Merge pull request from hrydgard/merge-glsl-hlsl-vertex-shadergens

Merge GLSL and HLSL vertex shader generators
This commit is contained in:
Henrik Rydgård 2020-11-01 21:51:25 +01:00 committed by GitHub
commit 34874238ac
No known key found for this signature in database
GPG Key ID: 4AEE18F83AFDEB23
24 changed files with 481 additions and 1060 deletions

View File

@ -1250,8 +1250,6 @@ set(GPU_GLES
GPU/GLES/TextureScalerGLES.h
GPU/GLES/DrawEngineGLES.cpp
GPU/GLES/DrawEngineGLES.h
GPU/GLES/VertexShaderGeneratorGLES.cpp
GPU/GLES/VertexShaderGeneratorGLES.h
)
set(GPU_VULKAN
@ -1299,7 +1297,7 @@ set(GPU_D3D9
GPU/Directx9/TextureScalerDX9.cpp
GPU/Directx9/TextureScalerDX9.h
GPU/Directx9/VertexShaderGeneratorHLSL.cpp
GPU/Directx9/VertexShaderGeneratorHLSL.h
GPU/GLES/VertexShaderGeneratorGLES.h
)
set(GPU_D3D11
@ -1340,6 +1338,8 @@ set(GPU_SOURCES
GPU/Common/DepalettizeShaderCommon.h
GPU/Common/FragmentShaderGenerator.cpp
GPU/Common/FragmentShaderGenerator.h
GPU/Common/VertexShaderGenerator.cpp
GPU/Common/VertexShaderGenerator.h
GPU/Common/FramebufferManagerCommon.cpp
GPU/Common/FramebufferManagerCommon.h
GPU/Common/GPUDebugInterface.cpp

View File

@ -51,6 +51,9 @@ const char *hlsl_preamble_fs =
"#define vec4 float4\n"
"#define uvec3 uint3\n"
"#define ivec3 int3\n"
"#define ivec4 int4\n"
"#define mat4 float4x4\n"
"#define mat3x4 float4x3\n" // note how the conventions are backwards
"#define splat3(x) float3(x, x, x)\n"
"#define mix lerp\n"
"#define mod(x, y) fmod(x, y)\n";
@ -246,7 +249,7 @@ bool GenerateFragmentShader(const FShaderID &id, char *buffer, const ShaderLangu
WRITE(p, "Texture2D<vec4> fboTex : register(t1);\n");
}
}
WRITE(p, "cbuffer base : register(b0) {\n%s};\n", cb_baseStr);
WRITE(p, "cbuffer base : register(b0) {\n%s};\n", ub_baseStr);
}
if (enableAlphaTest) {

View File

@ -171,6 +171,7 @@ ShaderLanguageDesc::ShaderLanguageDesc(ShaderLanguage lang) {
texelFetch = "texelFetch";
forceMatrix4x4 = false;
coefsFromBuffers = true;
vsOutPrefix = "Out.";
break;
}
}

View File

@ -157,6 +157,7 @@ struct ShaderLanguageDesc {
const char *texelFetch = nullptr;
const char *lastFragData = nullptr;
const char *framebufferFetchExtension = nullptr;
const char *vsOutPrefix = "";
bool glslES30 = false;
bool bitwiseOps = false;
bool forceMatrix4x4 = false;
@ -175,3 +176,16 @@ enum class PspAttributeLocation {
COUNT
};
// Pre-fetched attrs and uniforms (used by GL only).
enum {
ATTR_POSITION = 0,
ATTR_TEXCOORD = 1,
ATTR_NORMAL = 2,
ATTR_W1 = 3,
ATTR_W2 = 4,
ATTR_COLOR0 = 5,
ATTR_COLOR1 = 6,
ATTR_COUNT,
};

View File

@ -72,34 +72,6 @@ R"( mat4 u_proj;
vec2 u_texclampoff;
)";
// HLSL code is shared so these names are changed to match those in DX9.
static const char *cb_baseStr =
R"( float4x4 u_proj;
float4x4 u_proj_through;
float4x3 u_view;
float4x3 u_world;
float4x3 u_texmtx;
float4 u_uvscaleoffset;
float4 u_depthRange;
float2 u_fogcoef;
float u_stencilReplaceValue;
float4 u_matambientalpha;
uint u_spline_counts;
uint u_depal_mask_shift_off_fmt;
int pad2;
int pad3;
float4 u_cullRangeMin;
float4 u_cullRangeMax;
float3 u_fogcolor;
float3 u_texenv;
uint4 u_alphacolorref;
uint4 u_alphacolormask;
float3 u_blendFixA;
float3 u_blendFixB;
float4 u_texclamp;
float2 u_texclampoff;
)";
// 512 bytes. Would like to shrink more. Some colors only have 8-bit precision and we expand
// them to float unnecessarily, could just as well expand in the shader.
struct UB_VS_Lights {
@ -151,42 +123,6 @@ R"( vec4 u_ambient;
vec3 u_lightspecular3;
)";
// HLSL code is shared so these names are changed to match those in DX9.
static const char *cb_vs_lightsStr =
R"( float4 u_ambient;
float3 u_matdiffuse;
float4 u_matspecular;
float3 u_matemissive;
float3 u_lightpos0;
float3 u_lightpos1;
float3 u_lightpos2;
float3 u_lightpos3;
float3 u_lightdir0;
float3 u_lightdir1;
float3 u_lightdir2;
float3 u_lightdir3;
float3 u_lightatt0;
float3 u_lightatt1;
float3 u_lightatt2;
float3 u_lightatt3;
float4 u_lightangle_spotCoef0;
float4 u_lightangle_spotCoef1;
float4 u_lightangle_spotCoef2;
float4 u_lightangle_spotCoef3;
float3 u_lightambient0;
float3 u_lightambient1;
float3 u_lightambient2;
float3 u_lightambient3;
float3 u_lightdiffuse0;
float3 u_lightdiffuse1;
float3 u_lightdiffuse2;
float3 u_lightdiffuse3;
float3 u_lightspecular0;
float3 u_lightspecular1;
float3 u_lightspecular2;
float3 u_lightspecular3;
)";
// With some cleverness, we could get away with uploading just half this when only the four or five first
// bones are being used. This is 384b.
struct UB_VS_Bones {
@ -197,10 +133,6 @@ static const char *ub_vs_bonesStr =
R"( mat3x4 u_bone0; mat3x4 u_bone1; mat3x4 u_bone2; mat3x4 u_bone3; mat3x4 u_bone4; mat3x4 u_bone5; mat3x4 u_bone6; mat3x4 u_bone7; mat3x4 u_bone8;
)";
static const char *cb_vs_bonesStr =
R"( float4x3 u_bone[8];
)";
void CalcCullRange(float minValues[4], float maxValues[4], bool flipViewport, bool hasNegZ);
void BaseUpdateUniforms(UB_VS_FS_Base *ub, uint64_t dirtyUniforms, bool flipViewport, bool useBufferedRendering);

View File

@ -27,8 +27,7 @@
#include "GPU/Common/ShaderId.h"
#include "GPU/Common/ShaderUniforms.h"
#include "GPU/Common/VertexDecoderCommon.h"
#include "GPU/GLES/VertexShaderGeneratorGLES.h"
#include "GPU/GLES/ShaderManagerGLES.h"
#include "GPU/Common/VertexShaderGenerator.h"
#undef WRITE
@ -58,6 +57,30 @@ static const char * const boneWeightInDecl[9] = {
"in mediump vec4 w1, w2;\n",
};
const char *boneWeightAttrDeclHLSL[9] = {
"#ERROR boneWeightAttrDecl#\n",
"float a_w1:TEXCOORD1;\n",
"vec2 a_w1:TEXCOORD1;\n",
"vec3 a_w1:TEXCOORD1;\n",
"vec4 a_w1:TEXCOORD1;\n",
"vec4 a_w1:TEXCOORD1;\n float a_w2:TEXCOORD2;\n",
"vec4 a_w1:TEXCOORD1;\n vec2 a_w2:TEXCOORD2;\n",
"vec4 a_w1:TEXCOORD1;\n vec3 a_w2:TEXCOORD2;\n",
"vec4 a_w1:TEXCOORD1;\n vec4 a_w2:TEXCOORD2;\n",
};
const char *boneWeightAttrInitHLSL[9] = {
" #ERROR#\n",
" vec4 w1 = vec4(In.a_w1, 0.0, 0.0, 0.0);\n",
" vec4 w1 = vec4(In.a_w1.xy, 0.0, 0.0);\n",
" vec4 w1 = vec4(In.a_w1.xyz, 0.0);\n",
" vec4 w1 = In.a_w1;\n",
" vec4 w1 = In.a_w1;\n vec4 w2 = vec4(In.a_w2, 0.0, 0.0, 0.0);\n",
" vec4 w1 = In.a_w1;\n vec4 w2 = vec4(In.a_w2.xy, 0.0, 0.0);\n",
" vec4 w1 = In.a_w1;\n vec4 w2 = vec4(In.a_w2.xyz, 0.0);\n",
" vec4 w1 = In.a_w1;\n vec4 w2 = In.a_w2;\n",
};
// Depth range and viewport
//
// After the multiplication with the projection matrix, we have a 4D vector in clip space.
@ -103,9 +126,28 @@ const char *vulkan_glsl_preamble_vs =
"#version 450\n"
"#extension GL_ARB_separate_shader_objects : enable\n"
"#extension GL_ARB_shading_language_420pack : enable\n"
"#define splat3(x) vec3(x)\n\n";
"#define mul(x, y) ((x) * (y))\n"
"#define splat3(x) vec3(x)\n"
"#define lowp\n"
"#define mediump\n"
"#define highp\n"
"\n";
bool GenerateVertexShaderGLSL(const VShaderID &id, char *buffer, const ShaderLanguageDesc &compat, uint32_t *attrMask, uint64_t *uniformMask, std::string *errorString) {
const char *hlsl_preamble_vs =
"#define vec2 float2\n"
"#define vec3 float3\n"
"#define vec4 float4\n"
"#define ivec2 int2\n"
"#define ivec4 int4\n"
"#define mat4 float4x4\n"
"#define mat3x4 float4x3\n" // note how the conventions are backwards
"#define splat3(x) vec3(x, x, x)\n"
"#define lowp\n"
"#define mediump\n"
"#define highp\n"
"\n";
bool GenerateVertexShader(const VShaderID &id, char *buffer, const ShaderLanguageDesc &compat, uint32_t *attrMask, uint64_t *uniformMask, std::string *errorString) {
*attrMask = 0;
*uniformMask = 0;
@ -115,7 +157,9 @@ bool GenerateVertexShaderGLSL(const VShaderID &id, char *buffer, const ShaderLan
char *p = buffer;
if (compat.shaderLanguage == GLSL_VULKAN) {
WRITE(p, "%s", vulkan_glsl_preamble_vs);
} else {
} else if (compat.shaderLanguage == HLSL_D3D11 || compat.shaderLanguage == HLSL_D3D9) {
WRITE(p, "%s", hlsl_preamble_vs);
} else if (ShaderLanguageIsOpenGL(compat.shaderLanguage)) {
if (compat.gles) {
// PowerVR needs highp to do the fog in MHU correctly.
// Others don't, and some can't handle highp in the fragment shader.
@ -123,19 +167,18 @@ bool GenerateVertexShaderGLSL(const VShaderID &id, char *buffer, const ShaderLan
highpTexcoord = highpFog;
}
WRITE(p, "#version %d%s\n", compat.glslVersionNumber, compat.gles && compat.glslES30 ? " es" : "");
if (gl_extensions.EXT_gpu_shader4) {
WRITE(p, "#extension GL_EXT_gpu_shader4 : enable\n");
}
if (compat.gles) {
WRITE(p, "precision highp float;\n");
} else {
WRITE(p, "#define lowp\n");
WRITE(p, "#define mediump\n");
WRITE(p, "#define highp\n");
}
WRITE(p, "#define splat3(x) vec3(x)\n");
}
if (ShaderLanguageIsOpenGL(compat.shaderLanguage) && gl_extensions.EXT_gpu_shader4) {
WRITE(p, "#extension GL_EXT_gpu_shader4 : enable\n");
}
if (compat.gles) {
WRITE(p, "precision highp float;\n");
} else if (compat.shaderLanguage != GLSL_VULKAN) {
WRITE(p, "#define lowp\n");
WRITE(p, "#define mediump\n");
WRITE(p, "#define highp\n");
WRITE(p, "#define mul(x, y) ((x) * (y))\n");
}
bool isModeThrough = id.Bit(VS_BIT_IS_THROUGH);
@ -158,32 +201,29 @@ bool GenerateVertexShaderGLSL(const VShaderID &id, char *buffer, const ShaderLan
bool flipNormal = id.Bit(VS_BIT_NORM_REVERSE);
int ls0 = id.Bits(VS_BIT_LS0, 2);
int ls1 = id.Bits(VS_BIT_LS1, 2);
bool enableBones = id.Bit(VS_BIT_ENABLE_BONES);
bool enableBones = id.Bit(VS_BIT_ENABLE_BONES) && useHWTransform;
bool enableLighting = id.Bit(VS_BIT_LIGHTING_ENABLE);
int matUpdate = id.Bits(VS_BIT_MATERIAL_UPDATE, 3);
// Apparently we don't support bezier/spline together with bones.
bool doBezier = id.Bit(VS_BIT_BEZIER) && !enableBones && useHWTransform;
bool doSpline = id.Bit(VS_BIT_SPLINE) && !enableBones && useHWTransform;
if ((doBezier || doSpline) && !hasNormal) {
// Bad usage.
*errorString = "Invalid flags - tess requires normal.";
return false;
if (doBezier || doSpline) {
if (!hasNormal) {
// Bad usage.
*errorString = "Invalid flags - tess requires normal.";
return false;
}
if (compat.shaderLanguage == HLSL_D3D9 || compat.texelFetch == nullptr) {
*errorString = "Tess not supported on this shader language version";
return false;
}
}
bool hasColorTess = id.Bit(VS_BIT_HAS_COLOR_TESS);
bool hasTexcoordTess = id.Bit(VS_BIT_HAS_TEXCOORD_TESS);
bool hasNormalTess = id.Bit(VS_BIT_HAS_NORMAL_TESS);
bool flipNormalTess = id.Bit(VS_BIT_NORM_REVERSE_TESS);
if (compat.shaderLanguage == GLSL_VULKAN) {
WRITE(p, "\n");
WRITE(p, "layout (std140, set = 0, binding = 3) uniform baseVars {\n%s};\n", ub_baseStr);
if (enableLighting || doShadeMapping)
WRITE(p, "layout (std140, set = 0, binding = 4) uniform lightVars {\n%s};\n", ub_vs_lightsStr);
if (enableBones)
WRITE(p, "layout (std140, set = 0, binding = 5) uniform boneVars {\n%s};\n", ub_vs_bonesStr);
}
const char *shading = "";
if (compat.glslES30 || compat.shaderLanguage == GLSL_VULKAN)
shading = doFlatShading ? "flat " : "";
@ -202,11 +242,20 @@ bool GenerateVertexShaderGLSL(const VShaderID &id, char *buffer, const ShaderLan
int numBoneWeights = 0;
int boneWeightScale = id.Bits(VS_BIT_WEIGHT_FMTSCALE, 2);
bool texcoordInVec3 = false;
if (enableBones) {
numBoneWeights = 1 + id.Bits(VS_BIT_BONES, 3);
}
bool texCoordInVec3 = false;
if (compat.shaderLanguage == GLSL_VULKAN) {
WRITE(p, "\n");
WRITE(p, "layout (std140, set = 0, binding = 3) uniform baseVars {\n%s};\n", ub_baseStr);
if (enableLighting || doShadeMapping)
WRITE(p, "layout (std140, set = 0, binding = 4) uniform lightVars {\n%s};\n", ub_vs_lightsStr);
if (enableBones)
WRITE(p, "layout (std140, set = 0, binding = 5) uniform boneVars {\n%s};\n", ub_vs_bonesStr);
if (enableBones) {
numBoneWeights = 1 + id.Bits(VS_BIT_BONES, 3);
WRITE(p, "%s", boneWeightDecl[numBoneWeights]);
}
@ -222,7 +271,7 @@ bool GenerateVertexShaderGLSL(const VShaderID &id, char *buffer, const ShaderLan
if (doTexture && hasTexcoord) {
if (!useHWTransform && doTextureTransform && !isModeThrough) {
WRITE(p, "layout (location = %d) in vec3 texcoord;\n", (int)PspAttributeLocation::TEXCOORD);
texcoordInVec3 = true;
texCoordInVec3 = true;
} else
WRITE(p, "layout (location = %d) in vec2 texcoord;\n", (int)PspAttributeLocation::TEXCOORD);
}
@ -245,9 +294,148 @@ bool GenerateVertexShaderGLSL(const VShaderID &id, char *buffer, const ShaderLan
// See the fragment shader generator
WRITE(p, "layout (location = 3) out float v_fogdepth;\n");
}
} else if (compat.shaderLanguage == HLSL_D3D11 || compat.shaderLanguage == HLSL_D3D9) {
// Note: These two share some code after this hellishly large if/else.
if (compat.shaderLanguage == HLSL_D3D11) {
WRITE(p, "cbuffer base : register(b0) {\n%s};\n", ub_baseStr);
WRITE(p, "cbuffer lights: register(b1) {\n%s};\n", ub_vs_lightsStr);
WRITE(p, "cbuffer bones : register(b2) {\n%s};\n", ub_vs_bonesStr);
} else {
WRITE(p, "#pragma warning( disable : 3571 )\n");
if (isModeThrough) {
WRITE(p, "mat4 u_proj_through : register(c%i);\n", CONST_VS_PROJ_THROUGH);
} else {
WRITE(p, "mat4 u_proj : register(c%i);\n", CONST_VS_PROJ);
// Add all the uniforms we'll need to transform properly.
}
if (enableFog) {
WRITE(p, "vec2 u_fogcoef : register(c%i);\n", CONST_VS_FOGCOEF);
}
if (useHWTransform || !hasColor)
WRITE(p, "vec4 u_matambientalpha : register(c%i);\n", CONST_VS_MATAMBIENTALPHA); // matambient + matalpha
if (useHWTransform) {
// When transforming by hardware, we need a great deal more uniforms...
WRITE(p, "mat3x4 u_world : register(c%i);\n", CONST_VS_WORLD);
WRITE(p, "mat3x4 u_view : register(c%i);\n", CONST_VS_VIEW);
if (doTextureTransform)
WRITE(p, "mat3x4 u_texmtx : register(c%i);\n", CONST_VS_TEXMTX);
if (enableBones) {
#ifdef USE_BONE_ARRAY
WRITE(p, "mat3x4 u_bone[%i] : register(c%i);\n", numBones, CONST_VS_BONE0);
#else
for (int i = 0; i < numBoneWeights; i++) {
WRITE(p, "mat3x4 u_bone%i : register(c%i);\n", i, CONST_VS_BONE0 + i * 3);
}
#endif
}
if (doTexture) {
WRITE(p, "vec4 u_uvscaleoffset : register(c%i);\n", CONST_VS_UVSCALEOFFSET);
}
for (int i = 0; i < 4; i++) {
if (doLight[i] != LIGHT_OFF) {
// This is needed for shade mapping
WRITE(p, "vec3 u_lightpos%i : register(c%i);\n", i, CONST_VS_LIGHTPOS + i);
}
if (doLight[i] == LIGHT_FULL) {
GELightType type = static_cast<GELightType>(id.Bits(VS_BIT_LIGHT0_TYPE + 4 * i, 2));
GELightComputation comp = static_cast<GELightComputation>(id.Bits(VS_BIT_LIGHT0_COMP + 4 * i, 2));
if (type != GE_LIGHTTYPE_DIRECTIONAL)
WRITE(p, "vec3 u_lightatt%i : register(c%i);\n", i, CONST_VS_LIGHTATT + i);
if (type == GE_LIGHTTYPE_SPOT || type == GE_LIGHTTYPE_UNKNOWN) {
WRITE(p, "vec3 u_lightdir%i : register(c%i);\n", i, CONST_VS_LIGHTDIR + i);
WRITE(p, "vec4 u_lightangle_spotCoef%i : register(c%i);\n", i, CONST_VS_LIGHTANGLE_SPOTCOEF + i);
}
WRITE(p, "vec3 u_lightambient%i : register(c%i);\n", i, CONST_VS_LIGHTAMBIENT + i);
WRITE(p, "vec3 u_lightdiffuse%i : register(c%i);\n", i, CONST_VS_LIGHTDIFFUSE + i);
if (comp == GE_LIGHTCOMP_BOTH) {
WRITE(p, "vec3 u_lightspecular%i : register(c%i);\n", i, CONST_VS_LIGHTSPECULAR + i);
}
}
}
if (enableLighting) {
WRITE(p, "vec4 u_ambient : register(c%i);\n", CONST_VS_AMBIENT);
if ((matUpdate & 2) == 0 || !hasColor)
WRITE(p, "vec3 u_matdiffuse : register(c%i);\n", CONST_VS_MATDIFFUSE);
// if ((matUpdate & 4) == 0)
WRITE(p, "vec4 u_matspecular : register(c%i);\n", CONST_VS_MATSPECULAR); // Specular coef is contained in alpha
WRITE(p, "vec3 u_matemissive : register(c%i);\n", CONST_VS_MATEMISSIVE);
}
}
if (!isModeThrough && gstate_c.Supports(GPU_ROUND_DEPTH_TO_16BIT)) {
WRITE(p, "vec4 u_depthRange : register(c%i);\n", CONST_VS_DEPTHRANGE);
}
if (!isModeThrough) {
WRITE(p, "vec4 u_cullRangeMin : register(c%i);\n", CONST_VS_CULLRANGEMIN);
WRITE(p, "vec4 u_cullRangeMax : register(c%i);\n", CONST_VS_CULLRANGEMAX);
}
}
// And the "varyings".
if (useHWTransform) {
WRITE(p, "struct VS_IN { \n");
if ((doSpline || doBezier) && compat.shaderLanguage == HLSL_D3D11) {
WRITE(p, " uint instanceId : SV_InstanceID;\n");
}
if (enableBones) {
WRITE(p, " %s", boneWeightAttrDeclHLSL[numBoneWeights]);
}
if (doTexture && hasTexcoord) {
WRITE(p, " vec2 texcoord : TEXCOORD0;\n");
}
if (hasColor) {
WRITE(p, " vec4 color0 : COLOR0;\n");
}
if (hasNormal) {
WRITE(p, " vec3 normal : NORMAL;\n");
}
WRITE(p, " vec3 position : POSITION;\n");
WRITE(p, "};\n");
} else {
WRITE(p, "struct VS_IN {\n");
WRITE(p, " vec4 position : POSITION;\n");
if (doTexture && hasTexcoord) {
if (doTextureTransform && !isModeThrough) {
texCoordInVec3 = true;
WRITE(p, " vec3 texcoord : TEXCOORD0;\n");
} else
WRITE(p, " vec2 texcoord : TEXCOORD0;\n");
}
if (hasColor) {
WRITE(p, " vec4 color0 : COLOR0;\n");
}
// only software transform supplies color1 as vertex data
if (lmode) {
WRITE(p, " vec3 color1 : COLOR1;\n");
}
WRITE(p, "};\n");
}
WRITE(p, "struct VS_OUT {\n");
if (doTexture) {
WRITE(p, " vec3 v_texcoord : TEXCOORD0;\n");
}
const char *colorInterpolation = doFlatShading && compat.shaderLanguage == HLSL_D3D11 ? "nointerpolation " : "";
WRITE(p, " %svec4 v_color0 : COLOR0;\n", colorInterpolation);
if (lmode)
WRITE(p, " vec3 v_color1 : COLOR1;\n");
if (enableFog) {
WRITE(p, " float v_fogdepth: TEXCOORD1;\n");
}
if (compat.shaderLanguage == HLSL_D3D9) {
WRITE(p, " vec4 gl_Position : POSITION;\n");
} else {
WRITE(p, " vec4 gl_Position : SV_Position;\n");
}
WRITE(p, "};\n");
} else {
if (enableBones) {
numBoneWeights = 1 + id.Bits(VS_BIT_BONES, 3);
const char * const * boneWeightDecl = boneWeightAttrDecl;
if (!strcmp(compat.attribute, "in")) {
boneWeightDecl = boneWeightInDecl;
@ -272,7 +460,7 @@ bool GenerateVertexShaderGLSL(const VShaderID &id, char *buffer, const ShaderLan
if (doTexture && hasTexcoord) {
if (!useHWTransform && doTextureTransform && !isModeThrough) {
WRITE(p, "%s vec3 texcoord;\n", compat.attribute);
texcoordInVec3 = true;
texCoordInVec3 = true;
} else {
WRITE(p, "%s vec2 texcoord;\n", compat.attribute);
}
@ -413,35 +601,52 @@ bool GenerateVertexShaderGLSL(const VShaderID &id, char *buffer, const ShaderLan
if (compat.shaderLanguage == GLSL_VULKAN) {
WRITE(p, "struct TessData {\n");
WRITE(p, " vec4 pos;\n");
WRITE(p, " vec4 uv;\n");
WRITE(p, " vec4 color;\n");
WRITE(p, " vec4 tex;\n");
WRITE(p, " vec4 col;\n");
WRITE(p, "};\n");
WRITE(p, "layout (std430, set = 0, binding = 6) readonly buffer s_tess_data {\n");
WRITE(p, " TessData data[];\n");
WRITE(p, "} tess_data;\n");
WRITE(p, " TessData tess_data[];\n");
WRITE(p, "};\n");
WRITE(p, "struct TessWeight {\n");
WRITE(p, " vec4 basis;\n");
WRITE(p, " vec4 deriv;\n");
WRITE(p, "};\n");
WRITE(p, "layout (std430, set = 0, binding = 7) readonly buffer s_tess_weights_u {\n");
WRITE(p, " TessWeight data[];\n");
WRITE(p, "} tess_weights_u;\n");
WRITE(p, " TessWeight tess_weights_u[];\n");
WRITE(p, "};\n");
WRITE(p, "layout (std430, set = 0, binding = 8) readonly buffer s_tess_weights_v {\n");
WRITE(p, " TessWeight data[];\n");
WRITE(p, "} tess_weights_v;\n");
} else {
WRITE(p, " TessWeight tess_weights_v[];\n");
WRITE(p, "};\n");
} else if (ShaderLanguageIsOpenGL(compat.shaderLanguage)) {
WRITE(p, "uniform sampler2D u_tess_points;\n"); // Control Points
WRITE(p, "uniform sampler2D u_tess_weights_u;\n");
WRITE(p, "uniform sampler2D u_tess_weights_v;\n");
WRITE(p, "uniform int u_spline_counts;\n");
} else if (compat.shaderLanguage == HLSL_D3D11) {
WRITE(p, "struct TessData {\n");
WRITE(p, " vec3 pos; float pad1;\n");
WRITE(p, " vec2 tex; vec2 pad2;\n");
WRITE(p, " vec4 col;\n");
WRITE(p, "};\n");
WRITE(p, "StructuredBuffer<TessData> tess_data : register(t0);\n");
WRITE(p, "struct TessWeight {\n");
WRITE(p, " vec4 basis;\n");
WRITE(p, " vec4 deriv;\n");
WRITE(p, "};\n");
WRITE(p, "StructuredBuffer<TessWeight> tess_weights_u : register(t1);\n");
WRITE(p, "StructuredBuffer<TessWeight> tess_weights_v : register(t2);\n");
} else if (compat.shaderLanguage == HLSL_D3D9) {
}
const char *init[3] = { "0.0, 0.0", "0.0, 0.0, 0.0", "0.0, 0.0, 0.0, 0.0" };
for (int i = 2; i <= 4; i++) {
// Define 3 types vec2, vec3, vec4
WRITE(p, "vec%d tess_sample(in vec%d points[16], mat4 weights) {\n", i, i);
WRITE(p, " vec%d pos = vec%d(0.0);\n", i, i);
WRITE(p, " vec%d pos = vec%d(%s);\n", i, i, init[i - 2]);
for (int v = 0; v < 4; ++v) {
for (int u = 0; u < 4; ++u) {
WRITE(p, " pos += weights[%i][%i] * points[%i];\n", v, u, v * 4 + u);
@ -451,10 +656,14 @@ bool GenerateVertexShaderGLSL(const VShaderID &id, char *buffer, const ShaderLan
WRITE(p, "}\n");
}
if (compat.glslVersionNumber < 130) { // For glsl version 1.10
if (ShaderLanguageIsOpenGL(compat.shaderLanguage) && compat.glslVersionNumber < 130) { // For glsl version 1.10
WRITE(p, "mat4 outerProduct(vec4 u, vec4 v) {\n");
WRITE(p, " return mat4(u * v[0], u * v[1], u * v[2], u * v[3]);\n");
WRITE(p, "}\n");
} else if (compat.shaderLanguage == HLSL_D3D9 || compat.shaderLanguage == HLSL_D3D11) {
WRITE(p, "mat4 outerProduct(vec4 u, vec4 v) {\n");
WRITE(p, " return mul((float4x1)v, (float1x4)u);\n");
WRITE(p, "}\n");
}
WRITE(p, "struct Tess {\n");
@ -466,7 +675,13 @@ bool GenerateVertexShaderGLSL(const VShaderID &id, char *buffer, const ShaderLan
WRITE(p, " vec3 nrm;\n");
WRITE(p, "};\n");
WRITE(p, "void tessellate(out Tess tess) {\n");
if (compat.shaderLanguage == HLSL_D3D9 || compat.shaderLanguage == HLSL_D3D11) {
WRITE(p, "void tessellate(in VS_IN In, out Tess tess) {\n");
WRITE(p, " vec3 position = In.position;\n");
WRITE(p, " vec3 normal = In.normal;\n");
} else {
WRITE(p, "void tessellate(out Tess tess) {\n");
}
WRITE(p, " ivec2 point_pos = ivec2(position.z, normal.z)%s;\n", doBezier ? " * 3" : "");
WRITE(p, " ivec2 weight_idx = ivec2(position.xy);\n");
@ -479,17 +694,17 @@ bool GenerateVertexShaderGLSL(const VShaderID &id, char *buffer, const ShaderLan
for (int i = 0; i < 4; i++) {
for (int j = 0; j < 4; j++) {
WRITE(p, " index = (%i + point_pos.y) * int(u_spline_counts) + (%i + point_pos.x);\n", i, j);
WRITE(p, " _pos[%i] = tess_data.data[index].pos.xyz;\n", i * 4 + j);
WRITE(p, " _pos[%i] = tess_data[index].pos.xyz;\n", i * 4 + j);
if (doTexture && hasTexcoordTess)
WRITE(p, " _tex[%i] = tess_data.data[index].uv.xy;\n", i * 4 + j);
WRITE(p, " _tex[%i] = tess_data[index].tex.xy;\n", i * 4 + j);
if (hasColorTess)
WRITE(p, " _col[%i] = tess_data.data[index].color;\n", i * 4 + j);
WRITE(p, " _col[%i] = tess_data[index].col;\n", i * 4 + j);
}
}
// Basis polynomials as weight coefficients
WRITE(p, " vec4 basis_u = tess_weights_u.data[weight_idx.x].basis;\n");
WRITE(p, " vec4 basis_v = tess_weights_v.data[weight_idx.y].basis;\n");
WRITE(p, " vec4 basis_u = tess_weights_u[weight_idx.x].basis;\n");
WRITE(p, " vec4 basis_v = tess_weights_v[weight_idx.y].basis;\n");
WRITE(p, " mat4 basis = outerProduct(basis_u, basis_v);\n");
} else {
@ -527,8 +742,8 @@ bool GenerateVertexShaderGLSL(const VShaderID &id, char *buffer, const ShaderLan
if (hasNormalTess) {
if (compat.coefsFromBuffers) {
// Derivatives as weight coefficients
WRITE(p, " vec4 deriv_u = tess_weights_u.data[weight_idx.x].deriv;\n");
WRITE(p, " vec4 deriv_v = tess_weights_v.data[weight_idx.y].deriv;\n");
WRITE(p, " vec4 deriv_u = tess_weights_u[weight_idx.x].deriv;\n");
WRITE(p, " vec4 deriv_v = tess_weights_v[weight_idx.y].deriv;\n");
} else {
// Derivatives as weight coefficients
WRITE(p, " vec4 deriv_u = %s(u_tess_weights_u, %s, 0);\n", compat.texelFetch, "ivec2(weight_idx.x * 2 + 1, 0)");
@ -542,37 +757,66 @@ bool GenerateVertexShaderGLSL(const VShaderID &id, char *buffer, const ShaderLan
WRITE(p, "}\n");
}
WRITE(p, "void main() {\n");
if (ShaderLanguageIsOpenGL(compat.shaderLanguage) || compat.shaderLanguage == GLSL_VULKAN) {
WRITE(p, "void main() {\n");
} else if (compat.shaderLanguage == HLSL_D3D9 || compat.shaderLanguage == HLSL_D3D11) {
WRITE(p, "VS_OUT main(VS_IN In) {\n");
WRITE(p, " VS_OUT Out;\n");
if (doTexture && hasTexcoord) {
if (texCoordInVec3) {
WRITE(p, " vec3 texcoord = In.texcoord;\n");
} else {
WRITE(p, " vec2 texcoord = In.texcoord;\n");
}
}
if (hasColor) {
WRITE(p, " vec4 color0 = In.color0;\n");
if (lmode && !useHWTransform) {
WRITE(p, " vec3 color1 = In.color1;\n");
}
}
if (hasNormal) {
WRITE(p, " vec3 normal = In.normal;\n");
}
if (useHWTransform) {
WRITE(p, " vec3 position = In.position;\n");
} else {
WRITE(p, " vec4 position = In.position;\n");
}
if (enableBones) {
WRITE(p, "%s", boneWeightAttrInitHLSL[numBoneWeights]);
}
}
if (!useHWTransform) {
// Simple pass-through of vertex data to fragment shader
if (doTexture) {
if (texcoordInVec3) {
WRITE(p, " v_texcoord = texcoord;\n");
if (texCoordInVec3) {
WRITE(p, " %sv_texcoord = texcoord;\n", compat.vsOutPrefix);
} else {
WRITE(p, " v_texcoord = vec3(texcoord, 1.0);\n");
WRITE(p, " %sv_texcoord = vec3(texcoord, 1.0);\n", compat.vsOutPrefix);
}
}
if (hasColor) {
WRITE(p, " v_color0 = color0;\n");
WRITE(p, " %sv_color0 = color0;\n", compat.vsOutPrefix);
if (lmode)
WRITE(p, " v_color1 = color1;\n");
WRITE(p, " %sv_color1 = color1;\n", compat.vsOutPrefix);
} else {
WRITE(p, " v_color0 = u_matambientalpha;\n");
WRITE(p, " %sv_color0 = u_matambientalpha;\n", compat.vsOutPrefix);
if (lmode)
WRITE(p, " v_color1 = vec3(0.0);\n");
WRITE(p, " %sv_color1 = splat3(0.0);\n", compat.vsOutPrefix);
}
if (enableFog) {
WRITE(p, " v_fogdepth = position.w;\n");
WRITE(p, " %sv_fogdepth = position.w;\n", compat.vsOutPrefix);
}
if (isModeThrough) {
WRITE(p, " vec4 outPos = u_proj_through * vec4(position.xyz, 1.0);\n");
WRITE(p, " vec4 outPos = mul(u_proj_through, vec4(position.xyz, 1.0));\n");
} else {
// The viewport is used in this case, so need to compensate for that.
if (gstate_c.Supports(GPU_ROUND_DEPTH_TO_16BIT)) {
WRITE(p, " vec4 outPos = depthRoundZVP(u_proj * vec4(position.xyz, 1.0));\n");
WRITE(p, " vec4 outPos = depthRoundZVP(mul(u_proj, vec4(position.xyz, 1.0)));\n");
} else {
WRITE(p, " vec4 outPos = u_proj * vec4(position.xyz, 1.0);\n");
WRITE(p, " vec4 outPos = mul(u_proj, vec4(position.xyz, 1.0));\n");
}
}
} else {
@ -581,19 +825,23 @@ bool GenerateVertexShaderGLSL(const VShaderID &id, char *buffer, const ShaderLan
if (doBezier || doSpline) {
// Hardware tessellation
WRITE(p, " Tess tess;\n");
WRITE(p, " tessellate(tess);\n");
if (compat.shaderLanguage == HLSL_D3D9 || compat.shaderLanguage == HLSL_D3D11) {
WRITE(p, " tessellate(In, tess);\n");
} else {
WRITE(p, " tessellate(tess);\n");
}
WRITE(p, " vec3 worldpos = (vec4(tess.pos.xyz, 1.0) * u_world).xyz;\n");
WRITE(p, " vec3 worldpos = mul(vec4(tess.pos.xyz, 1.0), u_world).xyz;\n");
if (hasNormalTess) {
WRITE(p, " mediump vec3 worldnormal = normalize((vec4(%stess.nrm, 0.0) * u_world).xyz);\n", flipNormalTess ? "-" : "");
WRITE(p, " mediump vec3 worldnormal = normalize(mul(vec4(%stess.nrm, 0.0), u_world).xyz);\n", flipNormalTess ? "-" : "");
} else {
WRITE(p, " mediump vec3 worldnormal = vec3(0.0, 0.0, 1.0);\n");
}
} else {
// No skinning, just standard T&L.
WRITE(p, " vec3 worldpos = (vec4(position.xyz, 1.0) * u_world).xyz;\n");
WRITE(p, " vec3 worldpos = mul(vec4(position, 1.0), u_world).xyz;\n");
if (hasNormal)
WRITE(p, " mediump vec3 worldnormal = normalize((vec4(%snormal, 0.0) * u_world).xyz);\n", flipNormal ? "-" : "");
WRITE(p, " mediump vec3 worldnormal = normalize(mul(vec4(%snormal, 0.0), u_world).xyz);\n", flipNormal ? "-" : "");
else
WRITE(p, " mediump vec3 worldnormal = vec3(0.0, 0.0, 1.0);\n");
}
@ -626,24 +874,24 @@ bool GenerateVertexShaderGLSL(const VShaderID &id, char *buffer, const ShaderLan
WRITE(p, ";\n");
WRITE(p, " vec3 skinnedpos = (vec4(position, 1.0) * skinMatrix).xyz %s;\n", factor);
WRITE(p, " vec3 worldpos = (vec4(skinnedpos, 1.0) * u_world).xyz;\n");
WRITE(p, " vec3 skinnedpos = mul(vec4(position, 1.0), skinMatrix).xyz%s;\n", factor);
WRITE(p, " vec3 worldpos = mul(vec4(skinnedpos, 1.0), u_world).xyz;\n");
if (hasNormal) {
WRITE(p, " mediump vec3 skinnednormal = (vec4(%snormal, 0.0) * skinMatrix).xyz %s;\n", flipNormal ? "-" : "", factor);
WRITE(p, " mediump vec3 skinnednormal = mul(vec4(%snormal, 0.0), skinMatrix).xyz%s;\n", flipNormal ? "-" : "", factor);
} else {
WRITE(p, " mediump vec3 skinnednormal = (vec4(0.0, 0.0, %s1.0, 0.0) * skinMatrix).xyz %s;\n", flipNormal ? "-" : "", factor);
WRITE(p, " mediump vec3 skinnednormal = mul(vec4(0.0, 0.0, %s1.0, 0.0), skinMatrix).xyz%s;\n", flipNormal ? "-" : "", factor);
}
WRITE(p, " mediump vec3 worldnormal = normalize((vec4(skinnednormal, 0.0) * u_world).xyz);\n");
WRITE(p, " mediump vec3 worldnormal = normalize(mul(vec4(skinnednormal, 0.0), u_world).xyz);\n");
}
WRITE(p, " vec4 viewPos = vec4((vec4(worldpos, 1.0) * u_view).xyz, 1.0);\n");
WRITE(p, " vec4 viewPos = vec4(mul(vec4(worldpos, 1.0), u_view).xyz, 1.0);\n");
// Final view and projection transforms.
if (gstate_c.Supports(GPU_ROUND_DEPTH_TO_16BIT)) {
WRITE(p, " vec4 outPos = depthRoundZVP(u_proj * viewPos);\n");
WRITE(p, " vec4 outPos = depthRoundZVP(mul(u_proj, viewPos));\n");
} else {
WRITE(p, " vec4 outPos = u_proj * viewPos;\n");
WRITE(p, " vec4 outPos = mul(u_proj, viewPos);\n");
}
// TODO: Declare variables for dots for shade mapping if needed.
@ -680,7 +928,7 @@ bool GenerateVertexShaderGLSL(const VShaderID &id, char *buffer, const ShaderLan
}
if (!specularIsZero) {
WRITE(p, " lowp vec3 lightSum1 = vec3(0.0);\n");
WRITE(p, " lowp vec3 lightSum1 = splat3(0.0);\n");
}
if (!diffuseIsZero) {
WRITE(p, " vec3 toLight;\n");
@ -771,32 +1019,32 @@ bool GenerateVertexShaderGLSL(const VShaderID &id, char *buffer, const ShaderLan
if (enableLighting) {
// Sum up ambient, emissive here.
if (lmode) {
WRITE(p, " v_color0 = clamp(lightSum0, 0.0, 1.0);\n");
WRITE(p, " %sv_color0 = clamp(lightSum0, 0.0, 1.0);\n", compat.vsOutPrefix);
// v_color1 only exists when lmode = 1.
if (specularIsZero) {
WRITE(p, " v_color1 = vec3(0.0);\n");
WRITE(p, " %sv_color1 = splat3(0.0);\n", compat.vsOutPrefix);
} else {
WRITE(p, " v_color1 = clamp(lightSum1, 0.0, 1.0);\n");
WRITE(p, " %sv_color1 = clamp(lightSum1, 0.0, 1.0);\n", compat.vsOutPrefix);
}
} else {
if (specularIsZero) {
WRITE(p, " v_color0 = clamp(lightSum0, 0.0, 1.0);\n");
WRITE(p, " %sv_color0 = clamp(lightSum0, 0.0, 1.0);\n", compat.vsOutPrefix);
} else {
WRITE(p, " v_color0 = clamp(clamp(lightSum0, 0.0, 1.0) + vec4(lightSum1, 0.0), 0.0, 1.0);\n");
WRITE(p, " %sv_color0 = clamp(clamp(lightSum0, 0.0, 1.0) + vec4(lightSum1, 0.0), 0.0, 1.0);\n", compat.vsOutPrefix);
}
}
} else {
// Lighting doesn't affect color.
if (hasColor) {
if (doBezier || doSpline)
WRITE(p, " v_color0 = tess.col;\n");
WRITE(p, " %sv_color0 = tess.col;\n", compat.vsOutPrefix);
else
WRITE(p, " v_color0 = color0;\n");
WRITE(p, " %sv_color0 = color0;\n", compat.vsOutPrefix);
} else {
WRITE(p, " v_color0 = u_matambientalpha;\n");
WRITE(p, " %sv_color0 = u_matambientalpha;\n", compat.vsOutPrefix);
}
if (lmode)
WRITE(p, " v_color1 = vec3(0.0);\n");
WRITE(p, " %sv_color1 = splat3(0.0);\n", compat.vsOutPrefix);
}
bool scaleUV = !isModeThrough && (uvGenMode == GE_TEXMAP_TEXTURE_COORDS || uvGenMode == GE_TEXMAP_UNKNOWN);
@ -809,17 +1057,17 @@ bool GenerateVertexShaderGLSL(const VShaderID &id, char *buffer, const ShaderLan
if (scaleUV) {
if (hasTexcoord) {
if (doBezier || doSpline)
WRITE(p, " v_texcoord = vec3(tess.tex.xy * u_uvscaleoffset.xy + u_uvscaleoffset.zw, 0.0);\n");
WRITE(p, " %sv_texcoord = vec3(tess.tex.xy * u_uvscaleoffset.xy + u_uvscaleoffset.zw, 0.0);\n", compat.vsOutPrefix);
else
WRITE(p, " v_texcoord = vec3(texcoord.xy * u_uvscaleoffset.xy, 0.0);\n");
WRITE(p, " %sv_texcoord = vec3(texcoord.xy * u_uvscaleoffset.xy, 0.0);\n", compat.vsOutPrefix);
} else {
WRITE(p, " v_texcoord = vec3(0.0);\n");
WRITE(p, " %sv_texcoord = splat3(0.0);\n", compat.vsOutPrefix);
}
} else {
if (hasTexcoord) {
WRITE(p, " v_texcoord = vec3(texcoord.xy * u_uvscaleoffset.xy + u_uvscaleoffset.zw, 0.0);\n");
WRITE(p, " %sv_texcoord = vec3(texcoord.xy * u_uvscaleoffset.xy + u_uvscaleoffset.zw, 0.0);\n", compat.vsOutPrefix);
} else {
WRITE(p, " v_texcoord = vec3(u_uvscaleoffset.zw, 0.0);\n");
WRITE(p, " %sv_texcoord = vec3(u_uvscaleoffset.zw, 0.0);\n", compat.vsOutPrefix);
}
}
break;
@ -829,7 +1077,7 @@ bool GenerateVertexShaderGLSL(const VShaderID &id, char *buffer, const ShaderLan
std::string temp_tc;
switch (uvProjMode) {
case GE_PROJMAP_POSITION: // Use model space XYZ as source
temp_tc = "vec4(position.xyz, 1.0)";
temp_tc = "vec4(position, 1.0)";
break;
case GE_PROJMAP_UV: // Use unscaled UV as source
{
@ -855,7 +1103,7 @@ bool GenerateVertexShaderGLSL(const VShaderID &id, char *buffer, const ShaderLan
break;
}
// Transform by texture matrix. XYZ as we are doing projection mapping.
WRITE(p, " v_texcoord = (%s * u_texmtx).xyz * vec3(u_uvscaleoffset.xy, 1.0);\n", temp_tc.c_str());
WRITE(p, " %sv_texcoord = mul(%s, u_texmtx).xyz * vec3(u_uvscaleoffset.xy, 1.0);\n", compat.vsOutPrefix, temp_tc.c_str());
}
break;
@ -863,7 +1111,7 @@ bool GenerateVertexShaderGLSL(const VShaderID &id, char *buffer, const ShaderLan
{
std::string lightFactor0 = StringFromFormat("(length(u_lightpos%i) == 0.0 ? worldnormal.z : dot(normalize(u_lightpos%i), worldnormal))", ls0, ls0);
std::string lightFactor1 = StringFromFormat("(length(u_lightpos%i) == 0.0 ? worldnormal.z : dot(normalize(u_lightpos%i), worldnormal))", ls1, ls1);
WRITE(p, " v_texcoord = vec3(u_uvscaleoffset.xy * vec2(1.0 + %s, 1.0 + %s) * 0.5, 1.0);\n", lightFactor0.c_str(), lightFactor1.c_str());
WRITE(p, " %sv_texcoord = vec3(u_uvscaleoffset.xy * vec2(1.0 + %s, 1.0 + %s) * 0.5, 1.0);\n", compat.vsOutPrefix, lightFactor0.c_str(), lightFactor1.c_str());
}
break;
@ -875,7 +1123,7 @@ bool GenerateVertexShaderGLSL(const VShaderID &id, char *buffer, const ShaderLan
// Compute fogdepth
if (enableFog)
WRITE(p, " v_fogdepth = (viewPos.z + u_fogcoef.x) * u_fogcoef.y;\n");
WRITE(p, " %sv_fogdepth = (viewPos.z + u_fogcoef.x) * u_fogcoef.y;\n", compat.vsOutPrefix);
}
if (!isModeThrough && gstate_c.Supports(GPU_SUPPORTS_VS_RANGE_CULLING)) {
@ -885,15 +1133,19 @@ bool GenerateVertexShaderGLSL(const VShaderID &id, char *buffer, const ShaderLan
const char *outMin = "projPos.x < u_cullRangeMin.x || projPos.y < u_cullRangeMin.y || projPos.z < u_cullRangeMin.z";
const char *outMax = "projPos.x > u_cullRangeMax.x || projPos.y > u_cullRangeMax.y || projPos.z > u_cullRangeMax.z";
WRITE(p, " if (%s || %s) {\n", outMin, outMax);
WRITE(p, " outPos.xyzw = vec4(u_cullRangeMax.w);\n");
WRITE(p, " outPos.xyzw = u_cullRangeMax.wwww;\n");
WRITE(p, " }\n");
WRITE(p, " }\n");
}
WRITE(p, " gl_Position = outPos;\n");
if (compat.shaderLanguage == GLSL_VULKAN) {
WRITE(p, " gl_PointSize = 1.0;\n");
}
// We've named the output gl_Position in HLSL as well.
WRITE(p, " %sgl_Position = outPos;\n", compat.vsOutPrefix);
if (compat.shaderLanguage == GLSL_VULKAN) {
WRITE(p, " gl_PointSize = 1.0;\n");
}
if (compat.shaderLanguage == HLSL_D3D11 || compat.shaderLanguage == HLSL_D3D9) {
WRITE(p, " return Out;\n");
}
WRITE(p, "}\n");
return true;
}

View File

@ -17,11 +17,13 @@
#pragma once
#include <string>
#include "GPU/Common/ShaderID.h"
#include "Common/CommonTypes.h"
bool GenerateVertexShaderHLSL(const VShaderID &id, char *buffer, ShaderLanguage lang, std::string *errorString);
struct VShaderID;
bool GenerateVertexShader(const VShaderID &id, char *buffer, const ShaderLanguageDesc &compat, uint32_t *attrMask, uint64_t *uniformMask, std::string *errorString);
// D3D9 constants.
enum {
CONST_VS_PROJ = 0,
CONST_VS_PROJ_THROUGH = 4,

View File

@ -26,8 +26,13 @@ std::vector<uint8_t> CompileShaderToBytecodeD3D11(const char *code, size_t codeS
std::string errors;
if (errorMsgs) {
errors = std::string((const char *)errorMsgs->GetBufferPointer(), errorMsgs->GetBufferSize());
ERROR_LOG(G3D, "%s: %s", SUCCEEDED(result) ? "warnings" : "errors", errors.c_str());
if (SUCCEEDED(result)) {
WARN_LOG(G3D, "%s: %s", "warnings", errors.c_str());
} else {
ERROR_LOG(G3D, "%s: %s", "errors", errors.c_str());
}
OutputDebugStringA(LineNumberString(code).c_str());
OutputDebugStringA(errors.c_str());
errorMsgs->Release();
}
if (compiledCode) {

View File

@ -34,8 +34,8 @@
#include "GPU/Math3D.h"
#include "GPU/GPUState.h"
#include "GPU/ge_constants.h"
#include "GPU/Common/VertexShaderGenerator.h"
#include "GPU/D3D11/ShaderManagerD3D11.h"
#include "GPU/Directx9/VertexShaderGeneratorHLSL.h"
#include "GPU/D3D11/D3D11Util.h"
D3D11FragmentShader::D3D11FragmentShader(ID3D11Device *device, D3D_FEATURE_LEVEL featureLevel, FShaderID id, const char *code, bool useHWTransform)
@ -208,7 +208,9 @@ void ShaderManagerD3D11::GetShaders(int prim, u32 vertType, D3D11VertexShader **
if (vsIter == vsCache_.end()) {
// Vertex shader not in cache. Let's compile it.
std::string genErrorString;
GenerateVertexShaderHLSL(VSID, codeBuffer_, HLSL_D3D11, &genErrorString);
uint32_t attrMask;
uint64_t uniformMask;
GenerateVertexShader(VSID, codeBuffer_, compat_, &attrMask, &uniformMask, &genErrorString);
vs = new D3D11VertexShader(device_, featureLevel_, VSID, codeBuffer_, vertType, useHWTransform);
vsCache_[VSID] = vs;
} else {

View File

@ -580,7 +580,9 @@ VSShader *ShaderManagerDX9::ApplyShader(bool useHWTransform, bool useHWTessellat
if (vsIter == vsCache_.end()) {
// Vertex shader not in cache. Let's compile it.
std::string genErrorString;
if (GenerateVertexShaderHLSL(VSID, codeBuffer_, HLSL_D3D9, &genErrorString)) {
uint32_t attrMask;
uint64_t uniformMask;
if (GenerateVertexShader(VSID, codeBuffer_, compat_, &attrMask, &uniformMask, &genErrorString)) {
vs = new VSShader(device_, VSID, codeBuffer_, useHWTransform);
}
if (!vs || vs->Failed()) {
@ -603,7 +605,9 @@ VSShader *ShaderManagerDX9::ApplyShader(bool useHWTransform, bool useHWTessellat
// next time and we'll do this over and over...
// Can still work with software transform.
bool success = GenerateVertexShaderHLSL(VSID, codeBuffer_, HLSL_D3D9, &genErrorString);
uint32_t attrMask;
uint64_t uniformMask;
bool success = GenerateVertexShader(VSID, codeBuffer_, compat_, &attrMask, &uniformMask, &genErrorString);
_assert_(success);
vs = new VSShader(device_, VSID, codeBuffer_, false);
}

View File

@ -21,7 +21,7 @@
#include <cstdint>
#include "Common/Common.h"
#include "GPU/Directx9/VertexShaderGeneratorHLSL.h"
#include "GPU/Common/VertexShaderGenerator.h"
#include "GPU/Common/FragmentShaderGenerator.h"
#include "GPU/Common/ShaderCommon.h"
#include "GPU/Common/ShaderId.h"

View File

@ -1,717 +0,0 @@
// Copyright (c) 2012- PPSSPP 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 or later versions.
// 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 git repository and contact information can be found at
// https://github.com/hrydgard/ppsspp and http://www.ppsspp.org/.
#include <cstdio>
#include <locale.h>
#include "Common/StringUtils.h"
#include "GPU/ge_constants.h"
#include "GPU/GPUState.h"
#include "Core/Config.h"
#include "GPU/Directx9/VertexShaderGeneratorHLSL.h"
#include "GPU/Common/VertexDecoderCommon.h"
#include "GPU/Common/ShaderUniforms.h"
#undef WRITE
#define WRITE p+=sprintf
static const char * const boneWeightAttrDecl[9] = {
"#ERROR#",
"float a_w1:TEXCOORD1;\n",
"vec2 a_w1:TEXCOORD1;\n",
"vec3 a_w1:TEXCOORD1;\n",
"vec4 a_w1:TEXCOORD1;\n",
"vec4 a_w1:TEXCOORD1;\n float a_w2:TEXCOORD2;\n",
"vec4 a_w1:TEXCOORD1;\n vec2 a_w2:TEXCOORD2;\n",
"vec4 a_w1:TEXCOORD1;\n vec3 a_w2:TEXCOORD2;\n",
"vec4 a_w1:TEXCOORD1;\n vec4 a_w2:TEXCOORD2;\n",
};
bool GenerateVertexShaderHLSL(const VShaderID &id, char *buffer, ShaderLanguage lang, std::string *errorString) {
char *p = buffer;
bool isModeThrough = id.Bit(VS_BIT_IS_THROUGH);
bool lmode = id.Bit(VS_BIT_LMODE);
bool doTexture = id.Bit(VS_BIT_DO_TEXTURE);
GETexMapMode uvGenMode = static_cast<GETexMapMode>(id.Bits(VS_BIT_UVGEN_MODE, 2));
bool doTextureTransform = uvGenMode == GE_TEXMAP_TEXTURE_MATRIX;
// this is only valid for some settings of uvGenMode
GETexProjMapMode uvProjMode = static_cast<GETexProjMapMode>(id.Bits(VS_BIT_UVPROJ_MODE, 2));
bool doShadeMapping = uvGenMode == GE_TEXMAP_ENVIRONMENT_MAP;
bool doFlatShading = id.Bit(VS_BIT_FLATSHADE);
bool useHWTransform = id.Bit(VS_BIT_USE_HW_TRANSFORM);
bool hasColor = id.Bit(VS_BIT_HAS_COLOR) || !useHWTransform;
bool hasNormal = id.Bit(VS_BIT_HAS_NORMAL) && useHWTransform;
bool hasTexcoord = id.Bit(VS_BIT_HAS_TEXCOORD) || !useHWTransform;
bool enableFog = id.Bit(VS_BIT_ENABLE_FOG);
bool flipNormal = id.Bit(VS_BIT_NORM_REVERSE);
int ls0 = id.Bits(VS_BIT_LS0, 2);
int ls1 = id.Bits(VS_BIT_LS1, 2);
bool enableBones = id.Bit(VS_BIT_ENABLE_BONES);
bool enableLighting = id.Bit(VS_BIT_LIGHTING_ENABLE);
int matUpdate = id.Bits(VS_BIT_MATERIAL_UPDATE, 3);
bool doBezier = id.Bit(VS_BIT_BEZIER) && !enableBones && useHWTransform;
bool doSpline = id.Bit(VS_BIT_SPLINE) && !enableBones && useHWTransform;
if ((doBezier || doSpline) && !hasNormal) {
// Bad usage.
*errorString = "Invalid flags - tess requires normal.";
return false;
}
bool hasColorTess = id.Bit(VS_BIT_HAS_COLOR_TESS);
bool hasTexcoordTess = id.Bit(VS_BIT_HAS_TEXCOORD_TESS);
bool hasNormalTess = id.Bit(VS_BIT_HAS_NORMAL_TESS);
bool flipNormalTess = id.Bit(VS_BIT_NORM_REVERSE_TESS);
DoLightComputation doLight[4] = { LIGHT_OFF, LIGHT_OFF, LIGHT_OFF, LIGHT_OFF };
if (useHWTransform) {
int shadeLight0 = doShadeMapping ? ls0 : -1;
int shadeLight1 = doShadeMapping ? ls1 : -1;
for (int i = 0; i < 4; i++) {
if (i == shadeLight0 || i == shadeLight1)
doLight[i] = LIGHT_SHADE;
if (enableLighting && id.Bit(VS_BIT_LIGHT0_ENABLE + i))
doLight[i] = LIGHT_FULL;
}
}
int numBoneWeights = 0;
int boneWeightScale = id.Bits(VS_BIT_WEIGHT_FMTSCALE, 2);
if (enableBones) {
numBoneWeights = 1 + id.Bits(VS_BIT_BONES, 3);
}
// Output some compatibility defines
WRITE(p, "#define vec2 float2\n");
WRITE(p, "#define vec3 float3\n");
WRITE(p, "#define vec4 float4\n");
WRITE(p, "#define mat4 float4x4\n");
WRITE(p, "#define mat3x4 float4x3\n"); // note how the conventions are backwards
WRITE(p, "#define splat3(x) vec3(x, x, x)\n");
if (lang == HLSL_D3D9) {
WRITE(p, "#pragma warning( disable : 3571 )\n");
if (isModeThrough) {
WRITE(p, "float4x4 u_proj_through : register(c%i);\n", CONST_VS_PROJ_THROUGH);
} else {
WRITE(p, "float4x4 u_proj : register(c%i);\n", CONST_VS_PROJ);
// Add all the uniforms we'll need to transform properly.
}
if (enableFog) {
WRITE(p, "vec2 u_fogcoef : register(c%i);\n", CONST_VS_FOGCOEF);
}
if (useHWTransform || !hasColor)
WRITE(p, "vec4 u_matambientalpha : register(c%i);\n", CONST_VS_MATAMBIENTALPHA); // matambient + matalpha
if (useHWTransform) {
// When transforming by hardware, we need a great deal more uniforms...
WRITE(p, "mat3x4 u_world : register(c%i);\n", CONST_VS_WORLD);
WRITE(p, "mat3x4 u_view : register(c%i);\n", CONST_VS_VIEW);
if (doTextureTransform)
WRITE(p, "mat3x4 u_texmtx : register(c%i);\n", CONST_VS_TEXMTX);
if (enableBones) {
#ifdef USE_BONE_ARRAY
WRITE(p, "mat3x4 u_bone[%i] : register(c%i);\n", numBones, CONST_VS_BONE0);
#else
for (int i = 0; i < numBoneWeights; i++) {
WRITE(p, "mat3x4 u_bone%i : register(c%i);\n", i, CONST_VS_BONE0 + i * 3);
}
#endif
}
if (doTexture) {
WRITE(p, "vec4 u_uvscaleoffset : register(c%i);\n", CONST_VS_UVSCALEOFFSET);
}
for (int i = 0; i < 4; i++) {
if (doLight[i] != LIGHT_OFF) {
// This is needed for shade mapping
WRITE(p, "vec3 u_lightpos%i : register(c%i);\n", i, CONST_VS_LIGHTPOS + i);
}
if (doLight[i] == LIGHT_FULL) {
GELightType type = static_cast<GELightType>(id.Bits(VS_BIT_LIGHT0_TYPE + 4 * i, 2));
GELightComputation comp = static_cast<GELightComputation>(id.Bits(VS_BIT_LIGHT0_COMP + 4 * i, 2));
if (type != GE_LIGHTTYPE_DIRECTIONAL)
WRITE(p, "vec3 u_lightatt%i : register(c%i);\n", i, CONST_VS_LIGHTATT + i);
if (type == GE_LIGHTTYPE_SPOT || type == GE_LIGHTTYPE_UNKNOWN) {
WRITE(p, "vec3 u_lightdir%i : register(c%i);\n", i, CONST_VS_LIGHTDIR + i);
WRITE(p, "vec4 u_lightangle_spotCoef%i : register(c%i);\n", i, CONST_VS_LIGHTANGLE_SPOTCOEF + i);
}
WRITE(p, "vec3 u_lightambient%i : register(c%i);\n", i, CONST_VS_LIGHTAMBIENT + i);
WRITE(p, "vec3 u_lightdiffuse%i : register(c%i);\n", i, CONST_VS_LIGHTDIFFUSE + i);
if (comp == GE_LIGHTCOMP_BOTH) {
WRITE(p, "vec3 u_lightspecular%i : register(c%i);\n", i, CONST_VS_LIGHTSPECULAR + i);
}
}
}
if (enableLighting) {
WRITE(p, "vec4 u_ambient : register(c%i);\n", CONST_VS_AMBIENT);
if ((matUpdate & 2) == 0 || !hasColor)
WRITE(p, "vec3 u_matdiffuse : register(c%i);\n", CONST_VS_MATDIFFUSE);
// if ((matUpdate & 4) == 0)
WRITE(p, "vec4 u_matspecular : register(c%i);\n", CONST_VS_MATSPECULAR); // Specular coef is contained in alpha
WRITE(p, "vec3 u_matemissive : register(c%i);\n", CONST_VS_MATEMISSIVE);
}
}
if (!isModeThrough && gstate_c.Supports(GPU_ROUND_DEPTH_TO_16BIT)) {
WRITE(p, "vec4 u_depthRange : register(c%i);\n", CONST_VS_DEPTHRANGE);
}
if (!isModeThrough) {
WRITE(p, "vec4 u_cullRangeMin : register(c%i);\n", CONST_VS_CULLRANGEMIN);
WRITE(p, "vec4 u_cullRangeMax : register(c%i);\n", CONST_VS_CULLRANGEMAX);
}
} else {
WRITE(p, "cbuffer base : register(b0) {\n%s};\n", cb_baseStr);
WRITE(p, "cbuffer lights: register(b1) {\n%s};\n", cb_vs_lightsStr);
WRITE(p, "cbuffer bones : register(b2) {\n%s};\n", cb_vs_bonesStr);
}
bool scaleUV = !isModeThrough && (uvGenMode == GE_TEXMAP_TEXTURE_COORDS || uvGenMode == GE_TEXMAP_UNKNOWN);
// And the "varyings".
bool texCoordInVec3 = false;
if (useHWTransform) {
WRITE(p, "struct VS_IN { \n");
if ((doSpline || doBezier) && lang == HLSL_D3D11) {
WRITE(p, " uint instanceId : SV_InstanceID;\n");
}
if (enableBones) {
WRITE(p, " %s", boneWeightAttrDecl[numBoneWeights]);
}
if (doTexture && hasTexcoord) {
WRITE(p, " vec2 texcoord : TEXCOORD0;\n");
}
if (hasColor) {
WRITE(p, " vec4 color0 : COLOR0;\n");
}
if (hasNormal) {
WRITE(p, " vec3 normal : NORMAL;\n");
}
WRITE(p, " vec3 position : POSITION;\n");
WRITE(p, "};\n");
} else {
WRITE(p, "struct VS_IN {\n");
WRITE(p, " vec4 position : POSITION;\n");
if (doTexture && hasTexcoord) {
if (doTextureTransform && !isModeThrough) {
texCoordInVec3 = true;
WRITE(p, " vec3 texcoord : TEXCOORD0;\n");
}
else
WRITE(p, " vec2 texcoord : TEXCOORD0;\n");
}
if (hasColor) {
WRITE(p, " vec4 color0 : COLOR0;\n");
}
// only software transform supplies color1 as vertex data
if (lmode) {
WRITE(p, " vec4 color1 : COLOR1;\n");
}
WRITE(p, "};\n");
}
WRITE(p, "struct VS_OUT {\n");
if (doTexture) {
WRITE(p, " vec3 v_texcoord : TEXCOORD0;\n");
}
const char *colorInterpolation = doFlatShading && lang == HLSL_D3D11 ? "nointerpolation " : "";
WRITE(p, " %svec4 v_color0 : COLOR0;\n", colorInterpolation);
if (lmode)
WRITE(p, " vec3 v_color1 : COLOR1;\n");
if (enableFog) {
WRITE(p, " float v_fogdepth: TEXCOORD1;\n");
}
if (lang == HLSL_D3D9) {
WRITE(p, " vec4 gl_Position : POSITION;\n");
} else {
WRITE(p, " vec4 gl_Position : SV_Position;\n");
}
WRITE(p, "};\n");
// Confirmed: Through mode gets through exactly the same in GL and D3D in Phantasy Star: Text is 38023.0 in the test scene.
if (!isModeThrough && gstate_c.Supports(GPU_ROUND_DEPTH_TO_16BIT)) {
// Apply the projection and viewport to get the Z buffer value, floor to integer, undo the viewport and projection.
// The Z range in D3D is different but we compensate for that using parameters.
WRITE(p, "\nvec4 depthRoundZVP(vec4 v) {\n");
WRITE(p, " float z = v.z / v.w;\n");
WRITE(p, " z = (z * u_depthRange.x + u_depthRange.y);\n");
WRITE(p, " z = floor(z);\n");
WRITE(p, " z = (z - u_depthRange.z) * u_depthRange.w;\n");
WRITE(p, " return vec4(v.x, v.y, z * v.w, v.w);\n");
WRITE(p, "}\n\n");
}
// Hardware tessellation
if (doSpline || doBezier) {
if (lang == HLSL_D3D11) {
WRITE(p, "struct TessData {\n");
WRITE(p, " vec3 pos; float pad1;\n");
WRITE(p, " vec2 tex; vec2 pad2;\n");
WRITE(p, " vec4 col;\n");
WRITE(p, "};\n");
WRITE(p, "StructuredBuffer<TessData> tess_data : register(t0);\n");
WRITE(p, "struct TessWeight {\n");
WRITE(p, " vec4 basis;\n");
WRITE(p, " vec4 deriv;\n");
WRITE(p, "};\n");
WRITE(p, "StructuredBuffer<TessWeight> tess_weights_u : register(t1);\n");
WRITE(p, "StructuredBuffer<TessWeight> tess_weights_v : register(t2);\n");
}
const char *init[3] = { "0.0, 0.0", "0.0, 0.0, 0.0", "0.0, 0.0, 0.0, 0.0" };
for (int i = 2; i <= 4; i++) {
// Define 3 types vec2, vec3, vec4
WRITE(p, "float%d tess_sample(in float%d points[16], float4x4 weights) {\n", i, i);
WRITE(p, " float%d pos = float%d(%s);\n", i, i, init[i - 2]);
for (int v = 0; v < 4; ++v) {
for (int u = 0; u < 4; ++u) {
WRITE(p, " pos += weights[%i][%i] * points[%i];\n", v, u, v * 4 + u);
}
}
WRITE(p, " return pos;\n");
WRITE(p, "}\n");
}
WRITE(p, "float4x4 outerProduct(vec4 u, vec4 v) {\n");
WRITE(p, " return mul((float4x1)v, (float1x4)u);\n");
WRITE(p, "}\n");
WRITE(p, "struct Tess {\n");
WRITE(p, " vec3 pos;\n");
if (doTexture)
WRITE(p, " vec2 tex;\n");
WRITE(p, " vec4 col;\n");
if (hasNormalTess)
WRITE(p, " vec3 nrm;\n");
WRITE(p, "};\n");
WRITE(p, "void tessellate(in VS_IN In, out Tess tess) {\n");
WRITE(p, " int2 point_pos = int2(In.position.z, In.normal.z)%s;\n", doBezier ? " * 3" : "");
WRITE(p, " int2 weight_idx = int2(In.position.xy);\n");
// Load 4x4 control points
WRITE(p, " vec3 _pos[16];\n");
WRITE(p, " vec2 _tex[16];\n");
WRITE(p, " vec4 _col[16];\n");
WRITE(p, " int index;\n");
for (int i = 0; i < 4; i++) {
for (int j = 0; j < 4; j++) {
WRITE(p, " index = (%i + point_pos.y) * u_spline_counts + (%i + point_pos.x);\n", i, j);
WRITE(p, " _pos[%i] = tess_data[index].pos;\n", i * 4 + j);
if (doTexture && hasTexcoordTess)
WRITE(p, " _tex[%i] = tess_data[index].tex;\n", i * 4 + j);
if (hasColorTess)
WRITE(p, " _col[%i] = tess_data[index].col;\n", i * 4 + j);
}
}
// Basis polynomials as weight coefficients
WRITE(p, " vec4 basis_u = tess_weights_u[weight_idx.x].basis;\n");
WRITE(p, " vec4 basis_v = tess_weights_v[weight_idx.y].basis;\n");
WRITE(p, " float4x4 basis = outerProduct(basis_u, basis_v);\n");
// Tessellate
WRITE(p, " tess.pos = tess_sample(_pos, basis);\n");
if (doTexture) {
if (hasTexcoordTess)
WRITE(p, " tess.tex = tess_sample(_tex, basis);\n");
else
WRITE(p, " tess.tex = In.normal.xy;\n");
}
if (hasColorTess)
WRITE(p, " tess.col = tess_sample(_col, basis);\n");
else
WRITE(p, " tess.col = u_matambientalpha;\n");
if (hasNormalTess) {
// Derivatives as weight coefficients
WRITE(p, " vec4 deriv_u = tess_weights_u[weight_idx.x].deriv;\n");
WRITE(p, " vec4 deriv_v = tess_weights_v[weight_idx.y].deriv;\n");
WRITE(p, " vec3 du = tess_sample(_pos, outerProduct(deriv_u, basis_v));\n");
WRITE(p, " vec3 dv = tess_sample(_pos, outerProduct(basis_u, deriv_v));\n");
WRITE(p, " tess.nrm = normalize(cross(du, dv));\n");
}
WRITE(p, "}\n");
}
WRITE(p, "VS_OUT main(VS_IN In) {\n");
WRITE(p, " VS_OUT Out;\n");
if (!useHWTransform) {
// Simple pass-through of vertex data to fragment shader
if (doTexture) {
if (texCoordInVec3) {
WRITE(p, " Out.v_texcoord = In.texcoord;\n");
} else {
WRITE(p, " Out.v_texcoord = vec3(In.texcoord, 1.0);\n");
}
}
if (hasColor) {
WRITE(p, " Out.v_color0 = In.color0;\n");
if (lmode)
WRITE(p, " Out.v_color1 = In.color1.rgb;\n");
} else {
WRITE(p, " Out.v_color0 = In.u_matambientalpha;\n");
if (lmode)
WRITE(p, " Out.v_color1 = vec3(0.0);\n");
}
if (enableFog) {
WRITE(p, " Out.v_fogdepth = In.position.w;\n");
}
if (isModeThrough) {
WRITE(p, " float4 outPos = mul(u_proj_through, vec4(In.position.xyz, 1.0));\n");
} else {
if (gstate_c.Supports(GPU_ROUND_DEPTH_TO_16BIT)) {
WRITE(p, " vec4 outPos = depthRoundZVP(mul(u_proj, vec4(In.position.xyz, 1.0)));\n");
} else {
WRITE(p, " vec4 outPos = mul(u_proj, vec4(In.position.xyz, 1.0));\n");
}
}
} else {
// Step 1: World Transform / Skinning
if (!enableBones) {
if (doSpline || doBezier) {
// Hardware tessellation
WRITE(p, " Tess tess;\n");
WRITE(p, " tessellate(In, tess);\n");
WRITE(p, " vec3 worldpos = mul(vec4(tess.pos.xyz, 1.0), u_world);\n");
if (hasNormalTess)
WRITE(p, " vec3 worldnormal = normalize(mul(vec4(%stess.nrm, 0.0), u_world));\n", flipNormalTess ? "-" : "");
else
WRITE(p, " vec3 worldnormal = vec3(0.0, 0.0, 1.0);\n");
} else {
// No skinning, just standard T&L.
WRITE(p, " vec3 worldpos = mul(vec4(In.position.xyz, 1.0), u_world);\n");
if (hasNormal)
WRITE(p, " vec3 worldnormal = normalize(mul(vec4(%sIn.normal, 0.0), u_world));\n", flipNormal ? "-" : "");
else
WRITE(p, " vec3 worldnormal = vec3(0.0, 0.0, 1.0);\n");
}
} else {
static const char * const boneWeightAttr[8] = {
"a_w1.x", "a_w1.y", "a_w1.z", "a_w1.w",
"a_w2.x", "a_w2.y", "a_w2.z", "a_w2.w",
};
if (lang == HLSL_D3D11) {
if (numBoneWeights == 1)
WRITE(p, " mat3x4 skinMatrix = mul(In.a_w1, u_bone[0])");
else
WRITE(p, " float4x3 skinMatrix = mul(In.a_w1.x, u_bone[0])");
for (int i = 1; i < numBoneWeights; i++) {
const char *weightAttr = boneWeightAttr[i];
// workaround for "cant do .x of scalar" issue
if (numBoneWeights == 1 && i == 0) weightAttr = "a_w1";
if (numBoneWeights == 5 && i == 4) weightAttr = "a_w2";
WRITE(p, " + mul(In.%s, u_bone[%i])", weightAttr, i);
}
} else {
if (numBoneWeights == 1)
WRITE(p, " float4x3 skinMatrix = mul(In.a_w1, u_bone0)");
else
WRITE(p, " float4x3 skinMatrix = mul(In.a_w1.x, u_bone0)");
for (int i = 1; i < numBoneWeights; i++) {
const char *weightAttr = boneWeightAttr[i];
// workaround for "cant do .x of scalar" issue
if (numBoneWeights == 1 && i == 0) weightAttr = "a_w1";
if (numBoneWeights == 5 && i == 4) weightAttr = "a_w2";
WRITE(p, " + mul(In.%s, u_bone%i)", weightAttr, i);
}
}
WRITE(p, ";\n");
// Trying to simplify this results in bugs in LBP...
WRITE(p, " vec3 skinnedpos = mul(float4(In.position.xyz, 1.0), skinMatrix);\n");
WRITE(p, " vec3 worldpos = mul(float4(skinnedpos, 1.0), u_world);\n");
if (hasNormal) {
WRITE(p, " vec3 skinnednormal = mul(float4(%sIn.normal, 0.0), skinMatrix);\n", flipNormal ? "-" : "");
} else {
WRITE(p, " vec3 skinnednormal = mul(float4(0.0, 0.0, %s1.0, 0.0), skinMatrix);\n", flipNormal ? "-" : "");
}
WRITE(p, " vec3 worldnormal = normalize(mul(float4(skinnednormal, 0.0), u_world));\n");
}
WRITE(p, " vec4 viewPos = vec4(mul(vec4(worldpos, 1.0), u_view), 1.0);\n");
// Final view and projection transforms.
if (gstate_c.Supports(GPU_ROUND_DEPTH_TO_16BIT)) {
WRITE(p, " vec4 outPos = depthRoundZVP(mul(u_proj, viewPos));\n");
} else {
WRITE(p, " vec4 outPos = mul(u_proj, viewPos);\n");
}
// TODO: Declare variables for dots for shade mapping if needed.
const char *ambientStr = (matUpdate & 1) && hasColor ? "In.color0" : "u_matambientalpha";
const char *diffuseStr = (matUpdate & 2) && hasColor ? "In.color0.rgb" : "u_matdiffuse";
const char *specularStr = (matUpdate & 4) && hasColor ? "In.color0.rgb" : "u_matspecular.rgb";
if (doBezier || doSpline) {
// TODO: Probably, should use hasColorTess but FF4 has a problem with drawing the background.
ambientStr = (matUpdate & 1) && hasColor ? "tess.col" : "u_matambientalpha";
diffuseStr = (matUpdate & 2) && hasColor ? "tess.col.rgb" : "u_matdiffuse";
specularStr = (matUpdate & 4) && hasColor ? "tess.col.rgb" : "u_matspecular.rgb";
}
bool diffuseIsZero = true;
bool specularIsZero = true;
bool distanceNeeded = false;
bool anySpots = false;
if (enableLighting) {
WRITE(p, " vec4 lightSum0 = u_ambient * %s + float4(u_matemissive, 0.0);\n", ambientStr);
for (int i = 0; i < 4; i++) {
GELightType type = static_cast<GELightType>(id.Bits(VS_BIT_LIGHT0_TYPE + 4 * i, 2));
GELightComputation comp = static_cast<GELightComputation>(id.Bits(VS_BIT_LIGHT0_COMP + 4 * i, 2));
if (doLight[i] != LIGHT_FULL)
continue;
diffuseIsZero = false;
if (comp == GE_LIGHTCOMP_BOTH)
specularIsZero = false;
if (type != GE_LIGHTTYPE_DIRECTIONAL)
distanceNeeded = true;
if (type == GE_LIGHTTYPE_SPOT || type == GE_LIGHTTYPE_UNKNOWN)
anySpots = true;
}
if (!specularIsZero) {
WRITE(p, " vec3 lightSum1 = 0;\n");
}
if (!diffuseIsZero) {
WRITE(p, " vec3 toLight;\n");
WRITE(p, " vec3 diffuse;\n");
}
if (distanceNeeded) {
WRITE(p, " float distance;\n");
WRITE(p, " float lightScale;\n");
}
WRITE(p, " float ldot;\n");
if (anySpots) {
WRITE(p, " float angle;\n");
}
}
// Calculate lights if needed. If shade mapping is enabled, lights may need to be
// at least partially calculated.
for (int i = 0; i < 4; i++) {
if (doLight[i] != LIGHT_FULL)
continue;
GELightType type = static_cast<GELightType>(id.Bits(VS_BIT_LIGHT0_TYPE + 4 * i, 2));
GELightComputation comp = static_cast<GELightComputation>(id.Bits(VS_BIT_LIGHT0_COMP + 4 * i, 2));
if (type == GE_LIGHTTYPE_DIRECTIONAL) {
// We prenormalize light positions for directional lights.
WRITE(p, " toLight = u_lightpos%i;\n", i);
} else {
WRITE(p, " toLight = u_lightpos%i - worldpos;\n", i);
WRITE(p, " distance = length(toLight);\n");
WRITE(p, " toLight /= distance;\n");
}
bool doSpecular = comp == GE_LIGHTCOMP_BOTH;
bool poweredDiffuse = comp == GE_LIGHTCOMP_ONLYPOWDIFFUSE;
WRITE(p, " ldot = dot(toLight, worldnormal);\n");
if (poweredDiffuse) {
// pow(0.0, 0.0) may be undefined, but the PSP seems to treat it as 1.0.
// Seen in Tales of the World: Radiant Mythology (#2424.)
WRITE(p, " if (u_matspecular.a <= 0.0) {\n");
WRITE(p, " ldot = 1.0;\n");
WRITE(p, " } else {\n");
WRITE(p, " ldot = pow(max(ldot, 0.0), u_matspecular.a);\n");
WRITE(p, " }\n");
}
const char *timesLightScale = " * lightScale";
// Attenuation
switch (type) {
case GE_LIGHTTYPE_DIRECTIONAL:
timesLightScale = "";
break;
case GE_LIGHTTYPE_POINT:
WRITE(p, " lightScale = clamp(1.0 / dot(u_lightatt%i, vec3(1.0, distance, distance*distance)), 0.0, 1.0);\n", i);
break;
case GE_LIGHTTYPE_SPOT:
case GE_LIGHTTYPE_UNKNOWN:
WRITE(p, " angle = length(u_lightdir%i) == 0.0 ? 0.0 : dot(normalize(u_lightdir%i), toLight);\n", i, i);
WRITE(p, " if (angle >= u_lightangle_spotCoef%i.x) {\n", i);
WRITE(p, " lightScale = clamp(1.0 / dot(u_lightatt%i, vec3(1.0, distance, distance*distance)), 0.0, 1.0) * (u_lightangle_spotCoef%i.y <= 0.0 ? 1.0 : pow(angle, u_lightangle_spotCoef%i.y));\n", i, i, i);
WRITE(p, " } else {\n");
WRITE(p, " lightScale = 0.0;\n");
WRITE(p, " }\n");
break;
default:
// ILLEGAL
break;
}
WRITE(p, " diffuse = (u_lightdiffuse%i * %s) * max(ldot, 0.0);\n", i, diffuseStr);
if (doSpecular) {
WRITE(p, " if (ldot >= 0.0) {\n");
WRITE(p, " ldot = dot(normalize(toLight + vec3(0.0, 0.0, 1.0)), worldnormal);\n");
WRITE(p, " if (u_matspecular.a <= 0.0) {\n");
WRITE(p, " ldot = 1.0;\n");
WRITE(p, " } else {\n");
WRITE(p, " ldot = pow(max(ldot, 0.0), u_matspecular.a);\n");
WRITE(p, " }\n");
WRITE(p, " if (ldot > 0.0)\n");
WRITE(p, " lightSum1 += u_lightspecular%i * %s * ldot %s;\n", i, specularStr, timesLightScale);
WRITE(p, " }\n");
}
WRITE(p, " lightSum0.rgb += (u_lightambient%i * %s.rgb + diffuse)%s;\n", i, ambientStr, timesLightScale);
}
if (enableLighting) {
// Sum up ambient, emissive here.
if (lmode) {
WRITE(p, " Out.v_color0 = clamp(lightSum0, 0.0, 1.0);\n");
// v_color1 only exists when lmode = 1.
if (specularIsZero) {
WRITE(p, " Out.v_color1 = splat3(0.0);\n");
} else {
WRITE(p, " Out.v_color1 = clamp(lightSum1, 0.0, 1.0);\n");
}
} else {
if (specularIsZero) {
WRITE(p, " Out.v_color0 = clamp(lightSum0, 0.0, 1.0);\n");
} else {
WRITE(p, " Out.v_color0 = clamp(clamp(lightSum0, 0.0, 1.0) + float4(lightSum1, 0.0), 0.0, 1.0);\n");
}
}
} else {
// Lighting doesn't affect color.
if (hasColor) {
if (doBezier || doSpline)
WRITE(p, " Out.v_color0 = tess.col;\n");
else
WRITE(p, " Out.v_color0 = In.color0;\n");
} else {
WRITE(p, " Out.v_color0 = u_matambientalpha;\n");
}
if (lmode)
WRITE(p, " Out.v_color1 = splat3(0.0);\n");
}
// Step 3: UV generation
if (doTexture) {
switch (uvGenMode) {
case GE_TEXMAP_TEXTURE_COORDS: // Scale-offset. Easy.
case GE_TEXMAP_UNKNOWN: // Not sure what this is, but Riviera uses it. Treating as coords works.
if (scaleUV) {
if (hasTexcoord) {
if (doBezier || doSpline)
WRITE(p, " Out.v_texcoord = vec3(tess.tex.xy * u_uvscaleoffset.xy + u_uvscaleoffset.zw, 0.0);\n");
else
WRITE(p, " Out.v_texcoord = vec3(In.texcoord.xy * u_uvscaleoffset.xy, 0.0);\n");
} else {
WRITE(p, " Out.v_texcoord = splat3(0.0);\n");
}
} else {
if (hasTexcoord) {
WRITE(p, " Out.v_texcoord = vec3(In.texcoord.xy * u_uvscaleoffset.xy + u_uvscaleoffset.zw, 0.0);\n");
} else {
WRITE(p, " Out.v_texcoord = vec3(u_uvscaleoffset.zw, 0.0);\n");
}
}
break;
case GE_TEXMAP_TEXTURE_MATRIX: // Projection mapping.
{
std::string temp_tc;
switch (uvProjMode) {
case GE_PROJMAP_POSITION: // Use model space XYZ as source
temp_tc = "vec4(In.position.xyz, 1.0)";
break;
case GE_PROJMAP_UV: // Use unscaled UV as source
{
if (hasTexcoord) {
temp_tc = StringFromFormat("vec4(In.texcoord.xy, 0.0, 1.0)");
} else {
temp_tc = "vec4(0.0, 0.0, 0.0, 1.0)";
}
}
break;
case GE_PROJMAP_NORMALIZED_NORMAL: // Use normalized transformed normal as source
if (hasNormal)
temp_tc = flipNormal ? "vec4(normalize(-In.normal), 1.0)" : "vec4(normalize(In.normal), 1.0)";
else
temp_tc = "vec4(0.0, 0.0, 1.0, 1.0)";
break;
case GE_PROJMAP_NORMAL: // Use non-normalized transformed normal as source
if (hasNormal)
temp_tc = flipNormal ? "vec4(-In.normal, 1.0)" : "vec4(In.normal, 1.0)";
else
temp_tc = "vec4(0.0, 0.0, 1.0, 1.0)";
break;
}
// Transform by texture matrix. XYZ as we are doing projection mapping.
WRITE(p, " Out.v_texcoord.xyz = mul(%s, u_texmtx) * vec3(u_uvscaleoffset.xy, 1.0);\n", temp_tc.c_str());
}
break;
case GE_TEXMAP_ENVIRONMENT_MAP: // Shade mapping - use dots from light sources.
{
std::string lightFactor0 = StringFromFormat("(length(u_lightpos%i) == 0.0 ? worldnormal.z : dot(normalize(u_lightpos%i), worldnormal))", ls0, ls0);
std::string lightFactor1 = StringFromFormat("(length(u_lightpos%i) == 0.0 ? worldnormal.z : dot(normalize(u_lightpos%i), worldnormal))", ls1, ls1);
WRITE(p, " Out.v_texcoord = vec3(u_uvscaleoffset.xy * vec2(1.0 + %s, 1.0 + %s) * 0.5, 1.0);\n", lightFactor0.c_str(), lightFactor1.c_str());
}
break;
default:
// Should be unreachable.
_assert_(false);
return false;
}
}
// Compute fogdepth
if (enableFog) {
WRITE(p, " Out.v_fogdepth = (viewPos.z + u_fogcoef.x) * u_fogcoef.y;\n");
}
}
if (!isModeThrough && gstate_c.Supports(GPU_SUPPORTS_VS_RANGE_CULLING)) {
WRITE(p, " vec3 projPos = outPos.xyz / outPos.w;\n");
// Vertex range culling doesn't happen when depth is clamped, so only do this if in range.
WRITE(p, " if (u_cullRangeMin.w <= 0.0 || (projPos.z >= u_cullRangeMin.z && projPos.z <= u_cullRangeMax.z)) {\n");
const char *outMin = "projPos.x < u_cullRangeMin.x || projPos.y < u_cullRangeMin.y || projPos.z < u_cullRangeMin.z";
const char *outMax = "projPos.x > u_cullRangeMax.x || projPos.y > u_cullRangeMax.y || projPos.z > u_cullRangeMax.z";
WRITE(p, " if (%s || %s) {\n", outMin, outMax);
WRITE(p, " outPos.w = u_cullRangeMax.w;\n");
WRITE(p, " }\n");
WRITE(p, " }\n");
}
WRITE(p, " Out.gl_Position = outPos;\n");
WRITE(p, " return Out;\n");
WRITE(p, "}\n");
return true;
}

View File

@ -721,7 +721,7 @@ Shader *ShaderManagerGLES::CompileVertexShader(VShaderID VSID) {
uint32_t attrMask;
uint64_t uniformMask;
std::string errorString;
if (!GenerateVertexShaderGLSL(VSID, codeBuffer_, compat_, &attrMask, &uniformMask, &errorString)) {
if (!GenerateVertexShader(VSID, codeBuffer_, compat_, &attrMask, &uniformMask, &errorString)) {
ERROR_LOG(G3D, "Shader gen error: %s", errorString.c_str());
return nullptr;
}

View File

@ -23,24 +23,11 @@
#include "Common/GPU/OpenGL/GLRenderManager.h"
#include "GPU/Common/ShaderCommon.h"
#include "GPU/Common/ShaderId.h"
#include "GPU/GLES/VertexShaderGeneratorGLES.h"
#include "GPU/Common/VertexShaderGenerator.h"
#include "GPU/Common/FragmentShaderGenerator.h"
class Shader;
// Pre-fetched attrs and uniforms
enum {
ATTR_POSITION = 0,
ATTR_TEXCOORD = 1,
ATTR_NORMAL = 2,
ATTR_W1 = 3,
ATTR_W2 = 4,
ATTR_COLOR0 = 5,
ATTR_COLOR1 = 6,
ATTR_COUNT,
};
class LinkedShader {
public:
LinkedShader(GLRenderManager *render, VShaderID VSID, Shader *vs, FShaderID FSID, Shader *fs, bool useHWTransform, bool preloading = false);

View File

@ -1,24 +0,0 @@
// Copyright (c) 2012- PPSSPP 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 or later versions.
// 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 git repository and contact information can be found at
// https://github.com/hrydgard/ppsspp and http://www.ppsspp.org/.
#pragma once
#include "Common/CommonTypes.h"
struct VShaderID;
bool GenerateVertexShaderGLSL(const VShaderID &id, char *buffer, const ShaderLanguageDesc &compat, uint32_t *attrMask, uint64_t *uniformMask, std::string *errorString);

View File

@ -368,6 +368,7 @@
<ClInclude Include="Common\TextureScalerCommon.h" />
<ClInclude Include="Common\TransformCommon.h" />
<ClInclude Include="Common\VertexDecoderCommon.h" />
<ClInclude Include="Common\VertexShaderGenerator.h" />
<ClInclude Include="D3D11\D3D11Util.h" />
<ClInclude Include="D3D11\DepalettizeShaderD3D11.h" />
<ClInclude Include="D3D11\DrawEngineD3D11.h" />
@ -390,7 +391,6 @@
<ClInclude Include="Directx9\TextureCacheDX9.h" />
<ClInclude Include="Directx9\TextureScalerDX9.h" />
<ClInclude Include="Directx9\DrawEngineDX9.h" />
<ClInclude Include="Directx9\VertexShaderGeneratorHLSL.h" />
<ClInclude Include="ge_constants.h" />
<ClInclude Include="GeDisasm.h" />
<ClInclude Include="GLES\DepalettizeShaderGLES.h">
@ -447,12 +447,6 @@
<ExcludedFromBuild Condition="'$(Configuration)|$(Platform)'=='Release|ARM64'">true</ExcludedFromBuild>
<ExcludedFromBuild Condition="'$(Configuration)|$(Platform)'=='Release|ARM'">true</ExcludedFromBuild>
</ClInclude>
<ClInclude Include="GLES\VertexShaderGeneratorGLES.h">
<ExcludedFromBuild Condition="'$(Configuration)|$(Platform)'=='Debug|ARM64'">true</ExcludedFromBuild>
<ExcludedFromBuild Condition="'$(Configuration)|$(Platform)'=='Debug|ARM'">true</ExcludedFromBuild>
<ExcludedFromBuild Condition="'$(Configuration)|$(Platform)'=='Release|ARM64'">true</ExcludedFromBuild>
<ExcludedFromBuild Condition="'$(Configuration)|$(Platform)'=='Release|ARM'">true</ExcludedFromBuild>
</ClInclude>
<ClInclude Include="GPU.h" />
<ClInclude Include="GPUCommon.h" />
<ClInclude Include="GPUInterface.h" />
@ -537,6 +531,7 @@
<ExcludedFromBuild Condition="'$(Configuration)|$(Platform)'=='Release|ARM64'">true</ExcludedFromBuild>
<ExcludedFromBuild Condition="'$(Configuration)|$(Platform)'=='Release|ARM'">true</ExcludedFromBuild>
</ClCompile>
<ClCompile Include="Common\VertexShaderGenerator.cpp" />
<ClCompile Include="D3D11\D3D11Util.cpp" />
<ClCompile Include="D3D11\DepalettizeShaderD3D11.cpp" />
<ClCompile Include="D3D11\DrawEngineD3D11.cpp" />
@ -561,7 +556,6 @@
<ClCompile Include="Directx9\TextureCacheDX9.cpp" />
<ClCompile Include="Directx9\TextureScalerDX9.cpp" />
<ClCompile Include="Directx9\DrawEngineDX9.cpp" />
<ClCompile Include="Directx9\VertexShaderGeneratorHLSL.cpp" />
<ClCompile Include="GeDisasm.cpp" />
<ClCompile Include="GeConstants.cpp" />
<ClCompile Include="GLES\DepalettizeShaderGLES.cpp">
@ -630,12 +624,6 @@
<ExcludedFromBuild Condition="'$(Configuration)|$(Platform)'=='Release|ARM64'">true</ExcludedFromBuild>
<ExcludedFromBuild Condition="'$(Configuration)|$(Platform)'=='Release|ARM'">true</ExcludedFromBuild>
</ClCompile>
<ClCompile Include="GLES\VertexShaderGeneratorGLES.cpp">
<ExcludedFromBuild Condition="'$(Configuration)|$(Platform)'=='Debug|ARM64'">true</ExcludedFromBuild>
<ExcludedFromBuild Condition="'$(Configuration)|$(Platform)'=='Debug|ARM'">true</ExcludedFromBuild>
<ExcludedFromBuild Condition="'$(Configuration)|$(Platform)'=='Release|ARM64'">true</ExcludedFromBuild>
<ExcludedFromBuild Condition="'$(Configuration)|$(Platform)'=='Release|ARM'">true</ExcludedFromBuild>
</ClCompile>
<ClCompile Include="GPU.cpp" />
<ClCompile Include="GPUCommon.cpp" />
<ClCompile Include="GPUState.cpp" />

View File

@ -189,9 +189,6 @@
<ClInclude Include="GLES\DepalettizeShaderGLES.h">
<Filter>GLES</Filter>
</ClInclude>
<ClInclude Include="GLES\VertexShaderGeneratorGLES.h">
<Filter>GLES</Filter>
</ClInclude>
<ClInclude Include="GLES\FramebufferManagerGLES.h">
<Filter>GLES</Filter>
</ClInclude>
@ -267,12 +264,12 @@
<ClInclude Include="ge_constants.h">
<Filter>Common</Filter>
</ClInclude>
<ClInclude Include="Directx9\VertexShaderGeneratorHLSL.h">
<Filter>DirectX9</Filter>
</ClInclude>
<ClInclude Include="Common\FragmentShaderGenerator.h">
<Filter>Common</Filter>
</ClInclude>
<ClInclude Include="Common\VertexShaderGenerator.h">
<Filter>Common</Filter>
</ClInclude>
</ItemGroup>
<ItemGroup>
<ClCompile Include="Math3D.cpp">
@ -446,9 +443,6 @@
<ClCompile Include="GLES\DepalettizeShaderGLES.cpp">
<Filter>GLES</Filter>
</ClCompile>
<ClCompile Include="GLES\VertexShaderGeneratorGLES.cpp">
<Filter>GLES</Filter>
</ClCompile>
<ClCompile Include="GLES\FramebufferManagerGLES.cpp">
<Filter>GLES</Filter>
</ClCompile>
@ -539,11 +533,11 @@
<ClCompile Include="GeConstants.cpp">
<Filter>Common</Filter>
</ClCompile>
<ClCompile Include="Directx9\VertexShaderGeneratorHLSL.cpp">
<Filter>DirectX9</Filter>
</ClCompile>
<ClCompile Include="Common\FragmentShaderGenerator.cpp">
<Filter>Common</Filter>
</ClCompile>
<ClCompile Include="Common\VertexShaderGenerator.cpp">
<Filter>Common</Filter>
</ClCompile>
</ItemGroup>
</Project>

View File

@ -36,11 +36,11 @@
#include "GPU/Math3D.h"
#include "GPU/GPUState.h"
#include "GPU/ge_constants.h"
#include "GPU/Common/FragmentShaderGenerator.h"
#include "GPU/Common/VertexShaderGenerator.h"
#include "GPU/Vulkan/ShaderManagerVulkan.h"
#include "GPU/Vulkan/DrawEngineVulkan.h"
#include "GPU/Vulkan/FramebufferManagerVulkan.h"
#include "GPU/Common/FragmentShaderGenerator.h"
#include "GPU/GLES/VertexShaderGeneratorGLES.h"
VulkanFragmentShader::VulkanFragmentShader(VulkanContext *vulkan, FShaderID id, const char *code)
: vulkan_(vulkan), id_(id), failed_(false), module_(0) {
@ -265,7 +265,7 @@ void ShaderManagerVulkan::GetShaders(int prim, u32 vertType, VulkanVertexShader
std::string genErrorString;
uint64_t uniformMask = 0; // Not used
uint32_t attributeMask = 0; // Not used
bool success = GenerateVertexShaderGLSL(VSID, codeBuffer_, compat_, &attributeMask, &uniformMask, &genErrorString);
bool success = GenerateVertexShader(VSID, codeBuffer_, compat_, &attributeMask, &uniformMask, &genErrorString);
_assert_(success);
vs = new VulkanVertexShader(vulkan_, VSID, codeBuffer_, useHWTransform);
vsCache_.Insert(VSID, vs);
@ -398,7 +398,7 @@ bool ShaderManagerVulkan::LoadCache(FILE *f) {
std::string genErrorString;
uint32_t attributeMask = 0;
uint64_t uniformMask = 0;
if (!GenerateVertexShaderGLSL(id, codeBuffer_, compat_, &attributeMask, &uniformMask, &genErrorString)) {
if (!GenerateVertexShader(id, codeBuffer_, compat_, &attributeMask, &uniformMask, &genErrorString)) {
return false;
}
VulkanVertexShader *vs = new VulkanVertexShader(vulkan_, id, codeBuffer_, useHWTransform);

View File

@ -24,7 +24,7 @@
#include "Common/GPU/Vulkan/VulkanMemory.h"
#include "GPU/Common/ShaderCommon.h"
#include "GPU/Common/ShaderId.h"
#include "GPU/GLES/VertexShaderGeneratorGLES.h"
#include "GPU/Common/VertexShaderGenerator.h"
#include "GPU/Common/FragmentShaderGenerator.h"
#include "GPU/Vulkan/VulkanUtil.h"
#include "Common/Math/lin/matrix4x4.h"

View File

@ -402,6 +402,7 @@
<ClInclude Include="..\..\GPU\Common\TextureScalerCommon.h" />
<ClInclude Include="..\..\GPU\Common\TransformCommon.h" />
<ClInclude Include="..\..\GPU\Common\VertexDecoderCommon.h" />
<ClInclude Include="..\..\GPU\Common\VertexShaderGenerator.h" />
<ClInclude Include="..\..\GPU\D3D11\D3D11Util.h" />
<ClInclude Include="..\..\GPU\D3D11\DepalettizeShaderD3D11.h" />
<ClInclude Include="..\..\GPU\D3D11\DrawEngineD3D11.h" />
@ -417,7 +418,6 @@
<ClInclude Include="..\..\GPU\Debugger\Record.h" />
<ClInclude Include="..\..\GPU\Debugger\RecordFormat.h" />
<ClInclude Include="..\..\GPU\Debugger\Stepping.h" />
<ClInclude Include="..\..\GPU\Directx9\VertexShaderGeneratorHLSL.h" />
<ClInclude Include="..\..\GPU\GeDisasm.h" />
<ClInclude Include="..\..\GPU\ge_constants.h" />
<ClInclude Include="..\..\GPU\GPU.h" />
@ -462,6 +462,7 @@
<ClCompile Include="..\..\GPU\Common\VertexDecoderCommon.cpp" />
<ClCompile Include="..\..\GPU\Common\VertexDecoderFake.cpp" />
<ClCompile Include="..\..\GPU\Common\VertexDecoderX86.cpp" />
<ClCompile Include="..\..\GPU\Common\VertexShaderGenerator.cpp" />
<ClCompile Include="..\..\GPU\D3D11\D3D11Util.cpp" />
<ClCompile Include="..\..\GPU\D3D11\DepalettizeShaderD3D11.cpp" />
<ClCompile Include="..\..\GPU\D3D11\DrawEngineD3D11.cpp" />
@ -477,7 +478,6 @@
<ClCompile Include="..\..\GPU\Debugger\Playback.cpp" />
<ClCompile Include="..\..\GPU\Debugger\Record.cpp" />
<ClCompile Include="..\..\GPU\Debugger\Stepping.cpp" />
<ClCompile Include="..\..\GPU\Directx9\VertexShaderGeneratorHLSL.cpp" />
<ClCompile Include="..\..\GPU\GeConstants.cpp" />
<ClCompile Include="..\..\GPU\GeDisasm.cpp" />
<ClCompile Include="..\..\GPU\GPU.cpp" />

View File

@ -41,7 +41,6 @@
<ClCompile Include="..\..\GPU\Debugger\Playback.cpp" />
<ClCompile Include="..\..\GPU\Debugger\Record.cpp" />
<ClCompile Include="..\..\GPU\Debugger\Stepping.cpp" />
<ClCompile Include="..\..\GPU\Directx9\VertexShaderGeneratorHLSL.cpp" />
<ClCompile Include="..\..\GPU\GeConstants.cpp" />
<ClCompile Include="..\..\GPU\GeDisasm.cpp" />
<ClCompile Include="..\..\GPU\GPU.cpp" />
@ -57,6 +56,7 @@
<ClCompile Include="pch.cpp" />
<ClCompile Include="..\..\GPU\Software\RasterizerRectangle.cpp" />
<ClCompile Include="..\..\GPU\Common\FragmentShaderGenerator.cpp" />
<ClCompile Include="..\..\GPU\Common\VertexShaderGenerator.cpp" />
</ItemGroup>
<ItemGroup>
<ClInclude Include="..\..\GPU\Common\DepalettizeShaderCommon.h" />
@ -96,7 +96,6 @@
<ClInclude Include="..\..\GPU\Debugger\Record.h" />
<ClInclude Include="..\..\GPU\Debugger\RecordFormat.h" />
<ClInclude Include="..\..\GPU\Debugger\Stepping.h" />
<ClInclude Include="..\..\GPU\Directx9\VertexShaderGeneratorHLSL.h" />
<ClInclude Include="..\..\GPU\GeDisasm.h" />
<ClInclude Include="..\..\GPU\ge_constants.h" />
<ClInclude Include="..\..\GPU\GPU.h" />
@ -114,5 +113,6 @@
<ClInclude Include="targetver.h" />
<ClInclude Include="..\..\GPU\Software\RasterizerRectangle.h" />
<ClInclude Include="..\..\GPU\Common\FragmentShaderGenerator.h" />
<ClInclude Include="..\..\GPU\Common\VertexShaderGenerator.h" />
</ItemGroup>
</Project>

View File

@ -326,6 +326,7 @@ EXEC_AND_LIB_FILES := \
$(SRC)/GPU/Common/TextureDecoder.cpp \
$(SRC)/GPU/Common/PostShader.cpp \
$(SRC)/GPU/Common/ShaderUniforms.cpp \
$(SRC)/GPU/Common/VertexShaderGenerator.cpp \
$(SRC)/GPU/Debugger/Breakpoints.cpp \
$(SRC)/GPU/Debugger/Debugger.cpp \
$(SRC)/GPU/Debugger/Playback.cpp \
@ -333,14 +334,13 @@ EXEC_AND_LIB_FILES := \
$(SRC)/GPU/Debugger/Stepping.cpp \
$(SRC)/GPU/GLES/FramebufferManagerGLES.cpp \
$(SRC)/GPU/GLES/DepalettizeShaderGLES.cpp \
$(SRC)/GPU/GLES/DepthBufferGLES.cpp.arm \
$(SRC)/GPU/GLES/DepthBufferGLES.cpp \
$(SRC)/GPU/GLES/GPU_GLES.cpp.arm \
$(SRC)/GPU/GLES/StencilBufferGLES.cpp.arm \
$(SRC)/GPU/GLES/StencilBufferGLES.cpp \
$(SRC)/GPU/GLES/TextureCacheGLES.cpp.arm \
$(SRC)/GPU/GLES/DrawEngineGLES.cpp.arm \
$(SRC)/GPU/GLES/StateMappingGLES.cpp.arm \
$(SRC)/GPU/GLES/ShaderManagerGLES.cpp.arm \
$(SRC)/GPU/GLES/VertexShaderGeneratorGLES.cpp.arm \
$(SRC)/GPU/GLES/FragmentTestCacheGLES.cpp.arm \
$(SRC)/GPU/GLES/TextureScalerGLES.cpp \
$(SRC)/GPU/Null/NullGpu.cpp \

View File

@ -249,6 +249,7 @@ SOURCES_CXX += \
$(GPUDIR)/Debugger/Record.cpp \
$(GPUDIR)/Debugger/Stepping.cpp \
$(GPUDIR)/Common/FragmentShaderGenerator.cpp \
$(GPUDIR)/Common/VertexShaderGenerator.cpp \
$(GPUDIR)/Common/TextureCacheCommon.cpp \
$(GPUDIR)/Common/TextureScalerCommon.cpp \
$(GPUDIR)/Common/SoftwareTransformCommon.cpp \
@ -269,7 +270,6 @@ SOURCES_CXX += \
$(GPUDIR)/Software/RasterizerRectangle.cpp \
$(GPUDIR)/GLES/DepalettizeShaderGLES.cpp \
$(GPUDIR)/GLES/DepthBufferGLES.cpp \
$(GPUDIR)/GLES/VertexShaderGeneratorGLES.cpp \
$(GPUDIR)/GLES/DrawEngineGLES.cpp \
$(GPUDIR)/GLES/GPU_GLES.cpp \
$(GPUDIR)/GLES/FragmentTestCacheGLES.cpp \

View File

@ -10,9 +10,7 @@
#include "GPU/Vulkan/VulkanContext.h"
#include "GPU/Common/FragmentShaderGenerator.h"
#include "GPU/Directx9/VertexShaderGeneratorHLSL.h"
#include "GPU/GLES/VertexShaderGeneratorGLES.h"
#include "GPU/Common/VertexShaderGenerator.h"
#include "GPU/D3D11/D3D11Util.h"
#include "GPU/D3D11/D3D11Loader.h"
@ -60,27 +58,27 @@ bool GenerateVShader(VShaderID id, char *buffer, ShaderLanguage lang, std::strin
case ShaderLanguage::GLSL_VULKAN:
{
ShaderLanguageDesc compat(ShaderLanguage::GLSL_VULKAN);
return GenerateVertexShaderGLSL(id, buffer, compat, &attrMask, &uniformMask, errorString);
return GenerateVertexShader(id, buffer, compat, &attrMask, &uniformMask, errorString);
}
case ShaderLanguage::GLSL_140:
{
ShaderLanguageDesc compat(ShaderLanguage::GLSL_140);
return GenerateVertexShaderGLSL(id, buffer, compat, &attrMask, &uniformMask, errorString);
return GenerateVertexShader(id, buffer, compat, &attrMask, &uniformMask, errorString);
}
case ShaderLanguage::GLSL_300:
{
ShaderLanguageDesc compat(ShaderLanguage::GLSL_140);
return GenerateVertexShaderGLSL(id, buffer, compat, &attrMask, &uniformMask, errorString);
return GenerateVertexShader(id, buffer, compat, &attrMask, &uniformMask, errorString);
}
case ShaderLanguage::HLSL_D3D9:
{
//ShaderLanguageDesc compat(ShaderLanguage::HLSL_D3D9);
return GenerateVertexShaderHLSL(id, buffer, ShaderLanguage::HLSL_D3D9, errorString);
ShaderLanguageDesc compat(ShaderLanguage::HLSL_D3D9);
return GenerateVertexShader(id, buffer, compat, &attrMask, &uniformMask, errorString);
}
case ShaderLanguage::HLSL_D3D11:
{
//ShaderLanguageDesc compat(ShaderLanguage::HLSL_D3D11);
return GenerateVertexShaderHLSL(id, buffer, ShaderLanguage::HLSL_D3D11, errorString);
ShaderLanguageDesc compat(ShaderLanguage::HLSL_D3D11);
return GenerateVertexShader(id, buffer, compat, &attrMask, &uniformMask, errorString);
}
default:
return false;
@ -167,6 +165,56 @@ bool TestShaderGenerators() {
int successes = 0;
int count = 700;
// Generate a bunch of random vertex shader IDs, try to generate shader source.
// Then compile it and check that it's ok.
for (int i = 0; i < count; i++) {
uint32_t bottom = rng.R32();
uint32_t top = rng.R32();
VShaderID id;
id.d[0] = bottom;
id.d[1] = top;
// The generated bits need some adjustment:
// We don't use these bits in the HLSL shader generator.
id.SetBits(VS_BIT_WEIGHT_FMTSCALE, 2, 0);
// If mode is through, we won't do hardware transform.
if (id.Bit(VS_BIT_IS_THROUGH)) {
id.SetBit(VS_BIT_USE_HW_TRANSFORM, 0);
}
if (!id.Bit(VS_BIT_USE_HW_TRANSFORM)) {
id.SetBit(VS_BIT_ENABLE_BONES, 0);
}
bool generateSuccess[numLanguages]{};
std::string genErrorString[numLanguages];
for (int j = 0; j < numLanguages; j++) {
generateSuccess[j] = GenerateVShader(id, buffer[j], languages[j], &genErrorString[j]);
if (!genErrorString[j].empty()) {
printf("%s\n", genErrorString[j].c_str());
}
}
// Now that we have the strings ready for easy comparison (buffer,4 in the watch window),
// let's try to compile them.
for (int j = 0; j < numLanguages; j++) {
if (generateSuccess[j]) {
std::string errorMessage;
if (!TestCompileShader(buffer[j], languages[j], true, &errorMessage)) {
printf("Error compiling vertex shader %d:\n\n%s\n\n%s\n", (int)j, LineNumberString(buffer[j]).c_str(), errorMessage.c_str());
return false;
}
successes++;
}
}
}
printf("%d/%d vertex shaders generated (it's normal that it's not all, there are invalid bit combos)\n", successes, count * numLanguages);
successes = 0;
count = 200;
// Generate a bunch of random fragment shader IDs, try to generate shader source.
// Then compile it and check that it's ok.
for (int i = 0; i < count; i++) {
@ -195,24 +243,6 @@ bool TestShaderGenerators() {
// We ignore the contents of the error string here, not even gonna try to compile if it errors.
}
// KEEPING FOR REUSE LATER: Defunct temporary test.
/*
if (generateSuccess[0] != generateSuccess[1]) {
printf("mismatching success! %s %s\n", genErrorString[0].c_str(), genErrorString[1].c_str());
printf("%s\n", buffer[0]);
printf("%s\n", buffer[1]);
return 1;
}
if (generateSuccess[0] && strcmp(buffer[0], buffer[1])) {
printf("mismatching shaders! a=glsl b=hlsl\n");
PrintDiff(buffer[0], buffer[1]);
return 1;
}
if (generateSuccess[2] && strcmp(buffer[2], buffer[3])) {
printf("mismatching shaders! a=glsl b=hlsl\n");
PrintDiff(buffer[2], buffer[3]);
return 1;
}*/
// Now that we have the strings ready for easy comparison (buffer,4 in the watch window),
// let's try to compile them.
for (int j = 0; j < numLanguages; j++) {
@ -232,58 +262,6 @@ bool TestShaderGenerators() {
successes = 0;
count = 200;
/*
// Generate a bunch of random vertex shader IDs, try to generate shader source.
// Then compile it and check that it's ok.
for (int i = 0; i < count; i++) {
uint32_t bottom = rng.R32();
uint32_t top = rng.R32();
VShaderID id;
id.d[0] = bottom;
id.d[1] = top;
bool generateSuccess[numLanguages]{};
std::string genErrorString[numLanguages];
for (int j = 0; j < numLanguages; j++) {
generateSuccess[j] = GenerateVShader(id, buffer[j], languages[j], &genErrorString[j]);
if (!genErrorString[j].empty()) {
printf("%s\n", genErrorString[j].c_str());
}
}
// KEEPING FOR REUSE LATER: Defunct temporary test: Compare GLSL-in-Vulkan-mode vs Vulkan
if (generateSuccess[0] != generateSuccess[1]) {
printf("mismatching success! '%s' '%s'\n", genErrorString[0].c_str(), genErrorString[1].c_str());
printf("%s\n", buffer[0]);
printf("%s\n", buffer[1]);
return false;
}
if (generateSuccess[0] && strcmp(buffer[0], buffer[1])) {
printf("mismatching shaders!\n");
PrintDiff(buffer[0], buffer[1]);
return false;
}
// Now that we have the strings ready for easy comparison (buffer,4 in the watch window),
// let's try to compile them.
for (int j = 0; j < numLanguages; j++) {
if (generateSuccess[j]) {
std::string errorMessage;
if (!TestCompileShader(buffer[j], languages[j], true, &errorMessage)) {
printf("Error compiling vertex shader:\n\n%s\n\n%s\n", LineNumberString(buffer[j]).c_str(), errorMessage.c_str());
return false;
}
successes++;
}
}
}
printf("%d/%d vertex shaders generated (it's normal that it's not all, there are invalid bit combos)\n", successes, count * numLanguages);
successes = 0;
count = 200;
*/
_CrtCheckMemory();