ShuriZma
/
dolphin
mirror of https://github.com/dolphin-emu/dolphin.git
1
0
Fork 0
Code Issues Packages Projects Releases Wiki Activity
dolphin/Source/Core/VideoCommon/Src/VertexShaderGen.cpp

604 lines
23 KiB
C++
Raw Normal View History

Fixup line endings.
2011-09-29 19:52:13 +00:00
// 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 <math.h>
#include <locale.h>
#include "NativeVertexFormat.h"
#include "BPMemory.h"
#include "CPMemory.h"
#include "LightingShaderGen.h"
#include "VertexShaderGen.h"
#include "VideoConfig.h"
// Mash together all the inputs that contribute to the code of a generated vertex shader into
// a unique identifier, basically containing all the bits. Yup, it's a lot ....
void GetVertexShaderId(VERTEXSHADERUID *uid, u32 components)
{
Clean up various things.
2011-09-29 21:32:05 +00:00
memset(uid->values, 0, sizeof(uid->values));
Fixup line endings.
2011-09-29 19:52:13 +00:00
uid->values[0] = components |
(xfregs.numTexGen.numTexGens << 23) |
(xfregs.numChan.numColorChans << 27) |
(xfregs.dualTexTrans.enabled << 29);
// TODO: If pixel lighting is enabled, do we even have to bother about storing lighting related registers here?
GetLightingShaderId(&uid->values[1]);
uid->values[2] |= (g_ActiveConfig.bEnablePixelLighting && g_ActiveConfig.backend_info.bSupportsPixelLighting) << 31;
u32 *pcurvalue = &uid->values[3];
for (unsigned int i = 0; i < xfregs.numTexGen.numTexGens; ++i) {
TexMtxInfo tinfo = xfregs.texMtxInfo[i];
if (tinfo.texgentype != XF_TEXGEN_EMBOSS_MAP)
tinfo.hex &= 0x7ff;
if (tinfo.texgentype != XF_TEXGEN_REGULAR)
tinfo.projection = 0;
u32 val = ((tinfo.hex >> 1) & 0x1ffff);
if (xfregs.dualTexTrans.enabled && tinfo.texgentype == XF_TEXGEN_REGULAR) {
// rewrite normalization and post index
val |= ((u32)xfregs.postMtxInfo[i].index << 17) | ((u32)xfregs.postMtxInfo[i].normalize << 23);
}
switch (i & 3) {
case 0: pcurvalue[0] |= val; break;
case 1: pcurvalue[0] |= val << 24; pcurvalue[1] = val >> 8; ++pcurvalue; break;
case 2: pcurvalue[0] |= val << 16; pcurvalue[1] = val >> 16; ++pcurvalue; break;
case 3: pcurvalue[0] |= val << 8; ++pcurvalue; break;
}
}
}
void GetSafeVertexShaderId(VERTEXSHADERUIDSAFE *uid, u32 components)
{
// Just store all used registers here without caring whether we need all bits or less.
Clean up various things.
2011-09-29 21:32:05 +00:00
memset(uid->values, 0, sizeof(uid->values));
Fixup line endings.
2011-09-29 19:52:13 +00:00
u32* ptr = uid->values;
*ptr++ = components;
*ptr++ = xfregs.numTexGen.hex;
*ptr++ = xfregs.numChan.hex;
*ptr++ = xfregs.dualTexTrans.hex;
for (int i = 0; i < 2; ++i) {
*ptr++ = xfregs.color[i].hex;
*ptr++ = xfregs.alpha[i].hex;
}
*ptr++ = g_ActiveConfig.bEnablePixelLighting && g_ActiveConfig.backend_info.bSupportsPixelLighting;
for (unsigned int i = 0; i < 8; ++i) {
*ptr++ = xfregs.texMtxInfo[i].hex;
*ptr++ = xfregs.postMtxInfo[i].hex;
}
_assert_((ptr - uid->values) == uid->GetNumValues());
}
void ValidateVertexShaderIDs(API_TYPE api, VERTEXSHADERUIDSAFE old_id, const std::string& old_code, u32 components)
{
if (!g_ActiveConfig.bEnableShaderDebugging)
return;
VERTEXSHADERUIDSAFE new_id;
GetSafeVertexShaderId(&new_id, components);
if (!(old_id == new_id))
{
std::string new_code(GenerateVertexShaderCode(components, api));
if (old_code != new_code)
{
_assert_(old_id.GetNumValues() == new_id.GetNumValues());
char msg[8192];
char* ptr = msg;
ptr += sprintf(ptr, "Vertex 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, "%svsuid_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);
}
}
}
static char text[16384];
#define WRITE p+=sprintf
Add in GLSL setting again. PS and VS making. Untested and won't work for now. Add in program shader cache files. Readd NativeVertexFormat stuffs. Add in PS and VS cache things. SetShaders in places. Fixed EFB cache index computations in OpenGL renderer. The previous computation was very likely to go out of array bounds, which could result in crashes on EFB access. Also, the cache size was rounded down instead of up. This is a problem since EFB_HEIGHT (528) is not a multiple of EFB_CACHE_RECT_SIZE (64).
2011-12-01 03:00:21 +00:00
char* GenerateVSOutputStruct(char* p, u32 components, API_TYPE ApiType)
Fixup line endings.
2011-09-29 19:52:13 +00:00
{
More stuff
2011-12-03 02:17:26 +00:00
// This turned a bit ugly with GLSL
// Will be less ugly with GLSL 1.3...hopefully
Fixup line endings.
2011-09-29 19:52:13 +00:00
WRITE(p, "struct VS_OUTPUT {\n");
More stuff
2011-12-03 02:17:26 +00:00
WRITE(p, " float4 pos %s POSITION;\n", ApiType == API_GLSL ? ";//" : ":");
WRITE(p, " float4 colors_0 %s COLOR0;\n", ApiType == API_GLSL ? ";//" : ":");
WRITE(p, " float4 colors_1 %s COLOR1;\n", ApiType == API_GLSL ? ";//" : ":");
Fixup line endings.
2011-09-29 19:52:13 +00:00
if (xfregs.numTexGen.numTexGens < 7) {
for (unsigned int i = 0; i < xfregs.numTexGen.numTexGens; ++i)
More stuff
2011-12-03 02:17:26 +00:00
WRITE(p, " float3 tex%d %s TEXCOORD%d;\n", i, ApiType == API_GLSL ? ";//" : ":", i);
WRITE(p, " float4 clipPos %s TEXCOORD%d;\n", ApiType == API_GLSL ? ";//" : ":", xfregs.numTexGen.numTexGens);
Fixup line endings.
2011-09-29 19:52:13 +00:00
if(g_ActiveConfig.bEnablePixelLighting && g_ActiveConfig.backend_info.bSupportsPixelLighting)
More stuff
2011-12-03 02:17:26 +00:00
WRITE(p, " float4 Normal %s TEXCOORD%d;\n", ApiType == API_GLSL ? ";//" : ":", xfregs.numTexGen.numTexGens + 1);
Fixup line endings.
2011-09-29 19:52:13 +00:00
} 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)
More stuff
2011-12-03 02:17:26 +00:00
WRITE(p, " float4 tex%d %s TEXCOORD%d;\n", i, ApiType == API_GLSL ? ";//" : ":", i);
Fixup line endings.
2011-09-29 19:52:13 +00:00
}
else
{
for (unsigned int i = 0; i < xfregs.numTexGen.numTexGens; ++i)
More stuff
2011-12-03 02:17:26 +00:00
WRITE(p, " float%d tex%d %s TEXCOORD%d;\n", i < 4 ? 4 : 3 , i, ApiType == API_GLSL ? ";//" : ":", i);
Fixup line endings.
2011-09-29 19:52:13 +00:00
}
More stuff
2011-12-03 02:17:26 +00:00
}
Fixup line endings.
2011-09-29 19:52:13 +00:00
WRITE(p, "};\n");
return p;
}
Add in GLSL setting again. PS and VS making. Untested and won't work for now. Add in program shader cache files. Readd NativeVertexFormat stuffs. Add in PS and VS cache things. SetShaders in places. Fixed EFB cache index computations in OpenGL renderer. The previous computation was very likely to go out of array bounds, which could result in crashes on EFB access. Also, the cache size was rounded down instead of up. This is a problem since EFB_HEIGHT (528) is not a multiple of EFB_CACHE_RECT_SIZE (64).
2011-12-01 03:00:21 +00:00
extern const char* WriteRegister(API_TYPE ApiType, const char *prefix, const u32 num);
Add support for GL_ARB_shading_language_420pack so we don't have to binding sampler locations. Also add support for GL_ARB_separate_shader_objects which doesn't currently work for some reason....investigating.
2011-12-08 07:51:08 +00:00
extern const char* WriteBinding(API_TYPE ApiType, const u32 num);
Fix one error.
2011-12-10 03:15:15 +00:00
const char *WriteLocation(API_TYPE ApiType);
Add in GLSL setting again. PS and VS making. Untested and won't work for now. Add in program shader cache files. Readd NativeVertexFormat stuffs. Add in PS and VS cache things. SetShaders in places. Fixed EFB cache index computations in OpenGL renderer. The previous computation was very likely to go out of array bounds, which could result in crashes on EFB access. Also, the cache size was rounded down instead of up. This is a problem since EFB_HEIGHT (528) is not a multiple of EFB_CACHE_RECT_SIZE (64).
2011-12-01 03:00:21 +00:00
const char *GenerateVertexShaderCode(u32 components, API_TYPE ApiType)
Fixup line endings.
2011-09-29 19:52:13 +00:00
{
setlocale(LC_NUMERIC, "C"); // Reset locale for compilation
text[sizeof(text) - 1] = 0x7C; // canary
_assert_(bpmem.genMode.numtexgens == xfregs.numTexGen.numTexGens);
_assert_(bpmem.genMode.numcolchans == xfregs.numChan.numColorChans);
Add in GLSL setting again. PS and VS making. Untested and won't work for now. Add in program shader cache files. Readd NativeVertexFormat stuffs. Add in PS and VS cache things. SetShaders in places. Fixed EFB cache index computations in OpenGL renderer. The previous computation was very likely to go out of array bounds, which could result in crashes on EFB access. Also, the cache size was rounded down instead of up. This is a problem since EFB_HEIGHT (528) is not a multiple of EFB_CACHE_RECT_SIZE (64).
2011-12-01 03:00:21 +00:00
bool is_d3d = (ApiType & API_D3D9 || ApiType == API_D3D11);
Fixup line endings.
2011-09-29 19:52:13 +00:00
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();
char *p = text;
WRITE(p, "//Vertex Shader: comp:%x, \n", components);
Add in GLSL setting again. PS and VS making. Untested and won't work for now. Add in program shader cache files. Readd NativeVertexFormat stuffs. Add in PS and VS cache things. SetShaders in places. Fixed EFB cache index computations in OpenGL renderer. The previous computation was very likely to go out of array bounds, which could result in crashes on EFB access. Also, the cache size was rounded down instead of up. This is a problem since EFB_HEIGHT (528) is not a multiple of EFB_CACHE_RECT_SIZE (64).
2011-12-01 03:00:21 +00:00
if(ApiType == API_GLSL)
{
// A few required defines and ones that will make our lives a lot easier
Remove bSupportsGLSLLocation since it won't work how I expect it.
2011-12-10 01:14:02 +00:00
if (g_ActiveConfig.backend_info.bSupportsGLSLBinding || g_ActiveConfig.backend_info.bSupportsGLSLUBO)
Add support for GL_ARB_shading_language_420pack so we don't have to binding sampler locations. Also add support for GL_ARB_separate_shader_objects which doesn't currently work for some reason....investigating.
2011-12-08 07:51:08 +00:00
{
WRITE(p, "#version 330 compatibility\n");
if (g_ActiveConfig.backend_info.bSupportsGLSLBinding)
WRITE(p, "#extension GL_ARB_shading_language_420pack : enable\n");
Remove bSupportsGLSLLocation since it won't work how I expect it.
2011-12-10 01:14:02 +00:00
if (g_ActiveConfig.backend_info.bSupportsGLSLUBO)
WRITE(p, "#extension GL_ARB_uniform_buffer_object : enable\n");
Bit of cleanup. Clean up my massive hack in the ShaderManagerFiles. Almost feature parity with Nvidia CG now I think. Just need to do Alpha test with Dual source blending now.
2011-12-08 09:20:31 +00:00
WRITE(p, "#define ATTRIN in\n");
WRITE(p, "#define ATTROUT out\n");
Add support for GL_ARB_shading_language_420pack so we don't have to binding sampler locations. Also add support for GL_ARB_separate_shader_objects which doesn't currently work for some reason....investigating.
2011-12-08 07:51:08 +00:00
}
else
Bit of cleanup. Clean up my massive hack in the ShaderManagerFiles. Almost feature parity with Nvidia CG now I think. Just need to do Alpha test with Dual source blending now.
2011-12-08 09:20:31 +00:00
{
Add support for GL_ARB_shading_language_420pack so we don't have to binding sampler locations. Also add support for GL_ARB_separate_shader_objects which doesn't currently work for some reason....investigating.
2011-12-08 07:51:08 +00:00
WRITE(p, "#version 120\n");
Bit of cleanup. Clean up my massive hack in the ShaderManagerFiles. Almost feature parity with Nvidia CG now I think. Just need to do Alpha test with Dual source blending now.
2011-12-08 09:20:31 +00:00
WRITE(p, "#define ATTRIN attribute\n");
WRITE(p, "#define ATTROUT attribute\n");
}
Add in GLSL setting again. PS and VS making. Untested and won't work for now. Add in program shader cache files. Readd NativeVertexFormat stuffs. Add in PS and VS cache things. SetShaders in places. Fixed EFB cache index computations in OpenGL renderer. The previous computation was very likely to go out of array bounds, which could result in crashes on EFB access. Also, the cache size was rounded down instead of up. This is a problem since EFB_HEIGHT (528) is not a multiple of EFB_CACHE_RECT_SIZE (64).
2011-12-01 03:00:21 +00:00
// Silly differences
WRITE(p, "#define float2 vec2\n");
WRITE(p, "#define float3 vec3\n");
WRITE(p, "#define float4 vec4\n");
// cg to glsl function translation
WRITE(p, "#define frac(x) fract(x)\n");
WRITE(p, "#define saturate(x) clamp(x, 0.0f, 1.0f)\n");
WRITE(p, "#define lerp(x, y, z) mix(x, y, z)\n");
Fixup line endings.
2011-09-29 19:52:13 +00:00
}
Add in UBOs, doesn't work yet. Still debugging here.
2011-12-09 23:30:05 +00:00
Add in GLSL setting again. PS and VS making. Untested and won't work for now. Add in program shader cache files. Readd NativeVertexFormat stuffs. Add in PS and VS cache things. SetShaders in places. Fixed EFB cache index computations in OpenGL renderer. The previous computation was very likely to go out of array bounds, which could result in crashes on EFB access. Also, the cache size was rounded down instead of up. This is a problem since EFB_HEIGHT (528) is not a multiple of EFB_CACHE_RECT_SIZE (64).
2011-12-01 03:00:21 +00:00
// uniforms
yay, UBOs work 100% now.
2011-12-10 14:07:13 +00:00
if(ApiType == API_GLSL && g_ActiveConfig.backend_info.bSupportsGLSLUBO)
Make sure to support everything even if GPU doesn't.
2011-12-10 20:35:37 +00:00
WRITE(p, "layout(std140%s) uniform VSBlock {\n", g_ActiveConfig.backend_info.bSupportsGLSLBinding ? ", binding = 2" : "");
yay, UBOs work 100% now.
2011-12-10 14:07:13 +00:00
WRITE(p, "%sfloat4 "I_POSNORMALMATRIX"[6] %s;\n", WriteLocation(ApiType), WriteRegister(ApiType, "c", C_POSNORMALMATRIX));
WRITE(p, "%sfloat4 "I_PROJECTION"[4] %s;\n", WriteLocation(ApiType), WriteRegister(ApiType, "c", C_PROJECTION));
WRITE(p, "%sfloat4 "I_MATERIALS"[4] %s;\n", WriteLocation(ApiType), WriteRegister(ApiType, "c", C_MATERIALS));
WRITE(p, "%sfloat4 "I_LIGHTS"[40] %s;\n", WriteLocation(ApiType), WriteRegister(ApiType, "c", C_LIGHTS));
WRITE(p, "%sfloat4 "I_TEXMATRICES"[24] %s;\n", WriteLocation(ApiType), WriteRegister(ApiType, "c", C_TEXMATRICES)); // also using tex matrices
WRITE(p, "%sfloat4 "I_TRANSFORMMATRICES"[64] %s;\n", WriteLocation(ApiType),WriteRegister(ApiType, "c", C_TRANSFORMMATRICES));
WRITE(p, "%sfloat4 "I_NORMALMATRICES"[32] %s;\n", WriteLocation(ApiType), WriteRegister(ApiType, "c", C_NORMALMATRICES));
WRITE(p, "%sfloat4 "I_POSTTRANSFORMMATRICES"[64] %s;\n", WriteLocation(ApiType), WriteRegister(ApiType, "c", C_POSTTRANSFORMMATRICES));
WRITE(p, "%sfloat4 "I_DEPTHPARAMS" %s;\n", WriteLocation(ApiType), WriteRegister(ApiType, "c", C_DEPTHPARAMS));
Add in UBOs, doesn't work yet. Still debugging here.
2011-12-09 23:30:05 +00:00
yay, UBOs work 100% now.
2011-12-10 14:07:13 +00:00
if(ApiType == API_GLSL && g_ActiveConfig.backend_info.bSupportsGLSLUBO)
WRITE(p, "};\n");
UBO works for Pixel Shaders if Binding for UBO is zero, otherwise fails. Probably why Vertex shader UBO is failing. Too tired to investigate right now.
2011-12-10 13:38:30 +00:00
Add in UBOs, doesn't work yet. Still debugging here.
2011-12-09 23:30:05 +00:00
Add in GLSL setting again. PS and VS making. Untested and won't work for now. Add in program shader cache files. Readd NativeVertexFormat stuffs. Add in PS and VS cache things. SetShaders in places. Fixed EFB cache index computations in OpenGL renderer. The previous computation was very likely to go out of array bounds, which could result in crashes on EFB access. Also, the cache size was rounded down instead of up. This is a problem since EFB_HEIGHT (528) is not a multiple of EFB_CACHE_RECT_SIZE (64).
2011-12-01 03:00:21 +00:00
p = GenerateVSOutputStruct(p, components, ApiType);
if(ApiType == API_GLSL)
{
if (components & VB_HAS_NRM0)
WRITE(p, " float3 rawnorm0 = gl_Normal; // NORMAL0,\n");
if (components & VB_HAS_POSMTXIDX)
Bit of cleanup. Clean up my massive hack in the ShaderManagerFiles. Almost feature parity with Nvidia CG now I think. Just need to do Alpha test with Dual source blending now.
2011-12-08 09:20:31 +00:00
WRITE(p, "ATTRIN float fposmtx; // ATTR%d,\n", SHADER_POSMTX_ATTRIB);
Add in GLSL setting again. PS and VS making. Untested and won't work for now. Add in program shader cache files. Readd NativeVertexFormat stuffs. Add in PS and VS cache things. SetShaders in places. Fixed EFB cache index computations in OpenGL renderer. The previous computation was very likely to go out of array bounds, which could result in crashes on EFB access. Also, the cache size was rounded down instead of up. This is a problem since EFB_HEIGHT (528) is not a multiple of EFB_CACHE_RECT_SIZE (64).
2011-12-01 03:00:21 +00:00
if (components & VB_HAS_NRM1)
Bit of cleanup. Clean up my massive hack in the ShaderManagerFiles. Almost feature parity with Nvidia CG now I think. Just need to do Alpha test with Dual source blending now.
2011-12-08 09:20:31 +00:00
WRITE(p, "ATTRIN float3 rawnorm1; // ATTR%d,\n", SHADER_NORM1_ATTRIB);
Add in GLSL setting again. PS and VS making. Untested and won't work for now. Add in program shader cache files. Readd NativeVertexFormat stuffs. Add in PS and VS cache things. SetShaders in places. Fixed EFB cache index computations in OpenGL renderer. The previous computation was very likely to go out of array bounds, which could result in crashes on EFB access. Also, the cache size was rounded down instead of up. This is a problem since EFB_HEIGHT (528) is not a multiple of EFB_CACHE_RECT_SIZE (64).
2011-12-01 03:00:21 +00:00
if (components & VB_HAS_NRM2)
Bit of cleanup. Clean up my massive hack in the ShaderManagerFiles. Almost feature parity with Nvidia CG now I think. Just need to do Alpha test with Dual source blending now.
2011-12-08 09:20:31 +00:00
WRITE(p, "ATTRIN float3 rawnorm2; // ATTR%d,\n", SHADER_NORM2_ATTRIB);
Add in GLSL setting again. PS and VS making. Untested and won't work for now. Add in program shader cache files. Readd NativeVertexFormat stuffs. Add in PS and VS cache things. SetShaders in places. Fixed EFB cache index computations in OpenGL renderer. The previous computation was very likely to go out of array bounds, which could result in crashes on EFB access. Also, the cache size was rounded down instead of up. This is a problem since EFB_HEIGHT (528) is not a multiple of EFB_CACHE_RECT_SIZE (64).
2011-12-01 03:00:21 +00:00
if (components & VB_HAS_COL0)
WRITE(p, " float4 color0 = gl_Color; // COLOR0,\n");
if (components & VB_HAS_COL1)
WRITE(p, " float4 color1 = gl_SecondaryColor; // COLOR1,\n");
for (int i = 0; i < 8; ++i) {
u32 hastexmtx = (components & (VB_HAS_TEXMTXIDX0<<i));
if ((components & (VB_HAS_UV0<<i)) || hastexmtx)
WRITE(p, " float%d tex%d = gl_MultiTexCoord%d.xy%s; // TEXCOORD%d,\n", hastexmtx ? 3 : 2, i, i, hastexmtx ? "z" : "", i);
}
WRITE(p, " float4 rawpos = gl_Vertex;\n") ;
WRITE(p, "void main()\n{\n");
Fixup line endings.
2011-09-29 19:52:13 +00:00
}
Add in GLSL setting again. PS and VS making. Untested and won't work for now. Add in program shader cache files. Readd NativeVertexFormat stuffs. Add in PS and VS cache things. SetShaders in places. Fixed EFB cache index computations in OpenGL renderer. The previous computation was very likely to go out of array bounds, which could result in crashes on EFB access. Also, the cache size was rounded down instead of up. This is a problem since EFB_HEIGHT (528) is not a multiple of EFB_CACHE_RECT_SIZE (64).
2011-12-01 03:00:21 +00:00
else
{
WRITE(p, "VS_OUTPUT main(\n");
// inputs
if (components & VB_HAS_NRM0)
WRITE(p, " float3 rawnorm0 : NORMAL0,\n");
if (components & VB_HAS_NRM1) {
if (is_d3d)
WRITE(p, " float3 rawnorm1 : NORMAL1,\n");
else
WRITE(p, " float3 rawnorm1 : ATTR%d,\n", SHADER_NORM1_ATTRIB);
Fixup line endings.
2011-09-29 19:52:13 +00:00
}
Add in GLSL setting again. PS and VS making. Untested and won't work for now. Add in program shader cache files. Readd NativeVertexFormat stuffs. Add in PS and VS cache things. SetShaders in places. Fixed EFB cache index computations in OpenGL renderer. The previous computation was very likely to go out of array bounds, which could result in crashes on EFB access. Also, the cache size was rounded down instead of up. This is a problem since EFB_HEIGHT (528) is not a multiple of EFB_CACHE_RECT_SIZE (64).
2011-12-01 03:00:21 +00:00
if (components & VB_HAS_NRM2) {
if (is_d3d)
WRITE(p, " float3 rawnorm2 : NORMAL2,\n");
else
WRITE(p, " float3 rawnorm2 : ATTR%d,\n", SHADER_NORM2_ATTRIB);
}
if (components & VB_HAS_COL0)
WRITE(p, " float4 color0 : COLOR0,\n");
if (components & VB_HAS_COL1)
WRITE(p, " float4 color1 : COLOR1,\n");
for (int i = 0; i < 8; ++i) {
u32 hastexmtx = (components & (VB_HAS_TEXMTXIDX0<<i));
if ((components & (VB_HAS_UV0<<i)) || hastexmtx)
WRITE(p, " float%d tex%d : TEXCOORD%d,\n", hastexmtx ? 3 : 2, i, i);
}
if (components & VB_HAS_POSMTXIDX) {
if (is_d3d)
{
WRITE(p, " float4 blend_indices : BLENDINDICES,\n");
}
else
WRITE(p, " float fposmtx : ATTR%d,\n", SHADER_POSMTX_ATTRIB);
}
WRITE(p, " float4 rawpos : POSITION) {\n");
Fixup line endings.
2011-09-29 19:52:13 +00:00
}
WRITE(p, "VS_OUTPUT o;\n");
// transforms
if (components & VB_HAS_POSMTXIDX)
{
Add in GLSL setting again. PS and VS making. Untested and won't work for now. Add in program shader cache files. Readd NativeVertexFormat stuffs. Add in PS and VS cache things. SetShaders in places. Fixed EFB cache index computations in OpenGL renderer. The previous computation was very likely to go out of array bounds, which could result in crashes on EFB access. Also, the cache size was rounded down instead of up. This is a problem since EFB_HEIGHT (528) is not a multiple of EFB_CACHE_RECT_SIZE (64).
2011-12-01 03:00:21 +00:00
if (ApiType & API_D3D9)
Fixup line endings.
2011-09-29 19:52:13 +00:00
{
WRITE(p, "int4 indices = D3DCOLORtoUBYTE4(blend_indices);\n");
WRITE(p, "int posmtx = indices.x;\n");
}
Add in GLSL setting again. PS and VS making. Untested and won't work for now. Add in program shader cache files. Readd NativeVertexFormat stuffs. Add in PS and VS cache things. SetShaders in places. Fixed EFB cache index computations in OpenGL renderer. The previous computation was very likely to go out of array bounds, which could result in crashes on EFB access. Also, the cache size was rounded down instead of up. This is a problem since EFB_HEIGHT (528) is not a multiple of EFB_CACHE_RECT_SIZE (64).
2011-12-01 03:00:21 +00:00
else if (ApiType == API_D3D11)
Fixup line endings.
2011-09-29 19:52:13 +00:00
{
WRITE(p, "int posmtx = blend_indices.x * 255.0f;\n");
}
else
{
Almost there.
2011-12-08 04:04:34 +00:00
WRITE(p, "int posmtx = int(fposmtx);\n");
Fixup line endings.
2011-09-29 19:52:13 +00:00
}
Now CG plays nice with this new stuff.
2011-12-01 06:33:12 +00:00
WRITE(p, "float4 pos = float4(dot("I_TRANSFORMMATRICES"[posmtx], rawpos), dot("I_TRANSFORMMATRICES"[posmtx+1], rawpos), dot("I_TRANSFORMMATRICES"[posmtx+2], rawpos), 1);\n");
Fixup line endings.
2011-09-29 19:52:13 +00:00
if (components & VB_HAS_NRMALL) {
WRITE(p, "int normidx = posmtx >= 32 ? (posmtx-32) : posmtx;\n");
Now CG plays nice with this new stuff.
2011-12-01 06:33:12 +00:00
WRITE(p, "float3 N0 = "I_NORMALMATRICES"[normidx].xyz, N1 = "I_NORMALMATRICES"[normidx+1].xyz, N2 = "I_NORMALMATRICES"[normidx+2].xyz;\n");
Fixup line endings.
2011-09-29 19:52:13 +00:00
}
if (components & VB_HAS_NRM0)
WRITE(p, "float3 _norm0 = normalize(float3(dot(N0, rawnorm0), dot(N1, rawnorm0), dot(N2, rawnorm0)));\n");
if (components & VB_HAS_NRM1)
WRITE(p, "float3 _norm1 = float3(dot(N0, rawnorm1), dot(N1, rawnorm1), dot(N2, rawnorm1));\n");
if (components & VB_HAS_NRM2)
WRITE(p, "float3 _norm2 = float3(dot(N0, rawnorm2), dot(N1, rawnorm2), dot(N2, rawnorm2));\n");
}
else
{
Now CG plays nice with this new stuff.
2011-12-01 06:33:12 +00:00
WRITE(p, "float4 pos = float4(dot("I_POSNORMALMATRIX"[0], rawpos), dot("I_POSNORMALMATRIX"[1], rawpos), dot("I_POSNORMALMATRIX"[2], rawpos), 1.0f);\n");
Fixup line endings.
2011-09-29 19:52:13 +00:00
if (components & VB_HAS_NRM0)
Now CG plays nice with this new stuff.
2011-12-01 06:33:12 +00:00
WRITE(p, "float3 _norm0 = normalize(float3(dot("I_POSNORMALMATRIX"[3].xyz, rawnorm0), dot("I_POSNORMALMATRIX"[4].xyz, rawnorm0), dot("I_POSNORMALMATRIX"[5].xyz, rawnorm0)));\n");
Fixup line endings.
2011-09-29 19:52:13 +00:00
if (components & VB_HAS_NRM1)
Now CG plays nice with this new stuff.
2011-12-01 06:33:12 +00:00
WRITE(p, "float3 _norm1 = float3(dot("I_POSNORMALMATRIX"[3].xyz, rawnorm1), dot("I_POSNORMALMATRIX"[4].xyz, rawnorm1), dot("I_POSNORMALMATRIX"[5].xyz, rawnorm1));\n");
Fixup line endings.
2011-09-29 19:52:13 +00:00
if (components & VB_HAS_NRM2)
Playing through SSBM story made me find this.
2011-12-08 10:11:30 +00:00
WRITE(p, "float3 _norm2 = float3(dot("I_POSNORMALMATRIX"[3].xyz, rawnorm2), dot("I_POSNORMALMATRIX"[4].xyz, rawnorm2), dot("I_POSNORMALMATRIX"[5].xyz, rawnorm2));\n");
Fixup line endings.
2011-09-29 19:52:13 +00:00
}
if (!(components & VB_HAS_NRM0))
WRITE(p, "float3 _norm0 = float3(0.0f, 0.0f, 0.0f);\n");
Now CG plays nice with this new stuff.
2011-12-01 06:33:12 +00:00
WRITE(p, "o.pos = float4(dot("I_PROJECTION"[0], pos), dot("I_PROJECTION"[1], pos), dot("I_PROJECTION"[2], pos), dot("I_PROJECTION"[3], pos));\n");
Fixup line endings.
2011-09-29 19:52:13 +00:00
WRITE(p, "float4 mat, lacc;\n"
"float3 ldir, h;\n"
"float dist, dist2, attn;\n");
if(xfregs.numChan.numColorChans == 0)
{
if (components & VB_HAS_COL0)
WRITE(p, "o.colors_0 = color0;\n");
else
WRITE(p, "o.colors_0 = float4(1.0f, 1.0f, 1.0f, 1.0f);\n");
}
// TODO: This probably isn't necessary if pixel lighting is enabled.
p = GenerateLightingShader(p, components, I_MATERIALS, I_LIGHTS, "color", "o.colors_");
if(xfregs.numChan.numColorChans < 2)
{
if (components & VB_HAS_COL1)
WRITE(p, "o.colors_1 = color1;\n");
else
WRITE(p, "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
WRITE(p, "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];
WRITE(p, "{\n");
WRITE(p, "coord = float4(0.0f, 0.0f, 1.0f, 1.0f);\n");
switch (texinfo.sourcerow) {
case XF_SRCGEOM_INROW:
_assert_( texinfo.inputform == XF_TEXINPUT_ABC1 );
WRITE(p, "coord = rawpos;\n"); // pos.w is 1
break;
case XF_SRCNORMAL_INROW:
if (components & VB_HAS_NRM0) {
_assert_( texinfo.inputform == XF_TEXINPUT_ABC1 );
WRITE(p, "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 );
WRITE(p, "coord = float4(rawnorm1.xyz, 1.0f);\n");
}
break;
case XF_SRCBINORMAL_B_INROW:
if (components & VB_HAS_NRM2) {
_assert_( texinfo.inputform == XF_TEXINPUT_ABC1 );
WRITE(p, "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)) )
WRITE(p, "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
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
Add support for GL_ARB_shading_language_420pack so we don't have to binding sampler locations. Also add support for GL_ARB_separate_shader_objects which doesn't currently work for some reason....investigating.
2011-12-08 07:51:08 +00:00
WRITE(p, "ldir = normalize("I_LIGHTS"[%d + 3].xyz - pos.xyz);\n", texinfo.embosslightshift);
Fixup line endings.
2011-09-29 19:52:13 +00:00
WRITE(p, "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
WRITE(p, "o.tex%d.xyz = o.tex%d.xyz;\n", i, texinfo.embosssourceshift);
}
break;
case XF_TEXGEN_COLOR_STRGBC0:
_assert_(texinfo.sourcerow == XF_SRCCOLORS_INROW);
WRITE(p, "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);
WRITE(p, "o.tex%d.xyz = float3(o.colors_1.x, o.colors_1.y, 1);\n", i);
break;
case XF_TEXGEN_REGULAR:
default:
Now CG plays nice with this new stuff.
2011-12-01 06:33:12 +00:00
if (components & (VB_HAS_TEXMTXIDX0<<i))
{
WRITE(p, "int tmp = int(tex%d.z);\n", i);
Fixup line endings.
2011-09-29 19:52:13 +00:00
if (texinfo.projection == XF_TEXPROJ_STQ)
Now CG plays nice with this new stuff.
2011-12-01 06:33:12 +00:00
WRITE(p, "o.tex%d.xyz = float3(dot(coord, "I_TRANSFORMMATRICES"[tmp]), dot(coord, "I_TRANSFORMMATRICES"[tmp+1]), dot(coord, "I_TRANSFORMMATRICES"[tmp+2]));\n", i);
Fixup line endings.
2011-09-29 19:52:13 +00:00
else {
Now CG plays nice with this new stuff.
2011-12-01 06:33:12 +00:00
WRITE(p, "o.tex%d.xyz = float3(dot(coord, "I_TRANSFORMMATRICES"[tmp]), dot(coord, "I_TRANSFORMMATRICES"[tmp+1]), 1);\n", i);
Fixup line endings.
2011-09-29 19:52:13 +00:00
}
}
else {
if (texinfo.projection == XF_TEXPROJ_STQ)
Now CG plays nice with this new stuff.
2011-12-01 06:33:12 +00:00
WRITE(p, "o.tex%d.xyz = float3(dot(coord, "I_TEXMATRICES"[%d]), dot(coord, "I_TEXMATRICES"[%d]), dot(coord, "I_TEXMATRICES"[%d]));\n", i, 3*i, 3*i+1, 3*i+2);
Fixup line endings.
2011-09-29 19:52:13 +00:00
else
Now CG plays nice with this new stuff.
2011-12-01 06:33:12 +00:00
WRITE(p, "o.tex%d.xyz = float3(dot(coord, "I_TEXMATRICES"[%d]), dot(coord, "I_TEXMATRICES"[%d]), 1);\n", i, 3*i, 3*i+1);
Fixup line endings.
2011-09-29 19:52:13 +00:00
}
break;
}
if (xfregs.dualTexTrans.enabled && texinfo.texgentype == XF_TEXGEN_REGULAR) { // only works for regular tex gen types?
const PostMtxInfo& postInfo = xfregs.postMtxInfo[i];
int postidx = postInfo.index;
Now CG plays nice with this new stuff.
2011-12-01 06:33:12 +00:00
WRITE(p, "float4 P0 = "I_POSTTRANSFORMMATRICES"[%d];\n"
"float4 P1 = "I_POSTTRANSFORMMATRICES"[%d];\n"
"float4 P2 = "I_POSTTRANSFORMMATRICES"[%d];\n",
Fixup line endings.
2011-09-29 19:52:13 +00:00
postidx&0x3f, (postidx+1)&0x3f, (postidx+2)&0x3f);
if (texGenSpecialCase) {
// no normalization
// q of input is 1
// q of output is unknown
// multiply by postmatrix
WRITE(p, "o.tex%d.xyz = float3(dot(P0.xy, o.tex%d.xy) + P0.z + P0.w, dot(P1.xy, o.tex%d.xy) + P1.z + P1.w, 0.0f);\n", i, i, i);
}
else
{
if (postInfo.normalize)
WRITE(p, "o.tex%d.xyz = normalize(o.tex%d.xyz);\n", i, i);
// multiply by postmatrix
WRITE(p, "o.tex%d.xyz = float3(dot(P0.xyz, o.tex%d.xyz) + P0.w, dot(P1.xyz, o.tex%d.xyz) + P1.w, dot(P2.xyz, o.tex%d.xyz) + P2.w);\n", i, i, i, i);
}
}
WRITE(p, "}\n");
}
// clipPos/w needs to be done in pixel shader, not here
if (xfregs.numTexGen.numTexGens < 7) {
WRITE(p, "o.clipPos = float4(pos.x,pos.y,o.pos.z,o.pos.w);\n");
} else {
WRITE(p, "o.tex0.w = pos.x;\n");
WRITE(p, "o.tex1.w = pos.y;\n");
WRITE(p, "o.tex2.w = o.pos.z;\n");
WRITE(p, "o.tex3.w = o.pos.w;\n");
}
if(g_ActiveConfig.bEnablePixelLighting && g_ActiveConfig.backend_info.bSupportsPixelLighting)
{
if (xfregs.numTexGen.numTexGens < 7) {
WRITE(p, "o.Normal = float4(_norm0.x,_norm0.y,_norm0.z,pos.z);\n");
} else {
WRITE(p, "o.tex4.w = _norm0.x;\n");
WRITE(p, "o.tex5.w = _norm0.y;\n");
WRITE(p, "o.tex6.w = _norm0.z;\n");
if (xfregs.numTexGen.numTexGens < 8)
WRITE(p, "o.tex7 = pos.xyzz;\n");
else
WRITE(p, "o.tex7.w = pos.z;\n");
}
if (components & VB_HAS_COL0)
WRITE(p, "o.colors_0 = color0;\n");
if (components & VB_HAS_COL1)
WRITE(p, "o.colors_1 = color1;\n");
}
//write the true depth value, if the game uses depth textures pixel shaders will override with the correct values
//if not early z culling will improve speed
if (is_d3d)
{
Fix compilation with g++ 4.7 and some warnings. Fixes issue 5347.
2012-04-07 19:45:12 +00:00
WRITE(p, "o.pos.z = " I_DEPTHPARAMS".x * o.pos.w + o.pos.z * " I_DEPTHPARAMS".y;\n");
Fixup line endings.
2011-09-29 19:52:13 +00:00
}
else
{
// this results in a scale from -1..0 to -1..1 after perspective
// divide
WRITE(p, "o.pos.z = o.pos.w + o.pos.z * 2.0f;\n");
// Sonic Unleashed puts its final rendering at the near or
// far plane of the viewing frustrum(actually box, they use
// orthogonal projection for that), and we end up putting it
// just beyond, and the rendering gets clipped away. (The
// primitive gets dropped)
WRITE(p, "o.pos.z = o.pos.z * 1048575.0f/1048576.0f;\n");
// the next steps of the OGL pipeline are:
// (x_c,y_c,z_c,w_c) = o.pos //switch to OGL spec terminology
// clipping to -w_c <= (x_c,y_c,z_c) <= w_c
// (x_d,y_d,z_d) = (x_c,y_c,z_c)/w_c//perspective divide
// z_w = (f-n)/2*z_d + (n+f)/2
// z_w now contains the value to go to the 0..1 depth buffer
//trying to get the correct semantic while not using glDepthRange
//seems to get rather complicated
}
D3D9: Fix texel to pixel mapping when sampling textures properly.
2012-03-24 15:30:00 +00:00
if (api_type & API_D3D9)
{
// D3D9 is addressing pixel centers instead of pixel boundaries in clip space.
// Thus we need to offset the final position by half a pixel
Fix compilation with g++ 4.7 and some warnings. Fixes issue 5347.
2012-04-07 19:45:12 +00:00
WRITE(p, "o.pos = o.pos + float4(" I_DEPTHPARAMS".z, " I_DEPTHPARAMS".w, 0.f, 0.f);\n");
D3D9: Fix texel to pixel mapping when sampling textures properly.
2012-03-24 15:30:00 +00:00
}
Add in GLSL setting again. PS and VS making. Untested and won't work for now. Add in program shader cache files. Readd NativeVertexFormat stuffs. Add in PS and VS cache things. SetShaders in places. Fixed EFB cache index computations in OpenGL renderer. The previous computation was very likely to go out of array bounds, which could result in crashes on EFB access. Also, the cache size was rounded down instead of up. This is a problem since EFB_HEIGHT (528) is not a multiple of EFB_CACHE_RECT_SIZE (64).
2011-12-01 03:00:21 +00:00
if(ApiType == API_GLSL)
{
// Bit ugly here
// Will look better when we bind uniforms in GLSL 1.3
// clipPos/w needs to be done in pixel shader, not here
Fixup line endings.
2011-09-29 19:52:13 +00:00
Add in GLSL setting again. PS and VS making. Untested and won't work for now. Add in program shader cache files. Readd NativeVertexFormat stuffs. Add in PS and VS cache things. SetShaders in places. Fixed EFB cache index computations in OpenGL renderer. The previous computation was very likely to go out of array bounds, which could result in crashes on EFB access. Also, the cache size was rounded down instead of up. This is a problem since EFB_HEIGHT (528) is not a multiple of EFB_CACHE_RECT_SIZE (64).
2011-12-01 03:00:21 +00:00
if (xfregs.numTexGen.numTexGens < 7) {
for (unsigned int i = 0; i < xfregs.numTexGen.numTexGens; ++i)
WRITE(p, " gl_TexCoord[%d].xyz = o.tex%d;\n", i, i);
WRITE(p, " gl_TexCoord[%d] = o.clipPos;\n", xfregs.numTexGen.numTexGens);
if(g_ActiveConfig.bEnablePixelLighting && g_ActiveConfig.backend_info.bSupportsPixelLighting)
WRITE(p, " gl_TexCoord[%d] = o.Normal;\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)
WRITE(p, " gl_TexCoord[%d] = o.tex%d;\n", i, i);
}
else
{
for (unsigned int i = 0; i < xfregs.numTexGen.numTexGens; ++i)
WRITE(p, " gl_TexCoord[%d]%s = o.tex%d;\n", i, i < 4 ? ".xyzw" : ".xyz" , i);
}
}
WRITE(p, "gl_FrontColor = o.colors_0;\n");
WRITE(p, "gl_FrontSecondaryColor = o.colors_1;\n");
WRITE(p, "gl_Position = o.pos;\n");
WRITE(p, "}\n");
}
else
WRITE(p, "return o;\n}\n");
Fixup line endings.
2011-09-29 19:52:13 +00:00
if (text[sizeof(text) - 1] != 0x7C)
PanicAlert("VertexShader generator - buffer too small, canary has been eaten!");
setlocale(LC_NUMERIC, ""); // restore locale
return text;
}