(dot("I_INDTEXMTX"[%d].xyz, indtevcrd%d), dot("I_INDTEXMTX"[%d].xyz, indtevcrd%d));\n", - n, mtxidx, n, mtxidx+1, n); - } - else if (bpmem.tevind[n].mid <= 7 && bHasTexCoord) - { // s matrix - // TODO: Might become negative? - int mtxidx = 2*(bpmem.tevind[n].mid-5); - WRITE(p, "float2 indtevtrans%d = "I_INDTEXMTX"[%d].ww * uv%d.xy * indtevcrd%d.xx;\n", n, mtxidx, texcoord, n); - } - else if (bpmem.tevind[n].mid <= 11 && bHasTexCoord) - { // t matrix - // TODO: Might become negative? - int mtxidx = 2*(bpmem.tevind[n].mid-9); - WRITE(p, "float2 indtevtrans%d = "I_INDTEXMTX"[%d].ww * uv%d.xy * indtevcrd%d.yy;\n", n, mtxidx, texcoord, n); - } - else - WRITE(p, "float2 indtevtrans%d = 0;\n", n); - } - else - WRITE(p, "float2 indtevtrans%d = 0;\n", n); - - // --------- - // Wrapping - // --------- - - // wrap S - if (bpmem.tevind[n].sw == ITW_OFF) - WRITE(p, "wrappedcoord.x = uv%d.x;\n", texcoord); - else if (bpmem.tevind[n].sw == ITW_0) - WRITE(p, "wrappedcoord.x = 0.0f;\n"); - else - WRITE(p, "wrappedcoord.x = fmod( uv%d.x, %s );\n", texcoord, tevIndWrapStart[bpmem.tevind[n].sw]); - - // wrap T - if (bpmem.tevind[n].tw == ITW_OFF) - WRITE(p, "wrappedcoord.y = uv%d.y;\n", texcoord); - else if (bpmem.tevind[n].tw == ITW_0) - WRITE(p, "wrappedcoord.y = 0.0f;\n"); - else - WRITE(p, "wrappedcoord.y = fmod( uv%d.y, %s );\n", texcoord, tevIndWrapStart[bpmem.tevind[n].tw]); - - if (bpmem.tevind[n].fb_addprev) // add previous tevcoord - WRITE(p, "tevcoord.xy += wrappedcoord + indtevtrans%d;\n", n); - else - WRITE(p, "tevcoord.xy = wrappedcoord + indtevtrans%d;\n", n); - } - - TevStageCombiner::ColorCombiner &cc = bpmem.combiners[n].colorC; - TevStageCombiner::AlphaCombiner &ac = bpmem.combiners[n].alphaC; - - // blah1 - if(cc.a == TEVCOLORARG_RASA || cc.a == TEVCOLORARG_RASC - || cc.b == TEVCOLORARG_RASA || cc.b == TEVCOLORARG_RASC - || cc.c == TEVCOLORARG_RASA || cc.c == TEVCOLORARG_RASC - || cc.d == TEVCOLORARG_RASA || cc.d == TEVCOLORARG_RASC - || ac.a == TEVALPHAARG_RASA || ac.b == TEVALPHAARG_RASA - || ac.c == TEVALPHAARG_RASA || ac.d == TEVALPHAARG_RASA) - { - char *rasswap = swapModeTable[bpmem.combiners[n].alphaC.rswap]; - WRITE(p, "rastemp = %s.%s;\n", tevRasTable[bpmem.tevorders[n / 2].getColorChan(n & 1)], rasswap); - WRITE(p, "crastemp = frac(rastemp * (255.0f/256.0f)) * (256.0f/255.0f);\n"); - } - - - if (bpmem.tevorders[n/2].getEnable(n&1)) - { - if(!bHasIndStage) - { - // calc tevcord - if(bHasTexCoord) - WRITE(p, "tevcoord.xy = uv%d.xy;\n", texcoord); - else - WRITE(p, "tevcoord.xy = float2(0.0f, 0.0f);\n"); - } - - char *texswap = swapModeTable[bpmem.combiners[n].alphaC.tswap]; - int texmap = bpmem.tevorders[n/2].getTexMap(n&1); - SampleTexture(p, "textemp", "tevcoord", texswap, texmap, ApiType); - } - else - WRITE(p, "textemp = float4(1.0f, 1.0f, 1.0f, 1.0f);\n"); - - - // blah2 - if (cc.a == TEVCOLORARG_KONST || cc.b == TEVCOLORARG_KONST || cc.c == TEVCOLORARG_KONST || cc.d == TEVCOLORARG_KONST - || ac.a == TEVALPHAARG_KONST || ac.b == TEVALPHAARG_KONST || ac.c == TEVALPHAARG_KONST || ac.d == TEVALPHAARG_KONST) - { - int kc = bpmem.tevksel[n / 2].getKC(n & 1); - int ka = bpmem.tevksel[n / 2].getKA(n & 1); - WRITE(p, "konsttemp = float4(%s, %s);\n", tevKSelTableC[kc], tevKSelTableA[ka]); - if(kc > 7 || ka > 7) - { - WRITE(p, "ckonsttemp = frac(konsttemp * (255.0f/256.0f)) * (256.0f/255.0f);\n"); - } - else - { - WRITE(p, "ckonsttemp = konsttemp;\n"); - } - } - - if(cc.a == TEVCOLORARG_CPREV || cc.a == TEVCOLORARG_APREV - || cc.b == TEVCOLORARG_CPREV || cc.b == TEVCOLORARG_APREV - || cc.c == TEVCOLORARG_CPREV || cc.c == TEVCOLORARG_APREV - || ac.a == TEVALPHAARG_APREV || ac.b == TEVALPHAARG_APREV || ac.c == TEVALPHAARG_APREV) - WRITE(p, "cprev = frac(prev * (255.0f/256.0f)) * (256.0f/255.0f);\n"); - - - if(cc.a == TEVCOLORARG_C0 || cc.a == TEVCOLORARG_A0 - || cc.b == TEVCOLORARG_C0 || cc.b == TEVCOLORARG_A0 - || cc.c == TEVCOLORARG_C0 || cc.c == TEVCOLORARG_A0 - || ac.a == TEVALPHAARG_A0 || ac.b == TEVALPHAARG_A0 || ac.c == TEVALPHAARG_A0) - WRITE(p, "cc0 = frac(c0 * (255.0f/256.0f)) * (256.0f/255.0f);\n"); - - - if(cc.a == TEVCOLORARG_C1 || cc.a == TEVCOLORARG_A1 - || cc.b == TEVCOLORARG_C1 || cc.b == TEVCOLORARG_A1 - || cc.c == TEVCOLORARG_C1 || cc.c == TEVCOLORARG_A1 - || ac.a == TEVALPHAARG_A1 || ac.b == TEVALPHAARG_A1 || ac.c == TEVALPHAARG_A1) - WRITE(p, "cc1 = frac(c1 * (255.0f/256.0f)) * (256.0f/255.0f);\n"); - - - if(cc.a == TEVCOLORARG_C2 || cc.a == TEVCOLORARG_A2 - || cc.b == TEVCOLORARG_C2 || cc.b == TEVCOLORARG_A2 - || cc.c == TEVCOLORARG_C2 || cc.c == TEVCOLORARG_A2 - || ac.a == TEVALPHAARG_A2 || ac.b == TEVALPHAARG_A2 || ac.c == TEVALPHAARG_A2) - WRITE(p, "cc2 = frac(c2 * (255.0f/256.0f)) * (256.0f/255.0f);\n"); - - - WRITE(p, "// color combine\n"); - if (cc.clamp) - WRITE(p, "%s = saturate(", tevCOutputTable[cc.dest]); - else - WRITE(p, "%s = ", tevCOutputTable[cc.dest]); - - // combine the color channel - if (cc.bias != TevBias_COMPARE) // if not compare - { - //normal color combiner goes here - if (cc.shift > TEVSCALE_1) - WRITE(p, "%s*(", tevScaleTable[cc.shift]); - - if(!(cc.d == TEVCOLORARG_ZERO && cc.op == TEVOP_ADD)) - WRITE(p, "%s%s", tevCInputTable[cc.d], tevOpTable[cc.op]); - - if (cc.a == cc.b) - WRITE(p, "%s", tevCInputTable[cc.a + 16]); - else if (cc.c == TEVCOLORARG_ZERO) - WRITE(p, "%s", tevCInputTable[cc.a + 16]); - else if (cc.c == TEVCOLORARG_ONE) - WRITE(p, "%s", tevCInputTable[cc.b + 16]); - else if (cc.a == TEVCOLORARG_ZERO) - WRITE(p, "%s*%s", tevCInputTable[cc.b + 16], tevCInputTable[cc.c + 16]); - else if (cc.b == TEVCOLORARG_ZERO) - WRITE(p, "%s*(float3(1.0f, 1.0f, 1.0f)-%s)", tevCInputTable[cc.a + 16], tevCInputTable[cc.c + 16]); - else - WRITE(p, "lerp(%s, %s, %s)", tevCInputTable[cc.a + 16], tevCInputTable[cc.b + 16], tevCInputTable[cc.c + 16]); - - WRITE(p, "%s", tevBiasTable[cc.bias]); - - if (cc.shift > TEVSCALE_1) - WRITE(p, ")"); - } - else - { - int cmp = (cc.shift<<1)|cc.op|8; // comparemode stored here - WRITE(p, TEVCMPColorOPTable[cmp],//lookup the function from the op table - tevCInputTable[cc.d], - tevCInputTable[cc.a + 16], - tevCInputTable[cc.b + 16], - tevCInputTable[cc.c + 16]); - } - if (cc.clamp) - WRITE(p, ")"); - WRITE(p,";\n"); - - WRITE(p, "// alpha combine\n"); - // combine the alpha channel - if (ac.clamp) - WRITE(p, "%s = saturate(", tevAOutputTable[ac.dest]); - else - WRITE(p, "%s = ", tevAOutputTable[ac.dest]); - - if (ac.bias != TevBias_COMPARE) // if not compare - { - //normal alpha combiner goes here - if (ac.shift > TEVSCALE_1) - WRITE(p, "%s*(", tevScaleTable[ac.shift]); - - if(!(ac.d == TEVALPHAARG_ZERO && ac.op == TEVOP_ADD)) - WRITE(p, "%s.a%s", tevAInputTable[ac.d], tevOpTable[ac.op]); - - if (ac.a == ac.b) - WRITE(p, "%s.a", tevAInputTable[ac.a + 8]); - else if (ac.c == TEVALPHAARG_ZERO) - WRITE(p, "%s.a", tevAInputTable[ac.a + 8]); - else if (ac.a == TEVALPHAARG_ZERO) - WRITE(p, "%s.a*%s.a", tevAInputTable[ac.b + 8], tevAInputTable[ac.c + 8]); - else if (ac.b == TEVALPHAARG_ZERO) - WRITE(p, "%s.a*(1.0f-%s.a)", tevAInputTable[ac.a + 8], tevAInputTable[ac.c + 8]); - else - WRITE(p, "lerp(%s.a, %s.a, %s.a)", tevAInputTable[ac.a + 8], tevAInputTable[ac.b + 8], tevAInputTable[ac.c + 8]); - - WRITE(p, "%s",tevBiasTable[ac.bias]); - - if (ac.shift>0) - WRITE(p, ")"); - - } - else - { - //compare alpha combiner goes here - int cmp = (ac.shift<<1)|ac.op|8; // comparemode stored here - WRITE(p, TEVCMPAlphaOPTable[cmp], - tevAInputTable[ac.d], - tevAInputTable[ac.a + 8], - tevAInputTable[ac.b + 8], - tevAInputTable[ac.c + 8]); - } - if (ac.clamp) - WRITE(p, ")"); - WRITE(p, ";\n\n"); - WRITE(p, "// TEV done\n"); -} - -void SampleTexture(char *&p, const char *destination, const char *texcoords, const char *texswap, int texmap, API_TYPE ApiType) -{ - if (ApiType == API_D3D11) - WRITE(p, "%s=Tex%d.Sample(samp%d,%s.xy * "I_TEXDIMS"[%d].xy).%s;\n", destination, texmap,texmap, texcoords, texmap, texswap); - else - WRITE(p, "%s=tex2D(samp%d,%s.xy * "I_TEXDIMS"[%d].xy).%s;\n", destination, texmap, texcoords, texmap, texswap); -} - -static const char *tevAlphaFuncsTable[] = -{ - "(false)", //ALPHACMP_NEVER 0 - "(prev.a <= %s - (0.25f/255.0f))", //ALPHACMP_LESS 1 - "(abs( prev.a - %s ) < (0.5f/255.0f))", //ALPHACMP_EQUAL 2 - "(prev.a < %s + (0.25f/255.0f))", //ALPHACMP_LEQUAL 3 - "(prev.a >= %s + (0.25f/255.0f))", //ALPHACMP_GREATER 4 - "(abs( prev.a - %s ) >= (0.5f/255.0f))", //ALPHACMP_NEQUAL 5 - "(prev.a > %s - (0.25f/255.0f))", //ALPHACMP_GEQUAL 6 - "(true)" //ALPHACMP_ALWAYS 7 -}; - -static const char *tevAlphaFunclogicTable[] = -{ - " && ", // and - " || ", // or - " != ", // xor - " == " // xnor -}; -static int AlphaPreTest() -{ - u32 op = bpmem.alphaFunc.logic; - u32 comp[2] = {bpmem.alphaFunc.comp0, bpmem.alphaFunc.comp1}; - - // First kill all the simple cases - switch(op) - { - case 0: // AND - if (comp[0] == ALPHACMP_ALWAYS && comp[1] == ALPHACMP_ALWAYS) return true; - if (comp[0] == ALPHACMP_NEVER || comp[1] == ALPHACMP_NEVER) return false; - break; - case 1: // OR - if (comp[0] == ALPHACMP_ALWAYS || comp[1] == ALPHACMP_ALWAYS) return true; - if (comp[0] == ALPHACMP_NEVER && comp[1] == ALPHACMP_NEVER)return false; - break; - case 2: // XOR - if ((comp[0] == ALPHACMP_ALWAYS && comp[1] == ALPHACMP_NEVER) || (comp[0] == ALPHACMP_NEVER && comp[1] == ALPHACMP_ALWAYS)) - return true; - if ((comp[0] == ALPHACMP_ALWAYS && comp[1] == ALPHACMP_ALWAYS) || (comp[0] == ALPHACMP_NEVER && comp[1] == ALPHACMP_NEVER)) - return false; - break; - case 3: // XNOR - if ((comp[0] == ALPHACMP_ALWAYS && comp[1] == ALPHACMP_NEVER) || (comp[0] == ALPHACMP_NEVER && comp[1] == ALPHACMP_ALWAYS)) - return false; - if ((comp[0] == ALPHACMP_ALWAYS && comp[1] == ALPHACMP_ALWAYS) || (comp[0] == ALPHACMP_NEVER && comp[1] == ALPHACMP_NEVER)) - return true; - break; - default: PanicAlert("bad logic for alpha test? %08x", op); - } - return -1; -} - - -static bool WriteAlphaTest(char *&p, API_TYPE ApiType,DSTALPHA_MODE dstAlphaMode) -{ - static const char *alphaRef[2] = - { - I_ALPHA"[0].r", - I_ALPHA"[0].g" - }; - - int Pretest = AlphaPreTest(); - if(Pretest >= 0) - { - return Pretest != 0; - } - - // using discard then return works the same in cg and dx9 but not in dx11 - WRITE(p, "if(!( "); - - int compindex = bpmem.alphaFunc.comp0 % 8; - WRITE(p, tevAlphaFuncsTable[compindex],alphaRef[0]);//lookup the first component from the alpha function table - - WRITE(p, "%s", tevAlphaFunclogicTable[bpmem.alphaFunc.logic % 4]);//lookup the logic op - - compindex = bpmem.alphaFunc.comp1 % 8; - WRITE(p, tevAlphaFuncsTable[compindex],alphaRef[1]);//lookup the second component from the alpha function table - WRITE(p, ")){ocol0 = 0;%s%s discard;%s}\n", - dstAlphaMode == DSTALPHA_DUAL_SOURCE_BLEND ? "ocol1 = 0;" : "", - DepthTextureEnable ? "depth = 1.f;" : "", - (ApiType != API_D3D11) ? "return;" : ""); - return true; -} - -static const char *tevFogFuncsTable[] = -{ - "", //No Fog - "", //? - "", //Linear - "", //? - " fog = 1.0f - pow(2.0f, -8.0f * fog);\n", //exp - " fog = 1.0f - pow(2.0f, -8.0f * fog * fog);\n", //exp2 - " fog = pow(2.0f, -8.0f * (1.0f - fog));\n", //backward exp - " fog = 1.0f - fog;\n fog = pow(2.0f, -8.0f * fog * fog);\n" //backward exp2 -}; - -static void WriteFog(char *&p) -{ - if(bpmem.fog.c_proj_fsel.fsel == 0)return;//no Fog - - if (bpmem.fog.c_proj_fsel.proj == 0) - { - // perspective - // ze = A/(B - (Zs >> B_SHF) - WRITE (p, " float ze = "I_FOG"[1].x / ("I_FOG"[1].y - (zCoord / "I_FOG"[1].w));\n"); - } - else - { - // orthographic - // ze = a*Zs (here, no B_SHF) - WRITE (p, " float ze = "I_FOG"[1].x * zCoord;\n"); - } - - // x_adjust = sqrt((x-center)^2 + k^2)/k - // ze *= x_adjust - //this is complitly teorical as the real hard seems to use a table intead of calculate the values. - if(bpmem.fogRange.Base.Enabled) - { - WRITE (p, " float x_adjust = (2.0f * (clipPos.x / "I_FOG"[2].y)) - 1.0f - "I_FOG"[2].x;\n"); - WRITE (p, " x_adjust = sqrt(x_adjust * x_adjust + "I_FOG"[2].z * "I_FOG"[2].z) / "I_FOG"[2].z;\n"); - WRITE (p, " ze *= x_adjust;\n"); - } - - WRITE (p, " float fog = saturate(ze - "I_FOG"[1].z);\n"); - - if(bpmem.fog.c_proj_fsel.fsel > 3) - { - WRITE(p, "%s", tevFogFuncsTable[bpmem.fog.c_proj_fsel.fsel]); - } - else - { - if(bpmem.fog.c_proj_fsel.fsel != 2) - WARN_LOG(VIDEO, "Unknown Fog Type! %08x", bpmem.fog.c_proj_fsel.fsel); - } - - WRITE(p, " prev.rgb = lerp(prev.rgb,"I_FOG"[0].rgb,fog);\n"); - - +// Copyright (C) 2003 Dolphin Project. + +// This program is free software: you can redistribute it and/or modify +// it under the terms of the GNU General Public License as published by +// the Free Software Foundation, version 2.0. + +// This program is distributed in the hope that it will be useful, +// but WITHOUT ANY WARRANTY; without even the implied warranty of +// MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the +// GNU General Public License 2.0 for more details. + +// A copy of the GPL 2.0 should have been included with the program. +// If not, see http://www.gnu.org/licenses/ + +// Official SVN repository and contact information can be found at +// http://code.google.com/p/dolphin-emu/ + +#include +#include +#include +#include + +#include "LightingShaderGen.h" +#include "PixelShaderGen.h" +#include "XFMemory.h" // for texture projection mode +#include "BPMemory.h" +#include "VideoConfig.h" +#include "NativeVertexFormat.h" + +PIXELSHADERUID last_pixel_shader_uid; + +static int AlphaPreTest(); + +static void StageHash(int stage, u32* out) +{ + out[0] |= bpmem.combiners[stage].colorC.hex & 0xFFFFFF; // 24 + u32 alphaC = bpmem.combiners[stage].alphaC.hex & 0xFFFFF0; // 24, strip out tswap and rswap for now + out[0] |= (alphaC&0xF0) << 24; // 8 + out[1] |= alphaC >> 8; // 16 + + // reserve 3 bits for bpmem.tevorders[stage/2].getTexMap + out[1] |= bpmem.tevorders[stage/2].getTexCoord(stage&1) << 19; // 3 + out[1] |= bpmem.tevorders[stage/2].getEnable(stage&1) << 22; // 1 + // reserve 3 bits for bpmem.tevorders[stage/2].getColorChan + + bool bHasIndStage = bpmem.tevind[stage].IsActive() && bpmem.tevind[stage].bt < bpmem.genMode.numindstages; + out[2] |= bHasIndStage << 2; // 1 + + bool needstexcoord = false; + + if (bHasIndStage) + { + out[2] |= (bpmem.tevind[stage].hex & 0x17FFFF) << 3; // 21, TODO: needs an explanation + needstexcoord = true; + } + + + TevStageCombiner::ColorCombiner& cc = bpmem.combiners[stage].colorC; + TevStageCombiner::AlphaCombiner& ac = bpmem.combiners[stage].alphaC; + + if(cc.a == TEVCOLORARG_RASA || cc.a == TEVCOLORARG_RASC + || cc.b == TEVCOLORARG_RASA || cc.b == TEVCOLORARG_RASC + || cc.c == TEVCOLORARG_RASA || cc.c == TEVCOLORARG_RASC + || cc.d == TEVCOLORARG_RASA || cc.d == TEVCOLORARG_RASC + || ac.a == TEVALPHAARG_RASA || ac.b == TEVALPHAARG_RASA + || ac.c == TEVALPHAARG_RASA || ac.d == TEVALPHAARG_RASA) + { + out[0] |= bpmem.combiners[stage].alphaC.rswap; + out[2] |= bpmem.tevksel[bpmem.combiners[stage].alphaC.rswap*2].swap1 << 24; // 2 + out[2] |= bpmem.tevksel[bpmem.combiners[stage].alphaC.rswap*2].swap2 << 26; // 2 + out[2] |= bpmem.tevksel[bpmem.combiners[stage].alphaC.rswap*2+1].swap1 << 28; // 2 + out[2] |= bpmem.tevksel[bpmem.combiners[stage].alphaC.rswap*2+1].swap2 << 30; // 2 + out[1] |= (bpmem.tevorders[stage/2].getColorChan(stage&1)&1) << 23; + out[2] |= (bpmem.tevorders[stage/2].getColorChan(stage&1)&0x6) >> 1; + } + + out[3] |= bpmem.tevorders[stage/2].getEnable(stage&1); + if (bpmem.tevorders[stage/2].getEnable(stage&1)) + { + if (bHasIndStage) needstexcoord = true; + + out[0] |= bpmem.combiners[stage].alphaC.tswap; + out[3] |= bpmem.tevksel[bpmem.combiners[stage].alphaC.tswap*2].swap1 << 1; // 2 + out[3] |= bpmem.tevksel[bpmem.combiners[stage].alphaC.tswap*2].swap2 << 3; // 2 + out[3] |= bpmem.tevksel[bpmem.combiners[stage].alphaC.tswap*2+1].swap1 << 5; // 2 + out[3] |= bpmem.tevksel[bpmem.combiners[stage].alphaC.tswap*2+1].swap2 << 7; // 2 + out[1] |= bpmem.tevorders[stage/2].getTexMap(stage&1) << 16; + } + + if (cc.a == TEVCOLORARG_KONST || cc.b == TEVCOLORARG_KONST || cc.c == TEVCOLORARG_KONST || cc.d == TEVCOLORARG_KONST + || ac.a == TEVALPHAARG_KONST || ac.b == TEVALPHAARG_KONST || ac.c == TEVALPHAARG_KONST || ac.d == TEVALPHAARG_KONST) + { + out[3] |= bpmem.tevksel[stage/2].getKC(stage&1) << 9; // 5 + out[3] |= bpmem.tevksel[stage/2].getKA(stage&1) << 14; // 5 + } + + if (needstexcoord) + { + out[1] |= bpmem.tevorders[stage/2].getTexCoord(stage&1) << 16; + } +} + +// Mash together all the inputs that contribute to the code of a generated pixel shader into +// 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, DSTALPHA_MODE dstAlphaMode) +{ + uid->values[0] |= bpmem.genMode.numtevstages; // 4 + uid->values[0] |= bpmem.genMode.numtexgens << 4; // 4 + uid->values[0] |= dstAlphaMode << 8; // 2 + + bool DepthTextureEnable = (bpmem.ztex2.op != ZTEXTURE_DISABLE && !bpmem.zcontrol.zcomploc && bpmem.zmode.testenable && bpmem.zmode.updateenable) || g_ActiveConfig.bEnablePerPixelDepth; + + uid->values[0] |= DepthTextureEnable << 10; // 1 + + bool enablePL = g_ActiveConfig.bEnablePixelLighting && g_ActiveConfig.backend_info.bSupportsPixelLighting; + uid->values[0] |= enablePL << 11; // 1 + + if (!enablePL) uid->values[0] |= xfregs.numTexGen.numTexGens << 12; // 4 + u32 alphaPreTest = AlphaPreTest()+1; + + uid->values[0] |= alphaPreTest << 16; // 2 + + if (alphaPreTest == 1 || (alphaPreTest && !DepthTextureEnable && dstAlphaMode == DSTALPHA_ALPHA_PASS)) + { + // Courtesy of PreAlphaTest, we're done already ;) + // NOTE: The comment header of generated shaders depends on the value of bpmem.genmode.numindstages.. shouldnt really bother about that though. + uid->num_values = 1; + return; + } + + for (unsigned int i = 0; i < bpmem.genMode.numtexgens; ++i) + { + if (18+i < 32) + uid->values[0] |= xfregs.texMtxInfo[i].projection << (18+i); // 1 + else + uid->values[1] |= xfregs.texMtxInfo[i].projection << (i - 14); // 1 + } + + uid->values[1] = bpmem.genMode.numindstages << 2; // 3 + u32 indirectStagesUsed = 0; + for (unsigned int i = 0; i < bpmem.genMode.numindstages; ++i) + if (bpmem.tevind[i].IsActive() && bpmem.tevind[i].bt < bpmem.genMode.numindstages) + indirectStagesUsed |= (1 << bpmem.tevind[i].bt); + + assert(indirectStagesUsed == (indirectStagesUsed & 0xF)); + + uid->values[1] |= indirectStagesUsed << 5; // 4; + + for (unsigned int i = 0; i < bpmem.genMode.numindstages; ++i) + { + if (indirectStagesUsed & (1 << i)) + { + uid->values[1] |= (bpmem.tevindref.getTexCoord(i) < bpmem.genMode.numtexgens) << (9 + 3*i); // 1 + if (bpmem.tevindref.getTexCoord(i) < bpmem.genMode.numtexgens) + uid->values[1] |= bpmem.tevindref.getTexCoord(i) << (10 + 3*i); // 2 + } + } + + u32* ptr = &uid->values[2]; + for (int i = 0; i < bpmem.genMode.numtevstages+1; ++i) + { + StageHash(i, ptr); + ptr += 4; // max: ptr = &uid->values[66] + } + + ptr[0] |= bpmem.alphaFunc.comp0; // 3 + ptr[0] |= bpmem.alphaFunc.comp1 << 3; // 3 + ptr[0] |= bpmem.alphaFunc.logic << 6; // 2 + + if (alphaPreTest == 0 || alphaPreTest == 2) + { + ptr[0] |= bpmem.fog.c_proj_fsel.fsel << 8; // 3 + if (DepthTextureEnable) + { + ptr[0] |= bpmem.ztex2.op << 11; // 2 + ptr[0] |= bpmem.zcontrol.zcomploc << 13; // 1 + ptr[0] |= bpmem.zmode.testenable << 14; // 1 + ptr[0] |= bpmem.zmode.updateenable << 15; // 1 + } + } + + if (dstAlphaMode != DSTALPHA_ALPHA_PASS) + { + if (bpmem.fog.c_proj_fsel.fsel != 0) + { + ptr[0] |= bpmem.fog.c_proj_fsel.proj << 16; // 1 + ptr[0] |= bpmem.fogRange.Base.Enabled << 17; // 1 + } + } + + ++ptr; + if (enablePL) + ptr += GetLightingShaderId(ptr); + + uid->num_values = ptr - uid->values; +} + +void GetSafePixelShaderId(PIXELSHADERUIDSAFE *uid, DSTALPHA_MODE dstAlphaMode) +{ + u32* ptr = uid->values; + *ptr++ = dstAlphaMode; // 0 + *ptr++ = bpmem.genMode.hex; // 1 + *ptr++ = bpmem.ztex2.hex; // 2 + *ptr++ = bpmem.zcontrol.hex; // 3 + *ptr++ = bpmem.zmode.hex; // 4 + *ptr++ = g_ActiveConfig.bEnablePerPixelDepth; // 5 + *ptr++ = g_ActiveConfig.bEnablePixelLighting && g_ActiveConfig.backend_info.bSupportsPixelLighting; // 6 + *ptr++ = xfregs.numTexGen.hex; // 7 + + if (g_ActiveConfig.bEnablePixelLighting && g_ActiveConfig.backend_info.bSupportsPixelLighting) + { + // TODO: Include register states for lighting shader + *ptr++ = xfregs.color[0].hex; + *ptr++ = xfregs.alpha[0].hex; + *ptr++ = xfregs.color[1].hex; + *ptr++ = xfregs.alpha[1].hex; + } + + for (unsigned int i = 0; i < 8; ++i) + *ptr++ = xfregs.texMtxInfo[i].hex; // 8-15 + + for (unsigned int i = 0; i < 16; ++i) + *ptr++ = bpmem.tevind[i].hex; // 16-31 + + *ptr++ = bpmem.tevindref.hex; // 32 + + for (int i = 0; i < bpmem.genMode.numtevstages+1; ++i) // up to 16 times + { + *ptr++ = bpmem.combiners[i].colorC.hex; // 33+5*i + *ptr++ = bpmem.combiners[i].alphaC.hex; // 34+5*i + *ptr++ = bpmem.tevind[i].hex; // 35+5*i + *ptr++ = bpmem.tevksel[i/2].hex; // 36+5*i + *ptr++ = bpmem.tevorders[i/2].hex; // 37+5*i + } + + ptr = &uid->values[113]; + + *ptr++ = bpmem.alphaFunc.hex; // 113 + + *ptr++ = bpmem.fog.c_proj_fsel.hex; // 114 + *ptr++ = bpmem.fogRange.Base.hex; // 115 + + _assert_((ptr - uid->values) == uid->GetNumValues()); +} + +void ValidatePixelShaderIDs(API_TYPE api, PIXELSHADERUIDSAFE old_id, const std::string& old_code, DSTALPHA_MODE dstAlphaMode, u32 components) +{ + if (!g_ActiveConfig.bEnableShaderDebugging) + return; + + PIXELSHADERUIDSAFE new_id; + GetSafePixelShaderId(&new_id, dstAlphaMode); + + if (!(old_id == new_id)) + { + std::string new_code(GeneratePixelShaderCode(dstAlphaMode, api, components)); + if (old_code != new_code) + { + _assert_(old_id.GetNumValues() == new_id.GetNumValues()); + + char msg[8192]; + char* ptr = msg; + ptr += sprintf(ptr, "Pixel shader IDs matched but unique IDs did not!\nUnique IDs (old <-> new):\n"); + const int N = new_id.GetNumValues(); + for (int i = 0; i < N/2; ++i) + ptr += sprintf(ptr, "%02d, %08X %08X | %08X %08X\n", 2*i, old_id.values[2*i], old_id.values[2*i+1], + new_id.values[2*i], new_id.values[2*i+1]); + if (N % 2) + ptr += sprintf(ptr, "%02d, %08X | %08X\n", N-1, old_id.values[N-1], new_id.values[N-1]); + + static int num_failures = 0; + char szTemp[MAX_PATH]; + sprintf(szTemp, "%spsuid_mismatch_%04i.txt", File::GetUserPath(D_DUMP_IDX).c_str(), num_failures++); + std::ofstream file(szTemp); + file << msg; + file << "\n\nOld shader code:\n" << old_code; + file << "\n\nNew shader code:\n" << new_code; + file.close(); + + PanicAlert("Unique pixel shader ID mismatch!\n\nReport this to the devs, along with the contents of %s.", szTemp); + } + } +} + +// old tev->pixelshader notes +// +// color for this stage (alpha, color) is given by bpmem.tevorders[0].colorchan0 +// konstant for this stage (alpha, color) is given by bpmem.tevksel +// inputs are given by bpmem.combiners[0].colorC.a/b/c/d << could be current chan color +// according to GXTevColorArg table above +// output is given by .outreg +// tevtemp is set according to swapmodetables and + +static void WriteStage(char *&p, int n, API_TYPE ApiType); +static void SampleTexture(char *&p, const char *destination, const char *texcoords, const char *texswap, int texmap, API_TYPE ApiType); +// static void WriteAlphaCompare(char *&p, int num, int comp); +static bool WriteAlphaTest(char *&p, API_TYPE ApiType,DSTALPHA_MODE dstAlphaMode); +static void WriteFog(char *&p); + +static const char *tevKSelTableC[] = // KCSEL +{ + "1.0f,1.0f,1.0f", // 1 = 0x00 + "0.875f,0.875f,0.875f", // 7_8 = 0x01 + "0.75f,0.75f,0.75f", // 3_4 = 0x02 + "0.625f,0.625f,0.625f", // 5_8 = 0x03 + "0.5f,0.5f,0.5f", // 1_2 = 0x04 + "0.375f,0.375f,0.375f", // 3_8 = 0x05 + "0.25f,0.25f,0.25f", // 1_4 = 0x06 + "0.125f,0.125f,0.125f", // 1_8 = 0x07 + "ERROR", // 0x08 + "ERROR", // 0x09 + "ERROR", // 0x0a + "ERROR", // 0x0b + I_KCOLORS"[0].rgb", // K0 = 0x0C + I_KCOLORS"[1].rgb", // K1 = 0x0D + I_KCOLORS"[2].rgb", // K2 = 0x0E + I_KCOLORS"[3].rgb", // K3 = 0x0F + I_KCOLORS"[0].rrr", // K0_R = 0x10 + I_KCOLORS"[1].rrr", // K1_R = 0x11 + I_KCOLORS"[2].rrr", // K2_R = 0x12 + I_KCOLORS"[3].rrr", // K3_R = 0x13 + I_KCOLORS"[0].ggg", // K0_G = 0x14 + I_KCOLORS"[1].ggg", // K1_G = 0x15 + I_KCOLORS"[2].ggg", // K2_G = 0x16 + I_KCOLORS"[3].ggg", // K3_G = 0x17 + I_KCOLORS"[0].bbb", // K0_B = 0x18 + I_KCOLORS"[1].bbb", // K1_B = 0x19 + I_KCOLORS"[2].bbb", // K2_B = 0x1A + I_KCOLORS"[3].bbb", // K3_B = 0x1B + I_KCOLORS"[0].aaa", // K0_A = 0x1C + I_KCOLORS"[1].aaa", // K1_A = 0x1D + I_KCOLORS"[2].aaa", // K2_A = 0x1E + I_KCOLORS"[3].aaa", // K3_A = 0x1F +}; + +static const char *tevKSelTableA[] = // KASEL +{ + "1.0f", // 1 = 0x00 + "0.875f",// 7_8 = 0x01 + "0.75f", // 3_4 = 0x02 + "0.625f",// 5_8 = 0x03 + "0.5f", // 1_2 = 0x04 + "0.375f",// 3_8 = 0x05 + "0.25f", // 1_4 = 0x06 + "0.125f",// 1_8 = 0x07 + "ERROR", // 0x08 + "ERROR", // 0x09 + "ERROR", // 0x0a + "ERROR", // 0x0b + "ERROR", // 0x0c + "ERROR", // 0x0d + "ERROR", // 0x0e + "ERROR", // 0x0f + I_KCOLORS"[0].r", // K0_R = 0x10 + I_KCOLORS"[1].r", // K1_R = 0x11 + I_KCOLORS"[2].r", // K2_R = 0x12 + I_KCOLORS"[3].r", // K3_R = 0x13 + I_KCOLORS"[0].g", // K0_G = 0x14 + I_KCOLORS"[1].g", // K1_G = 0x15 + I_KCOLORS"[2].g", // K2_G = 0x16 + I_KCOLORS"[3].g", // K3_G = 0x17 + I_KCOLORS"[0].b", // K0_B = 0x18 + I_KCOLORS"[1].b", // K1_B = 0x19 + I_KCOLORS"[2].b", // K2_B = 0x1A + I_KCOLORS"[3].b", // K3_B = 0x1B + I_KCOLORS"[0].a", // K0_A = 0x1C + I_KCOLORS"[1].a", // K1_A = 0x1D + I_KCOLORS"[2].a", // K2_A = 0x1E + I_KCOLORS"[3].a", // K3_A = 0x1F +}; + +static const char *tevScaleTable[] = // CS +{ + "1.0f", // SCALE_1 + "2.0f", // SCALE_2 + "4.0f", // SCALE_4 + "0.5f", // DIVIDE_2 +}; + +static const char *tevBiasTable[] = // TB +{ + "", // ZERO, + "+0.5f", // ADDHALF, + "-0.5f", // SUBHALF, + "", +}; + +static const char *tevOpTable[] = { // TEV + "+", // TEVOP_ADD = 0, + "-", // TEVOP_SUB = 1, +}; + +static const char *tevCInputTable[] = // CC +{ + "(prev.rgb)", // CPREV, + "(prev.aaa)", // APREV, + "(c0.rgb)", // C0, + "(c0.aaa)", // A0, + "(c1.rgb)", // C1, + "(c1.aaa)", // A1, + "(c2.rgb)", // C2, + "(c2.aaa)", // A2, + "(textemp.rgb)", // TEXC, + "(textemp.aaa)", // TEXA, + "(rastemp.rgb)", // RASC, + "(rastemp.aaa)", // RASA, + "float3(1.0f, 1.0f, 1.0f)", // ONE + "float3(0.5f, 0.5f, 0.5f)", // HALF + "(konsttemp.rgb)", //"konsttemp.rgb", // KONST + "float3(0.0f, 0.0f, 0.0f)", // ZERO + ///aded extra values to map clamped values + "(cprev.rgb)", // CPREV, + "(cprev.aaa)", // APREV, + "(cc0.rgb)", // C0, + "(cc0.aaa)", // A0, + "(cc1.rgb)", // C1, + "(cc1.aaa)", // A1, + "(cc2.rgb)", // C2, + "(cc2.aaa)", // A2, + "(textemp.rgb)", // TEXC, + "(textemp.aaa)", // TEXA, + "(crastemp.rgb)", // RASC, + "(crastemp.aaa)", // RASA, + "float3(1.0f, 1.0f, 1.0f)", // ONE + "float3(0.5f, 0.5f, 0.5f)", // HALF + "(ckonsttemp.rgb)", //"konsttemp.rgb", // KONST + "float3(0.0f, 0.0f, 0.0f)", // ZERO + "PADERROR", "PADERROR", "PADERROR", "PADERROR" +}; + +static const char *tevAInputTable[] = // CA +{ + "prev", // APREV, + "c0", // A0, + "c1", // A1, + "c2", // A2, + "textemp", // TEXA, + "rastemp", // RASA, + "konsttemp", // KONST, (hw1 had quarter) + "float4(0.0f, 0.0f, 0.0f, 0.0f)", // ZERO + ///aded extra values to map clamped values + "cprev", // APREV, + "cc0", // A0, + "cc1", // A1, + "cc2", // A2, + "textemp", // TEXA, + "crastemp", // RASA, + "ckonsttemp", // KONST, (hw1 had quarter) + "float4(0.0f, 0.0f, 0.0f, 0.0f)", // ZERO + "PADERROR", "PADERROR", "PADERROR", "PADERROR", + "PADERROR", "PADERROR", "PADERROR", "PADERROR", +}; + +static const char *tevRasTable[] = +{ + "colors_0", + "colors_1", + "ERROR", //2 + "ERROR", //3 + "ERROR", //4 + "alphabump", // use bump alpha + "(alphabump*(255.0f/248.0f))", //normalized + "float4(0.0f, 0.0f, 0.0f, 0.0f)", // zero +}; + +//static const char *tevTexFunc[] = { "tex2D", "texRECT" }; + +static const char *tevCOutputTable[] = { "prev.rgb", "c0.rgb", "c1.rgb", "c2.rgb" }; +static const char *tevAOutputTable[] = { "prev.a", "c0.a", "c1.a", "c2.a" }; +static const char *tevIndAlphaSel[] = {"", "x", "y", "z"}; +//static const char *tevIndAlphaScale[] = {"", "*32", "*16", "*8"}; +static const char *tevIndAlphaScale[] = {"*(248.0f/255.0f)", "*(224.0f/255.0f)", "*(240.0f/255.0f)", "*(248.0f/255.0f)"}; +static const char *tevIndBiasField[] = {"", "x", "y", "xy", "z", "xz", "yz", "xyz"}; // indexed by bias +static const char *tevIndBiasAdd[] = {"-128.0f", "1.0f", "1.0f", "1.0f" }; // indexed by fmt +static const char *tevIndWrapStart[] = {"0.0f", "256.0f", "128.0f", "64.0f", "32.0f", "16.0f", "0.001f" }; +static const char *tevIndFmtScale[] = {"255.0f", "31.0f", "15.0f", "7.0f" }; + +#define WRITE p+=sprintf + +static char swapModeTable[4][5]; + +static char text[16384]; +static bool DepthTextureEnable; + +static void BuildSwapModeTable() +{ + static const char *swapColors = "rgba"; + for (int i = 0; i < 4; i++) + { + swapModeTable[i][0] = swapColors[bpmem.tevksel[i*2].swap1]; + swapModeTable[i][1] = swapColors[bpmem.tevksel[i*2].swap2]; + swapModeTable[i][2] = swapColors[bpmem.tevksel[i*2+1].swap1]; + swapModeTable[i][3] = swapColors[bpmem.tevksel[i*2+1].swap2]; + swapModeTable[i][4] = 0; + } +} + +const char *GeneratePixelShaderCode(DSTALPHA_MODE dstAlphaMode, API_TYPE ApiType, u32 components) +{ + setlocale(LC_NUMERIC, "C"); // Reset locale for compilation + text[sizeof(text) - 1] = 0x7C; // canary + + BuildSwapModeTable(); // Needed for WriteStage + int numStages = bpmem.genMode.numtevstages + 1; + int numTexgen = bpmem.genMode.numtexgens; + + char *p = text; + WRITE(p, "//Pixel Shader for TEV stages\n"); + WRITE(p, "//%i TEV stages, %i texgens, XXX IND stages\n", + numStages, numTexgen/*, bpmem.genMode.numindstages*/); + + int nIndirectStagesUsed = 0; + if (bpmem.genMode.numindstages > 0) + { + for (int i = 0; i < numStages; ++i) + { + if (bpmem.tevind[i].IsActive() && bpmem.tevind[i].bt < bpmem.genMode.numindstages) + nIndirectStagesUsed |= 1 << bpmem.tevind[i].bt; + } + } + DepthTextureEnable = (bpmem.ztex2.op != ZTEXTURE_DISABLE && !bpmem.zcontrol.zcomploc && bpmem.zmode.testenable && bpmem.zmode.updateenable) || g_ActiveConfig.bEnablePerPixelDepth ; + // Declare samplers + + if(ApiType != API_D3D11) + { + WRITE(p, "uniform sampler2D "); + } + else + { + WRITE(p, "sampler "); + } + + bool bfirst = true; + for (int i = 0; i < 8; ++i) + { + WRITE(p, "%s samp%d : register(s%d)", bfirst?"":",", i, i); + bfirst = false; + } + WRITE(p, ";\n"); + if(ApiType == API_D3D11) + { + WRITE(p, "Texture2D "); + bfirst = true; + for (int i = 0; i < 8; ++i) + { + WRITE(p, "%s Tex%d : register(t%d)", bfirst?"":",", i, i); + bfirst = false; + } + WRITE(p, ";\n"); + } + + WRITE(p, "\n"); + + WRITE(p, "uniform float4 "I_COLORS"[4] : register(c%d);\n", C_COLORS); + WRITE(p, "uniform float4 "I_KCOLORS"[4] : register(c%d);\n", C_KCOLORS); + WRITE(p, "uniform float4 "I_ALPHA"[1] : register(c%d);\n", C_ALPHA); + WRITE(p, "uniform float4 "I_TEXDIMS"[8] : register(c%d);\n", C_TEXDIMS); + WRITE(p, "uniform float4 "I_ZBIAS"[2] : register(c%d);\n", C_ZBIAS); + WRITE(p, "uniform float4 "I_INDTEXSCALE"[2] : register(c%d);\n", C_INDTEXSCALE); + WRITE(p, "uniform float4 "I_INDTEXMTX"[6] : register(c%d);\n", C_INDTEXMTX); + WRITE(p, "uniform float4 "I_FOG"[3] : register(c%d);\n", C_FOG); + + if(g_ActiveConfig.bEnablePixelLighting && g_ActiveConfig.backend_info.bSupportsPixelLighting) + { + WRITE(p,"typedef struct { float4 col; float4 cosatt; float4 distatt; float4 pos; float4 dir; } Light;\n"); + WRITE(p,"typedef struct { Light lights[8]; } s_"I_PLIGHTS";\n"); + WRITE(p, "uniform s_"I_PLIGHTS" "I_PLIGHTS" : register(c%d);\n", C_PLIGHTS); + WRITE(p, "typedef struct { float4 C0, C1, C2, C3; } s_"I_PMATERIALS";\n"); + WRITE(p, "uniform s_"I_PMATERIALS" "I_PMATERIALS" : register(c%d);\n", C_PMATERIALS); + } + + WRITE(p, "void main(\n"); + if(ApiType != API_D3D11) + { + WRITE(p, " out float4 ocol0 : COLOR0,%s%s\n in float4 rawpos : %s,\n", + dstAlphaMode == DSTALPHA_DUAL_SOURCE_BLEND ? "\n out float4 ocol1 : COLOR1," : "", + DepthTextureEnable ? "\n out float depth : DEPTH," : "", + ApiType & API_OPENGL ? "WPOS" : ApiType & API_D3D9_SM20 ? "POSITION" : "VPOS"); + } + else + { + WRITE(p, " out float4 ocol0 : SV_Target0,%s%s\n in float4 rawpos : SV_Position,\n", + dstAlphaMode == DSTALPHA_DUAL_SOURCE_BLEND ? "\n out float4 ocol1 : SV_Target1," : "", + DepthTextureEnable ? "\n out float depth : SV_Depth," : ""); + } + + WRITE(p, " in float4 colors_0 : COLOR0,\n"); + WRITE(p, " in float4 colors_1 : COLOR1"); + + // compute window position if needed because binding semantic WPOS is not widely supported + if (numTexgen < 7) + { + for (int i = 0; i < numTexgen; ++i) + WRITE(p, ",\n in float3 uv%d : TEXCOORD%d", i, i); + WRITE(p, ",\n in float4 clipPos : TEXCOORD%d", numTexgen); + if(g_ActiveConfig.bEnablePixelLighting && g_ActiveConfig.backend_info.bSupportsPixelLighting) + WRITE(p, ",\n in float4 Normal : TEXCOORD%d", numTexgen + 1); + } + else + { + // wpos is in w of first 4 texcoords + if(g_ActiveConfig.bEnablePixelLighting && g_ActiveConfig.backend_info.bSupportsPixelLighting) + { + for (int i = 0; i < 8; ++i) + WRITE(p, ",\n in float4 uv%d : TEXCOORD%d", i, i); + } + else + { + for (unsigned int i = 0; i < xfregs.numTexGen.numTexGens; ++i) + WRITE(p, ",\n in float%d uv%d : TEXCOORD%d", i < 4 ? 4 : 3 , i, i); + } + } + WRITE(p, " ) {\n"); + + char* pmainstart = p; + int Pretest = AlphaPreTest(); + if (dstAlphaMode == DSTALPHA_ALPHA_PASS && !DepthTextureEnable && Pretest >= 0) + { + if (!Pretest) + { + // alpha test will always fail, so restart the shader and just make it an empty function + WRITE(p, "ocol0 = 0;\n"); + WRITE(p, "discard;\n"); + if(ApiType != API_D3D11) + WRITE(p, "return;\n"); + } + else + { + WRITE(p, " ocol0 = "I_ALPHA"[0].aaaa;\n"); + } + WRITE(p, "}\n"); + return text; + } + + WRITE(p, " float4 c0 = "I_COLORS"[1], c1 = "I_COLORS"[2], c2 = "I_COLORS"[3], prev = float4(0.0f, 0.0f, 0.0f, 0.0f), textemp = float4(0.0f, 0.0f, 0.0f, 0.0f), rastemp = float4(0.0f, 0.0f, 0.0f, 0.0f), konsttemp = float4(0.0f, 0.0f, 0.0f, 0.0f);\n" + " float3 comp16 = float3(1.0f, 255.0f, 0.0f), comp24 = float3(1.0f, 255.0f, 255.0f*255.0f);\n" + " float4 alphabump=float4(0.0f,0.0f,0.0f,0.0f);\n" + " float3 tevcoord=float3(0.0f, 0.0f, 0.0f);\n" + " float2 wrappedcoord=float2(0.0f,0.0f), tempcoord=float2(0.0f,0.0f);\n" + " float4 cc0=float4(0.0f,0.0f,0.0f,0.0f), cc1=float4(0.0f,0.0f,0.0f,0.0f);\n" + " float4 cc2=float4(0.0f,0.0f,0.0f,0.0f), cprev=float4(0.0f,0.0f,0.0f,0.0f);\n" + " float4 crastemp=float4(0.0f,0.0f,0.0f,0.0f),ckonsttemp=float4(0.0f,0.0f,0.0f,0.0f);\n\n"); + + if(g_ActiveConfig.bEnablePixelLighting && g_ActiveConfig.backend_info.bSupportsPixelLighting) + { + if (xfregs.numTexGen.numTexGens < 7) + { + WRITE(p,"float3 _norm0 = normalize(Normal.xyz);\n\n"); + WRITE(p,"float3 pos = float3(clipPos.x,clipPos.y,Normal.w);\n"); + } + else + { + WRITE(p," float3 _norm0 = normalize(float3(uv4.w,uv5.w,uv6.w));\n\n"); + WRITE(p,"float3 pos = float3(uv0.w,uv1.w,uv7.w);\n"); + } + + + WRITE(p, "float4 mat, lacc;\n" + "float3 ldir, h;\n" + "float dist, dist2, attn;\n"); + + p = GenerateLightingShader(p, components, I_PMATERIALS, I_PLIGHTS, "colors_", "colors_"); + } + + if (numTexgen < 7) + WRITE(p, "clipPos = float4(rawpos.x, rawpos.y, clipPos.z, clipPos.w);\n"); + else + WRITE(p, "float4 clipPos = float4(rawpos.x, rawpos.y, uv2.w, uv3.w);\n"); + + // HACK to handle cases where the tex gen is not enabled + if (numTexgen == 0) + { + WRITE(p, "float3 uv0 = float3(0.0f, 0.0f, 0.0f);\n"); + } + else + { + for (int i = 0; i < numTexgen; ++i) + { + // optional perspective divides + if (xfregs.texMtxInfo[i].projection == XF_TEXPROJ_STQ) + { + WRITE(p, "if (uv%d.z)", i); + WRITE(p, " uv%d.xy = uv%d.xy / uv%d.z;\n", i, i, i); + } + + WRITE(p, "uv%d.xy = uv%d.xy * "I_TEXDIMS"[%d].zw;\n", i, i, i); + } + } + + // indirect texture map lookup + for(u32 i = 0; i < bpmem.genMode.numindstages; ++i) + { + if (nIndirectStagesUsed & (1<= %s.r + (0.25f/255.0f)) ? %s : float3(0.0f, 0.0f, 0.0f))",//#define TEVCMP_R8_GT 8 + " %s + ((abs(%s.r - %s.r) < (0.5f/255.0f)) ? %s : float3(0.0f, 0.0f, 0.0f))",//#define TEVCMP_R8_EQ 9 + " %s + (( dot(%s.rgb, comp16) >= (dot(%s.rgb, comp16) + (0.25f/255.0f))) ? %s : float3(0.0f, 0.0f, 0.0f))",//#define TEVCMP_GR16_GT 10 + " %s + (abs(dot(%s.rgb, comp16) - dot(%s.rgb, comp16)) < (0.5f/255.0f) ? %s : float3(0.0f, 0.0f, 0.0f))",//#define TEVCMP_GR16_EQ 11 + " %s + (( dot(%s.rgb, comp24) >= (dot(%s.rgb, comp24) + (0.25f/255.0f))) ? %s : float3(0.0f, 0.0f, 0.0f))",//#define TEVCMP_BGR24_GT 12 + " %s + (abs(dot(%s.rgb, comp24) - dot(%s.rgb, comp24)) < (0.5f/255.0f) ? %s : float3(0.0f, 0.0f, 0.0f))",//#define TEVCMP_BGR24_EQ 13 + " %s + (max(sign(%s.rgb - %s.rgb - (0.25f/255.0f)), float3(0.0f, 0.0f, 0.0f)) * %s)",//#define TEVCMP_RGB8_GT 14 + " %s + ((float3(1.0f, 1.0f, 1.0f) - max(sign(abs(%s.rgb - %s.rgb) - (0.5f/255.0f)), float3(0.0f, 0.0f, 0.0f))) * %s)"//#define TEVCMP_RGB8_EQ 15 +}; + +//table with the alpha compare operations +static const char *TEVCMPAlphaOPTable[16] = +{ + "0.0f",//0 + "0.0f",//1 + "0.0f",//2 + "0.0f",//3 + "0.0f",//4 + "0.0f",//5 + "0.0f",//6 + "0.0f",//7 + " %s.a + ((%s.r >= (%s.r + (0.25f/255.0f))) ? %s.a : 0.0f)",//#define TEVCMP_R8_GT 8 + " %s.a + (abs(%s.r - %s.r) < (0.5f/255.0f) ? %s.a : 0.0f)",//#define TEVCMP_R8_EQ 9 + " %s.a + ((dot(%s.rgb, comp16) >= (dot(%s.rgb, comp16) + (0.25f/255.0f))) ? %s.a : 0.0f)",//#define TEVCMP_GR16_GT 10 + " %s.a + (abs(dot(%s.rgb, comp16) - dot(%s.rgb, comp16)) < (0.5f/255.0f) ? %s.a : 0.0f)",//#define TEVCMP_GR16_EQ 11 + " %s.a + ((dot(%s.rgb, comp24) >= (dot(%s.rgb, comp24) + (0.25f/255.0f))) ? %s.a : 0.0f)",//#define TEVCMP_BGR24_GT 12 + " %s.a + (abs(dot(%s.rgb, comp24) - dot(%s.rgb, comp24)) < (0.5f/255.0f) ? %s.a : 0.0f)",//#define TEVCMP_BGR24_EQ 13 + " %s.a + ((%s.a >= (%s.a + (0.25f/255.0f))) ? %s.a : 0.0f)",//#define TEVCMP_A8_GT 14 + " %s.a + (abs(%s.a - %s.a) < (0.5f/255.0f) ? %s.a : 0.0f)"//#define TEVCMP_A8_EQ 15 + +}; + + +static void WriteStage(char *&p, int n, API_TYPE ApiType) +{ + int texcoord = bpmem.tevorders[n/2].getTexCoord(n&1); + bool bHasTexCoord = (u32)texcoord < bpmem.genMode.numtexgens; + bool bHasIndStage = bpmem.tevind[n].IsActive() && bpmem.tevind[n].bt < bpmem.genMode.numindstages; + + // HACK to handle cases where the tex gen is not enabled + if (!bHasTexCoord) + texcoord = 0; + + WRITE(p, "// TEV stage %d\n", n); + + if (bHasIndStage) + { + WRITE(p, "// indirect op\n"); + // perform the indirect op on the incoming regular coordinates using indtex%d as the offset coords + if (bpmem.tevind[n].bs != ITBA_OFF) + { + WRITE(p, "alphabump = indtex%d.%s %s;\n", + bpmem.tevind[n].bt, + tevIndAlphaSel[bpmem.tevind[n].bs], + tevIndAlphaScale[bpmem.tevind[n].fmt]); + } + // format + WRITE(p, "float3 indtevcrd%d = indtex%d * %s;\n", n, bpmem.tevind[n].bt, tevIndFmtScale[bpmem.tevind[n].fmt]); + + // bias + if (bpmem.tevind[n].bias != ITB_NONE ) + WRITE(p, "indtevcrd%d.%s += %s;\n", n, tevIndBiasField[bpmem.tevind[n].bias], tevIndBiasAdd[bpmem.tevind[n].fmt]); + + // multiply by offset matrix and scale + if (bpmem.tevind[n].mid != 0) + { + if (bpmem.tevind[n].mid <= 3) + { + int mtxidx = 2*(bpmem.tevind[n].mid-1); + WRITE(p, "float2 indtevtrans%d = float2(dot("I_INDTEXMTX"[%d].xyz, indtevcrd%d), dot("I_INDTEXMTX"[%d].xyz, indtevcrd%d));\n", + n, mtxidx, n, mtxidx+1, n); + } + else if (bpmem.tevind[n].mid <= 7 && bHasTexCoord) + { // s matrix + // TODO: Might become negative? + int mtxidx = 2*(bpmem.tevind[n].mid-5); + WRITE(p, "float2 indtevtrans%d = "I_INDTEXMTX"[%d].ww * uv%d.xy * indtevcrd%d.xx;\n", n, mtxidx, texcoord, n); + } + else if (bpmem.tevind[n].mid <= 11 && bHasTexCoord) + { // t matrix + // TODO: Might become negative? + int mtxidx = 2*(bpmem.tevind[n].mid-9); + WRITE(p, "float2 indtevtrans%d = "I_INDTEXMTX"[%d].ww * uv%d.xy * indtevcrd%d.yy;\n", n, mtxidx, texcoord, n); + } + else + WRITE(p, "float2 indtevtrans%d = 0;\n", n); + } + else + WRITE(p, "float2 indtevtrans%d = 0;\n", n); + + // --------- + // Wrapping + // --------- + + // wrap S + if (bpmem.tevind[n].sw == ITW_OFF) + WRITE(p, "wrappedcoord.x = uv%d.x;\n", texcoord); + else if (bpmem.tevind[n].sw == ITW_0) + WRITE(p, "wrappedcoord.x = 0.0f;\n"); + else + WRITE(p, "wrappedcoord.x = fmod( uv%d.x, %s );\n", texcoord, tevIndWrapStart[bpmem.tevind[n].sw]); + + // wrap T + if (bpmem.tevind[n].tw == ITW_OFF) + WRITE(p, "wrappedcoord.y = uv%d.y;\n", texcoord); + else if (bpmem.tevind[n].tw == ITW_0) + WRITE(p, "wrappedcoord.y = 0.0f;\n"); + else + WRITE(p, "wrappedcoord.y = fmod( uv%d.y, %s );\n", texcoord, tevIndWrapStart[bpmem.tevind[n].tw]); + + if (bpmem.tevind[n].fb_addprev) // add previous tevcoord + WRITE(p, "tevcoord.xy += wrappedcoord + indtevtrans%d;\n", n); + else + WRITE(p, "tevcoord.xy = wrappedcoord + indtevtrans%d;\n", n); + } + + TevStageCombiner::ColorCombiner &cc = bpmem.combiners[n].colorC; + TevStageCombiner::AlphaCombiner &ac = bpmem.combiners[n].alphaC; + + // blah1 + if(cc.a == TEVCOLORARG_RASA || cc.a == TEVCOLORARG_RASC + || cc.b == TEVCOLORARG_RASA || cc.b == TEVCOLORARG_RASC + || cc.c == TEVCOLORARG_RASA || cc.c == TEVCOLORARG_RASC + || cc.d == TEVCOLORARG_RASA || cc.d == TEVCOLORARG_RASC + || ac.a == TEVALPHAARG_RASA || ac.b == TEVALPHAARG_RASA + || ac.c == TEVALPHAARG_RASA || ac.d == TEVALPHAARG_RASA) + { + char *rasswap = swapModeTable[bpmem.combiners[n].alphaC.rswap]; + WRITE(p, "rastemp = %s.%s;\n", tevRasTable[bpmem.tevorders[n / 2].getColorChan(n & 1)], rasswap); + WRITE(p, "crastemp = frac(rastemp * (255.0f/256.0f)) * (256.0f/255.0f);\n"); + } + + + if (bpmem.tevorders[n/2].getEnable(n&1)) + { + if(!bHasIndStage) + { + // calc tevcord + if(bHasTexCoord) + WRITE(p, "tevcoord.xy = uv%d.xy;\n", texcoord); + else + WRITE(p, "tevcoord.xy = float2(0.0f, 0.0f);\n"); + } + + char *texswap = swapModeTable[bpmem.combiners[n].alphaC.tswap]; + int texmap = bpmem.tevorders[n/2].getTexMap(n&1); + SampleTexture(p, "textemp", "tevcoord", texswap, texmap, ApiType); + } + else + WRITE(p, "textemp = float4(1.0f, 1.0f, 1.0f, 1.0f);\n"); + + + // blah2 + if (cc.a == TEVCOLORARG_KONST || cc.b == TEVCOLORARG_KONST || cc.c == TEVCOLORARG_KONST || cc.d == TEVCOLORARG_KONST + || ac.a == TEVALPHAARG_KONST || ac.b == TEVALPHAARG_KONST || ac.c == TEVALPHAARG_KONST || ac.d == TEVALPHAARG_KONST) + { + int kc = bpmem.tevksel[n / 2].getKC(n & 1); + int ka = bpmem.tevksel[n / 2].getKA(n & 1); + WRITE(p, "konsttemp = float4(%s, %s);\n", tevKSelTableC[kc], tevKSelTableA[ka]); + if(kc > 7 || ka > 7) + { + WRITE(p, "ckonsttemp = frac(konsttemp * (255.0f/256.0f)) * (256.0f/255.0f);\n"); + } + else + { + WRITE(p, "ckonsttemp = konsttemp;\n"); + } + } + + if(cc.a == TEVCOLORARG_CPREV || cc.a == TEVCOLORARG_APREV + || cc.b == TEVCOLORARG_CPREV || cc.b == TEVCOLORARG_APREV + || cc.c == TEVCOLORARG_CPREV || cc.c == TEVCOLORARG_APREV + || ac.a == TEVALPHAARG_APREV || ac.b == TEVALPHAARG_APREV || ac.c == TEVALPHAARG_APREV) + WRITE(p, "cprev = frac(prev * (255.0f/256.0f)) * (256.0f/255.0f);\n"); + + + if(cc.a == TEVCOLORARG_C0 || cc.a == TEVCOLORARG_A0 + || cc.b == TEVCOLORARG_C0 || cc.b == TEVCOLORARG_A0 + || cc.c == TEVCOLORARG_C0 || cc.c == TEVCOLORARG_A0 + || ac.a == TEVALPHAARG_A0 || ac.b == TEVALPHAARG_A0 || ac.c == TEVALPHAARG_A0) + WRITE(p, "cc0 = frac(c0 * (255.0f/256.0f)) * (256.0f/255.0f);\n"); + + + if(cc.a == TEVCOLORARG_C1 || cc.a == TEVCOLORARG_A1 + || cc.b == TEVCOLORARG_C1 || cc.b == TEVCOLORARG_A1 + || cc.c == TEVCOLORARG_C1 || cc.c == TEVCOLORARG_A1 + || ac.a == TEVALPHAARG_A1 || ac.b == TEVALPHAARG_A1 || ac.c == TEVALPHAARG_A1) + WRITE(p, "cc1 = frac(c1 * (255.0f/256.0f)) * (256.0f/255.0f);\n"); + + + if(cc.a == TEVCOLORARG_C2 || cc.a == TEVCOLORARG_A2 + || cc.b == TEVCOLORARG_C2 || cc.b == TEVCOLORARG_A2 + || cc.c == TEVCOLORARG_C2 || cc.c == TEVCOLORARG_A2 + || ac.a == TEVALPHAARG_A2 || ac.b == TEVALPHAARG_A2 || ac.c == TEVALPHAARG_A2) + WRITE(p, "cc2 = frac(c2 * (255.0f/256.0f)) * (256.0f/255.0f);\n"); + + + WRITE(p, "// color combine\n"); + if (cc.clamp) + WRITE(p, "%s = saturate(", tevCOutputTable[cc.dest]); + else + WRITE(p, "%s = ", tevCOutputTable[cc.dest]); + + // combine the color channel + if (cc.bias != TevBias_COMPARE) // if not compare + { + //normal color combiner goes here + if (cc.shift > TEVSCALE_1) + WRITE(p, "%s*(", tevScaleTable[cc.shift]); + + if(!(cc.d == TEVCOLORARG_ZERO && cc.op == TEVOP_ADD)) + WRITE(p, "%s%s", tevCInputTable[cc.d], tevOpTable[cc.op]); + + if (cc.a == cc.b) + WRITE(p, "%s", tevCInputTable[cc.a + 16]); + else if (cc.c == TEVCOLORARG_ZERO) + WRITE(p, "%s", tevCInputTable[cc.a + 16]); + else if (cc.c == TEVCOLORARG_ONE) + WRITE(p, "%s", tevCInputTable[cc.b + 16]); + else if (cc.a == TEVCOLORARG_ZERO) + WRITE(p, "%s*%s", tevCInputTable[cc.b + 16], tevCInputTable[cc.c + 16]); + else if (cc.b == TEVCOLORARG_ZERO) + WRITE(p, "%s*(float3(1.0f, 1.0f, 1.0f)-%s)", tevCInputTable[cc.a + 16], tevCInputTable[cc.c + 16]); + else + WRITE(p, "lerp(%s, %s, %s)", tevCInputTable[cc.a + 16], tevCInputTable[cc.b + 16], tevCInputTable[cc.c + 16]); + + WRITE(p, "%s", tevBiasTable[cc.bias]); + + if (cc.shift > TEVSCALE_1) + WRITE(p, ")"); + } + else + { + int cmp = (cc.shift<<1)|cc.op|8; // comparemode stored here + WRITE(p, TEVCMPColorOPTable[cmp],//lookup the function from the op table + tevCInputTable[cc.d], + tevCInputTable[cc.a + 16], + tevCInputTable[cc.b + 16], + tevCInputTable[cc.c + 16]); + } + if (cc.clamp) + WRITE(p, ")"); + WRITE(p,";\n"); + + WRITE(p, "// alpha combine\n"); + // combine the alpha channel + if (ac.clamp) + WRITE(p, "%s = saturate(", tevAOutputTable[ac.dest]); + else + WRITE(p, "%s = ", tevAOutputTable[ac.dest]); + + if (ac.bias != TevBias_COMPARE) // if not compare + { + //normal alpha combiner goes here + if (ac.shift > TEVSCALE_1) + WRITE(p, "%s*(", tevScaleTable[ac.shift]); + + if(!(ac.d == TEVALPHAARG_ZERO && ac.op == TEVOP_ADD)) + WRITE(p, "%s.a%s", tevAInputTable[ac.d], tevOpTable[ac.op]); + + if (ac.a == ac.b) + WRITE(p, "%s.a", tevAInputTable[ac.a + 8]); + else if (ac.c == TEVALPHAARG_ZERO) + WRITE(p, "%s.a", tevAInputTable[ac.a + 8]); + else if (ac.a == TEVALPHAARG_ZERO) + WRITE(p, "%s.a*%s.a", tevAInputTable[ac.b + 8], tevAInputTable[ac.c + 8]); + else if (ac.b == TEVALPHAARG_ZERO) + WRITE(p, "%s.a*(1.0f-%s.a)", tevAInputTable[ac.a + 8], tevAInputTable[ac.c + 8]); + else + WRITE(p, "lerp(%s.a, %s.a, %s.a)", tevAInputTable[ac.a + 8], tevAInputTable[ac.b + 8], tevAInputTable[ac.c + 8]); + + WRITE(p, "%s",tevBiasTable[ac.bias]); + + if (ac.shift>0) + WRITE(p, ")"); + + } + else + { + //compare alpha combiner goes here + int cmp = (ac.shift<<1)|ac.op|8; // comparemode stored here + WRITE(p, TEVCMPAlphaOPTable[cmp], + tevAInputTable[ac.d], + tevAInputTable[ac.a + 8], + tevAInputTable[ac.b + 8], + tevAInputTable[ac.c + 8]); + } + if (ac.clamp) + WRITE(p, ")"); + WRITE(p, ";\n\n"); + WRITE(p, "// TEV done\n"); +} + +void SampleTexture(char *&p, const char *destination, const char *texcoords, const char *texswap, int texmap, API_TYPE ApiType) +{ + if (ApiType == API_D3D11) + WRITE(p, "%s=Tex%d.Sample(samp%d,%s.xy * "I_TEXDIMS"[%d].xy).%s;\n", destination, texmap,texmap, texcoords, texmap, texswap); + else + WRITE(p, "%s=tex2D(samp%d,%s.xy * "I_TEXDIMS"[%d].xy).%s;\n", destination, texmap, texcoords, texmap, texswap); +} + +static const char *tevAlphaFuncsTable[] = +{ + "(false)", //ALPHACMP_NEVER 0 + "(prev.a <= %s - (0.25f/255.0f))", //ALPHACMP_LESS 1 + "(abs( prev.a - %s ) < (0.5f/255.0f))", //ALPHACMP_EQUAL 2 + "(prev.a < %s + (0.25f/255.0f))", //ALPHACMP_LEQUAL 3 + "(prev.a >= %s + (0.25f/255.0f))", //ALPHACMP_GREATER 4 + "(abs( prev.a - %s ) >= (0.5f/255.0f))", //ALPHACMP_NEQUAL 5 + "(prev.a > %s - (0.25f/255.0f))", //ALPHACMP_GEQUAL 6 + "(true)" //ALPHACMP_ALWAYS 7 +}; + +static const char *tevAlphaFunclogicTable[] = +{ + " && ", // and + " || ", // or + " != ", // xor + " == " // xnor +}; +static int AlphaPreTest() +{ + u32 op = bpmem.alphaFunc.logic; + u32 comp[2] = {bpmem.alphaFunc.comp0, bpmem.alphaFunc.comp1}; + + // First kill all the simple cases + switch(op) + { + case 0: // AND + if (comp[0] == ALPHACMP_ALWAYS && comp[1] == ALPHACMP_ALWAYS) return true; + if (comp[0] == ALPHACMP_NEVER || comp[1] == ALPHACMP_NEVER) return false; + break; + case 1: // OR + if (comp[0] == ALPHACMP_ALWAYS || comp[1] == ALPHACMP_ALWAYS) return true; + if (comp[0] == ALPHACMP_NEVER && comp[1] == ALPHACMP_NEVER)return false; + break; + case 2: // XOR + if ((comp[0] == ALPHACMP_ALWAYS && comp[1] == ALPHACMP_NEVER) || (comp[0] == ALPHACMP_NEVER && comp[1] == ALPHACMP_ALWAYS)) + return true; + if ((comp[0] == ALPHACMP_ALWAYS && comp[1] == ALPHACMP_ALWAYS) || (comp[0] == ALPHACMP_NEVER && comp[1] == ALPHACMP_NEVER)) + return false; + break; + case 3: // XNOR + if ((comp[0] == ALPHACMP_ALWAYS && comp[1] == ALPHACMP_NEVER) || (comp[0] == ALPHACMP_NEVER && comp[1] == ALPHACMP_ALWAYS)) + return false; + if ((comp[0] == ALPHACMP_ALWAYS && comp[1] == ALPHACMP_ALWAYS) || (comp[0] == ALPHACMP_NEVER && comp[1] == ALPHACMP_NEVER)) + return true; + break; + default: PanicAlert("bad logic for alpha test? %08x", op); + } + return -1; +} + + +static bool WriteAlphaTest(char *&p, API_TYPE ApiType,DSTALPHA_MODE dstAlphaMode) +{ + static const char *alphaRef[2] = + { + I_ALPHA"[0].r", + I_ALPHA"[0].g" + }; + + int Pretest = AlphaPreTest(); + if(Pretest >= 0) + { + return Pretest != 0; + } + + // using discard then return works the same in cg and dx9 but not in dx11 + WRITE(p, "if(!( "); + + int compindex = bpmem.alphaFunc.comp0 % 8; + WRITE(p, tevAlphaFuncsTable[compindex],alphaRef[0]);//lookup the first component from the alpha function table + + WRITE(p, "%s", tevAlphaFunclogicTable[bpmem.alphaFunc.logic % 4]);//lookup the logic op + + compindex = bpmem.alphaFunc.comp1 % 8; + WRITE(p, tevAlphaFuncsTable[compindex],alphaRef[1]);//lookup the second component from the alpha function table + WRITE(p, ")){ocol0 = 0;%s%s discard;%s}\n", + dstAlphaMode == DSTALPHA_DUAL_SOURCE_BLEND ? "ocol1 = 0;" : "", + DepthTextureEnable ? "depth = 1.f;" : "", + (ApiType != API_D3D11) ? "return;" : ""); + return true; +} + +static const char *tevFogFuncsTable[] = +{ + "", //No Fog + "", //? + "", //Linear + "", //? + " fog = 1.0f - pow(2.0f, -8.0f * fog);\n", //exp + " fog = 1.0f - pow(2.0f, -8.0f * fog * fog);\n", //exp2 + " fog = pow(2.0f, -8.0f * (1.0f - fog));\n", //backward exp + " fog = 1.0f - fog;\n fog = pow(2.0f, -8.0f * fog * fog);\n" //backward exp2 +}; + +static void WriteFog(char *&p) +{ + if(bpmem.fog.c_proj_fsel.fsel == 0)return;//no Fog + + if (bpmem.fog.c_proj_fsel.proj == 0) + { + // perspective + // ze = A/(B - (Zs >> B_SHF) + WRITE (p, " float ze = "I_FOG"[1].x / ("I_FOG"[1].y - (zCoord / "I_FOG"[1].w));\n"); + } + else + { + // orthographic + // ze = a*Zs (here, no B_SHF) + WRITE (p, " float ze = "I_FOG"[1].x * zCoord;\n"); + } + + // x_adjust = sqrt((x-center)^2 + k^2)/k + // ze *= x_adjust + //this is complitly teorical as the real hard seems to use a table intead of calculate the values. + if(bpmem.fogRange.Base.Enabled) + { + WRITE (p, " float x_adjust = (2.0f * (clipPos.x / "I_FOG"[2].y)) - 1.0f - "I_FOG"[2].x;\n"); + WRITE (p, " x_adjust = sqrt(x_adjust * x_adjust + "I_FOG"[2].z * "I_FOG"[2].z) / "I_FOG"[2].z;\n"); + WRITE (p, " ze *= x_adjust;\n"); + } + + WRITE (p, " float fog = saturate(ze - "I_FOG"[1].z);\n"); + + if(bpmem.fog.c_proj_fsel.fsel > 3) + { + WRITE(p, "%s", tevFogFuncsTable[bpmem.fog.c_proj_fsel.fsel]); + } + else + { + if(bpmem.fog.c_proj_fsel.fsel != 2) + WARN_LOG(VIDEO, "Unknown Fog Type! %08x", bpmem.fog.c_proj_fsel.fsel); + } + + WRITE(p, " prev.rgb = lerp(prev.rgb,"I_FOG"[0].rgb,fog);\n"); + + } \ No newline at end of file -- cgit v1.2.3 From 81c614fa07dd4949351d1fd13f9bbf03a956cdac Mon Sep 17 00:00:00 2001 From: NeoBrainX Date: Thu, 29 Sep 2011 23:32:05 +0200 Subject: Clean up various things. --- Source/Core/VideoCommon/Src/PixelShaderGen.cpp | 19 +++++++++++-------- 1 file changed, 11 insertions(+), 8 deletions(-) (limited to 'Source/Core/VideoCommon/Src/PixelShaderGen.cpp') diff --git a/Source/Core/VideoCommon/Src/PixelShaderGen.cpp b/Source/Core/VideoCommon/Src/PixelShaderGen.cpp index f52c6b4190..1726855d54 100644 --- a/Source/Core/VideoCommon/Src/PixelShaderGen.cpp +++ b/Source/Core/VideoCommon/Src/PixelShaderGen.cpp @@ -27,8 +27,6 @@ #include "VideoConfig.h" #include "NativeVertexFormat.h" -PIXELSHADERUID last_pixel_shader_uid; - static int AlphaPreTest(); static void StageHash(int stage, u32* out) @@ -104,8 +102,9 @@ static void StageHash(int stage, u32* out) // 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, DSTALPHA_MODE dstAlphaMode) +void GetPixelShaderId(PIXELSHADERUID *uid, DSTALPHA_MODE dstAlphaMode, u32 components) { + memset(uid->values, 0, sizeof(uid->values)); uid->values[0] |= bpmem.genMode.numtevstages; // 4 uid->values[0] |= bpmem.genMode.numtexgens << 4; // 4 uid->values[0] |= dstAlphaMode << 8; // 2 @@ -192,13 +191,17 @@ void GetPixelShaderId(PIXELSHADERUID *uid, DSTALPHA_MODE dstAlphaMode) ++ptr; if (enablePL) + { ptr += GetLightingShaderId(ptr); + *ptr++ = components; + } uid->num_values = ptr - uid->values; } -void GetSafePixelShaderId(PIXELSHADERUIDSAFE *uid, DSTALPHA_MODE dstAlphaMode) +void GetSafePixelShaderId(PIXELSHADERUIDSAFE *uid, DSTALPHA_MODE dstAlphaMode, u32 components) { + memset(uid->values, 0, sizeof(uid->values)); u32* ptr = uid->values; *ptr++ = dstAlphaMode; // 0 *ptr++ = bpmem.genMode.hex; // 1 @@ -211,11 +214,11 @@ void GetSafePixelShaderId(PIXELSHADERUIDSAFE *uid, DSTALPHA_MODE dstAlphaMode) if (g_ActiveConfig.bEnablePixelLighting && g_ActiveConfig.backend_info.bSupportsPixelLighting) { - // TODO: Include register states for lighting shader *ptr++ = xfregs.color[0].hex; *ptr++ = xfregs.alpha[0].hex; *ptr++ = xfregs.color[1].hex; *ptr++ = xfregs.alpha[1].hex; + *ptr++ = components; } for (unsigned int i = 0; i < 8; ++i) @@ -251,7 +254,7 @@ void ValidatePixelShaderIDs(API_TYPE api, PIXELSHADERUIDSAFE old_id, const std:: return; PIXELSHADERUIDSAFE new_id; - GetSafePixelShaderId(&new_id, dstAlphaMode); + GetSafePixelShaderId(&new_id, dstAlphaMode, components); if (!(old_id == new_id)) { @@ -874,13 +877,13 @@ static void WriteStage(char *&p, int n, API_TYPE ApiType) } else if (bpmem.tevind[n].mid <= 7 && bHasTexCoord) { // s matrix - // TODO: Might become negative? + _assert_(bpmem.tevind[n].mid >= 5); int mtxidx = 2*(bpmem.tevind[n].mid-5); WRITE(p, "float2 indtevtrans%d = "I_INDTEXMTX"[%d].ww * uv%d.xy * indtevcrd%d.xx;\n", n, mtxidx, texcoord, n); } else if (bpmem.tevind[n].mid <= 11 && bHasTexCoord) { // t matrix - // TODO: Might become negative? + _assert_(bpmem.tevind[n].mid >= 9); int mtxidx = 2*(bpmem.tevind[n].mid-9); WRITE(p, "float2 indtevtrans%d = "I_INDTEXMTX"[%d].ww * uv%d.xy * indtevcrd%d.yy;\n", n, mtxidx, texcoord, n); } -- cgit v1.2.3 From 8f31968466fb3cbb1ca033cf3baf5ba843640f6f Mon Sep 17 00:00:00 2001 From: Pierre Bourdon Date: Tue, 1 Nov 2011 01:37:54 +0100 Subject: Revert "Merge branch 'zcomploc-support'" This reverts commit 9dad9ebe899b4f2c36923c9721a09ff45b4e573c, reversing changes made to e76bc71efe8c3c724e14a2f741fe4f34669d1bea. --- Source/Core/VideoCommon/Src/PixelShaderGen.cpp | 27 +++++++++----------------- 1 file changed, 9 insertions(+), 18 deletions(-) (limited to 'Source/Core/VideoCommon/Src/PixelShaderGen.cpp') diff --git a/Source/Core/VideoCommon/Src/PixelShaderGen.cpp b/Source/Core/VideoCommon/Src/PixelShaderGen.cpp index f9aa7fd551..1726855d54 100644 --- a/Source/Core/VideoCommon/Src/PixelShaderGen.cpp +++ b/Source/Core/VideoCommon/Src/PixelShaderGen.cpp @@ -143,6 +143,8 @@ void GetPixelShaderId(PIXELSHADERUID *uid, DSTALPHA_MODE dstAlphaMode, u32 compo if (bpmem.tevind[i].IsActive() && bpmem.tevind[i].bt < bpmem.genMode.numindstages) indirectStagesUsed |= (1 << bpmem.tevind[i].bt); + assert(indirectStagesUsed == (indirectStagesUsed & 0xF)); + uid->values[1] |= indirectStagesUsed << 5; // 4; for (unsigned int i = 0; i < bpmem.genMode.numindstages; ++i) @@ -169,10 +171,10 @@ void GetPixelShaderId(PIXELSHADERUID *uid, DSTALPHA_MODE dstAlphaMode, u32 compo if (alphaPreTest == 0 || alphaPreTest == 2) { ptr[0] |= bpmem.fog.c_proj_fsel.fsel << 8; // 3 - ptr[0] |= bpmem.zcontrol.zcomploc << 11; // 1 if (DepthTextureEnable) { - ptr[0] |= bpmem.ztex2.op << 12; // 2 + ptr[0] |= bpmem.ztex2.op << 11; // 2 + ptr[0] |= bpmem.zcontrol.zcomploc << 13; // 1 ptr[0] |= bpmem.zmode.testenable << 14; // 1 ptr[0] |= bpmem.zmode.updateenable << 15; // 1 } @@ -1179,21 +1181,10 @@ static bool WriteAlphaTest(char *&p, API_TYPE ApiType,DSTALPHA_MODE dstAlphaMode compindex = bpmem.alphaFunc.comp1 % 8; WRITE(p, tevAlphaFuncsTable[compindex],alphaRef[1]);//lookup the second component from the alpha function table - WRITE(p, ")) {\n"); - - WRITE(p, "ocol0 = 0;\n"); - if (dstAlphaMode == DSTALPHA_DUAL_SOURCE_BLEND) - WRITE(p, "ocol1 = 0;\n"); - if (DepthTextureEnable) - WRITE(p, "depth = 1.f;\n"); - if (!bpmem.zcontrol.zcomploc) - { - WRITE(p, "discard;\n"); - if (ApiType != API_D3D11) - WRITE(p, "return;\n"); - } - - WRITE(p, "}\n"); + WRITE(p, ")){ocol0 = 0;%s%s discard;%s}\n", + dstAlphaMode == DSTALPHA_DUAL_SOURCE_BLEND ? "ocol1 = 0;" : "", + DepthTextureEnable ? "depth = 1.f;" : "", + (ApiType != API_D3D11) ? "return;" : ""); return true; } @@ -1251,4 +1242,4 @@ static void WriteFog(char *&p) WRITE(p, " prev.rgb = lerp(prev.rgb,"I_FOG"[0].rgb,fog);\n"); -} +} \ No newline at end of file -- cgit v1.2.3 From e58692653afda21f7593d340025c9d4328689244 Mon Sep 17 00:00:00 2001 From: NeoBrainX Date: Fri, 17 Feb 2012 17:09:17 +0100 Subject: PixelShaderGen: Use correct texel to pixel mapping when sampling textures (D3D9 only) Fixes issue 2068. Fixes issue 5158. --- Source/Core/VideoCommon/Src/PixelShaderGen.cpp | 10 ++++++++-- 1 file changed, 8 insertions(+), 2 deletions(-) (limited to 'Source/Core/VideoCommon/Src/PixelShaderGen.cpp') diff --git a/Source/Core/VideoCommon/Src/PixelShaderGen.cpp b/Source/Core/VideoCommon/Src/PixelShaderGen.cpp index 1726855d54..5e54b01129 100644 --- a/Source/Core/VideoCommon/Src/PixelShaderGen.cpp +++ b/Source/Core/VideoCommon/Src/PixelShaderGen.cpp @@ -1099,9 +1099,15 @@ static void WriteStage(char *&p, int n, API_TYPE ApiType) void SampleTexture(char *&p, const char *destination, const char *texcoords, const char *texswap, int texmap, API_TYPE ApiType) { if (ApiType == API_D3D11) - WRITE(p, "%s=Tex%d.Sample(samp%d,%s.xy * "I_TEXDIMS"[%d].xy).%s;\n", destination, texmap,texmap, texcoords, texmap, texswap); + WRITE(p, "%s=Tex%d.Sample(samp%d, %s.xy * "I_TEXDIMS"[%d].xy).%s;\n", destination, texmap,texmap, texcoords, texmap, texswap); + else if (ApiType & API_D3D9) + { + // D3D9 uses different pixel to texel mapping, so we need to offset our base address by half a pixel. + // Read the MSDN article "Directly Mapping Texels to Pixels (Direct3D 9)" for further info. + WRITE(p, "%s=tex2D(samp%d, (%s.xy + float2(0.5f,0.5f)) * "I_TEXDIMS"[%d].xy).%s;\n", destination, texmap, texcoords, texmap, texswap); + } else - WRITE(p, "%s=tex2D(samp%d,%s.xy * "I_TEXDIMS"[%d].xy).%s;\n", destination, texmap, texcoords, texmap, texswap); + WRITE(p, "%s=tex2D(samp%d, %s.xy * "I_TEXDIMS"[%d].xy).%s;\n", destination, texmap, texcoords, texmap, texswap); } static const char *tevAlphaFuncsTable[] = -- cgit v1.2.3 From 9adc119e3c2f845c0172b012badbbc86addad4a5 Mon Sep 17 00:00:00 2001 From: NeoBrainX Date: Mon, 20 Feb 2012 17:55:18 +0100 Subject: PixelShaderGen: For custom textures and scaled EFB copies, use correct texel to pixel mapping when sampling textures (D3D9 only) This is basically the same as revision e58692653afd, just for scaled textures and with a LOT more work... --- Source/Core/VideoCommon/Src/PixelShaderGen.cpp | 5 +++-- 1 file changed, 3 insertions(+), 2 deletions(-) (limited to 'Source/Core/VideoCommon/Src/PixelShaderGen.cpp') diff --git a/Source/Core/VideoCommon/Src/PixelShaderGen.cpp b/Source/Core/VideoCommon/Src/PixelShaderGen.cpp index 5e54b01129..ed504f1c82 100644 --- a/Source/Core/VideoCommon/Src/PixelShaderGen.cpp +++ b/Source/Core/VideoCommon/Src/PixelShaderGen.cpp @@ -560,6 +560,7 @@ const char *GeneratePixelShaderCode(DSTALPHA_MODE dstAlphaMode, API_TYPE ApiType WRITE(p, "uniform float4 "I_KCOLORS"[4] : register(c%d);\n", C_KCOLORS); WRITE(p, "uniform float4 "I_ALPHA"[1] : register(c%d);\n", C_ALPHA); WRITE(p, "uniform float4 "I_TEXDIMS"[8] : register(c%d);\n", C_TEXDIMS); + WRITE(p, "uniform float4 "I_VTEXSCALE"[8] : register(c%d);\n", C_VTEXSCALE); WRITE(p, "uniform float4 "I_ZBIAS"[2] : register(c%d);\n", C_ZBIAS); WRITE(p, "uniform float4 "I_INDTEXSCALE"[2] : register(c%d);\n", C_INDTEXSCALE); WRITE(p, "uniform float4 "I_INDTEXMTX"[6] : register(c%d);\n", C_INDTEXMTX); @@ -1102,9 +1103,9 @@ void SampleTexture(char *&p, const char *destination, const char *texcoords, con WRITE(p, "%s=Tex%d.Sample(samp%d, %s.xy * "I_TEXDIMS"[%d].xy).%s;\n", destination, texmap,texmap, texcoords, texmap, texswap); else if (ApiType & API_D3D9) { - // D3D9 uses different pixel to texel mapping, so we need to offset our base address by half a pixel. + // D3D9 uses different pixel to texel mapping, so we need to offset our sampling address by half a pixel (assuming native and virtual texture dimensions match each other, otherwise some math is involved). // Read the MSDN article "Directly Mapping Texels to Pixels (Direct3D 9)" for further info. - WRITE(p, "%s=tex2D(samp%d, (%s.xy + float2(0.5f,0.5f)) * "I_TEXDIMS"[%d].xy).%s;\n", destination, texmap, texcoords, texmap, texswap); + WRITE(p, "%s=tex2D(samp%d, (%s.xy + 0.5f*"I_VTEXSCALE"[%d].%s) * "I_TEXDIMS"[%d].xy).%s;\n", destination, texmap, texcoords, texmap/2, (texmap&1)?"zw":"xy", texmap, texswap); } else WRITE(p, "%s=tex2D(samp%d, %s.xy * "I_TEXDIMS"[%d].xy).%s;\n", destination, texmap, texcoords, texmap, texswap); -- cgit v1.2.3 From 3a9fed0ba285b3c2bb80de408ac1a9da9ca774b6 Mon Sep 17 00:00:00 2001 From: NeoBrainX Date: Tue, 21 Feb 2012 12:51:41 +0100 Subject: PixelShaderGen: Fix a bug introduced in revision 9adc119e3c2f. D3D11: Fix shader constant mapping. --- Source/Core/VideoCommon/Src/PixelShaderGen.cpp | 5 ++++- 1 file changed, 4 insertions(+), 1 deletion(-) (limited to 'Source/Core/VideoCommon/Src/PixelShaderGen.cpp') diff --git a/Source/Core/VideoCommon/Src/PixelShaderGen.cpp b/Source/Core/VideoCommon/Src/PixelShaderGen.cpp index ed504f1c82..fe5827de05 100644 --- a/Source/Core/VideoCommon/Src/PixelShaderGen.cpp +++ b/Source/Core/VideoCommon/Src/PixelShaderGen.cpp @@ -560,7 +560,10 @@ const char *GeneratePixelShaderCode(DSTALPHA_MODE dstAlphaMode, API_TYPE ApiType WRITE(p, "uniform float4 "I_KCOLORS"[4] : register(c%d);\n", C_KCOLORS); WRITE(p, "uniform float4 "I_ALPHA"[1] : register(c%d);\n", C_ALPHA); WRITE(p, "uniform float4 "I_TEXDIMS"[8] : register(c%d);\n", C_TEXDIMS); - WRITE(p, "uniform float4 "I_VTEXSCALE"[8] : register(c%d);\n", C_VTEXSCALE); + if (ApiType & API_D3D9) + { + WRITE(p, "uniform float4 "I_VTEXSCALE"[4] : register(c%d);\n", C_VTEXSCALE); + } WRITE(p, "uniform float4 "I_ZBIAS"[2] : register(c%d);\n", C_ZBIAS); WRITE(p, "uniform float4 "I_INDTEXSCALE"[2] : register(c%d);\n", C_INDTEXSCALE); WRITE(p, "uniform float4 "I_INDTEXMTX"[6] : register(c%d);\n", C_INDTEXMTX); -- cgit v1.2.3 From f5d4fe0bfe13937fa3736aa6b78974ceb6e190ad Mon Sep 17 00:00:00 2001 From: Henrik Rydgard Date: Sat, 3 Mar 2012 20:07:20 +0100 Subject: Fix some minor bugs pointed out by PVS Studio (thanks!) --- Source/Core/VideoCommon/Src/PixelShaderGen.cpp | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'Source/Core/VideoCommon/Src/PixelShaderGen.cpp') diff --git a/Source/Core/VideoCommon/Src/PixelShaderGen.cpp b/Source/Core/VideoCommon/Src/PixelShaderGen.cpp index fe5827de05..3ef923dc3e 100644 --- a/Source/Core/VideoCommon/Src/PixelShaderGen.cpp +++ b/Source/Core/VideoCommon/Src/PixelShaderGen.cpp @@ -158,7 +158,7 @@ void GetPixelShaderId(PIXELSHADERUID *uid, DSTALPHA_MODE dstAlphaMode, u32 compo } u32* ptr = &uid->values[2]; - for (int i = 0; i < bpmem.genMode.numtevstages+1; ++i) + for (unsigned int i = 0; i < bpmem.genMode.numtevstages+1; ++i) { StageHash(i, ptr); ptr += 4; // max: ptr = &uid->values[66] -- cgit v1.2.3