From 7eb353b3bd05d7a24362c58ac61dcc40a04c3deb Mon Sep 17 00:00:00 2001 From: Jules Blok Date: Sun, 21 Dec 2014 12:52:14 +0100 Subject: [PATCH] VideoCommon: Don't pass structs between shaders, use the interface blocks instead. --- Source/Core/VideoBackends/OGL/Render.cpp | 3 +- Source/Core/VideoCommon/GeometryShaderGen.cpp | 36 ++++++++++++----- Source/Core/VideoCommon/PixelShaderGen.cpp | 29 ++++++-------- Source/Core/VideoCommon/ShaderGenCommon.h | 40 +++++++++++++------ Source/Core/VideoCommon/VertexShaderGen.cpp | 10 +++-- 5 files changed, 74 insertions(+), 44 deletions(-) diff --git a/Source/Core/VideoBackends/OGL/Render.cpp b/Source/Core/VideoBackends/OGL/Render.cpp index 949c0c707b..8bb1ecc8f2 100644 --- a/Source/Core/VideoBackends/OGL/Render.cpp +++ b/Source/Core/VideoBackends/OGL/Render.cpp @@ -462,8 +462,7 @@ Renderer::Renderer() g_Config.backend_info.bSupportsEarlyZ = GLExtensions::Supports("GL_ARB_shader_image_load_store"); g_Config.backend_info.bSupportsBBox = GLExtensions::Supports("GL_ARB_shader_storage_buffer_object"); g_Config.backend_info.bSupportsGSInstancing = GLExtensions::Supports("GL_ARB_gpu_shader5"); - g_Config.backend_info.bSupportsGeometryShaders = (GLExtensions::Version() >= 320) && - !DriverDetails::HasBug(DriverDetails::BUG_INTELBROKENSTRUCTS); + g_Config.backend_info.bSupportsGeometryShaders = GLExtensions::Version() >= 320; // Desktop OpenGL supports the binding layout if it supports 420pack // OpenGL ES 3.1 supports it implicitly without an extension diff --git a/Source/Core/VideoCommon/GeometryShaderGen.cpp b/Source/Core/VideoCommon/GeometryShaderGen.cpp index 7e8fa080e2..0a49487bf8 100644 --- a/Source/Core/VideoCommon/GeometryShaderGen.cpp +++ b/Source/Core/VideoCommon/GeometryShaderGen.cpp @@ -84,7 +84,9 @@ static inline void GenerateGeometryShader(T& out, u32 primitive_type, API_TYPE A uid_data->numTexGens = bpmem.genMode.numtexgens; uid_data->pixel_lighting = g_ActiveConfig.bEnablePixelLighting; - GenerateVSOutputStruct(out, ApiType); + out.Write("struct VS_OUTPUT {\n"); + GenerateVSOutputMembers(out, ApiType); + out.Write("};\n"); if (ApiType == API_OPENGL) { @@ -92,11 +94,11 @@ static inline void GenerateGeometryShader(T& out, u32 primitive_type, API_TYPE A out.Write("#define InstanceID gl_InvocationID\n"); out.Write("in VertexData {\n"); - out.Write("\tcentroid %s VS_OUTPUT o;\n", g_ActiveConfig.backend_info.bSupportsBindingLayout ? "" : "in"); + GenerateVSOutputMembers(out, ApiType, g_ActiveConfig.backend_info.bSupportsBindingLayout ? "centroid" : "centroid in"); out.Write("} vs[%d];\n", vertex_in); out.Write("out VertexData {\n"); - out.Write("\tcentroid %s VS_OUTPUT o;\n", g_ActiveConfig.backend_info.bSupportsBindingLayout ? "" : "out"); + GenerateVSOutputMembers(out, ApiType, g_ActiveConfig.backend_info.bSupportsBindingLayout ? "centroid" : "centroid out"); if (g_ActiveConfig.iStereoMode > 0) out.Write("\tflat int layer;\n"); @@ -133,8 +135,9 @@ static inline void GenerateGeometryShader(T& out, u32 primitive_type, API_TYPE A { if (ApiType == API_OPENGL) { - out.Write("\tVS_OUTPUT start = vs[0].o;\n"); - out.Write("\tVS_OUTPUT end = vs[1].o;\n"); + out.Write("\tVS_OUTPUT start, end;\n"); + AssignVSOutputMembers(out, "start", "vs[0]"); + AssignVSOutputMembers(out, "end", "vs[1]"); } else { @@ -163,9 +166,14 @@ static inline void GenerateGeometryShader(T& out, u32 primitive_type, API_TYPE A else if (primitive_type == PRIMITIVE_POINTS) { if (ApiType == API_OPENGL) - out.Write("\tVS_OUTPUT center = vs[0].o;\n"); + { + out.Write("\tVS_OUTPUT center;\n"); + AssignVSOutputMembers(out, "center", "vs[0]"); + } else + { out.Write("\tVS_OUTPUT center = o[0];\n"); + } // Offset from center to upper right vertex // Lerp PointSize/2 from [0,0..VpWidth,VpHeight] to [-1,1..1,-1] @@ -188,9 +196,14 @@ static inline void GenerateGeometryShader(T& out, u32 primitive_type, API_TYPE A out.Write("\tfor (int i = 0; i < %d; ++i) {\n", vertex_in); if (ApiType == API_OPENGL) - out.Write("\tVS_OUTPUT f = vs[i].o;\n"); + { + out.Write("\tVS_OUTPUT f;\n"); + AssignVSOutputMembers(out, "f", "vs[i]"); + } else + { out.Write("\tVS_OUTPUT f = o[i];\n"); + } if (g_ActiveConfig.iStereoMode > 0) { @@ -289,9 +302,14 @@ static inline void EmitVertex(T& out, const char* vertex, API_TYPE ApiType, bool out.Write("\tif (i == 0) first = %s;\n", vertex); if (ApiType == API_OPENGL) + { out.Write("\tgl_Position = %s.pos;\n", vertex); - - out.Write("\tps.o = %s;\n", vertex); + AssignVSOutputMembers(out, "ps", vertex); + } + else + { + out.Write("\tps.o = %s;\n", vertex); + } if (ApiType == API_OPENGL) out.Write("\tEmitVertex();\n"); diff --git a/Source/Core/VideoCommon/PixelShaderGen.cpp b/Source/Core/VideoCommon/PixelShaderGen.cpp index e2c0168d13..f7b63bfb30 100644 --- a/Source/Core/VideoCommon/PixelShaderGen.cpp +++ b/Source/Core/VideoCommon/PixelShaderGen.cpp @@ -264,7 +264,9 @@ static inline void GeneratePixelShader(T& out, DSTALPHA_MODE dstAlphaMode, API_T } } - GenerateVSOutputStruct(out, ApiType); + out.Write("struct VS_OUTPUT {\n"); + GenerateVSOutputMembers(out, ApiType); + out.Write("};\n"); const bool forced_early_z = g_ActiveConfig.backend_info.bSupportsEarlyZ && bpmem.UseEarlyDepthTest() && (g_ActiveConfig.bFastDepthCalc || bpmem.alpha_test.TestResult() == AlphaTest::UNDETERMINED); const bool per_pixel_depth = (bpmem.ztex2.op != ZTEXTURE_DISABLE && bpmem.UseLateDepthTest()) || (!g_ActiveConfig.bFastDepthCalc && bpmem.zmode.testenable && !forced_early_z); @@ -320,7 +322,7 @@ static inline void GeneratePixelShader(T& out, DSTALPHA_MODE dstAlphaMode, API_T if (g_ActiveConfig.backend_info.bSupportsGeometryShaders) { out.Write("in VertexData {\n"); - out.Write("\tcentroid %s VS_OUTPUT o;\n", g_ActiveConfig.backend_info.bSupportsBindingLayout ? "" : "in"); + GenerateVSOutputMembers(out, ApiType, g_ActiveConfig.backend_info.bSupportsBindingLayout ? "centroid" : "centroid in"); if (g_ActiveConfig.iStereoMode > 0) out.Write("\tflat int layer;\n"); @@ -348,23 +350,16 @@ static inline void GeneratePixelShader(T& out, DSTALPHA_MODE dstAlphaMode, API_T if (g_ActiveConfig.backend_info.bSupportsGeometryShaders) { - // compute window position if needed because binding semantic WPOS is not widely supported - // Let's set up attributes for (unsigned int i = 0; i < numTexgen; ++i) - { - out.Write("\tfloat3 uv%d = o.tex%d;\n", i, i); - } - out.Write("\tfloat4 clipPos = o.clipPos;\n"); - if (g_ActiveConfig.bEnablePixelLighting) - { - out.Write("\tfloat4 Normal = o.Normal;\n"); - } + out.Write("\tfloat3 uv%d = tex%d;\n", i, i); + } + else + { + // On Mali, global variables must be initialized as constants. + // This is why we initialize these variables locally instead. + out.Write("\tfloat4 colors_0 = colors_02;\n"); + out.Write("\tfloat4 colors_1 = colors_12;\n"); } - - // On Mali, global variables must be initialized as constants. - // This is why we initialize these variables locally instead. - out.Write("\tfloat4 colors_0 = %s;\n", g_ActiveConfig.backend_info.bSupportsGeometryShaders ? "o.colors_0" : "colors_02"); - out.Write("\tfloat4 colors_1 = %s;\n", g_ActiveConfig.backend_info.bSupportsGeometryShaders ? "o.colors_1" : "colors_12"); out.Write("\tfloat4 rawpos = gl_FragCoord;\n"); } diff --git a/Source/Core/VideoCommon/ShaderGenCommon.h b/Source/Core/VideoCommon/ShaderGenCommon.h index 444abad82c..505ef2f116 100644 --- a/Source/Core/VideoCommon/ShaderGenCommon.h +++ b/Source/Core/VideoCommon/ShaderGenCommon.h @@ -221,9 +221,13 @@ private: }; template -static void DefineOutputStructMember(T& object, API_TYPE api_type, const char* type, const char* name, int var_index, const char* semantic = "", int semantic_index = -1) +static void DefineOutputMember(T& object, API_TYPE api_type, const char* qualifier, const char* type, const char* name, int var_index, const char* semantic = "", int semantic_index = -1) { - object.Write(" %s %s", type, name); + if (qualifier != nullptr) + object.Write("\t%s %s %s", qualifier, type, name); + else + object.Write("\t%s %s", type, name); + if (var_index != -1) object.Write("%d", var_index); @@ -239,23 +243,35 @@ static void DefineOutputStructMember(T& object, API_TYPE api_type, const char* t } template -static inline void GenerateVSOutputStruct(T& object, API_TYPE api_type) +static inline void GenerateVSOutputMembers(T& object, API_TYPE api_type, const char* qualifier = nullptr) { - object.Write("struct VS_OUTPUT {\n"); - - DefineOutputStructMember(object, api_type, "float4", "pos", -1, "POSITION"); - DefineOutputStructMember(object, api_type, "float4", "colors_", 0, "COLOR", 0); - DefineOutputStructMember(object, api_type, "float4", "colors_", 1, "COLOR", 1); + DefineOutputMember(object, api_type, qualifier, "float4", "pos", -1, "POSITION"); + DefineOutputMember(object, api_type, qualifier, "float4", "colors_", 0, "COLOR", 0); + DefineOutputMember(object, api_type, qualifier, "float4", "colors_", 1, "COLOR", 1); for (unsigned int i = 0; i < xfmem.numTexGen.numTexGens; ++i) - DefineOutputStructMember(object, api_type, "float3", "tex", i, "TEXCOORD", i); + DefineOutputMember(object, api_type, qualifier, "float3", "tex", i, "TEXCOORD", i); - DefineOutputStructMember(object, api_type, "float4", "clipPos", -1, "TEXCOORD", xfmem.numTexGen.numTexGens); + DefineOutputMember(object, api_type, qualifier, "float4", "clipPos", -1, "TEXCOORD", xfmem.numTexGen.numTexGens); if (g_ActiveConfig.bEnablePixelLighting) - DefineOutputStructMember(object, api_type, "float4", "Normal", -1, "TEXCOORD", xfmem.numTexGen.numTexGens + 1); + DefineOutputMember(object, api_type, qualifier, "float4", "Normal", -1, "TEXCOORD", xfmem.numTexGen.numTexGens + 1); +} - object.Write("};\n"); +template +static inline void AssignVSOutputMembers(T& object, const char* a, const char* b) +{ + object.Write("\t%s.pos = %s.pos;\n", a, b); + object.Write("\t%s.colors_0 = %s.colors_0;\n", a, b); + object.Write("\t%s.colors_1 = %s.colors_1;\n", a, b); + + for (unsigned int i = 0; i < xfmem.numTexGen.numTexGens; ++i) + object.Write("\t%s.tex%d = %s.tex%d;\n", a, i, b, i); + + object.Write("\t%s.clipPos = %s.clipPos;\n", a, b); + + if (g_ActiveConfig.bEnablePixelLighting) + object.Write("\t%s.Normal = %s.Normal;\n", a, b); } // Constant variable names diff --git a/Source/Core/VideoCommon/VertexShaderGen.cpp b/Source/Core/VideoCommon/VertexShaderGen.cpp index 10ebc89f86..b48d1a1738 100644 --- a/Source/Core/VideoCommon/VertexShaderGen.cpp +++ b/Source/Core/VideoCommon/VertexShaderGen.cpp @@ -42,7 +42,9 @@ static inline void GenerateVertexShader(T& out, u32 components, API_TYPE api_typ out.Write(s_shader_uniforms); out.Write("};\n"); - GenerateVSOutputStruct(out, api_type); + out.Write("struct VS_OUTPUT {\n"); + GenerateVSOutputMembers(out, api_type); + out.Write("};\n"); uid_data->numTexGens = xfmem.numTexGen.numTexGens; uid_data->components = components; @@ -74,9 +76,9 @@ static inline void GenerateVertexShader(T& out, u32 components, API_TYPE api_typ if (g_ActiveConfig.backend_info.bSupportsGeometryShaders) { - out.Write("out VertexData {\n" - "\tcentroid %s VS_OUTPUT o;\n" - "};\n", g_ActiveConfig.backend_info.bSupportsBindingLayout ? "" : "out"); + out.Write("out VertexData {\n"); + GenerateVSOutputMembers(out, api_type, g_ActiveConfig.backend_info.bSupportsBindingLayout ? "centroid" : "centroid out"); + out.Write("} o;\n"); } else {