diff --git a/Source/Core/VideoCommon/Src/ShaderGenCommon.h b/Source/Core/VideoCommon/Src/ShaderGenCommon.h new file mode 100644 index 0000000000..28fb8e8956 --- /dev/null +++ b/Source/Core/VideoCommon/Src/ShaderGenCommon.h @@ -0,0 +1,94 @@ +// 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/ + +#ifndef _SHADERGENCOMMON_H +#define _SHADERGENCOMMON_H + +#include +#include +#include +#include "CommonTypes.h" + +template +class ShaderUid +{ +public: + ShaderUid() + { + memset(values, 0, sizeof(values)); + } + + void Write(const char* fmt, ...) {} + const char* GetBuffer() { return NULL; } + void SetBuffer(char* buffer) { } + + bool operator == (const ShaderUid& obj) const + { + return memcmp(this->values, obj.values, sizeof(values)) == 0; + } + + // TODO: Store last frame used and order by that? makes much more sense anyway... + bool operator < (const ShaderUid& obj) const + { + for (int i = 0; i < sizeof(uid_data) / sizeof(u32); ++i) + { + if (this->values[i] < obj.values[i]) + return true; + else if (this->values[i] > obj.values[i]) + return false; + } + return false; + } + + uid_data& GetUidData() { return data; } + +private: + union + { + uid_data data; + u32 values[sizeof(uid_data) / sizeof(u32)]; + }; +}; + +// Needs to be a template for hacks... +template +class ShaderCode +{ +public: + ShaderCode() : buf(NULL), write_ptr(NULL) + { + + } + + void Write(const char* fmt, ...) + { + va_list arglist; + va_start(arglist, fmt); + write_ptr += vsprintf(write_ptr, fmt, arglist); + va_end(arglist); + } + + const char* GetBuffer() { return buf; } + void SetBuffer(char* buffer) { buf = buffer; write_ptr = buffer; } + uid_data& GetUidData() { return *(uid_data*)NULL; } + +private: + const char* buf; + char* write_ptr; +}; + +#endif // _SHADERGENCOMMON_H diff --git a/Source/Core/VideoCommon/Src/VertexShaderGen.cpp b/Source/Core/VideoCommon/Src/VertexShaderGen.cpp index d7134b6dcb..acc89c6a54 100644 --- a/Source/Core/VideoCommon/Src/VertexShaderGen.cpp +++ b/Source/Core/VideoCommon/Src/VertexShaderGen.cpp @@ -267,7 +267,7 @@ void _GenerateLightingShader(T& object, int components, const char* materialsNam // TODO: Problem: this one uses copy constructors or sth for uids when returning... template -void GenerateShader(T& out, u32 components, API_TYPE api_type) +void GenerateVertexShader(T& out, u32 components, API_TYPE api_type) { #undef SetUidField #define SetUidField(name, value) if (type == GO_ShaderUid) {out.GetUidData().name = value; }; @@ -635,12 +635,12 @@ void GenerateShader(T& out, u32 components, API_TYPE api_type) setlocale(LC_NUMERIC, ""); // restore locale } -void GenerateShaderUid(ShaderUid& object, u32 components, API_TYPE api_type) +void GetVertexShaderUid(VertexShaderUid& object, u32 components, API_TYPE api_type) { - return GenerateShader(object, components, api_type); + GenerateVertexShader(object, components, api_type); } -void GenerateShaderCode(ShaderCode& object, u32 components, API_TYPE api_type) +void GenerateVertexShaderCode(VertexShaderCode& object, u32 components, API_TYPE api_type) { - return GenerateShader(object, components, api_type); + GenerateVertexShader(object, components, api_type); } diff --git a/Source/Core/VideoCommon/Src/VertexShaderGen.cpp~ b/Source/Core/VideoCommon/Src/VertexShaderGen.cpp~ new file mode 100644 index 0000000000..be57103918 --- /dev/null +++ b/Source/Core/VideoCommon/Src/VertexShaderGen.cpp~ @@ -0,0 +1,645 @@ +// 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 "NativeVertexFormat.h" + +#include "BPMemory.h" +#include "CPMemory.h" +#include "LightingShaderGen.h" +#include "VertexShaderGen.h" +#include "VideoConfig.h" + +static char text[16768]; + +enum GenOutput +{ + GO_ShaderCode, + GO_ShaderUid, +}; +// TODO: Check if something goes wrong if the cached shaders used pixel lighting but it's disabled later?? +template +void GenerateVSOutputStruct(T& object, u32 components, API_TYPE api_type) +{ + object.Write("struct VS_OUTPUT {\n"); + object.Write(" float4 pos : POSITION;\n"); + object.Write(" float4 colors_0 : COLOR0;\n"); + object.Write(" float4 colors_1 : COLOR1;\n"); + + if (xfregs.numTexGen.numTexGens < 7) + { + for (unsigned int i = 0; i < xfregs.numTexGen.numTexGens; ++i) + object.Write(" float3 tex%d : TEXCOORD%d;\n", i, i); + + object.Write(" float4 clipPos : TEXCOORD%d;\n", xfregs.numTexGen.numTexGens); +/// if(g_ActiveConfig.bEnablePixelLighting && g_ActiveConfig.backend_info.bSupportsPixelLighting) +/// object.Write(" float4 Normal : TEXCOORD%d;\n", xfregs.numTexGen.numTexGens + 1); + } + else + { + // clip position is in w of first 4 texcoords +/// if(g_ActiveConfig.bEnablePixelLighting && g_ActiveConfig.backend_info.bSupportsPixelLighting) +/// { +/// for (int i = 0; i < 8; ++i) +/// object.Write(" float4 tex%d : TEXCOORD%d;\n", i, i); +/// } +/// else + { + for (unsigned int i = 0; i < xfregs.numTexGen.numTexGens; ++i) + object.Write(" float%d tex%d : TEXCOORD%d;\n", i < 4 ? 4 : 3 , i, i); + } + } + object.Write("};\n"); +} + +template +void _GenerateLightShader(T& object, int index, int litchan_index, const char* lightsName, int coloralpha) +{ +#define SetUidField(name, value) if (type == GO_ShaderUid) { object.GetUidData().name = value; }; + const LitChannel& chan = (litchan_index > 1) ? xfregs.alpha[litchan_index-2] : xfregs.color[litchan_index]; + const char* swizzle = "xyzw"; + if (coloralpha == 1 ) swizzle = "xyz"; + else if (coloralpha == 2 ) swizzle = "w"; + + SetUidField(lit_chans[litchan_index].attnfunc, chan.attnfunc); + SetUidField(lit_chans[litchan_index].diffusefunc, chan.diffusefunc); + if (!(chan.attnfunc & 1)) { + // atten disabled + switch (chan.diffusefunc) { + case LIGHTDIF_NONE: + object.Write("lacc.%s += %s.lights[%d].col.%s;\n", swizzle, lightsName, index, swizzle); + break; + case LIGHTDIF_SIGN: + case LIGHTDIF_CLAMP: + object.Write("ldir = normalize(%s.lights[%d].pos.xyz - pos.xyz);\n", lightsName, index); + object.Write("lacc.%s += %sdot(ldir, _norm0)) * %s.lights[%d].col.%s;\n", + swizzle, chan.diffusefunc != LIGHTDIF_SIGN ? "max(0.0f," :"(", lightsName, index, swizzle); + break; + default: _assert_(0); + } + } + else { // spec and spot + + if (chan.attnfunc == 3) + { // spot + object.Write("ldir = %s.lights[%d].pos.xyz - pos.xyz;\n", lightsName, index); + object.Write("dist2 = dot(ldir, ldir);\n" + "dist = sqrt(dist2);\n" + "ldir = ldir / dist;\n" + "attn = max(0.0f, dot(ldir, %s.lights[%d].dir.xyz));\n", lightsName, index); + object.Write("attn = max(0.0f, dot(%s.lights[%d].cosatt.xyz, float3(1.0f, attn, attn*attn))) / dot(%s.lights[%d].distatt.xyz, float3(1.0f,dist,dist2));\n", lightsName, index, lightsName, index); + } + else if (chan.attnfunc == 1) + { // specular + object.Write("ldir = normalize(%s.lights[%d].pos.xyz);\n", lightsName, index); + object.Write("attn = (dot(_norm0,ldir) >= 0.0f) ? max(0.0f, dot(_norm0, %s.lights[%d].dir.xyz)) : 0.0f;\n", lightsName, index); + object.Write("attn = max(0.0f, dot(%s.lights[%d].cosatt.xyz, float3(1,attn,attn*attn))) / dot(%s.lights[%d].distatt.xyz, float3(1,attn,attn*attn));\n", lightsName, index, lightsName, index); + } + + switch (chan.diffusefunc) + { + case LIGHTDIF_NONE: + object.Write("lacc.%s += attn * %s.lights[%d].col.%s;\n", swizzle, lightsName, index, swizzle); + break; + case LIGHTDIF_SIGN: + case LIGHTDIF_CLAMP: + object.Write("lacc.%s += attn * %sdot(ldir, _norm0)) * %s.lights[%d].col.%s;\n", + swizzle, + chan.diffusefunc != LIGHTDIF_SIGN ? "max(0.0f," :"(", + lightsName, + index, + swizzle); + break; + default: _assert_(0); + } + } + object.Write("\n"); +} + +// vertex shader +// lights/colors +// materials name is I_MATERIALS in vs and I_PMATERIALS in ps +// inColorName is color in vs and colors_ in ps +// dest is o.colors_ in vs and colors_ in ps +template +void _GenerateLightingShader(T& object, int components, const char* materialsName, const char* lightsName, const char* inColorName, const char* dest) +{ + for (unsigned int j = 0; j < xfregs.numChan.numColorChans; j++) + { + const LitChannel& color = xfregs.color[j]; + const LitChannel& alpha = xfregs.alpha[j]; + + object.Write("{\n"); + + SetUidField(lit_chans[j].matsource, xfregs.color[j].matsource); + if (color.matsource) {// from vertex + if (components & (VB_HAS_COL0 << j)) + object.Write("mat = %s%d;\n", inColorName, j); + else if (components & VB_HAS_COL0) + object.Write("mat = %s0;\n", inColorName); + else + object.Write("mat = float4(1.0f, 1.0f, 1.0f, 1.0f);\n"); + } + else // from color + object.Write("mat = %s.C%d;\n", materialsName, j+2); + + SetUidField(lit_chans[j].enablelighting, xfregs.color[j].enablelighting); + if (color.enablelighting) { + SetUidField(lit_chans[j].ambsource, xfregs.color[j].ambsource); + if (color.ambsource) { // from vertex + if (components & (VB_HAS_COL0<(object, i, j, lightsName, 3); + } + } + } + } + + // no shared lights + for (int i = 0; i < 8; ++i) + { + if (!(mask&(1<(object, i, j, lightsName, 1); + if (!(mask&(1<(object, i, j+2, lightsName, 2); + } + } + else if (color.enablelighting || alpha.enablelighting) + { + // lights are disabled on one channel so process only the active ones + const LitChannel& workingchannel = color.enablelighting ? color : alpha; + const int lit_index = color.enablelighting ? j : (j+2); + int coloralpha = color.enablelighting ? 1 : 2; + + SetUidField(lit_chans[lit_index].light_mask, workingchannel.GetFullLightMask()); + for (int i = 0; i < 8; ++i) + { + if (workingchannel.GetFullLightMask() & (1<(object, i, lit_index, lightsName, coloralpha); + } + } + object.Write("%s%d = mat * saturate(lacc);\n", dest, j); + object.Write("}\n"); + } +} + +// TODO: Problem: this one uses copy constructors or sth for uids when returning... +template +void GenerateVertexShader(T& out, u32 components, API_TYPE api_type) +{ +#undef SetUidField +#define SetUidField(name, value) if (type == GO_ShaderUid) {out.GetUidData().name = value; }; + + if (type == GO_ShaderCode) + { + out.SetBuffer(text); + setlocale(LC_NUMERIC, "C"); // Reset locale for compilation + } + + /// text[sizeof(text) - 1] = 0x7C; // canary + + bool is_d3d = (api_type & API_D3D9 || api_type == API_D3D11); + u32 lightMask = 0; + if (xfregs.numChan.numColorChans > 0) + lightMask |= xfregs.color[0].GetFullLightMask() | xfregs.alpha[0].GetFullLightMask(); + if (xfregs.numChan.numColorChans > 1) + lightMask |= xfregs.color[1].GetFullLightMask() | xfregs.alpha[1].GetFullLightMask(); + + out.Write("//Vertex Shader: comp:%x, \n", components); + out.Write("typedef struct { float4 T0, T1, T2; float4 N0, N1, N2; } s_" I_POSNORMALMATRIX";\n" + "typedef struct { float4 t; } FLT4;\n" + "typedef struct { FLT4 T[24]; } s_" I_TEXMATRICES";\n" + "typedef struct { FLT4 T[64]; } s_" I_TRANSFORMMATRICES";\n" + "typedef struct { FLT4 T[32]; } s_" I_NORMALMATRICES";\n" + "typedef struct { FLT4 T[64]; } s_" I_POSTTRANSFORMMATRICES";\n" + "typedef struct { float4 col; float4 cosatt; float4 distatt; float4 pos; float4 dir; } Light;\n" + "typedef struct { Light lights[8]; } s_" I_LIGHTS";\n" + "typedef struct { float4 C0, C1, C2, C3; } s_" I_MATERIALS";\n" + "typedef struct { float4 T0, T1, T2, T3; } s_" I_PROJECTION";\n" + ); + +/// p = GenerateVSOutputStruct(p, components, api_type); + GenerateVSOutputStruct(out, components, api_type); + + // uniforms + + out.Write("uniform s_" I_TRANSFORMMATRICES" " I_TRANSFORMMATRICES" : register(c%d);\n", C_TRANSFORMMATRICES); + out.Write("uniform s_" I_TEXMATRICES" " I_TEXMATRICES" : register(c%d);\n", C_TEXMATRICES); + out.Write("uniform s_" I_NORMALMATRICES" " I_NORMALMATRICES" : register(c%d);\n", C_NORMALMATRICES); + out.Write("uniform s_" I_POSNORMALMATRIX" " I_POSNORMALMATRIX" : register(c%d);\n", C_POSNORMALMATRIX); + out.Write("uniform s_" I_POSTTRANSFORMMATRICES" " I_POSTTRANSFORMMATRICES" : register(c%d);\n", C_POSTTRANSFORMMATRICES); + out.Write("uniform s_" I_LIGHTS" " I_LIGHTS" : register(c%d);\n", C_LIGHTS); + out.Write("uniform s_" I_MATERIALS" " I_MATERIALS" : register(c%d);\n", C_MATERIALS); + out.Write("uniform s_" I_PROJECTION" " I_PROJECTION" : register(c%d);\n", C_PROJECTION); + out.Write("uniform float4 " I_DEPTHPARAMS" : register(c%d);\n", C_DEPTHPARAMS); + + out.Write("VS_OUTPUT main(\n"); + + SetUidField(numTexGens, xfregs.numTexGen.numTexGens); + SetUidField(components, components); + // inputs + if (components & VB_HAS_NRM0) + out.Write(" float3 rawnorm0 : NORMAL0,\n"); + if (components & VB_HAS_NRM1) + { + if (is_d3d) + out.Write(" float3 rawnorm1 : NORMAL1,\n"); + else + out.Write(" float3 rawnorm1 : ATTR%d,\n", SHADER_NORM1_ATTRIB); + } + if (components & VB_HAS_NRM2) + { + if (is_d3d) + out.Write(" float3 rawnorm2 : NORMAL2,\n"); + else + out.Write(" float3 rawnorm2 : ATTR%d,\n", SHADER_NORM2_ATTRIB); + } + if (components & VB_HAS_COL0) + out.Write(" float4 color0 : COLOR0,\n"); + if (components & VB_HAS_COL1) + out.Write(" float4 color1 : COLOR1,\n"); + for (int i = 0; i < 8; ++i) { + u32 hastexmtx = (components & (VB_HAS_TEXMTXIDX0<= 32 ? (posmtx-32) : posmtx;\n"); + out.Write("float3 N0 = " I_NORMALMATRICES".T[normidx].t.xyz, N1 = " I_NORMALMATRICES".T[normidx+1].t.xyz, N2 = " I_NORMALMATRICES".T[normidx+2].t.xyz;\n"); + } + + if (components & VB_HAS_NRM0) + out.Write("float3 _norm0 = normalize(float3(dot(N0, rawnorm0), dot(N1, rawnorm0), dot(N2, rawnorm0)));\n"); + if (components & VB_HAS_NRM1) + out.Write("float3 _norm1 = float3(dot(N0, rawnorm1), dot(N1, rawnorm1), dot(N2, rawnorm1));\n"); + if (components & VB_HAS_NRM2) + out.Write("float3 _norm2 = float3(dot(N0, rawnorm2), dot(N1, rawnorm2), dot(N2, rawnorm2));\n"); + } + else + { + out.Write("float4 pos = float4(dot(" I_POSNORMALMATRIX".T0, rawpos), dot(" I_POSNORMALMATRIX".T1, rawpos), dot(" I_POSNORMALMATRIX".T2, rawpos), 1.0f);\n"); + if (components & VB_HAS_NRM0) + out.Write("float3 _norm0 = normalize(float3(dot(" I_POSNORMALMATRIX".N0.xyz, rawnorm0), dot(" I_POSNORMALMATRIX".N1.xyz, rawnorm0), dot(" I_POSNORMALMATRIX".N2.xyz, rawnorm0)));\n"); + if (components & VB_HAS_NRM1) + out.Write("float3 _norm1 = float3(dot(" I_POSNORMALMATRIX".N0.xyz, rawnorm1), dot(" I_POSNORMALMATRIX".N1.xyz, rawnorm1), dot(" I_POSNORMALMATRIX".N2.xyz, rawnorm1));\n"); + if (components & VB_HAS_NRM2) + out.Write("float3 _norm2 = float3(dot(" I_POSNORMALMATRIX".N0.xyz, rawnorm2), dot(" I_POSNORMALMATRIX".N1.xyz, rawnorm2), dot(" I_POSNORMALMATRIX".N2.xyz, rawnorm2));\n"); + } + + if (!(components & VB_HAS_NRM0)) + out.Write("float3 _norm0 = float3(0.0f, 0.0f, 0.0f);\n"); + + + + out.Write("o.pos = float4(dot(" I_PROJECTION".T0, pos), dot(" I_PROJECTION".T1, pos), dot(" I_PROJECTION".T2, pos), dot(" I_PROJECTION".T3, pos));\n"); + + out.Write("float4 mat, lacc;\n" + "float3 ldir, h;\n" + "float dist, dist2, attn;\n"); + + SetUidField(numColorChans, xfregs.numChan.numColorChans); + if(xfregs.numChan.numColorChans == 0) + { + if (components & VB_HAS_COL0) + out.Write("o.colors_0 = color0;\n"); + else + out.Write("o.colors_0 = float4(1.0f, 1.0f, 1.0f, 1.0f);\n"); + } + + // TODO: This probably isn't necessary if pixel lighting is enabled. + _GenerateLightingShader(out, components, I_MATERIALS, I_LIGHTS, "color", "o.colors_"); + + if(xfregs.numChan.numColorChans < 2) + { + if (components & VB_HAS_COL1) + out.Write("o.colors_1 = color1;\n"); + else + out.Write("o.colors_1 = o.colors_0;\n"); + } + // special case if only pos and tex coord 0 and tex coord input is AB11 + // donko - this has caused problems in some games. removed for now. + bool texGenSpecialCase = false; + /*bool texGenSpecialCase = + ((g_VtxDesc.Hex & 0x60600L) == g_VtxDesc.Hex) && // only pos and tex coord 0 + (g_VtxDesc.Tex0Coord != NOT_PRESENT) && + (xfregs.texcoords[0].texmtxinfo.inputform == XF_TEXINPUT_AB11); + */ + + // transform texcoords + out.Write("float4 coord = float4(0.0f, 0.0f, 1.0f, 1.0f);\n"); + for (unsigned int i = 0; i < xfregs.numTexGen.numTexGens; ++i) { + TexMtxInfo& texinfo = xfregs.texMtxInfo[i]; + + out.Write("{\n"); + out.Write("coord = float4(0.0f, 0.0f, 1.0f, 1.0f);\n"); + SetUidField(texMtxInfo[i].sourcerow, xfregs.texMtxInfo[i].sourcerow); + switch (texinfo.sourcerow) { + case XF_SRCGEOM_INROW: + _assert_( texinfo.inputform == XF_TEXINPUT_ABC1 ); + out.Write("coord = rawpos;\n"); // pos.w is 1 + break; + case XF_SRCNORMAL_INROW: + if (components & VB_HAS_NRM0) { + _assert_( texinfo.inputform == XF_TEXINPUT_ABC1 ); + out.Write("coord = float4(rawnorm0.xyz, 1.0f);\n"); + } + break; + case XF_SRCCOLORS_INROW: + _assert_( texinfo.texgentype == XF_TEXGEN_COLOR_STRGBC0 || texinfo.texgentype == XF_TEXGEN_COLOR_STRGBC1 ); + break; + case XF_SRCBINORMAL_T_INROW: + if (components & VB_HAS_NRM1) { + _assert_( texinfo.inputform == XF_TEXINPUT_ABC1 ); + out.Write("coord = float4(rawnorm1.xyz, 1.0f);\n"); + } + break; + case XF_SRCBINORMAL_B_INROW: + if (components & VB_HAS_NRM2) { + _assert_( texinfo.inputform == XF_TEXINPUT_ABC1 ); + out.Write("coord = float4(rawnorm2.xyz, 1.0f);\n"); + } + break; + default: + _assert_(texinfo.sourcerow <= XF_SRCTEX7_INROW); + if (components & (VB_HAS_UV0<<(texinfo.sourcerow - XF_SRCTEX0_INROW)) ) + out.Write("coord = float4(tex%d.x, tex%d.y, 1.0f, 1.0f);\n", texinfo.sourcerow - XF_SRCTEX0_INROW, texinfo.sourcerow - XF_SRCTEX0_INROW); + break; + } + + // first transformation + SetUidField(texMtxInfo[i].texgentype, xfregs.texMtxInfo[i].texgentype); + switch (texinfo.texgentype) { + case XF_TEXGEN_EMBOSS_MAP: // calculate tex coords into bump map + + if (components & (VB_HAS_NRM1|VB_HAS_NRM2)) { + // transform the light dir into tangent space + SetUidField(texMtxInfo[i].embosslightshift, xfregs.texMtxInfo[i].embosslightshift); + SetUidField(texMtxInfo[i].embosssourceshift, xfregs.texMtxInfo[i].embosssourceshift); + out.Write("ldir = normalize(" I_LIGHTS".lights[%d].pos.xyz - pos.xyz);\n", texinfo.embosslightshift); + out.Write("o.tex%d.xyz = o.tex%d.xyz + float3(dot(ldir, _norm1), dot(ldir, _norm2), 0.0f);\n", i, texinfo.embosssourceshift); + } + else + { + _assert_(0); // should have normals + SetUidField(texMtxInfo[i].embosssourceshift, xfregs.texMtxInfo[i].embosssourceshift); + out.Write("o.tex%d.xyz = o.tex%d.xyz;\n", i, texinfo.embosssourceshift); + } + + break; + case XF_TEXGEN_COLOR_STRGBC0: + _assert_(texinfo.sourcerow == XF_SRCCOLORS_INROW); + out.Write("o.tex%d.xyz = float3(o.colors_0.x, o.colors_0.y, 1);\n", i); + break; + case XF_TEXGEN_COLOR_STRGBC1: + _assert_(texinfo.sourcerow == XF_SRCCOLORS_INROW); + out.Write("o.tex%d.xyz = float3(o.colors_1.x, o.colors_1.y, 1);\n", i); + break; + case XF_TEXGEN_REGULAR: + default: + SetUidField(texMtxInfo[i].projection, xfregs.texMtxInfo[i].projection); + if (components & (VB_HAS_TEXMTXIDX0<(object, components, api_type); +} + +void GenerateVertexShaderCode(VertexShaderCode& object, u32 components, API_TYPE api_type) +{ + GenerateVertexShader(object, components, api_type); +} diff --git a/Source/Core/VideoCommon/Src/VertexShaderGen.h b/Source/Core/VideoCommon/Src/VertexShaderGen.h index 11bb5c29f6..32d3eff91b 100644 --- a/Source/Core/VideoCommon/Src/VertexShaderGen.h +++ b/Source/Core/VideoCommon/Src/VertexShaderGen.h @@ -21,6 +21,7 @@ #include #include "XFMemory.h" #include "VideoCommon.h" +#include "ShaderGenCommon.h" #define SHADER_POSMTX_ATTRIB 1 #define SHADER_NORM1_ATTRIB 6 @@ -81,74 +82,11 @@ struct uid_data } lit_chans[4]; }; +typedef ShaderUid VertexShaderUid; +typedef ShaderCode VertexShaderCode; -class ShaderUid -{ -public: - ShaderUid() - { - memset(values, 0, sizeof(values)); - } - - void Write(const char* fmt, ...) {} - const char* GetBuffer() { return NULL; } - void SetBuffer(char* buffer) { } - - bool operator == (const ShaderUid& obj) const - { - return memcmp(this->values, obj.values, sizeof(values)) == 0; - } - - // TODO: Store last frame used and order by that? makes much more sense anyway... - bool operator < (const ShaderUid& obj) const - { - for (int i = 0; i < 24; ++i) - { - if (this->values[i] < obj.values[i]) - return true; - else if (this->values[i] > obj.values[i]) - return false; - } - return false; - } - - uid_data& GetUidData() { return data; } - -private: - union - { - uid_data data; - u32 values[24]; // TODO: Length? - }; -}; - -class ShaderCode -{ -public: - ShaderCode() : buf(NULL), write_ptr(NULL) - { - - } - - void Write(const char* fmt, ...) - { - va_list arglist; - va_start(arglist, fmt); - write_ptr += vsprintf(write_ptr, fmt, arglist); - va_end(arglist); - } - - const char* GetBuffer() { return buf; } - void SetBuffer(char* buffer) { buf = buffer; write_ptr = buffer; } - uid_data& GetUidData() { return *(uid_data*)NULL; } - -private: - const char* buf; - char* write_ptr; -}; - -void GenerateShaderUid(ShaderUid& object, u32 components, API_TYPE api_type); -void GenerateShaderCode(ShaderCode& object, u32 components, API_TYPE api_type); +void GetVertexShaderUid(VertexShaderUid& object, u32 components, API_TYPE api_type); +void GenerateVertexShaderCode(VertexShaderCode& object, u32 components, API_TYPE api_type); #endif // GCOGL_VERTEXSHADER_H diff --git a/Source/Core/VideoCommon/Src/VertexShaderGen.h~ b/Source/Core/VideoCommon/Src/VertexShaderGen.h~ new file mode 100644 index 0000000000..050ed76649 --- /dev/null +++ b/Source/Core/VideoCommon/Src/VertexShaderGen.h~ @@ -0,0 +1,92 @@ +// 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/ + +#ifndef GCOGL_VERTEXSHADER_H +#define GCOGL_VERTEXSHADER_H + +#include +#include "XFMemory.h" +#include "VideoCommon.h" +#include "ShaderGenCommon.h" + +#define SHADER_POSMTX_ATTRIB 1 +#define SHADER_NORM1_ATTRIB 6 +#define SHADER_NORM2_ATTRIB 7 + + +// shader variables +#define I_POSNORMALMATRIX "cpnmtx" +#define I_PROJECTION "cproj" +#define I_MATERIALS "cmtrl" +#define I_LIGHTS "clights" +#define I_TEXMATRICES "ctexmtx" +#define I_TRANSFORMMATRICES "ctrmtx" +#define I_NORMALMATRICES "cnmtx" +#define I_POSTTRANSFORMMATRICES "cpostmtx" +#define I_DEPTHPARAMS "cDepth" // farZ, zRange, scaled viewport width, scaled viewport height + +#define C_POSNORMALMATRIX 0 +#define C_PROJECTION (C_POSNORMALMATRIX + 6) +#define C_MATERIALS (C_PROJECTION + 4) +#define C_LIGHTS (C_MATERIALS + 4) +#define C_TEXMATRICES (C_LIGHTS + 40) +#define C_TRANSFORMMATRICES (C_TEXMATRICES + 24) +#define C_NORMALMATRICES (C_TRANSFORMMATRICES + 64) +#define C_POSTTRANSFORMMATRICES (C_NORMALMATRICES + 32) +#define C_DEPTHPARAMS (C_POSTTRANSFORMMATRICES + 64) +#define C_VENVCONST_END (C_DEPTHPARAMS + 4) + +// TODO: Need packing? +struct uid_data +{ + u32 components; + u32 numColorChans : 2; + u32 numTexGens : 4; + + struct { + u32 projection : 1; // XF_TEXPROJ_X + u32 inputform : 2; // XF_TEXINPUT_X + u32 texgentype : 3; // XF_TEXGEN_X + u32 sourcerow : 5; // XF_SRCGEOM_X + u32 embosssourceshift : 3; // what generated texcoord to use + u32 embosslightshift : 3; // light index that is used + } texMtxInfo[8]; + struct { + u32 index : 6; // base row of dual transform matrix + u32 normalize : 1; // normalize before send operation + } postMtxInfo[8]; + struct { + u32 enabled : 1; + } dualTexTrans; + struct { + u32 matsource : 1; + u32 enablelighting : 1; + u32 ambsource : 1; + u32 diffusefunc : 2; + u32 attnfunc : 2; + u32 light_mask : 8; + } lit_chans[4]; +}; + +typedef ShaderUid VertexShaderUid; +typedef ShaderCode VertexShaderCode; + +void GetVertexShaderUid(VertexShaderUid& object, u32 components, API_TYPE api_type); +void GenerateVertexShaderCode(VertexShaderCode& object, u32 components, API_TYPE api_type); + + +#endif // GCOGL_VERTEXSHADER_H diff --git a/Source/Plugins/Plugin_VideoOGL/Src/VertexShaderCache.cpp b/Source/Plugins/Plugin_VideoOGL/Src/VertexShaderCache.cpp index 7eaf1546da..29dd76fe6d 100644 --- a/Source/Plugins/Plugin_VideoOGL/Src/VertexShaderCache.cpp +++ b/Source/Plugins/Plugin_VideoOGL/Src/VertexShaderCache.cpp @@ -42,7 +42,7 @@ GLuint VertexShaderCache::CurrentShader; bool VertexShaderCache::ShaderEnabled; VertexShaderCache::VSCacheEntry* VertexShaderCache::last_entry = NULL; -ShaderUid VertexShaderCache::last_uid; +VertexShaderUid VertexShaderCache::last_uid; static int s_nMaxVertexInstructions; @@ -75,8 +75,8 @@ void VertexShaderCache::Shutdown() VERTEXSHADER* VertexShaderCache::SetShader(u32 components) { // Possible optimization: Don't always generate the shader uid, but keep track of changes in BPStructs instead - ShaderUid uid; - GenerateShaderUid(uid, components, API_OPENGL); + VertexShaderUid uid; + GetVertexShaderUid(uid, components, API_OPENGL); if (last_entry) { if (uid == last_uid) @@ -101,8 +101,8 @@ VERTEXSHADER* VertexShaderCache::SetShader(u32 components) // Make an entry in the table VSCacheEntry& entry = vshaders[uid]; last_entry = &entry; - ShaderCode code; - GenerateShaderCode(code, components, API_OPENGL); + VertexShaderCode code; + GenerateVertexShaderCode(code, components, API_OPENGL); #if defined(_DEBUG) || defined(DEBUGFAST) if (g_ActiveConfig.iLog & CONF_SAVESHADERS && code) { diff --git a/Source/Plugins/Plugin_VideoOGL/Src/VertexShaderCache.cpp~ b/Source/Plugins/Plugin_VideoOGL/Src/VertexShaderCache.cpp~ new file mode 100644 index 0000000000..f42cf09581 --- /dev/null +++ b/Source/Plugins/Plugin_VideoOGL/Src/VertexShaderCache.cpp~ @@ -0,0 +1,271 @@ +// 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 "Globals.h" +#include "VideoConfig.h" +#include "Statistics.h" + +#include "GLUtil.h" + +#include "Render.h" +#include "VertexShaderGen.h" +#include "VertexShaderManager.h" +#include "VertexShaderCache.h" +#include "VertexManager.h" +#include "VertexLoader.h" +#include "XFMemory.h" +#include "ImageWrite.h" +#include "FileUtil.h" +#include "Debugger.h" + +namespace OGL +{ + +VertexShaderCache::VSCache VertexShaderCache::vshaders; +GLuint VertexShaderCache::CurrentShader; +bool VertexShaderCache::ShaderEnabled; + +VertexShaderCache::VSCacheEntry* VertexShaderCache::last_entry = NULL; +VertexShaderUid VertexShaderCache::last_uid; + +static int s_nMaxVertexInstructions; + + +void VertexShaderCache::Init() +{ + glEnable(GL_VERTEX_PROGRAM_ARB); + ShaderEnabled = true; + CurrentShader = 0; + last_entry = NULL; + + glGetProgramivARB(GL_VERTEX_PROGRAM_ARB, GL_MAX_PROGRAM_NATIVE_INSTRUCTIONS_ARB, (GLint *)&s_nMaxVertexInstructions); + if (strstr((const char*)glGetString(GL_VENDOR), "Humper") != NULL) s_nMaxVertexInstructions = 4096; +#if CG_VERSION_NUM == 2100 + if (strstr((const char*)glGetString(GL_VENDOR), "ATI") != NULL) + { + s_nMaxVertexInstructions = 4096; + } +#endif +} + +void VertexShaderCache::Shutdown() +{ + for (VSCache::iterator iter = vshaders.begin(); iter != vshaders.end(); ++iter) + iter->second.Destroy(); + vshaders.clear(); +} + + +VERTEXSHADER* VertexShaderCache::SetShader(u32 components) +{ + // Possible optimization: Don't always generate the shader uid, but keep track of changes in BPStructs instead + VertexShaderUid uid; + GetVertexShaderUid(uid, components, API_OPENGL); + if (last_entry) + { + if (uid == last_uid) + { + GFX_DEBUGGER_PAUSE_AT(NEXT_VERTEX_SHADER_CHANGE, true); + return &last_entry->shader; + } + } + + last_uid = uid; + + VSCache::iterator iter = vshaders.find(uid); + if (iter != vshaders.end()) + { + VSCacheEntry &entry = iter->second; + last_entry = &entry; + + GFX_DEBUGGER_PAUSE_AT(NEXT_VERTEX_SHADER_CHANGE, true); + return &last_entry->shader; + } + + // Make an entry in the table + VSCacheEntry& entry = vshaders[uid]; + last_entry = &entry; + ShaderCode code; + GenerateShaderCode(code, components, API_OPENGL); + +#if defined(_DEBUG) || defined(DEBUGFAST) + if (g_ActiveConfig.iLog & CONF_SAVESHADERS && code) { + static int counter = 0; + char szTemp[MAX_PATH]; + sprintf(szTemp, "%svs_%04i.txt", File::GetUserPath(D_DUMP_IDX).c_str(), counter++); + + SaveData(szTemp, code.GetBuffer()); + } +#endif + + if (!code.GetBuffer() || !VertexShaderCache::CompileVertexShader(entry.shader, code.GetBuffer())) { + GFX_DEBUGGER_PAUSE_AT(NEXT_ERROR, true); + return NULL; + } + + INCSTAT(stats.numVertexShadersCreated); + SETSTAT(stats.numVertexShadersAlive, vshaders.size()); + GFX_DEBUGGER_PAUSE_AT(NEXT_VERTEX_SHADER_CHANGE, true); + return &last_entry->shader; +} + +bool VertexShaderCache::CompileVertexShader(VERTEXSHADER& vs, const char* pstrprogram) +{ + // Reset GL error before compiling shaders. Yeah, we need to investigate the causes of these. + GLenum err = GL_REPORT_ERROR(); + if (err != GL_NO_ERROR) + { + ERROR_LOG(VIDEO, "glError %08x before VS!", err); + } + +#if defined HAVE_CG && HAVE_CG + char stropt[64]; + sprintf(stropt, "MaxLocalParams=256,MaxInstructions=%d", s_nMaxVertexInstructions); + const char *opts[] = {"-profileopts", stropt, "-O2", "-q", NULL}; + CGprogram tempprog = cgCreateProgram(g_cgcontext, CG_SOURCE, pstrprogram, g_cgvProf, "main", opts); + if (!cgIsProgram(tempprog)) { + static int num_failures = 0; + char szTemp[MAX_PATH]; + sprintf(szTemp, "%sbad_vs_%04i.txt", File::GetUserPath(D_DUMP_IDX).c_str(), num_failures++); + std::ofstream file(szTemp); + file << pstrprogram; + file.close(); + + PanicAlert("Failed to compile vertex shader!\nThis usually happens when trying to use Dolphin with an outdated GPU or integrated GPU like the Intel GMA series.\n\nIf you're sure this is Dolphin's error anyway, post the contents of %s along with this error message at the forums.\n\nDebug info (%d):\n%s", + szTemp, + g_cgfProf, + cgGetLastListing(g_cgcontext)); + + cgDestroyProgram(tempprog); + ERROR_LOG(VIDEO, "Failed to load vs %s:", cgGetLastListing(g_cgcontext)); + ERROR_LOG(VIDEO, "%s", pstrprogram); + return false; + } + + if (cgGetError() != CG_NO_ERROR) + { + WARN_LOG(VIDEO, "Failed to load vs %s:", cgGetLastListing(g_cgcontext)); + WARN_LOG(VIDEO, "%s", pstrprogram); + } + + // This looks evil - we modify the program through the const char * we got from cgGetProgramString! + // It SHOULD not have any nasty side effects though - but you never know... + char *pcompiledprog = (char*)cgGetProgramString(tempprog, CG_COMPILED_PROGRAM); + char *plocal = strstr(pcompiledprog, "program.local"); + while (plocal != NULL) { + const char* penv = " program.env"; + memcpy(plocal, penv, 13); + plocal = strstr(plocal + 13, "program.local"); + } + glGenProgramsARB(1, &vs.glprogid); + SetCurrentShader(vs.glprogid); + + glProgramStringARB(GL_VERTEX_PROGRAM_ARB, GL_PROGRAM_FORMAT_ASCII_ARB, (GLsizei)strlen(pcompiledprog), pcompiledprog); + err = GL_REPORT_ERROR(); + if (err != GL_NO_ERROR) { + ERROR_LOG(VIDEO, "%s", pstrprogram); + ERROR_LOG(VIDEO, "%s", pcompiledprog); + } + + cgDestroyProgram(tempprog); +#endif + + if (g_ActiveConfig.bEnableShaderDebugging) + vs.strprog = pstrprogram; + + return true; +} + +void VertexShaderCache::DisableShader() +{ + if (ShaderEnabled) + { + glDisable(GL_VERTEX_PROGRAM_ARB); + ShaderEnabled = false; + } +} + + +void VertexShaderCache::SetCurrentShader(GLuint Shader) +{ + if (!ShaderEnabled) + { + glEnable(GL_VERTEX_PROGRAM_ARB); + ShaderEnabled= true; + } + if (CurrentShader != Shader) + { + if(Shader != 0) + CurrentShader = Shader; + glBindProgramARB(GL_VERTEX_PROGRAM_ARB, CurrentShader); + } +} + +void Renderer::SetVSConstant4f(unsigned int const_number, float f1, float f2, float f3, float f4) +{ + glProgramEnvParameter4fARB(GL_VERTEX_PROGRAM_ARB, const_number, f1, f2, f3, f4); +} + +void Renderer::SetVSConstant4fv(unsigned int const_number, const float *f) +{ + glProgramEnvParameter4fvARB(GL_VERTEX_PROGRAM_ARB, const_number, f); +} + +void Renderer::SetMultiVSConstant4fv(unsigned int const_number, unsigned int count, const float *f) +{ + if(GLEW_EXT_gpu_program_parameters) + { + glProgramEnvParameters4fvEXT(GL_VERTEX_PROGRAM_ARB, const_number, count, f); + } + else + { + for (unsigned int i = 0; i < count; i++,f+=4) + glProgramEnvParameter4fvARB(GL_VERTEX_PROGRAM_ARB, const_number + i, f); + } +} + +void Renderer::SetMultiVSConstant3fv(unsigned int const_number, unsigned int count, const float *f) +{ + if(GLEW_EXT_gpu_program_parameters) + { + float buf[4 * C_VENVCONST_END]; + for (unsigned int i = 0; i < count; i++) + { + buf[4*i ] = *f++; + buf[4*i+1] = *f++; + buf[4*i+2] = *f++; + buf[4*i+3] = 0.f; + } + glProgramEnvParameters4fvEXT(GL_VERTEX_PROGRAM_ARB, const_number, count, buf); + } + else + { + for (unsigned int i = 0; i < count; i++) + { + float buf[4]; + buf[0] = *f++; + buf[1] = *f++; + buf[2] = *f++; + buf[3] = 0.f; + glProgramEnvParameter4fvARB(GL_VERTEX_PROGRAM_ARB, const_number + i, buf); + } + } +} + +} // namespace OGL diff --git a/Source/Plugins/Plugin_VideoOGL/Src/VertexShaderCache.h b/Source/Plugins/Plugin_VideoOGL/Src/VertexShaderCache.h index f3831b3f7b..019abf6b6f 100644 --- a/Source/Plugins/Plugin_VideoOGL/Src/VertexShaderCache.h +++ b/Source/Plugins/Plugin_VideoOGL/Src/VertexShaderCache.h @@ -40,7 +40,7 @@ class VertexShaderCache struct VSCacheEntry { VERTEXSHADER shader; - ShaderUid safe_uid; + VertexShaderUid safe_uid; VSCacheEntry() {} void Destroy() { // printf("Destroying vs %i\n", shader.glprogid); @@ -49,12 +49,12 @@ class VertexShaderCache } }; - typedef std::map VSCache; + typedef std::map VSCache; static VSCache vshaders; static VSCacheEntry* last_entry; - static ShaderUid last_uid; // TODO: Use reference instead.. + static VertexShaderUid last_uid; // TODO: Use reference instead.. static GLuint CurrentShader; static bool ShaderEnabled; diff --git a/Source/Plugins/Plugin_VideoOGL/Src/VertexShaderCache.h~ b/Source/Plugins/Plugin_VideoOGL/Src/VertexShaderCache.h~ new file mode 100644 index 0000000000..859645c59a --- /dev/null +++ b/Source/Plugins/Plugin_VideoOGL/Src/VertexShaderCache.h~ @@ -0,0 +1,76 @@ +// 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/ + +#ifndef _VERTEXSHADERCACHE_H_ +#define _VERTEXSHADERCACHE_H_ + +#include +#include + +#include "BPMemory.h" +#include "VertexShaderGen.h" + +namespace OGL +{ + +struct VERTEXSHADER +{ + VERTEXSHADER() : glprogid(0) {} + GLuint glprogid; // opengl program id + + std::string strprog; +}; + +class VertexShaderCache +{ + struct VSCacheEntry + { + VERTEXSHADER shader; + VertexShaderUid safe_uid; + VSCacheEntry() {} + void Destroy() { + // printf("Destroying vs %i\n", shader.glprogid); + glDeleteProgramsARB(1, &shader.glprogid); + shader.glprogid = 0; + } + }; + + typedef std::map VSCache; + + static VSCache vshaders; + + static VSCacheEntry* last_entry; + static ShaderUid last_uid; // TODO: Use reference instead.. + + static GLuint CurrentShader; + static bool ShaderEnabled; + +public: + static void Init(); + static void Shutdown(); + + static VERTEXSHADER* SetShader(u32 components); + static bool CompileVertexShader(VERTEXSHADER& ps, const char* pstrprogram); + + static void SetCurrentShader(GLuint Shader); + static void DisableShader(); + +}; + +} // namespace OGL + +#endif // _VERTEXSHADERCACHE_H_