diff options
39 files changed, 1377 insertions, 1826 deletions
diff --git a/indra/llrender/llgl.cpp b/indra/llrender/llgl.cpp index e15157af05..67bd9e277e 100644 --- a/indra/llrender/llgl.cpp +++ b/indra/llrender/llgl.cpp @@ -2413,7 +2413,7 @@ void LLGLState::checkStates(const std::string& msg) BOOL error = FALSE; - if (src != GL_SRC_ALPHA || dst != GL_ONE_MINUS_SRC_ALPHA) + /*if (src != GL_SRC_ALPHA || dst != GL_ONE_MINUS_SRC_ALPHA) { if (gDebugSession) { @@ -2424,7 +2424,7 @@ void LLGLState::checkStates(const std::string& msg) { LL_GL_ERRS << "Blend function corrupted: " << std::hex << src << " " << std::hex << dst << " " << msg << std::dec << LL_ENDL; } - } + }*/ for (boost::unordered_map<LLGLenum, LLGLboolean>::iterator iter = sStateMap.begin(); iter != sStateMap.end(); ++iter) diff --git a/indra/llrender/llglslshader.cpp b/indra/llrender/llglslshader.cpp index e3b29dc812..c5f4efd2c0 100644 --- a/indra/llrender/llglslshader.cpp +++ b/indra/llrender/llglslshader.cpp @@ -1,25 +1,25 @@ -/** +/** * @file llglslshader.cpp * @brief GLSL helper functions and state. * * $LicenseInfo:firstyear=2005&license=viewerlgpl$ * Second Life Viewer Source Code * Copyright (C) 2010, Linden Research, Inc. - * + * * This library is free software; you can redistribute it and/or * modify it under the terms of the GNU Lesser General Public * License as published by the Free Software Foundation; * version 2.1 of the License only. - * + * * This library 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 * Lesser General Public License for more details. - * + * * You should have received a copy of the GNU Lesser General Public * License along with this library; if not, write to the Free Software * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA - * + * * Linden Research, Inc., 945 Battery Street, San Francisco, CA 94111 USA * $/LicenseInfo$ */ @@ -37,8 +37,8 @@ #include "OpenGL/OpenGL.h" #endif -// Print-print list of shader included source files that are linked together via glAttachShader() -// i.e. On macOS / OSX the AMD GLSL linker will display an error if a varying is left in an undefined state. + // Print-print list of shader included source files that are linked together via glAttachShader() + // i.e. On macOS / OSX the AMD GLSL linker will display an error if a varying is left in an undefined state. #define DEBUG_SHADER_INCLUDES 0 // Lots of STL stuff in here, using namespace std to keep things more readable @@ -55,24 +55,24 @@ std::set<LLGLSLShader*> LLGLSLShader::sInstances; U64 LLGLSLShader::sTotalTimeElapsed = 0; U32 LLGLSLShader::sTotalTrianglesDrawn = 0; U64 LLGLSLShader::sTotalSamplesDrawn = 0; -U32 LLGLSLShader::sTotalDrawCalls = 0; +U32 LLGLSLShader::sTotalBinds = 0; //UI shader -- declared here so llui_libtest will link properly LLGLSLShader gUIProgram; LLGLSLShader gSolidColorProgram; // NOTE: Keep gShaderConsts* and LLGLSLShader::ShaderConsts_e in sync! -const std::string gShaderConstsKey[ LLGLSLShader::NUM_SHADER_CONSTS ] = +const std::string gShaderConstsKey[LLGLSLShader::NUM_SHADER_CONSTS] = { - "LL_SHADER_CONST_CLOUD_MOON_DEPTH" - , "LL_SHADER_CONST_STAR_DEPTH" + "LL_SHADER_CONST_CLOUD_MOON_DEPTH" + , "LL_SHADER_CONST_STAR_DEPTH" }; // NOTE: Keep gShaderConsts* and LLGLSLShader::ShaderConsts_e in sync! -const std::string gShaderConstsVal[ LLGLSLShader::NUM_SHADER_CONSTS ] = +const std::string gShaderConstsVal[LLGLSLShader::NUM_SHADER_CONSTS] = { - "0.99998" // SHADER_CONST_CLOUD_MOON_DEPTH // SL-14113 - , "0.99999" // SHADER_CONST_STAR_DEPTH // SL-14113 + "0.99998" // SHADER_CONST_CLOUD_MOON_DEPTH // SL-14113 + , "0.99999" // SHADER_CONST_STAR_DEPTH // SL-14113 }; @@ -121,7 +121,7 @@ void LLGLSLShader::initProfile() sTotalTimeElapsed = 0; sTotalTrianglesDrawn = 0; sTotalSamplesDrawn = 0; - sTotalDrawCalls = 0; + sTotalBinds = 0; for (std::set<LLGLSLShader*>::iterator iter = sInstances.begin(); iter != sInstances.end(); ++iter) { @@ -132,10 +132,10 @@ void LLGLSLShader::initProfile() struct LLGLSLShaderCompareTimeElapsed { - bool operator()(const LLGLSLShader* const& lhs, const LLGLSLShader* const& rhs) - { - return lhs->mTimeElapsed < rhs->mTimeElapsed; - } + bool operator()(const LLGLSLShader* const& lhs, const LLGLSLShader* const& rhs) + { + return lhs->mTimeElapsed < rhs->mTimeElapsed; + } }; //static @@ -158,11 +158,11 @@ void LLGLSLShader::finishProfile(bool emit_report) { (*iter)->dumpStats(); } - - LL_INFOS() << "-----------------------------------" << LL_ENDL; - LL_INFOS() << "Total rendering time: " << llformat("%.4f ms", sTotalTimeElapsed/1000000.f) << LL_ENDL; - LL_INFOS() << "Total samples drawn: " << llformat("%.4f million", sTotalSamplesDrawn/1000000.f) << LL_ENDL; - LL_INFOS() << "Total triangles drawn: " << llformat("%.3f million", sTotalTrianglesDrawn/1000000.f) << LL_ENDL; + + LL_INFOS() << "-----------------------------------" << LL_ENDL; + LL_INFOS() << "Total rendering time: " << llformat("%.4f ms", sTotalTimeElapsed / 1000000.f) << LL_ENDL; + LL_INFOS() << "Total samples drawn: " << llformat("%.4f million", sTotalSamplesDrawn / 1000000.f) << LL_ENDL; + LL_INFOS() << "Total triangles drawn: " << llformat("%.3f million", sTotalTrianglesDrawn / 1000000.f) << LL_ENDL; } } @@ -171,15 +171,12 @@ void LLGLSLShader::clearStats() mTrianglesDrawn = 0; mTimeElapsed = 0; mSamplesDrawn = 0; - mDrawCalls = 0; - mTextureStateFetched = false; - mTextureMagFilter.clear(); - mTextureMinFilter.clear(); + mBinds = 0; } void LLGLSLShader::dumpStats() { - if (mDrawCalls > 0) + if (mBinds > 0) { LL_INFOS() << "=============================================" << LL_ENDL; LL_INFOS() << mName << LL_ENDL; @@ -187,36 +184,25 @@ void LLGLSLShader::dumpStats() { LL_INFOS() << mShaderFiles[i].first << LL_ENDL; } - for (U32 i = 0; i < mTexture.size(); ++i) - { - GLint idx = mTexture[i]; - - if (idx >= 0) - { - GLint uniform_idx = getUniformLocation(i); - LL_INFOS() << mUniformNameMap[uniform_idx] << " - " << std::hex << mTextureMagFilter[i] << "/" << mTextureMinFilter[i] << std::dec << LL_ENDL; - } - } LL_INFOS() << "=============================================" << LL_ENDL; - - F32 ms = mTimeElapsed/1000000.f; - F32 seconds = ms/1000.f; - F32 pct_tris = (F32) mTrianglesDrawn/(F32)sTotalTrianglesDrawn*100.f; - F32 tris_sec = (F32) (mTrianglesDrawn/1000000.0); + F32 ms = mTimeElapsed / 1000000.f; + F32 seconds = ms / 1000.f; + + F32 pct_tris = (F32)mTrianglesDrawn / (F32)sTotalTrianglesDrawn * 100.f; + F32 tris_sec = (F32)(mTrianglesDrawn / 1000000.0); tris_sec /= seconds; - F32 pct_samples = (F32) ((F64)mSamplesDrawn/(F64)sTotalSamplesDrawn)*100.f; - F32 samples_sec = (F32) mSamplesDrawn/1000000000.0; + F32 pct_samples = (F32)((F64)mSamplesDrawn / (F64)sTotalSamplesDrawn) * 100.f; + F32 samples_sec = (F32)mSamplesDrawn / 1000000000.0; samples_sec /= seconds; - F32 pct_calls = (F32) mDrawCalls/(F32)sTotalDrawCalls*100.f; - U32 avg_batch = mTrianglesDrawn/mDrawCalls; + F32 pct_binds = (F32)mBinds / (F32)sTotalBinds * 100.f; - LL_INFOS() << "Triangles Drawn: " << mTrianglesDrawn << " " << llformat("(%.2f pct of total, %.3f million/sec)", pct_tris, tris_sec ) << LL_ENDL; - LL_INFOS() << "Draw Calls: " << mDrawCalls << " " << llformat("(%.2f pct of total, avg %d tris/call)", pct_calls, avg_batch) << LL_ENDL; + LL_INFOS() << "Triangles Drawn: " << mTrianglesDrawn << " " << llformat("(%.2f pct of total, %.3f million/sec)", pct_tris, tris_sec) << LL_ENDL; + LL_INFOS() << "Binds: " << mBinds << " " << llformat("(%.2f pct of total)", pct_binds) << LL_ENDL; LL_INFOS() << "SamplesDrawn: " << mSamplesDrawn << " " << llformat("(%.2f pct of total, %.3f billion/sec)", pct_samples, samples_sec) << LL_ENDL; - LL_INFOS() << "Time Elapsed: " << mTimeElapsed << " " << llformat("(%.2f pct of total, %.5f ms)\n", (F32) ((F64)mTimeElapsed/(F64)sTotalTimeElapsed)*100.f, ms) << LL_ENDL; + LL_INFOS() << "Time Elapsed: " << mTimeElapsed << " " << llformat("(%.2f pct of total, %.5f ms)\n", (F32)((F64)mTimeElapsed / (F64)sTotalTimeElapsed) * 100.f, ms) << LL_ENDL; } } @@ -232,108 +218,80 @@ void LLGLSLShader::startProfile() } //static -void LLGLSLShader::stopProfile(U32 count, U32 mode) +void LLGLSLShader::stopProfile() { LL_PROFILE_ZONE_SCOPED_CATEGORY_SHADER; + if (sProfileEnabled && sCurBoundShaderPtr) { - sCurBoundShaderPtr->readProfileQuery(count, mode); + sCurBoundShaderPtr->unbind(); } } void LLGLSLShader::placeProfileQuery() { - if (mTimerQuery == 0) + if (sProfileEnabled) { - glGenQueries(1, &mSamplesQuery); - glGenQueries(1, &mTimerQuery); - } - - if (!mTextureStateFetched) - { - mTextureStateFetched = true; - mTextureMagFilter.resize(mTexture.size()); - mTextureMinFilter.resize(mTexture.size()); - - U32 cur_active = gGL.getCurrentTexUnitIndex(); - - for (U32 i = 0; i < mTexture.size(); ++i) + if (mTimerQuery == 0) { - GLint idx = mTexture[i]; - - if (idx >= 0) - { - gGL.getTexUnit(idx)->activate(); - - U32 mag = 0xFFFFFFFF; - U32 min = 0xFFFFFFFF; - - U32 type = LLTexUnit::getInternalType(gGL.getTexUnit(idx)->getCurrType()); - - glGetTexParameteriv(type, GL_TEXTURE_MAG_FILTER, (GLint*) &mag); - glGetTexParameteriv(type, GL_TEXTURE_MIN_FILTER, (GLint*) &min); - - mTextureMagFilter[i] = mag; - mTextureMinFilter[i] = min; - } + glGenQueries(1, &mSamplesQuery); + glGenQueries(1, &mTimerQuery); + glGenQueries(1, &mPrimitivesQuery); } - gGL.getTexUnit(cur_active)->activate(); + glBeginQuery(GL_SAMPLES_PASSED, mSamplesQuery); + glBeginQuery(GL_TIME_ELAPSED, mTimerQuery); + glBeginQuery(GL_PRIMITIVES_GENERATED, mPrimitivesQuery); } - - - glBeginQuery(GL_SAMPLES_PASSED, mSamplesQuery); - glBeginQuery(GL_TIME_ELAPSED, mTimerQuery); } -void LLGLSLShader::readProfileQuery(U32 count, U32 mode) +void LLGLSLShader::readProfileQuery() { - glEndQuery(GL_TIME_ELAPSED); - glEndQuery(GL_SAMPLES_PASSED); - - U64 time_elapsed = 0; - glGetQueryObjectui64v(mTimerQuery, GL_QUERY_RESULT, &time_elapsed); + if (sProfileEnabled) + { + glEndQuery(GL_TIME_ELAPSED); + glEndQuery(GL_SAMPLES_PASSED); + glEndQuery(GL_PRIMITIVES_GENERATED); - U64 samples_passed = 0; - glGetQueryObjectui64v(mSamplesQuery, GL_QUERY_RESULT, &samples_passed); + U64 time_elapsed = 0; + glGetQueryObjectui64v(mTimerQuery, GL_QUERY_RESULT, &time_elapsed); - sTotalTimeElapsed += time_elapsed; - mTimeElapsed += time_elapsed; + U64 samples_passed = 0; + glGetQueryObjectui64v(mSamplesQuery, GL_QUERY_RESULT, &samples_passed); - sTotalSamplesDrawn += samples_passed; - mSamplesDrawn += samples_passed; + U64 primitives_generated = 0; + glGetQueryObjectui64v(mPrimitivesQuery, GL_QUERY_RESULT, &primitives_generated); + sTotalTimeElapsed += time_elapsed; + mTimeElapsed += time_elapsed; - U32 tri_count = 0; - switch (mode) - { - case LLRender::TRIANGLES: tri_count = count/3; break; - case LLRender::TRIANGLE_FAN: tri_count = count-2; break; - case LLRender::TRIANGLE_STRIP: tri_count = count-2; break; - default: tri_count = count; break; //points lines etc just use primitive count - } + sTotalSamplesDrawn += samples_passed; + mSamplesDrawn += samples_passed; + + U32 tri_count = (U32)primitives_generated / 3; - mTrianglesDrawn += tri_count; - sTotalTrianglesDrawn += tri_count; + mTrianglesDrawn += tri_count; + sTotalTrianglesDrawn += tri_count; - sTotalDrawCalls++; - mDrawCalls++; + sTotalBinds++; + mBinds++; + } } LLGLSLShader::LLGLSLShader() - : mProgramObject(0), - mAttributeMask(0), - mTotalUniformSize(0), - mActiveTextureChannels(0), - mShaderLevel(0), - mShaderGroup(SG_DEFAULT), - mUniformsDirty(FALSE), - mTimerQuery(0), - mSamplesQuery(0) - + : mProgramObject(0), + mAttributeMask(0), + mTotalUniformSize(0), + mActiveTextureChannels(0), + mShaderLevel(0), + mShaderGroup(SG_DEFAULT), + mUniformsDirty(FALSE), + mTimerQuery(0), + mSamplesQuery(0), + mPrimitivesQuery(0) { - + } LLGLSLShader::~LLGLSLShader() @@ -399,10 +357,10 @@ void LLGLSLShader::unloadInternal() stop_glerror(); } -BOOL LLGLSLShader::createShader(std::vector<LLStaticHashedString> * attributes, - std::vector<LLStaticHashedString> * uniforms, - U32 varying_count, - const char** varyings) +BOOL LLGLSLShader::createShader(std::vector<LLStaticHashedString>* attributes, + std::vector<LLStaticHashedString>* uniforms, + U32 varying_count, + const char** varyings) { LL_PROFILE_ZONE_SCOPED_CATEGORY_SHADER; @@ -430,19 +388,19 @@ BOOL LLGLSLShader::createShader(std::vector<LLStaticHashedString> * attributes, } BOOL success = TRUE; - + #if LL_DARWIN // work-around missing mix(vec3,vec3,bvec3) mDefines["OLD_SELECT"] = "1"; #endif - + #if DEBUG_SHADER_INCLUDES fprintf(stderr, "--- %s ---\n", mName.c_str()); #endif // DEBUG_SHADER_INCLUDES //compile new source - vector< pair<string,GLenum> >::iterator fileIter = mShaderFiles.begin(); - for ( ; fileIter != mShaderFiles.end(); fileIter++ ) + vector< pair<string, GLenum> >::iterator fileIter = mShaderFiles.begin(); + for (; fileIter != mShaderFiles.end(); fileIter++) { GLuint shaderhandle = LLShaderMgr::instance()->loadShaderFile((*fileIter).first, mShaderLevel, (*fileIter).second, &mDefines, mFeatures.mIndexedTextureChannels); LL_DEBUGS("ShaderLoading") << "SHADER FILE: " << (*fileIter).first << " mShaderLevel=" << mShaderLevel << LL_ENDL; @@ -471,7 +429,7 @@ BOOL LLGLSLShader::createShader(std::vector<LLStaticHashedString> * attributes, #ifdef GL_INTERLEAVED_ATTRIBS if (varying_count > 0 && varyings) { - glTransformFeedbackVaryings((GLuint64) mProgramObject, varying_count, varyings, GL_INTERLEAVED_ATTRIBS); + glTransformFeedbackVaryings((GLuint64)mProgramObject, varying_count, varyings, GL_INTERLEAVED_ATTRIBS); } #endif @@ -484,7 +442,7 @@ BOOL LLGLSLShader::createShader(std::vector<LLStaticHashedString> * attributes, { success = mapUniforms(uniforms); } - if( !success ) + if (!success) { LL_SHADER_LOADING_WARNS() << "Failed to link shader: " << mName << LL_ENDL; @@ -493,7 +451,7 @@ BOOL LLGLSLShader::createShader(std::vector<LLStaticHashedString> * attributes, { LL_SHADER_LOADING_WARNS() << "Failed to link using shader level " << mShaderLevel << " trying again using shader level " << (mShaderLevel - 1) << LL_ENDL; mShaderLevel--; - return createShader(attributes,uniforms); + return createShader(attributes, uniforms); } } else if (mFeatures.mIndexedTextureChannels > 0) @@ -528,22 +486,22 @@ BOOL LLGLSLShader::createShader(std::vector<LLStaticHashedString> * attributes, } #if DEBUG_SHADER_INCLUDES -void dumpAttachObject( const char *func_name, GLuint program_object, const std::string &object_path ) +void dumpAttachObject(const char* func_name, GLuint program_object, const std::string& object_path) { GLchar* info_log; GLint info_len_expect = 0; GLint info_len_actual = 0; - glGetShaderiv(program_object, GL_INFO_LOG_LENGTH,, &info_len_expect); + glGetShaderiv(program_object, GL_INFO_LOG_LENGTH, , &info_len_expect); fprintf(stderr, " * %-20s(), log size: %d, %s\n", func_name, info_len_expect, object_path.c_str()); if (info_len_expect > 0) { fprintf(stderr, " ========== %s() ========== \n", func_name); - info_log = new GLchar [ info_len_expect ]; - glGetProgramInfoLog(program_object, info_len_expect, &info_len_actual, info_log); - fprintf(stderr, "%s\n", info_log); - delete [] info_log; + info_log = new GLchar[info_len_expect]; + glGetProgramInfoLog(program_object, info_len_expect, &info_len_actual, info_log); + fprintf(stderr, "%s\n", info_log); + delete[] info_log; } } #endif // DEBUG_SHADER_INCLUDES @@ -612,7 +570,7 @@ void LLGLSLShader::attachObjects(GLuint* objects, S32 count) } } -BOOL LLGLSLShader::mapAttributes(const std::vector<LLStaticHashedString> * attributes) +BOOL LLGLSLShader::mapAttributes(const std::vector<LLStaticHashedString>* attributes) { LL_PROFILE_ZONE_SCOPED_CATEGORY_SHADER; @@ -620,9 +578,9 @@ BOOL LLGLSLShader::mapAttributes(const std::vector<LLStaticHashedString> * attri for (U32 i = 0; i < LLShaderMgr::instance()->mReservedAttribs.size(); i++) { const char* name = LLShaderMgr::instance()->mReservedAttribs[i].c_str(); - glBindAttribLocation(mProgramObject, i, (const GLchar *) name); + glBindAttribLocation(mProgramObject, i, (const GLchar*)name); } - + //link the program BOOL res = link(); @@ -633,7 +591,7 @@ BOOL LLGLSLShader::mapAttributes(const std::vector<LLStaticHashedString> * attri #else mAttribute.resize(LLShaderMgr::instance()->mReservedAttribs.size() + numAttributes, -1); #endif - + if (res) { //read back channel locations @@ -643,7 +601,7 @@ BOOL LLGLSLShader::mapAttributes(const std::vector<LLStaticHashedString> * attri for (U32 i = 0; i < LLShaderMgr::instance()->mReservedAttribs.size(); i++) { const char* name = LLShaderMgr::instance()->mReservedAttribs[i].c_str(); - S32 index = glGetAttribLocation(mProgramObject, (const GLchar *)name); + S32 index = glGetAttribLocation(mProgramObject, (const GLchar*)name); if (index != -1) { #if LL_RELEASE_WITH_DEBUG_INFO @@ -671,11 +629,11 @@ BOOL LLGLSLShader::mapAttributes(const std::vector<LLStaticHashedString> * attri return TRUE; } - + return FALSE; } -void LLGLSLShader::mapUniform(GLint index, const vector<LLStaticHashedString> * uniforms) +void LLGLSLShader::mapUniform(GLint index, const vector<LLStaticHashedString>* uniforms) { LL_PROFILE_ZONE_SCOPED_CATEGORY_SHADER; @@ -691,45 +649,45 @@ void LLGLSLShader::mapUniform(GLint index, const vector<LLStaticHashedString> * name[0] = 0; - glGetActiveUniform(mProgramObject, index, 1024, &length, &size, &type, (GLchar *)name); + glGetActiveUniform(mProgramObject, index, 1024, &length, &size, &type, (GLchar*)name); if (size > 0) { - switch(type) - { - case GL_FLOAT_VEC2: size *= 2; break; - case GL_FLOAT_VEC3: size *= 3; break; - case GL_FLOAT_VEC4: size *= 4; break; - case GL_DOUBLE: size *= 2; break; - case GL_DOUBLE_VEC2: size *= 2; break; - case GL_DOUBLE_VEC3: size *= 6; break; - case GL_DOUBLE_VEC4: size *= 8; break; - case GL_INT_VEC2: size *= 2; break; - case GL_INT_VEC3: size *= 3; break; - case GL_INT_VEC4: size *= 4; break; - case GL_UNSIGNED_INT_VEC2: size *= 2; break; - case GL_UNSIGNED_INT_VEC3: size *= 3; break; - case GL_UNSIGNED_INT_VEC4: size *= 4; break; - case GL_BOOL_VEC2: size *= 2; break; - case GL_BOOL_VEC3: size *= 3; break; - case GL_BOOL_VEC4: size *= 4; break; - case GL_FLOAT_MAT2: size *= 4; break; - case GL_FLOAT_MAT3: size *= 9; break; - case GL_FLOAT_MAT4: size *= 16; break; - case GL_FLOAT_MAT2x3: size *= 6; break; - case GL_FLOAT_MAT2x4: size *= 8; break; - case GL_FLOAT_MAT3x2: size *= 6; break; - case GL_FLOAT_MAT3x4: size *= 12; break; - case GL_FLOAT_MAT4x2: size *= 8; break; - case GL_FLOAT_MAT4x3: size *= 12; break; - case GL_DOUBLE_MAT2: size *= 8; break; - case GL_DOUBLE_MAT3: size *= 18; break; - case GL_DOUBLE_MAT4: size *= 32; break; - case GL_DOUBLE_MAT2x3: size *= 12; break; - case GL_DOUBLE_MAT2x4: size *= 16; break; - case GL_DOUBLE_MAT3x2: size *= 12; break; - case GL_DOUBLE_MAT3x4: size *= 24; break; - case GL_DOUBLE_MAT4x2: size *= 16; break; - case GL_DOUBLE_MAT4x3: size *= 24; break; + switch (type) + { + case GL_FLOAT_VEC2: size *= 2; break; + case GL_FLOAT_VEC3: size *= 3; break; + case GL_FLOAT_VEC4: size *= 4; break; + case GL_DOUBLE: size *= 2; break; + case GL_DOUBLE_VEC2: size *= 2; break; + case GL_DOUBLE_VEC3: size *= 6; break; + case GL_DOUBLE_VEC4: size *= 8; break; + case GL_INT_VEC2: size *= 2; break; + case GL_INT_VEC3: size *= 3; break; + case GL_INT_VEC4: size *= 4; break; + case GL_UNSIGNED_INT_VEC2: size *= 2; break; + case GL_UNSIGNED_INT_VEC3: size *= 3; break; + case GL_UNSIGNED_INT_VEC4: size *= 4; break; + case GL_BOOL_VEC2: size *= 2; break; + case GL_BOOL_VEC3: size *= 3; break; + case GL_BOOL_VEC4: size *= 4; break; + case GL_FLOAT_MAT2: size *= 4; break; + case GL_FLOAT_MAT3: size *= 9; break; + case GL_FLOAT_MAT4: size *= 16; break; + case GL_FLOAT_MAT2x3: size *= 6; break; + case GL_FLOAT_MAT2x4: size *= 8; break; + case GL_FLOAT_MAT3x2: size *= 6; break; + case GL_FLOAT_MAT3x4: size *= 12; break; + case GL_FLOAT_MAT4x2: size *= 8; break; + case GL_FLOAT_MAT4x3: size *= 12; break; + case GL_DOUBLE_MAT2: size *= 8; break; + case GL_DOUBLE_MAT3: size *= 18; break; + case GL_DOUBLE_MAT4: size *= 32; break; + case GL_DOUBLE_MAT2x3: size *= 12; break; + case GL_DOUBLE_MAT2x4: size *= 16; break; + case GL_DOUBLE_MAT3x2: size *= 12; break; + case GL_DOUBLE_MAT3x4: size *= 24; break; + case GL_DOUBLE_MAT4x2: size *= 16; break; + case GL_DOUBLE_MAT4x3: size *= 24; break; } mTotalUniformSize += size; } @@ -750,11 +708,11 @@ void LLGLSLShader::mapUniform(GLint index, const vector<LLStaticHashedString> * mUniformMap[hashedName] = location; LL_DEBUGS("ShaderUniform") << "Uniform " << name << " is at location " << location << LL_ENDL; - + //find the index of this uniform - for (S32 i = 0; i < (S32) LLShaderMgr::instance()->mReservedUniforms.size(); i++) + for (S32 i = 0; i < (S32)LLShaderMgr::instance()->mReservedUniforms.size(); i++) { - if ( (mUniform[i] == -1) + if ((mUniform[i] == -1) && (LLShaderMgr::instance()->mReservedUniforms[i] == name)) { //found it @@ -768,12 +726,12 @@ void LLGLSLShader::mapUniform(GLint index, const vector<LLStaticHashedString> * { for (U32 i = 0; i < uniforms->size(); i++) { - if ( (mUniform[i+LLShaderMgr::instance()->mReservedUniforms.size()] == -1) + if ((mUniform[i + LLShaderMgr::instance()->mReservedUniforms.size()] == -1) && ((*uniforms)[i].String() == name)) { //found it - mUniform[i+LLShaderMgr::instance()->mReservedUniforms.size()] = location; - mTexture[i+LLShaderMgr::instance()->mReservedUniforms.size()] = mapUniformTextureChannel(location, type, size); + mUniform[i + LLShaderMgr::instance()->mReservedUniforms.size()] = location; + mTexture[i + LLShaderMgr::instance()->mReservedUniforms.size()] = mapUniformTextureChannel(location, type, size); return; } } @@ -791,9 +749,9 @@ void LLGLSLShader::addPermutation(std::string name, std::string value) mDefines[name] = value; } -void LLGLSLShader::addConstant( const LLGLSLShader::eShaderConsts shader_const ) +void LLGLSLShader::addConstant(const LLGLSLShader::eShaderConsts shader_const) { - addPermutation( gShaderConstsKey[ shader_const ], gShaderConstsVal[ shader_const ] ); + addPermutation(gShaderConstsKey[shader_const], gShaderConstsVal[shader_const]); } void LLGLSLShader::removePermutation(std::string name) @@ -805,7 +763,7 @@ GLint LLGLSLShader::mapUniformTextureChannel(GLint location, GLenum type, GLint { LL_PROFILE_ZONE_SCOPED_CATEGORY_SHADER; - if ((type >= GL_SAMPLER_1D && type <= GL_SAMPLER_2D_RECT_SHADOW) || + if ((type >= GL_SAMPLER_1D && type <= GL_SAMPLER_2D_RECT_SHADOW) || type == GL_SAMPLER_2D_MULTISAMPLE || type == GL_SAMPLER_CUBE_MAP_ARRAY) { //this here is a texture @@ -827,8 +785,8 @@ GLint LLGLSLShader::mapUniformTextureChannel(GLint location, GLenum type, GLint channel[i] = mActiveTextureChannels++; } glUniform1iv(location, size, channel); - LL_DEBUGS("ShaderUniform") << "Assigned to texture channel " << - (mActiveTextureChannels-size) << " through " << (mActiveTextureChannels-1) << LL_ENDL; + LL_DEBUGS("ShaderUniform") << "Assigned to texture channel " << + (mActiveTextureChannels - size) << " through " << (mActiveTextureChannels - 1) << LL_ENDL; } llassert(mActiveTextureChannels <= 32); // too many textures (probably) @@ -837,102 +795,102 @@ GLint LLGLSLShader::mapUniformTextureChannel(GLint location, GLenum type, GLint return -1; } -BOOL LLGLSLShader::mapUniforms(const vector<LLStaticHashedString> * uniforms) +BOOL LLGLSLShader::mapUniforms(const vector<LLStaticHashedString>* uniforms) { LL_PROFILE_ZONE_SCOPED_CATEGORY_SHADER; BOOL res = TRUE; - mTotalUniformSize = 0; - mActiveTextureChannels = 0; - mUniform.clear(); - mUniformMap.clear(); - mUniformNameMap.clear(); - mTexture.clear(); - mValue.clear(); - //initialize arrays - U32 numUniforms = (uniforms == NULL) ? 0 : uniforms->size(); - mUniform.resize(numUniforms + LLShaderMgr::instance()->mReservedUniforms.size(), -1); - mTexture.resize(numUniforms + LLShaderMgr::instance()->mReservedUniforms.size(), -1); - - bind(); - - //get the number of active uniforms - GLint activeCount; + mTotalUniformSize = 0; + mActiveTextureChannels = 0; + mUniform.clear(); + mUniformMap.clear(); + mUniformNameMap.clear(); + mTexture.clear(); + mValue.clear(); + //initialize arrays + U32 numUniforms = (uniforms == NULL) ? 0 : uniforms->size(); + mUniform.resize(numUniforms + LLShaderMgr::instance()->mReservedUniforms.size(), -1); + mTexture.resize(numUniforms + LLShaderMgr::instance()->mReservedUniforms.size(), -1); + + bind(); + + //get the number of active uniforms + GLint activeCount; glGetProgramiv(mProgramObject, GL_ACTIVE_UNIFORMS, &activeCount); - //........................................................................................................................................ - //........................................................................................ + //........................................................................................................................................ + //........................................................................................ - /* - EXPLANATION: - This is part of code is temporary because as the final result the mapUniform() should be rewrited. - But it's a huge a volume of work which is need to be a more carefully performed for avoid possible - regression's (i.e. it should be formalized a separate ticket in JIRA). + /* + EXPLANATION: + This is part of code is temporary because as the final result the mapUniform() should be rewrited. + But it's a huge a volume of work which is need to be a more carefully performed for avoid possible + regression's (i.e. it should be formalized a separate ticket in JIRA). - RESON: - The reason of this code is that SL engine is very sensitive to fact that "diffuseMap" should be appear - first as uniform parameter which is should get 0-"texture channel" index (see mapUniformTextureChannel() and mActiveTextureChannels) - it influence to which is texture matrix will be updated during rendering. + RESON: + The reason of this code is that SL engine is very sensitive to fact that "diffuseMap" should be appear + first as uniform parameter which is should get 0-"texture channel" index (see mapUniformTextureChannel() and mActiveTextureChannels) + it influence to which is texture matrix will be updated during rendering. - But, order of indexe's of uniform variables is not defined and GLSL compiler can change it as want - , even if the "diffuseMap" will be appear and use first in shader code. + But, order of indexe's of uniform variables is not defined and GLSL compiler can change it as want + , even if the "diffuseMap" will be appear and use first in shader code. - As example where this situation appear see: "Deferred Material Shader 28/29/30/31" - And tickets: MAINT-4165, MAINT-4839, MAINT-3568, MAINT-6437 + As example where this situation appear see: "Deferred Material Shader 28/29/30/31" + And tickets: MAINT-4165, MAINT-4839, MAINT-3568, MAINT-6437 --- davep TODO -- pretty sure the entire block here is superstitious and that the uniform index has nothing to do with the texture channel texture channel should follow the uniform VALUE - */ + */ - S32 diffuseMap = glGetUniformLocation(mProgramObject, "diffuseMap"); - S32 specularMap = glGetUniformLocation(mProgramObject, "specularMap"); - S32 bumpMap = glGetUniformLocation(mProgramObject, "bumpMap"); + S32 diffuseMap = glGetUniformLocation(mProgramObject, "diffuseMap"); + S32 specularMap = glGetUniformLocation(mProgramObject, "specularMap"); + S32 bumpMap = glGetUniformLocation(mProgramObject, "bumpMap"); S32 altDiffuseMap = glGetUniformLocation(mProgramObject, "altDiffuseMap"); - S32 environmentMap = glGetUniformLocation(mProgramObject, "environmentMap"); + S32 environmentMap = glGetUniformLocation(mProgramObject, "environmentMap"); S32 reflectionMap = glGetUniformLocation(mProgramObject, "reflectionMap"); - std::set<S32> skip_index; + std::set<S32> skip_index; - if (-1 != diffuseMap && (-1 != specularMap || -1 != bumpMap || -1 != environmentMap || -1 != altDiffuseMap)) - { - GLenum type; - GLsizei length; - GLint size = -1; - char name[1024]; + if (-1 != diffuseMap && (-1 != specularMap || -1 != bumpMap || -1 != environmentMap || -1 != altDiffuseMap)) + { + GLenum type; + GLsizei length; + GLint size = -1; + char name[1024]; - diffuseMap = altDiffuseMap = specularMap = bumpMap = environmentMap = -1; + diffuseMap = altDiffuseMap = specularMap = bumpMap = environmentMap = -1; - for (S32 i = 0; i < activeCount; i++) - { - name[0] = '\0'; + for (S32 i = 0; i < activeCount; i++) + { + name[0] = '\0'; - glGetActiveUniform(mProgramObject, i, 1024, &length, &size, &type, (GLchar *)name); + glGetActiveUniform(mProgramObject, i, 1024, &length, &size, &type, (GLchar*)name); - if (-1 == diffuseMap && std::string(name) == "diffuseMap") - { - diffuseMap = i; - continue; - } + if (-1 == diffuseMap && std::string(name) == "diffuseMap") + { + diffuseMap = i; + continue; + } - if (-1 == specularMap && std::string(name) == "specularMap") - { - specularMap = i; - continue; - } + if (-1 == specularMap && std::string(name) == "specularMap") + { + specularMap = i; + continue; + } - if (-1 == bumpMap && std::string(name) == "bumpMap") - { - bumpMap = i; - continue; - } + if (-1 == bumpMap && std::string(name) == "bumpMap") + { + bumpMap = i; + continue; + } - if (-1 == environmentMap && std::string(name) == "environmentMap") - { - environmentMap = i; - continue; - } + if (-1 == environmentMap && std::string(name) == "environmentMap") + { + environmentMap = i; + continue; + } if (-1 == reflectionMap && std::string(name) == "reflectionMap") { @@ -941,55 +899,55 @@ BOOL LLGLSLShader::mapUniforms(const vector<LLStaticHashedString> * uniforms) } if (-1 == altDiffuseMap && std::string(name) == "altDiffuseMap") - { - altDiffuseMap = i; - continue; - } - } - - bool specularDiff = specularMap < diffuseMap && -1 != specularMap; - bool bumpLessDiff = bumpMap < diffuseMap && -1 != bumpMap; - bool envLessDiff = environmentMap < diffuseMap && -1 != environmentMap; + { + altDiffuseMap = i; + continue; + } + } + + bool specularDiff = specularMap < diffuseMap && -1 != specularMap; + bool bumpLessDiff = bumpMap < diffuseMap && -1 != bumpMap; + bool envLessDiff = environmentMap < diffuseMap && -1 != environmentMap; bool refLessDiff = reflectionMap < diffuseMap && -1 != reflectionMap; - if (specularDiff || bumpLessDiff || envLessDiff || refLessDiff) - { - mapUniform(diffuseMap, uniforms); - skip_index.insert(diffuseMap); + if (specularDiff || bumpLessDiff || envLessDiff || refLessDiff) + { + mapUniform(diffuseMap, uniforms); + skip_index.insert(diffuseMap); - if (-1 != specularMap) { - mapUniform(specularMap, uniforms); - skip_index.insert(specularMap); - } + if (-1 != specularMap) { + mapUniform(specularMap, uniforms); + skip_index.insert(specularMap); + } - if (-1 != bumpMap) { - mapUniform(bumpMap, uniforms); - skip_index.insert(bumpMap); - } + if (-1 != bumpMap) { + mapUniform(bumpMap, uniforms); + skip_index.insert(bumpMap); + } - if (-1 != environmentMap) { - mapUniform(environmentMap, uniforms); - skip_index.insert(environmentMap); - } + if (-1 != environmentMap) { + mapUniform(environmentMap, uniforms); + skip_index.insert(environmentMap); + } if (-1 != reflectionMap) { mapUniform(reflectionMap, uniforms); skip_index.insert(reflectionMap); } - } - } + } + } - //........................................................................................ + //........................................................................................ - for (S32 i = 0; i < activeCount; i++) - { - //........................................................................................ - if (skip_index.end() != skip_index.find(i)) continue; - //........................................................................................ + for (S32 i = 0; i < activeCount; i++) + { + //........................................................................................ + if (skip_index.end() != skip_index.find(i)) continue; + //........................................................................................ - mapUniform(i, uniforms); - } - //........................................................................................................................................ + mapUniform(i, uniforms); + } + //........................................................................................................................................ if (mFeatures.hasReflectionProbes) // Set up block binding, in a way supported by Apple (rather than binding = 1 in .glsl). { // See slide 35 and more of https://docs.huihoo.com/apple/wwdc/2011/session_420__advances_in_opengl_for_mac_os_x_lion.pdf @@ -1002,10 +960,10 @@ BOOL LLGLSLShader::mapUniforms(const vector<LLStaticHashedString> * uniforms) glUniformBlockBinding(mProgramObject, UBOBlockIndex, BLOCKBINDING); } } - unbind(); + unbind(); - LL_DEBUGS("ShaderUniform") << "Total Uniform Size: " << mTotalUniformSize << LL_ENDL; - return res; + LL_DEBUGS("ShaderUniform") << "Total Uniform Size: " << mTotalUniformSize << LL_ENDL; + return res; } @@ -1031,10 +989,15 @@ void LLGLSLShader::bind() if (sCurBoundShader != mProgramObject) // Don't re-bind current shader { + if (sCurBoundShaderPtr) + { + sCurBoundShaderPtr->readProfileQuery(); + } LLVertexBuffer::unbind(); glUseProgram(mProgramObject); sCurBoundShader = mProgramObject; sCurBoundShaderPtr = this; + placeProfileQuery(); } if (mUniformsDirty) @@ -1060,8 +1023,11 @@ void LLGLSLShader::bind(bool rigged) void LLGLSLShader::unbind() { LL_PROFILE_ZONE_SCOPED_CATEGORY_SHADER; - gGL.flush(); + if (sCurBoundShaderPtr) + { + sCurBoundShaderPtr->readProfileQuery(); + } stop_glerror(); LLVertexBuffer::unbind(); glUseProgram(0); @@ -1075,22 +1041,28 @@ void LLGLSLShader::bindNoShader(void) LL_PROFILE_ZONE_SCOPED_CATEGORY_SHADER; LLVertexBuffer::unbind(); + + if (sCurBoundShaderPtr) + { + sCurBoundShaderPtr->readProfileQuery(); + } + glUseProgram(0); sCurBoundShader = 0; sCurBoundShaderPtr = NULL; } -S32 LLGLSLShader::bindTexture(const std::string &uniform, LLTexture *texture, LLTexUnit::eTextureType mode, LLTexUnit::eTextureColorSpace colorspace) +S32 LLGLSLShader::bindTexture(const std::string& uniform, LLTexture* texture, LLTexUnit::eTextureType mode, LLTexUnit::eTextureColorSpace colorspace) { LL_PROFILE_ZONE_SCOPED_CATEGORY_SHADER; S32 channel = 0; channel = getUniformLocation(uniform); - + return bindTexture(channel, texture, mode, colorspace); } -S32 LLGLSLShader::bindTexture(S32 uniform, LLTexture *texture, LLTexUnit::eTextureType mode, LLTexUnit::eTextureColorSpace colorspace) +S32 LLGLSLShader::bindTexture(S32 uniform, LLTexture* texture, LLTexUnit::eTextureType mode, LLTexUnit::eTextureColorSpace colorspace) { LL_PROFILE_ZONE_SCOPED_CATEGORY_SHADER; @@ -1099,25 +1071,25 @@ S32 LLGLSLShader::bindTexture(S32 uniform, LLTexture *texture, LLTexUnit::eTextu LL_SHADER_UNIFORM_ERRS() << "Uniform out of range: " << uniform << LL_ENDL; return -1; } - + uniform = mTexture[uniform]; - + if (uniform > -1) { gGL.getTexUnit(uniform)->bindFast(texture); gGL.getTexUnit(uniform)->setTextureColorSpace(colorspace); } - + return uniform; } -S32 LLGLSLShader::unbindTexture(const std::string &uniform, LLTexUnit::eTextureType mode) +S32 LLGLSLShader::unbindTexture(const std::string& uniform, LLTexUnit::eTextureType mode) { LL_PROFILE_ZONE_SCOPED_CATEGORY_SHADER; S32 channel = 0; channel = getUniformLocation(uniform); - + return unbindTexture(channel); } @@ -1130,14 +1102,14 @@ S32 LLGLSLShader::unbindTexture(S32 uniform, LLTexUnit::eTextureType mode) LL_SHADER_UNIFORM_ERRS() << "Uniform out of range: " << uniform << LL_ENDL; return -1; } - + uniform = mTexture[uniform]; - + if (uniform > -1) { gGL.getTexUnit(uniform)->unbindFast(mode); } - + return uniform; } @@ -1194,7 +1166,7 @@ void LLGLSLShader::uniform1i(U32 index, GLint x) LL_PROFILE_ZONE_SCOPED_CATEGORY_SHADER; llassert(sCurBoundShaderPtr == this); if (mProgramObject) - { + { if (mUniform.size() <= index) { LL_SHADER_UNIFORM_ERRS() << "Uniform index out of bounds." << LL_ENDL; @@ -1207,7 +1179,7 @@ void LLGLSLShader::uniform1i(U32 index, GLint x) if (iter == mValue.end() || iter->second.mV[0] != x) { glUniform1i(mUniform[index], x); - mValue[mUniform[index]] = LLVector4(x,0.f,0.f,0.f); + mValue[mUniform[index]] = LLVector4(x, 0.f, 0.f, 0.f); } } } @@ -1219,7 +1191,7 @@ void LLGLSLShader::uniform1f(U32 index, GLfloat x) llassert(sCurBoundShaderPtr == this); if (mProgramObject) - { + { if (mUniform.size() <= index) { LL_SHADER_UNIFORM_ERRS() << "Uniform index out of bounds." << LL_ENDL; @@ -1232,7 +1204,7 @@ void LLGLSLShader::uniform1f(U32 index, GLfloat x) if (iter == mValue.end() || iter->second.mV[0] != x) { glUniform1f(mUniform[index], x); - mValue[mUniform[index]] = LLVector4(x,0.f,0.f,0.f); + mValue[mUniform[index]] = LLVector4(x, 0.f, 0.f, 0.f); } } } @@ -1254,7 +1226,7 @@ void LLGLSLShader::uniform2f(U32 index, GLfloat x, GLfloat y) llassert(sCurBoundShaderPtr == this); if (mProgramObject) - { + { if (mUniform.size() <= index) { LL_SHADER_UNIFORM_ERRS() << "Uniform index out of bounds." << LL_ENDL; @@ -1264,8 +1236,8 @@ void LLGLSLShader::uniform2f(U32 index, GLfloat x, GLfloat y) if (mUniform[index] >= 0) { const auto& iter = mValue.find(mUniform[index]); - LLVector4 vec(x,y,0.f,0.f); - if (iter == mValue.end() || shouldChange(iter->second,vec)) + LLVector4 vec(x, y, 0.f, 0.f); + if (iter == mValue.end() || shouldChange(iter->second, vec)) { glUniform2f(mUniform[index], x, y); mValue[mUniform[index]] = vec; @@ -1280,7 +1252,7 @@ void LLGLSLShader::uniform3f(U32 index, GLfloat x, GLfloat y, GLfloat z) llassert(sCurBoundShaderPtr == this); if (mProgramObject) - { + { if (mUniform.size() <= index) { LL_SHADER_UNIFORM_ERRS() << "Uniform index out of bounds." << LL_ENDL; @@ -1290,8 +1262,8 @@ void LLGLSLShader::uniform3f(U32 index, GLfloat x, GLfloat y, GLfloat z) if (mUniform[index] >= 0) { const auto& iter = mValue.find(mUniform[index]); - LLVector4 vec(x,y,z,0.f); - if (iter == mValue.end() || shouldChange(iter->second,vec)) + LLVector4 vec(x, y, z, 0.f); + if (iter == mValue.end() || shouldChange(iter->second, vec)) { glUniform3f(mUniform[index], x, y, z); mValue[mUniform[index]] = vec; @@ -1306,7 +1278,7 @@ void LLGLSLShader::uniform4f(U32 index, GLfloat x, GLfloat y, GLfloat z, GLfloat llassert(sCurBoundShaderPtr == this); if (mProgramObject) - { + { if (mUniform.size() <= index) { LL_SHADER_UNIFORM_ERRS() << "Uniform index out of bounds." << LL_ENDL; @@ -1316,8 +1288,8 @@ void LLGLSLShader::uniform4f(U32 index, GLfloat x, GLfloat y, GLfloat z, GLfloat if (mUniform[index] >= 0) { const auto& iter = mValue.find(mUniform[index]); - LLVector4 vec(x,y,z,w); - if (iter == mValue.end() || shouldChange(iter->second,vec)) + LLVector4 vec(x, y, z, w); + if (iter == mValue.end() || shouldChange(iter->second, vec)) { glUniform4f(mUniform[index], x, y, z, w); mValue[mUniform[index]] = vec; @@ -1332,7 +1304,7 @@ void LLGLSLShader::uniform1iv(U32 index, U32 count, const GLint* v) llassert(sCurBoundShaderPtr == this); if (mProgramObject) - { + { if (mUniform.size() <= index) { LL_SHADER_UNIFORM_ERRS() << "Uniform index out of bounds." << LL_ENDL; @@ -1342,8 +1314,8 @@ void LLGLSLShader::uniform1iv(U32 index, U32 count, const GLint* v) if (mUniform[index] >= 0) { const auto& iter = mValue.find(mUniform[index]); - LLVector4 vec(v[0],0.f,0.f,0.f); - if (iter == mValue.end() || shouldChange(iter->second,vec) || count != 1) + LLVector4 vec(v[0], 0.f, 0.f, 0.f); + if (iter == mValue.end() || shouldChange(iter->second, vec) || count != 1) { glUniform1iv(mUniform[index], count, v); mValue[mUniform[index]] = vec; @@ -1385,7 +1357,7 @@ void LLGLSLShader::uniform1fv(U32 index, U32 count, const GLfloat* v) llassert(sCurBoundShaderPtr == this); if (mProgramObject) - { + { if (mUniform.size() <= index) { LL_SHADER_UNIFORM_ERRS() << "Uniform index out of bounds." << LL_ENDL; @@ -1395,8 +1367,8 @@ void LLGLSLShader::uniform1fv(U32 index, U32 count, const GLfloat* v) if (mUniform[index] >= 0) { const auto& iter = mValue.find(mUniform[index]); - LLVector4 vec(v[0],0.f,0.f,0.f); - if (iter == mValue.end() || shouldChange(iter->second,vec) || count != 1) + LLVector4 vec(v[0], 0.f, 0.f, 0.f); + if (iter == mValue.end() || shouldChange(iter->second, vec) || count != 1) { glUniform1fv(mUniform[index], count, v); mValue[mUniform[index]] = vec; @@ -1411,7 +1383,7 @@ void LLGLSLShader::uniform2fv(U32 index, U32 count, const GLfloat* v) llassert(sCurBoundShaderPtr == this); if (mProgramObject) - { + { if (mUniform.size() <= index) { LL_SHADER_UNIFORM_ERRS() << "Uniform index out of bounds." << LL_ENDL; @@ -1421,8 +1393,8 @@ void LLGLSLShader::uniform2fv(U32 index, U32 count, const GLfloat* v) if (mUniform[index] >= 0) { const auto& iter = mValue.find(mUniform[index]); - LLVector4 vec(v[0],v[1],0.f,0.f); - if (iter == mValue.end() || shouldChange(iter->second,vec) || count != 1) + LLVector4 vec(v[0], v[1], 0.f, 0.f); + if (iter == mValue.end() || shouldChange(iter->second, vec) || count != 1) { glUniform2fv(mUniform[index], count, v); mValue[mUniform[index]] = vec; @@ -1437,7 +1409,7 @@ void LLGLSLShader::uniform3fv(U32 index, U32 count, const GLfloat* v) llassert(sCurBoundShaderPtr == this); if (mProgramObject) - { + { if (mUniform.size() <= index) { LL_SHADER_UNIFORM_ERRS() << "Uniform index out of bounds." << LL_ENDL; @@ -1447,8 +1419,8 @@ void LLGLSLShader::uniform3fv(U32 index, U32 count, const GLfloat* v) if (mUniform[index] >= 0) { const auto& iter = mValue.find(mUniform[index]); - LLVector4 vec(v[0],v[1],v[2],0.f); - if (iter == mValue.end() || shouldChange(iter->second,vec) || count != 1) + LLVector4 vec(v[0], v[1], v[2], 0.f); + if (iter == mValue.end() || shouldChange(iter->second, vec) || count != 1) { glUniform3fv(mUniform[index], count, v); mValue[mUniform[index]] = vec; @@ -1463,7 +1435,7 @@ void LLGLSLShader::uniform4fv(U32 index, U32 count, const GLfloat* v) llassert(sCurBoundShaderPtr == this); if (mProgramObject) - { + { if (mUniform.size() <= index) { LL_SHADER_UNIFORM_ERRS() << "Uniform index out of bounds." << LL_ENDL; @@ -1473,8 +1445,8 @@ void LLGLSLShader::uniform4fv(U32 index, U32 count, const GLfloat* v) if (mUniform[index] >= 0) { const auto& iter = mValue.find(mUniform[index]); - LLVector4 vec(v[0],v[1],v[2],v[3]); - if (iter == mValue.end() || shouldChange(iter->second,vec) || count != 1) + LLVector4 vec(v[0], v[1], v[2], v[3]); + if (iter == mValue.end() || shouldChange(iter->second, vec) || count != 1) { LL_PROFILE_ZONE_SCOPED_CATEGORY_SHADER; glUniform4fv(mUniform[index], count, v); @@ -1484,13 +1456,13 @@ void LLGLSLShader::uniform4fv(U32 index, U32 count, const GLfloat* v) } } -void LLGLSLShader::uniformMatrix2fv(U32 index, U32 count, GLboolean transpose, const GLfloat *v) +void LLGLSLShader::uniformMatrix2fv(U32 index, U32 count, GLboolean transpose, const GLfloat* v) { LL_PROFILE_ZONE_SCOPED_CATEGORY_SHADER; llassert(sCurBoundShaderPtr == this); if (mProgramObject) - { + { if (mUniform.size() <= index) { LL_SHADER_UNIFORM_ERRS() << "Uniform index out of bounds." << LL_ENDL; @@ -1504,13 +1476,13 @@ void LLGLSLShader::uniformMatrix2fv(U32 index, U32 count, GLboolean transpose, c } } -void LLGLSLShader::uniformMatrix3fv(U32 index, U32 count, GLboolean transpose, const GLfloat *v) +void LLGLSLShader::uniformMatrix3fv(U32 index, U32 count, GLboolean transpose, const GLfloat* v) { LL_PROFILE_ZONE_SCOPED_CATEGORY_SHADER; llassert(sCurBoundShaderPtr == this); if (mProgramObject) - { + { if (mUniform.size() <= index) { LL_SHADER_UNIFORM_ERRS() << "Uniform index out of bounds." << LL_ENDL; @@ -1524,33 +1496,33 @@ void LLGLSLShader::uniformMatrix3fv(U32 index, U32 count, GLboolean transpose, c } } -void LLGLSLShader::uniformMatrix3x4fv(U32 index, U32 count, GLboolean transpose, const GLfloat *v) +void LLGLSLShader::uniformMatrix3x4fv(U32 index, U32 count, GLboolean transpose, const GLfloat* v) { LL_PROFILE_ZONE_SCOPED_CATEGORY_SHADER; llassert(sCurBoundShaderPtr == this); - if (mProgramObject) - { - if (mUniform.size() <= index) - { - LL_SHADER_UNIFORM_ERRS() << "Uniform index out of bounds." << LL_ENDL; - return; - } + if (mProgramObject) + { + if (mUniform.size() <= index) + { + LL_SHADER_UNIFORM_ERRS() << "Uniform index out of bounds." << LL_ENDL; + return; + } - if (mUniform[index] >= 0) - { - glUniformMatrix3x4fv(mUniform[index], count, transpose, v); - } - } + if (mUniform[index] >= 0) + { + glUniformMatrix3x4fv(mUniform[index], count, transpose, v); + } + } } -void LLGLSLShader::uniformMatrix4fv(U32 index, U32 count, GLboolean transpose, const GLfloat *v) +void LLGLSLShader::uniformMatrix4fv(U32 index, U32 count, GLboolean transpose, const GLfloat* v) { LL_PROFILE_ZONE_SCOPED_CATEGORY_SHADER; llassert(sCurBoundShaderPtr == this); if (mProgramObject) - { + { if (mUniform.size() <= index) { LL_SHADER_UNIFORM_ERRS() << "Uniform index out of bounds." << LL_ENDL; @@ -1626,12 +1598,12 @@ void LLGLSLShader::uniform1i(const LLStaticHashedString& uniform, GLint v) { LL_PROFILE_ZONE_SCOPED_CATEGORY_SHADER; GLint location = getUniformLocation(uniform); - + if (location >= 0) { const auto& iter = mValue.find(location); - LLVector4 vec(v,0.f,0.f,0.f); - if (iter == mValue.end() || shouldChange(iter->second,vec)) + LLVector4 vec(v, 0.f, 0.f, 0.f); + if (iter == mValue.end() || shouldChange(iter->second, vec)) { glUniform1i(location, v); mValue[location] = vec; @@ -1679,12 +1651,12 @@ void LLGLSLShader::uniform2i(const LLStaticHashedString& uniform, GLint i, GLint { LL_PROFILE_ZONE_SCOPED_CATEGORY_SHADER; GLint location = getUniformLocation(uniform); - + if (location >= 0) { const auto& iter = mValue.find(location); - LLVector4 vec(i,j,0.f,0.f); - if (iter == mValue.end() || shouldChange(iter->second,vec)) + LLVector4 vec(i, j, 0.f, 0.f); + if (iter == mValue.end() || shouldChange(iter->second, vec)) { glUniform2i(location, i, j); mValue[location] = vec; @@ -1697,12 +1669,12 @@ void LLGLSLShader::uniform1f(const LLStaticHashedString& uniform, GLfloat v) { LL_PROFILE_ZONE_SCOPED_CATEGORY_SHADER; GLint location = getUniformLocation(uniform); - + if (location >= 0) { const auto& iter = mValue.find(location); - LLVector4 vec(v,0.f,0.f,0.f); - if (iter == mValue.end() || shouldChange(iter->second,vec)) + LLVector4 vec(v, 0.f, 0.f, 0.f); + if (iter == mValue.end() || shouldChange(iter->second, vec)) { glUniform1f(location, v); mValue[location] = vec; @@ -1714,14 +1686,14 @@ void LLGLSLShader::uniform2f(const LLStaticHashedString& uniform, GLfloat x, GLf { LL_PROFILE_ZONE_SCOPED_CATEGORY_SHADER; GLint location = getUniformLocation(uniform); - + if (location >= 0) { const auto& iter = mValue.find(location); - LLVector4 vec(x,y,0.f,0.f); - if (iter == mValue.end() || shouldChange(iter->second,vec)) + LLVector4 vec(x, y, 0.f, 0.f); + if (iter == mValue.end() || shouldChange(iter->second, vec)) { - glUniform2f(location, x,y); + glUniform2f(location, x, y); mValue[location] = vec; } } @@ -1732,14 +1704,14 @@ void LLGLSLShader::uniform3f(const LLStaticHashedString& uniform, GLfloat x, GLf { LL_PROFILE_ZONE_SCOPED_CATEGORY_SHADER; GLint location = getUniformLocation(uniform); - + if (location >= 0) { const auto& iter = mValue.find(location); - LLVector4 vec(x,y,z,0.f); - if (iter == mValue.end() || shouldChange(iter->second,vec)) + LLVector4 vec(x, y, z, 0.f); + if (iter == mValue.end() || shouldChange(iter->second, vec)) { - glUniform3f(location, x,y,z); + glUniform3f(location, x, y, z); mValue[location] = vec; } } @@ -1753,8 +1725,8 @@ void LLGLSLShader::uniform1fv(const LLStaticHashedString& uniform, U32 count, co if (location >= 0) { const auto& iter = mValue.find(location); - LLVector4 vec(v[0],0.f,0.f,0.f); - if (iter == mValue.end() || shouldChange(iter->second,vec) || count != 1) + LLVector4 vec(v[0], 0.f, 0.f, 0.f); + if (iter == mValue.end() || shouldChange(iter->second, vec) || count != 1) { glUniform1fv(location, count, v); mValue[location] = vec; @@ -1766,12 +1738,12 @@ void LLGLSLShader::uniform2fv(const LLStaticHashedString& uniform, U32 count, co { LL_PROFILE_ZONE_SCOPED_CATEGORY_SHADER; GLint location = getUniformLocation(uniform); - + if (location >= 0) { const auto& iter = mValue.find(location); - LLVector4 vec(v[0],v[1],0.f,0.f); - if (iter == mValue.end() || shouldChange(iter->second,vec) || count != 1) + LLVector4 vec(v[0], v[1], 0.f, 0.f); + if (iter == mValue.end() || shouldChange(iter->second, vec) || count != 1) { glUniform2fv(location, count, v); mValue[location] = vec; @@ -1783,12 +1755,12 @@ void LLGLSLShader::uniform3fv(const LLStaticHashedString& uniform, U32 count, co { LL_PROFILE_ZONE_SCOPED_CATEGORY_SHADER; GLint location = getUniformLocation(uniform); - + if (location >= 0) { const auto& iter = mValue.find(location); - LLVector4 vec(v[0],v[1],v[2],0.f); - if (iter == mValue.end() || shouldChange(iter->second,vec) || count != 1) + LLVector4 vec(v[0], v[1], v[2], 0.f); + if (iter == mValue.end() || shouldChange(iter->second, vec) || count != 1) { glUniform3fv(location, count, v); mValue[location] = vec; @@ -1805,7 +1777,7 @@ void LLGLSLShader::uniform4fv(const LLStaticHashedString& uniform, U32 count, co { LLVector4 vec(v); const auto& iter = mValue.find(location); - if (iter == mValue.end() || shouldChange(iter->second,vec) || count != 1) + if (iter == mValue.end() || shouldChange(iter->second, vec) || count != 1) { LL_PROFILE_ZONE_SCOPED_CATEGORY_SHADER; glUniform4fv(location, count, v); @@ -1818,7 +1790,7 @@ void LLGLSLShader::uniformMatrix4fv(const LLStaticHashedString& uniform, U32 cou { LL_PROFILE_ZONE_SCOPED_CATEGORY_SHADER; GLint location = getUniformLocation(uniform); - + if (location >= 0) { stop_glerror(); diff --git a/indra/llrender/llglslshader.h b/indra/llrender/llglslshader.h index 23db1a8549..0522f78614 100644 --- a/indra/llrender/llglslshader.h +++ b/indra/llrender/llglslshader.h @@ -1,25 +1,25 @@ -/** +/** * @file llglslshader.h * @brief GLSL shader wrappers * * $LicenseInfo:firstyear=2001&license=viewerlgpl$ * Second Life Viewer Source Code * Copyright (C) 2010, Linden Research, Inc. - * + * * This library is free software; you can redistribute it and/or * modify it under the terms of the GNU Lesser General Public * License as published by the Free Software Foundation; * version 2.1 of the License only. - * + * * This library 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 * Lesser General Public License for more details. - * + * * You should have received a copy of the GNU Lesser General Public * License along with this library; if not, write to the Free Software * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA - * + * * Linden Research, Inc., 945 Battery Street, San Francisco, CA 94111 USA * $/LicenseInfo$ */ @@ -35,36 +35,36 @@ class LLShaderFeatures { public: - bool atmosphericHelpers; - bool calculatesLighting; - bool calculatesAtmospherics; - bool hasLighting; // implies no transport (it's possible to have neither though) - bool isAlphaLighting; // indicates lighting shaders need not be linked in (lighting performed directly in alpha shader to match deferred lighting functions) - bool isShiny; - bool isFullbright; // implies no lighting - bool isSpecular; - bool hasWaterFog; // implies no gamma - bool hasTransport; // implies no lighting (it's possible to have neither though) - bool hasSkinning; - bool hasObjectSkinning; - bool hasAtmospherics; - bool hasGamma; + bool atmosphericHelpers; + bool calculatesLighting; + bool calculatesAtmospherics; + bool hasLighting; // implies no transport (it's possible to have neither though) + bool isAlphaLighting; // indicates lighting shaders need not be linked in (lighting performed directly in alpha shader to match deferred lighting functions) + bool isShiny; + bool isFullbright; // implies no lighting + bool isSpecular; + bool hasWaterFog; // implies no gamma + bool hasTransport; // implies no lighting (it's possible to have neither though) + bool hasSkinning; + bool hasObjectSkinning; + bool hasAtmospherics; + bool hasGamma; bool hasShadows; bool hasAmbientOcclusion; - bool hasSrgb; + bool hasSrgb; bool encodesNormal; // include: shaders\class1\environment\encodeNormF.glsl bool isDeferred; bool hasScreenSpaceReflections; bool hasIndirect; - S32 mIndexedTextureChannels; - bool disableTextureIndex; - bool hasAlphaMask; + S32 mIndexedTextureChannels; + bool disableTextureIndex; + bool hasAlphaMask; bool hasReflectionProbes = false; - bool attachNothing; + bool attachNothing; - // char numLights; - - LLShaderFeatures(); + // char numLights; + + LLShaderFeatures(); }; // ============= Structure for caching shader uniforms =============== @@ -108,7 +108,7 @@ public: { mVectors.push_back({ index, value }); } - + void uniform4fv(S32 index, const F32* value) { mVectors.push_back({ index, LLVector4(value) }); @@ -125,7 +125,7 @@ public: } void apply(LLGLSLShader* shader); - + std::vector<IntSetting> mIntegers; std::vector<FloatSetting> mFloats; @@ -138,184 +138,182 @@ public: // NOTE: Keep gShaderConsts and LLGLSLShader::ShaderConsts_e in sync! enum eShaderConsts { - SHADER_CONST_CLOUD_MOON_DEPTH + SHADER_CONST_CLOUD_MOON_DEPTH , SHADER_CONST_STAR_DEPTH , NUM_SHADER_CONSTS }; // enum primarily used to control application sky settings uniforms - typedef enum - { - SG_DEFAULT = 0, // not sky or water specific - SG_SKY, // - SG_WATER, + typedef enum + { + SG_DEFAULT = 0, // not sky or water specific + SG_SKY, // + SG_WATER, SG_ANY, SG_COUNT - } eGroup; - - static std::set<LLGLSLShader*> sInstances; - static bool sProfileEnabled; - - LLGLSLShader(); - ~LLGLSLShader(); - - static GLuint sCurBoundShader; - static LLGLSLShader* sCurBoundShaderPtr; - static S32 sIndexedTextureChannels; - - static void initProfile(); - static void finishProfile(bool emit_report = true); - - static void startProfile(); - static void stopProfile(U32 count, U32 mode); - - void unload(); - void clearStats(); - void dumpStats(); - void placeProfileQuery(); - void readProfileQuery(U32 count, U32 mode); - - BOOL createShader(std::vector<LLStaticHashedString> * attributes, - std::vector<LLStaticHashedString> * uniforms, - U32 varying_count = 0, - const char** varyings = NULL); + } eGroup; + + static std::set<LLGLSLShader*> sInstances; + static bool sProfileEnabled; + + LLGLSLShader(); + ~LLGLSLShader(); + + static GLuint sCurBoundShader; + static LLGLSLShader* sCurBoundShaderPtr; + static S32 sIndexedTextureChannels; + + static void initProfile(); + static void finishProfile(bool emit_report = true); + + static void startProfile(); + static void stopProfile(); + + void unload(); + void clearStats(); + void dumpStats(); + void placeProfileQuery(); + void readProfileQuery(); + + BOOL createShader(std::vector<LLStaticHashedString>* attributes, + std::vector<LLStaticHashedString>* uniforms, + U32 varying_count = 0, + const char** varyings = NULL); BOOL attachFragmentObject(std::string object); BOOL attachVertexObject(std::string object); - void attachObject(GLuint object); - void attachObjects(GLuint* objects = NULL, S32 count = 0); - BOOL mapAttributes(const std::vector<LLStaticHashedString> * attributes); - BOOL mapUniforms(const std::vector<LLStaticHashedString> *); - void mapUniform(GLint index, const std::vector<LLStaticHashedString> *); - void uniform1i(U32 index, GLint i); - void uniform1f(U32 index, GLfloat v); + void attachObject(GLuint object); + void attachObjects(GLuint* objects = NULL, S32 count = 0); + BOOL mapAttributes(const std::vector<LLStaticHashedString>* attributes); + BOOL mapUniforms(const std::vector<LLStaticHashedString>*); + void mapUniform(GLint index, const std::vector<LLStaticHashedString>*); + void uniform1i(U32 index, GLint i); + void uniform1f(U32 index, GLfloat v); void fastUniform1f(U32 index, GLfloat v); - void uniform2f(U32 index, GLfloat x, GLfloat y); - void uniform3f(U32 index, GLfloat x, GLfloat y, GLfloat z); - void uniform4f(U32 index, GLfloat x, GLfloat y, GLfloat z, GLfloat w); - void uniform1iv(U32 index, U32 count, const GLint* i); + void uniform2f(U32 index, GLfloat x, GLfloat y); + void uniform3f(U32 index, GLfloat x, GLfloat y, GLfloat z); + void uniform4f(U32 index, GLfloat x, GLfloat y, GLfloat z, GLfloat w); + void uniform1iv(U32 index, U32 count, const GLint* i); void uniform4iv(U32 index, U32 count, const GLint* i); - void uniform1fv(U32 index, U32 count, const GLfloat* v); - void uniform2fv(U32 index, U32 count, const GLfloat* v); - void uniform3fv(U32 index, U32 count, const GLfloat* v); - void uniform4fv(U32 index, U32 count, const GLfloat* v); - void uniform2i(const LLStaticHashedString& uniform, GLint i, GLint j); - void uniformMatrix2fv(U32 index, U32 count, GLboolean transpose, const GLfloat *v); - void uniformMatrix3fv(U32 index, U32 count, GLboolean transpose, const GLfloat *v); - void uniformMatrix3x4fv(U32 index, U32 count, GLboolean transpose, const GLfloat *v); - void uniformMatrix4fv(U32 index, U32 count, GLboolean transpose, const GLfloat *v); - void uniform1i(const LLStaticHashedString& uniform, GLint i); + void uniform1fv(U32 index, U32 count, const GLfloat* v); + void uniform2fv(U32 index, U32 count, const GLfloat* v); + void uniform3fv(U32 index, U32 count, const GLfloat* v); + void uniform4fv(U32 index, U32 count, const GLfloat* v); + void uniform2i(const LLStaticHashedString& uniform, GLint i, GLint j); + void uniformMatrix2fv(U32 index, U32 count, GLboolean transpose, const GLfloat* v); + void uniformMatrix3fv(U32 index, U32 count, GLboolean transpose, const GLfloat* v); + void uniformMatrix3x4fv(U32 index, U32 count, GLboolean transpose, const GLfloat* v); + void uniformMatrix4fv(U32 index, U32 count, GLboolean transpose, const GLfloat* v); + void uniform1i(const LLStaticHashedString& uniform, GLint i); void uniform1iv(const LLStaticHashedString& uniform, U32 count, const GLint* v); void uniform4iv(const LLStaticHashedString& uniform, U32 count, const GLint* v); - void uniform1f(const LLStaticHashedString& uniform, GLfloat v); - void uniform2f(const LLStaticHashedString& uniform, GLfloat x, GLfloat y); - void uniform3f(const LLStaticHashedString& uniform, GLfloat x, GLfloat y, GLfloat z); - void uniform1fv(const LLStaticHashedString& uniform, U32 count, const GLfloat* v); - void uniform2fv(const LLStaticHashedString& uniform, U32 count, const GLfloat* v); - void uniform3fv(const LLStaticHashedString& uniform, U32 count, const GLfloat* v); - void uniform4fv(const LLStaticHashedString& uniform, U32 count, const GLfloat* v); - void uniformMatrix4fv(const LLStaticHashedString& uniform, U32 count, GLboolean transpose, const GLfloat *v); - - void setMinimumAlpha(F32 minimum); - - void vertexAttrib4f(U32 index, GLfloat x, GLfloat y, GLfloat z, GLfloat w); - void vertexAttrib4fv(U32 index, GLfloat* v); - - //GLint getUniformLocation(const std::string& uniform); - GLint getUniformLocation(const LLStaticHashedString& uniform); - GLint getUniformLocation(U32 index); - - GLint getAttribLocation(U32 attrib); - GLint mapUniformTextureChannel(GLint location, GLenum type, GLint size); - + void uniform1f(const LLStaticHashedString& uniform, GLfloat v); + void uniform2f(const LLStaticHashedString& uniform, GLfloat x, GLfloat y); + void uniform3f(const LLStaticHashedString& uniform, GLfloat x, GLfloat y, GLfloat z); + void uniform1fv(const LLStaticHashedString& uniform, U32 count, const GLfloat* v); + void uniform2fv(const LLStaticHashedString& uniform, U32 count, const GLfloat* v); + void uniform3fv(const LLStaticHashedString& uniform, U32 count, const GLfloat* v); + void uniform4fv(const LLStaticHashedString& uniform, U32 count, const GLfloat* v); + void uniformMatrix4fv(const LLStaticHashedString& uniform, U32 count, GLboolean transpose, const GLfloat* v); + + void setMinimumAlpha(F32 minimum); + + void vertexAttrib4f(U32 index, GLfloat x, GLfloat y, GLfloat z, GLfloat w); + void vertexAttrib4fv(U32 index, GLfloat* v); + + //GLint getUniformLocation(const std::string& uniform); + GLint getUniformLocation(const LLStaticHashedString& uniform); + GLint getUniformLocation(U32 index); + + GLint getAttribLocation(U32 attrib); + GLint mapUniformTextureChannel(GLint location, GLenum type, GLint size); + void clearPermutations(); - void addPermutation(std::string name, std::string value); - void removePermutation(std::string name); - - void addConstant( const LLGLSLShader::eShaderConsts shader_const ); - - //enable/disable texture channel for specified uniform - //if given texture uniform is active in the shader, - //the corresponding channel will be active upon return - //returns channel texture is enabled in from [0-MAX) - S32 enableTexture(S32 uniform, LLTexUnit::eTextureType mode = LLTexUnit::TT_TEXTURE, LLTexUnit::eTextureColorSpace space = LLTexUnit::TCS_LINEAR); + void addPermutation(std::string name, std::string value); + void removePermutation(std::string name); + + void addConstant(const LLGLSLShader::eShaderConsts shader_const); + + //enable/disable texture channel for specified uniform + //if given texture uniform is active in the shader, + //the corresponding channel will be active upon return + //returns channel texture is enabled in from [0-MAX) + S32 enableTexture(S32 uniform, LLTexUnit::eTextureType mode = LLTexUnit::TT_TEXTURE, LLTexUnit::eTextureColorSpace space = LLTexUnit::TCS_LINEAR); S32 disableTexture(S32 uniform, LLTexUnit::eTextureType mode = LLTexUnit::TT_TEXTURE, LLTexUnit::eTextureColorSpace space = LLTexUnit::TCS_LINEAR); - - // bindTexture returns the texture unit we've bound the texture to. - // You can reuse the return value to unbind a texture when required. - S32 bindTexture(const std::string& uniform, LLTexture *texture, LLTexUnit::eTextureType mode = LLTexUnit::TT_TEXTURE, LLTexUnit::eTextureColorSpace space = LLTexUnit::TCS_LINEAR); - S32 bindTexture(S32 uniform, LLTexture *texture, LLTexUnit::eTextureType mode = LLTexUnit::TT_TEXTURE, LLTexUnit::eTextureColorSpace space = LLTexUnit::TCS_LINEAR); - S32 unbindTexture(const std::string& uniform, LLTexUnit::eTextureType mode = LLTexUnit::TT_TEXTURE); - S32 unbindTexture(S32 uniform, LLTexUnit::eTextureType mode = LLTexUnit::TT_TEXTURE); - + + // bindTexture returns the texture unit we've bound the texture to. + // You can reuse the return value to unbind a texture when required. + S32 bindTexture(const std::string& uniform, LLTexture* texture, LLTexUnit::eTextureType mode = LLTexUnit::TT_TEXTURE, LLTexUnit::eTextureColorSpace space = LLTexUnit::TCS_LINEAR); + S32 bindTexture(S32 uniform, LLTexture* texture, LLTexUnit::eTextureType mode = LLTexUnit::TT_TEXTURE, LLTexUnit::eTextureColorSpace space = LLTexUnit::TCS_LINEAR); + S32 unbindTexture(const std::string& uniform, LLTexUnit::eTextureType mode = LLTexUnit::TT_TEXTURE); + S32 unbindTexture(S32 uniform, LLTexUnit::eTextureType mode = LLTexUnit::TT_TEXTURE); + BOOL link(BOOL suppress_errors = FALSE); - void bind(); + void bind(); //helper to conditionally bind mRiggedVariant instead of this void bind(bool rigged); - void unbind(); + void unbind(); - // Unbinds any previously bound shader by explicitly binding no shader. - static void bindNoShader(void); + // Unbinds any previously bound shader by explicitly binding no shader. + static void bindNoShader(void); - U32 mMatHash[LLRender::NUM_MATRIX_MODES]; - U32 mLightHash; + U32 mMatHash[LLRender::NUM_MATRIX_MODES]; + U32 mLightHash; - GLuint mProgramObject; + GLuint mProgramObject; #if LL_RELEASE_WITH_DEBUG_INFO - struct attr_name - { - GLint loc; - const char *name; - void operator = (GLint _loc) { loc = _loc; } - operator GLint () { return loc; } - }; - std::vector<attr_name> mAttribute; //lookup table of attribute enum to attribute channel + struct attr_name + { + GLint loc; + const char* name; + void operator = (GLint _loc) { loc = _loc; } + operator GLint () { return loc; } + }; + std::vector<attr_name> mAttribute; //lookup table of attribute enum to attribute channel #else - std::vector<GLint> mAttribute; //lookup table of attribute enum to attribute channel + std::vector<GLint> mAttribute; //lookup table of attribute enum to attribute channel #endif - U32 mAttributeMask; //mask of which reserved attributes are set (lines up with LLVertexBuffer::getTypeMask()) - std::vector<GLint> mUniform; //lookup table of uniform enum to uniform location - LLStaticStringTable<GLint> mUniformMap; //lookup map of uniform name to uniform location + U32 mAttributeMask; //mask of which reserved attributes are set (lines up with LLVertexBuffer::getTypeMask()) + std::vector<GLint> mUniform; //lookup table of uniform enum to uniform location + LLStaticStringTable<GLint> mUniformMap; //lookup map of uniform name to uniform location typedef std::unordered_map<GLint, std::string> uniform_name_map_t; typedef std::unordered_map<GLint, LLVector4> uniform_value_map_t; uniform_name_map_t mUniformNameMap; //lookup map of uniform location to uniform name - uniform_value_map_t mValue; //lookup map of uniform location to last known value - std::vector<GLint> mTexture; - S32 mTotalUniformSize; - S32 mActiveTextureChannels; - S32 mShaderLevel; - S32 mShaderGroup; // see LLGLSLShader::eGroup - BOOL mUniformsDirty; - LLShaderFeatures mFeatures; - std::vector< std::pair< std::string, GLenum > > mShaderFiles; - std::string mName; + uniform_value_map_t mValue; //lookup map of uniform location to last known value + std::vector<GLint> mTexture; + S32 mTotalUniformSize; + S32 mActiveTextureChannels; + S32 mShaderLevel; + S32 mShaderGroup; // see LLGLSLShader::eGroup + BOOL mUniformsDirty; + LLShaderFeatures mFeatures; + std::vector< std::pair< std::string, GLenum > > mShaderFiles; + std::string mName; typedef std::unordered_map<std::string, std::string> defines_map_t; - defines_map_t mDefines; - - //statistcis for profiling shader performance - U32 mTimerQuery; - U32 mSamplesQuery; - U64 mTimeElapsed; - static U64 sTotalTimeElapsed; - U32 mTrianglesDrawn; - static U32 sTotalTrianglesDrawn; - U64 mSamplesDrawn; - static U64 sTotalSamplesDrawn; - U32 mDrawCalls; - static U32 sTotalDrawCalls; - - bool mTextureStateFetched; - std::vector<U32> mTextureMagFilter; - std::vector<U32> mTextureMinFilter; + defines_map_t mDefines; + + //statistics for profiling shader performance + U32 mTimerQuery; + U32 mSamplesQuery; + U32 mPrimitivesQuery; + + U64 mTimeElapsed; + static U64 sTotalTimeElapsed; + U32 mTrianglesDrawn; + static U32 sTotalTrianglesDrawn; + U64 mSamplesDrawn; + static U64 sTotalSamplesDrawn; + U32 mBinds; + static U32 sTotalBinds; // this pointer should be set to whichever shader represents this shader's rigged variant LLGLSLShader* mRiggedVariant = nullptr; - #ifdef LL_PROFILER_ENABLE_RENDER_DOC +#ifdef LL_PROFILER_ENABLE_RENDER_DOC void setLabel(const char* label); - #endif +#endif private: - void unloadInternal(); + void unloadInternal(); }; //UI shader (declared here so llui_libtest will link properly) diff --git a/indra/llrender/llrender.cpp b/indra/llrender/llrender.cpp index c58fbe6c8e..02942eb292 100644 --- a/indra/llrender/llrender.cpp +++ b/indra/llrender/llrender.cpp @@ -35,6 +35,7 @@ #include "llrendertarget.h" #include "lltexture.h" #include "llshadermgr.h" +#include "llmd5.h" #if LL_WINDOWS extern void APIENTRY gl_debug_callback(GLenum source, @@ -66,6 +67,14 @@ LLVector2 LLRender::sUIGLScaleFactor = LLVector2(1.f, 1.f); static const U32 LL_NUM_TEXTURE_LAYERS = 32; static const U32 LL_NUM_LIGHT_UNITS = 8; +struct LLVBCache +{ + LLPointer<LLVertexBuffer> vb; + std::chrono::steady_clock::time_point touched; +}; + +static std::unordered_map<std::size_t, LLVBCache> sVBCache; + static const GLenum sGLTextureType[] = { GL_TEXTURE_2D, @@ -370,9 +379,10 @@ bool LLTexUnit::bind(LLRenderTarget* renderTarget, bool bindDepth) if (bindDepth) { - if (renderTarget->hasStencil()) + + if (renderTarget->getDepth() && !renderTarget->canSampleDepth()) { - LL_ERRS() << "Cannot bind a render buffer for sampling. Allocate render target without a stencil buffer if sampling of depth buffer is required." << LL_ENDL; + LL_ERRS() << "Cannot bind a render buffer for sampling. Allocate render target with depth buffer sampling enabled." << LL_ENDL; } bindManual(renderTarget->getUsage(), renderTarget->getDepth()); @@ -1634,24 +1644,105 @@ void LLRender::flush() if (mBuffer) { - if (mBuffer->useVBOs() && !mBuffer->isLocked()) - { //hack to only flush the part of the buffer that was updated (relies on stream draw using buffersubdata) - mBuffer->getVertexStrider(mVerticesp, 0, count); - mBuffer->getTexCoord0Strider(mTexcoordsp, 0, count); - mBuffer->getColorStrider(mColorsp, 0, count); + + LLMD5 hash; + U32 attribute_mask = LLGLSLShader::sCurBoundShaderPtr->mAttributeMask; + + { + LL_PROFILE_ZONE_NAMED_CATEGORY_VERTEX("vb cache hash"); + + hash.update((U8*)mVerticesp.get(), count * sizeof(LLVector4a)); + if (attribute_mask & LLVertexBuffer::MAP_TEXCOORD0) + { + hash.update((U8*)mTexcoordsp.get(), count * sizeof(LLVector2)); + } + + if (attribute_mask & LLVertexBuffer::MAP_COLOR) + { + hash.update((U8*)mColorsp.get(), count * sizeof(LLColor4U)); + } + + hash.finalize(); + } + + size_t vhash[2]; + hash.raw_digest((unsigned char*) vhash); + + // check the VB cache before making a new vertex buffer + // This is a giant hack to deal with (mostly) our terrible UI rendering code + // that was built on top of OpenGL immediate mode. Huge performance wins + // can be had by not uploading geometry to VRAM unless absolutely necessary. + // Most of our usage of the "immediate mode" style draw calls is actually + // sending the same geometry over and over again. + // To leverage this, we maintain a running hash of the vertex stream being + // built up before a flush, and then check that hash against a VB + // cache just before creating a vertex buffer in VRAM + auto& cache = sVBCache.find(vhash[0]); + + LLPointer<LLVertexBuffer> vb; + + if (cache != sVBCache.end()) + { + LL_PROFILE_ZONE_NAMED_CATEGORY_VERTEX("vb cache hit"); + // cache hit, just use the cached buffer + vb = cache->second.vb; + cache->second.touched = std::chrono::steady_clock::now(); + } + else + { + LL_PROFILE_ZONE_NAMED_CATEGORY_VERTEX("vb cache miss"); + vb = new LLVertexBuffer(attribute_mask, GL_STATIC_DRAW); + vb->allocateBuffer(count, 0, true); + vb->setPositionData((LLVector4a*) mVerticesp.get()); + + if (attribute_mask & LLVertexBuffer::MAP_TEXCOORD0) + { + vb->setTexCoordData(mTexcoordsp.get()); + } + + if (attribute_mask & LLVertexBuffer::MAP_COLOR) + { + vb->setColorData(mColorsp.get()); + } + + vb->unbind(); + + sVBCache[vhash[0]] = { vb , std::chrono::steady_clock::now() }; + + static U32 miss_count = 0; + miss_count++; + if (miss_count > 1024) + { + LL_PROFILE_ZONE_NAMED_CATEGORY_VERTEX("vb cache clean"); + miss_count = 0; + auto now = std::chrono::steady_clock::now(); + + using namespace std::chrono_literals; + // every 1024 misses, clean the cache of any VBs that haven't been touched in the last second + for (auto& iter = sVBCache.begin(); iter != sVBCache.end(); ) + { + if (now - iter->second.touched > 1s) + { + iter = sVBCache.erase(iter); + } + else + { + ++iter; + } + } + } } - mBuffer->flush(); - mBuffer->setBuffer(immediate_mask); + vb->setBuffer(attribute_mask); if (mMode == LLRender::QUADS && sGLCoreProfile) { - mBuffer->drawArrays(LLRender::TRIANGLES, 0, count); + vb->drawArrays(LLRender::TRIANGLES, 0, count); mQuadCycle = 1; } else { - mBuffer->drawArrays(mMode, 0, count); + vb->drawArrays(mMode, 0, count); } } else diff --git a/indra/llrender/llrendertarget.cpp b/indra/llrender/llrendertarget.cpp index 01ccf3d314..9827db8084 100644 --- a/indra/llrender/llrendertarget.cpp +++ b/indra/llrender/llrendertarget.cpp @@ -67,9 +67,8 @@ LLRenderTarget::LLRenderTarget() : mPreviousResX(0), mPreviousResY(0), mDepth(0), - mStencil(0), mUseDepth(false), - mRenderDepth(false), + mSampleDepth(false), mUsage(LLTexUnit::TT_TEXTURE) { } @@ -98,11 +97,11 @@ void LLRenderTarget::resize(U32 resx, U32 resy) if (mDepth) { //resize depth attachment - if (mStencil) + if (!mSampleDepth) { //use render buffers where stencil buffers are in play glBindRenderbuffer(GL_RENDERBUFFER, mDepth); - glRenderbufferStorage(GL_RENDERBUFFER, GL_DEPTH24_STENCIL8, mResX, mResY); + glRenderbufferStorage(GL_RENDERBUFFER, GL_DEPTH_COMPONENT24, mResX, mResY); glBindRenderbuffer(GL_RENDERBUFFER, 0); } else @@ -117,7 +116,7 @@ void LLRenderTarget::resize(U32 resx, U32 resy) } -bool LLRenderTarget::allocate(U32 resx, U32 resy, U32 color_fmt, bool depth, bool stencil, LLTexUnit::eTextureType usage, bool use_fbo, S32 samples) +bool LLRenderTarget::allocate(U32 resx, U32 resy, U32 color_fmt, bool depth, bool sample_depth, LLTexUnit::eTextureType usage, bool use_fbo, S32 samples) { LL_PROFILE_ZONE_SCOPED_CATEGORY_DISPLAY; resx = llmin(resx, (U32) gGLManager.mGLMaxTextureSize); @@ -130,9 +129,9 @@ bool LLRenderTarget::allocate(U32 resx, U32 resy, U32 color_fmt, bool depth, boo mResX = resx; mResY = resy; - mStencil = stencil; mUsage = usage; - mUseDepth = depth; + mUseDepth = depth; + mSampleDepth = sample_depth; if ((sUseFBO || use_fbo)) { @@ -150,13 +149,10 @@ bool LLRenderTarget::allocate(U32 resx, U32 resy, U32 color_fmt, bool depth, boo if (mDepth) { glBindFramebuffer(GL_FRAMEBUFFER, mFBO); - llassert(!mStencil); // use of stencil buffer is deprecated (performance penalty) - if (mStencil) + + if (!canSampleDepth()) { glFramebufferRenderbuffer(GL_FRAMEBUFFER, GL_DEPTH_ATTACHMENT, GL_RENDERBUFFER, mDepth); - stop_glerror(); - glFramebufferRenderbuffer(GL_FRAMEBUFFER, GL_STENCIL_ATTACHMENT, GL_RENDERBUFFER, mDepth); - stop_glerror(); } else { @@ -315,14 +311,12 @@ bool LLRenderTarget::addColorAttachment(U32 color_fmt) bool LLRenderTarget::allocateDepth() { LL_PROFILE_ZONE_SCOPED_CATEGORY_DISPLAY; - if (mStencil) + if (!mSampleDepth) { - //use render buffers where stencil buffers are in play + //use render buffers if depth buffer won't be sampled glGenRenderbuffers(1, (GLuint *) &mDepth); glBindRenderbuffer(GL_RENDERBUFFER, mDepth); - stop_glerror(); - clear_glerror(); - glRenderbufferStorage(GL_RENDERBUFFER, GL_DEPTH24_STENCIL8, mResX, mResY); + glRenderbufferStorage(GL_RENDERBUFFER, GL_DEPTH_COMPONENT24, mResX, mResY); glBindRenderbuffer(GL_RENDERBUFFER, 0); } else @@ -367,23 +361,15 @@ void LLRenderTarget::shareDepthBuffer(LLRenderTarget& target) if (mDepth) { - stop_glerror(); glBindFramebuffer(GL_FRAMEBUFFER, target.mFBO); - stop_glerror(); - - llassert(!mStencil); // deprecated -- performance penalty - if (mStencil) + + if (!mSampleDepth) { glFramebufferRenderbuffer(GL_FRAMEBUFFER, GL_DEPTH_ATTACHMENT, GL_RENDERBUFFER, mDepth); - stop_glerror(); - glFramebufferRenderbuffer(GL_FRAMEBUFFER, GL_STENCIL_ATTACHMENT, GL_RENDERBUFFER, mDepth); - stop_glerror(); - target.mStencil = true; } else { glFramebufferTexture2D(GL_FRAMEBUFFER, GL_DEPTH_ATTACHMENT, LLTexUnit::getInternalType(mUsage), mDepth, 0); - stop_glerror(); } check_framebuffer_status(); @@ -399,15 +385,13 @@ void LLRenderTarget::release() LL_PROFILE_ZONE_SCOPED_CATEGORY_DISPLAY; if (mDepth) { - if (mStencil) + if (!mSampleDepth) { glDeleteRenderbuffers(1, (GLuint*) &mDepth); - stop_glerror(); } else { LLImageGL::deleteTextures(1, &mDepth); - stop_glerror(); } mDepth = 0; @@ -419,12 +403,10 @@ void LLRenderTarget::release() if (mUseDepth) { //detach shared depth buffer - llassert(!mStencil); //deprecated, performance penalty - if (mStencil) + if (!mSampleDepth) { //attached as a renderbuffer - glFramebufferRenderbuffer(GL_FRAMEBUFFER, GL_STENCIL_ATTACHMENT, GL_RENDERBUFFER, 0); glFramebufferRenderbuffer(GL_FRAMEBUFFER, GL_DEPTH_ATTACHMENT, GL_RENDERBUFFER, 0); - mStencil = false; + mSampleDepth = false; } else { //attached as a texture @@ -624,84 +606,6 @@ void LLRenderTarget::flush(bool fetch_depth) } } -void LLRenderTarget::copyContents(LLRenderTarget& source, S32 srcX0, S32 srcY0, S32 srcX1, S32 srcY1, - S32 dstX0, S32 dstY0, S32 dstX1, S32 dstY1, U32 mask, U32 filter) -{ - LL_PROFILE_GPU_ZONE("LLRenderTarget::copyContents"); - - GLboolean write_depth = mask & GL_DEPTH_BUFFER_BIT ? TRUE : FALSE; - - LLGLDepthTest depth(write_depth, write_depth); - - gGL.flush(); - if (!source.mFBO || !mFBO) - { - LL_WARNS() << "Cannot copy framebuffer contents for non FBO render targets." << LL_ENDL; - return; - } - - - if (mask == GL_DEPTH_BUFFER_BIT && source.mStencil != mStencil) - { - stop_glerror(); - - glBindFramebuffer(GL_FRAMEBUFFER, source.mFBO); - check_framebuffer_status(); - gGL.getTexUnit(0)->bind(this, true); - stop_glerror(); - glCopyTexSubImage2D(LLTexUnit::getInternalType(mUsage), 0, srcX0, srcY0, dstX0, dstY0, dstX1, dstY1); - stop_glerror(); - glBindFramebuffer(GL_FRAMEBUFFER, sCurFBO); - stop_glerror(); - } - else - { - glBindFramebuffer(GL_READ_FRAMEBUFFER, source.mFBO); - stop_glerror(); - glBindFramebuffer(GL_DRAW_FRAMEBUFFER, mFBO); - stop_glerror(); - check_framebuffer_status(); - stop_glerror(); - glBlitFramebuffer(srcX0, srcY0, srcX1, srcY1, dstX0, dstY0, dstX1, dstY1, mask, filter); - stop_glerror(); - glBindFramebuffer(GL_READ_FRAMEBUFFER, 0); - stop_glerror(); - glBindFramebuffer(GL_DRAW_FRAMEBUFFER, 0); - stop_glerror(); - glBindFramebuffer(GL_FRAMEBUFFER, sCurFBO); - stop_glerror(); - } -} - -//static -void LLRenderTarget::copyContentsToFramebuffer(LLRenderTarget& source, S32 srcX0, S32 srcY0, S32 srcX1, S32 srcY1, - S32 dstX0, S32 dstY0, S32 dstX1, S32 dstY1, U32 mask, U32 filter) -{ - if (!source.mFBO) - { - LL_WARNS() << "Cannot copy framebuffer contents for non FBO render targets." << LL_ENDL; - return; - } - - { - LL_PROFILE_GPU_ZONE("copyContentsToFramebuffer"); - GLboolean write_depth = mask & GL_DEPTH_BUFFER_BIT ? TRUE : FALSE; - - LLGLDepthTest depth(write_depth, write_depth); - - glBindFramebuffer(GL_READ_FRAMEBUFFER, source.mFBO); - stop_glerror(); - glBindFramebuffer(GL_DRAW_FRAMEBUFFER, 0); - stop_glerror(); - check_framebuffer_status(); - stop_glerror(); - glBlitFramebuffer(srcX0, srcY0, srcX1, srcY1, dstX0, dstY0, dstX1, dstY1, mask, filter); - stop_glerror(); - glBindFramebuffer(GL_FRAMEBUFFER, sCurFBO); - stop_glerror(); - } -} - bool LLRenderTarget::isComplete() const { return (!mTex.empty() || mDepth) ? true : false; diff --git a/indra/llrender/llrendertarget.h b/indra/llrender/llrendertarget.h index 584f224dca..5f3214add3 100644 --- a/indra/llrender/llrendertarget.h +++ b/indra/llrender/llrendertarget.h @@ -73,7 +73,7 @@ public: //allocate resources for rendering //must be called before use //multiple calls will release previously allocated resources - bool allocate(U32 resx, U32 resy, U32 color_fmt, bool depth, bool stencil, LLTexUnit::eTextureType usage = LLTexUnit::TT_TEXTURE, bool use_fbo = false, S32 samples = 0); + bool allocate(U32 resx, U32 resy, U32 color_fmt, bool depth, bool sample_depth, LLTexUnit::eTextureType usage = LLTexUnit::TT_TEXTURE, bool use_fbo = false, S32 samples = 0); //resize existing attachments to use new resolution and color format // CAUTION: if the GL runs out of memory attempting to resize, this render target will be undefined @@ -136,7 +136,7 @@ public: U32 getNumTextures() const; U32 getDepth(void) const { return mDepth; } - bool hasStencil() const { return mStencil; } + bool canSampleDepth() const { return mSampleDepth; } void bindTexture(U32 index, S32 channel, LLTexUnit::eTextureFilterOptions filter_options = LLTexUnit::TFO_BILINEAR); @@ -148,12 +148,6 @@ public: // the current depth texture. A depth texture will be allocated if needed. void flush(bool fetch_depth = FALSE); - void copyContents(LLRenderTarget& source, S32 srcX0, S32 srcY0, S32 srcX1, S32 srcY1, - S32 dstX0, S32 dstY0, S32 dstX1, S32 dstY1, U32 mask, U32 filter); - - static void copyContentsToFramebuffer(LLRenderTarget& source, S32 srcX0, S32 srcY0, S32 srcX1, S32 srcY1, - S32 dstX0, S32 dstY0, S32 dstX1, S32 dstY1, U32 mask, U32 filter); - //Returns TRUE if target is ready to be rendered into. //That is, if the target has been allocated with at least //one renderable attachment (i.e. color buffer, depth buffer). @@ -172,9 +166,9 @@ protected: U32 mPreviousResY; U32 mDepth; - bool mStencil; - bool mUseDepth; - bool mRenderDepth; + bool mUseDepth; + bool mSampleDepth; + LLTexUnit::eTextureType mUsage; static LLRenderTarget* sBoundTarget; diff --git a/indra/llrender/llvertexbuffer.cpp b/indra/llrender/llvertexbuffer.cpp index 7b8f85acba..e1352691d4 100644 --- a/indra/llrender/llvertexbuffer.cpp +++ b/indra/llrender/llvertexbuffer.cpp @@ -62,6 +62,14 @@ U32 wpo2(U32 i) return r; } +struct CompareMappedRegion +{ + bool operator()(const LLVertexBuffer::MappedRegion& lhs, const LLVertexBuffer::MappedRegion& rhs) + { + return lhs.mStart < rhs.mStart; + } +}; + const U32 LL_VBO_BLOCK_SIZE = 2048; const U32 LL_VBO_POOL_MAX_SEED_SIZE = 256*1024; @@ -81,266 +89,217 @@ U32 vbo_block_index(U32 size) const U32 LL_VBO_POOL_SEED_COUNT = vbo_block_index(LL_VBO_POOL_MAX_SEED_SIZE) + 1; +#define ENABLE_GL_WORK_QUEUE 0 + +#if ENABLE_GL_WORK_QUEUE + +#define THREAD_COUNT 1 //============================================================================ -//static -LLVBOPool LLVertexBuffer::sStreamVBOPool(GL_STREAM_DRAW, GL_ARRAY_BUFFER); -LLVBOPool LLVertexBuffer::sDynamicVBOPool(GL_DYNAMIC_DRAW, GL_ARRAY_BUFFER); -LLVBOPool LLVertexBuffer::sDynamicCopyVBOPool(GL_DYNAMIC_COPY, GL_ARRAY_BUFFER); -LLVBOPool LLVertexBuffer::sStreamIBOPool(GL_STREAM_DRAW, GL_ELEMENT_ARRAY_BUFFER); -LLVBOPool LLVertexBuffer::sDynamicIBOPool(GL_DYNAMIC_DRAW, GL_ELEMENT_ARRAY_BUFFER); +// High performance WorkQueue for usage in real-time rendering work +class GLWorkQueue +{ +public: + using Work = std::function<void()>; -U32 LLVBOPool::sBytesPooled = 0; -U32 LLVBOPool::sIndexBytesPooled = 0; -U32 LLVBOPool::sNameIdx = 0; -U32 LLVBOPool::sNamePool[1024]; + GLWorkQueue(); -std::list<U32> LLVertexBuffer::sAvailableVAOName; -U32 LLVertexBuffer::sCurVAOName = 1; + void post(const Work& value); -U32 LLVertexBuffer::sAllocatedIndexBytes = 0; -U32 LLVertexBuffer::sIndexCount = 0; + size_t size(); -U32 LLVertexBuffer::sBindCount = 0; -U32 LLVertexBuffer::sSetCount = 0; -S32 LLVertexBuffer::sCount = 0; -S32 LLVertexBuffer::sGLCount = 0; -S32 LLVertexBuffer::sMappedCount = 0; -bool LLVertexBuffer::sDisableVBOMapping = false; -bool LLVertexBuffer::sEnableVBOs = true; -U32 LLVertexBuffer::sGLRenderBuffer = 0; -U32 LLVertexBuffer::sGLRenderArray = 0; -U32 LLVertexBuffer::sGLRenderIndices = 0; -U32 LLVertexBuffer::sLastMask = 0; -bool LLVertexBuffer::sVBOActive = false; -bool LLVertexBuffer::sIBOActive = false; -U32 LLVertexBuffer::sAllocatedBytes = 0; -U32 LLVertexBuffer::sVertexCount = 0; -bool LLVertexBuffer::sMapped = false; -bool LLVertexBuffer::sUseStreamDraw = true; -bool LLVertexBuffer::sUseVAO = false; -bool LLVertexBuffer::sPreferStreamDraw = false; + bool done(); -U32 LLVBOPool::genBuffer() -{ - LL_PROFILE_ZONE_SCOPED_CATEGORY_VERTEX + // Get the next element from the queue + Work pop(); - if (sNameIdx == 0) - { - glGenBuffers(1024, sNamePool); - sNameIdx = 1024; - } + void runOne(); - return sNamePool[--sNameIdx]; -} + bool runPending(); -void LLVBOPool::deleteBuffer(U32 name) -{ - LL_PROFILE_ZONE_SCOPED_CATEGORY_VERTEX - if (gGLManager.mInited) - { - LLVertexBuffer::unbind(); + void runUntilClose(); - glBindBuffer(mType, name); - glBufferData(mType, 0, NULL, mUsage); - glBindBuffer(mType, 0); + void close(); - glDeleteBuffers(1, &name); - } -} + bool isClosed(); + void syncGL(); -LLVBOPool::LLVBOPool(U32 vboUsage, U32 vboType) -: mUsage(vboUsage), mType(vboType), mMissCountDirty(true) -{ - mFreeList.resize(LL_VBO_POOL_SEED_COUNT); - mMissCount.resize(LL_VBO_POOL_SEED_COUNT); - std::fill(mMissCount.begin(), mMissCount.end(), 0); -} +private: + std::mutex mMutex; + std::condition_variable mCondition; + std::queue<Work> mQueue; + bool mClosed = false; +}; -U8* LLVBOPool::allocate(U32& name, U32 size, bool for_seed) +GLWorkQueue::GLWorkQueue() { - LL_PROFILE_ZONE_SCOPED_CATEGORY_VERTEX - llassert(vbo_block_size(size) == size); - - U8* ret = NULL; - U32 i = vbo_block_index(size); +} - if (mFreeList.size() <= i) - { - mFreeList.resize(i+1); - } +void GLWorkQueue::syncGL() +{ + /*if (mSync) + { + std::lock_guard<std::mutex> lock(mMutex); + glWaitSync(mSync, 0, GL_TIMEOUT_IGNORED); + mSync = 0; + }*/ +} - if (mFreeList[i].empty() || for_seed) - { - //make a new buffer - name = genBuffer(); +size_t GLWorkQueue::size() +{ + LL_PROFILE_ZONE_SCOPED_CATEGORY_THREAD; + std::lock_guard<std::mutex> lock(mMutex); + return mQueue.size(); +} - glBindBuffer(mType, name); +bool GLWorkQueue::done() +{ + return size() == 0 && isClosed(); +} - if (!for_seed && i < LL_VBO_POOL_SEED_COUNT) - { //record this miss - mMissCount[i]++; - mMissCountDirty = true; // signal to ::seedPool() - } +void GLWorkQueue::post(const GLWorkQueue::Work& value) +{ + LL_PROFILE_ZONE_SCOPED_CATEGORY_THREAD; + { + std::lock_guard<std::mutex> lock(mMutex); + mQueue.push(std::move(value)); + } - if (mType == GL_ARRAY_BUFFER) - { - LLVertexBuffer::sAllocatedBytes += size; - } - else - { - LLVertexBuffer::sAllocatedIndexBytes += size; - } + mCondition.notify_one(); +} - if (LLVertexBuffer::sDisableVBOMapping || mUsage != GL_DYNAMIC_DRAW) - { - glBufferData(mType, size, 0, mUsage); - if (mUsage != GL_DYNAMIC_COPY) - { //data will be provided by application - ret = (U8*) ll_aligned_malloc<64>(size); - if (!ret) - { - LL_ERRS() - << "Failed to allocate " << size << " bytes for LLVBOPool buffer " << name << "." << LL_NEWLINE - << "Free list size: " - << mFreeList.size() // this happens if we are out of memory so a solution might be to clear some from freelist - << " Allocated Bytes: " << LLVertexBuffer::sAllocatedBytes - << " Allocated Index Bytes: " << LLVertexBuffer::sAllocatedIndexBytes << " Pooled Bytes: " << sBytesPooled - << " Pooled Index Bytes: " << sIndexBytesPooled << LL_ENDL; - } - } - } - else - { //always use a true hint of static draw when allocating non-client-backed buffers - glBufferData(mType, size, 0, GL_STATIC_DRAW); - } +// Get the next element from the queue +GLWorkQueue::Work GLWorkQueue::pop() +{ + LL_PROFILE_ZONE_SCOPED_CATEGORY_THREAD; + // Lock the mutex + { + std::unique_lock<std::mutex> lock(mMutex); - glBindBuffer(mType, 0); + // Wait for a new element to become available or for the queue to close + { + mCondition.wait(lock, [=] { return !mQueue.empty() || mClosed; }); + } + } - if (for_seed) - { //put into pool for future use - llassert(mFreeList.size() > i); + Work ret; - Record rec; - rec.mGLName = name; - rec.mClientData = ret; - - if (mType == GL_ARRAY_BUFFER) - { - sBytesPooled += size; - } - else - { - sIndexBytesPooled += size; - } - mFreeList[i].push_back(rec); - mMissCountDirty = true; // signal to ::seedPool() - } - } - else - { - name = mFreeList[i].front().mGLName; - ret = mFreeList[i].front().mClientData; + { + std::lock_guard<std::mutex> lock(mMutex); - if (mType == GL_ARRAY_BUFFER) - { - sBytesPooled -= size; - } - else - { - sIndexBytesPooled -= size; - } + // Get the next element from the queue + if (mQueue.size() > 0) + { + ret = mQueue.front(); + mQueue.pop(); + } + else + { + ret = []() {}; + } + } - mFreeList[i].pop_front(); - mMissCountDirty = true; // signal to ::seedPool() - } + return ret; +} - return ret; +void GLWorkQueue::runOne() +{ + LL_PROFILE_ZONE_SCOPED_CATEGORY_THREAD; + Work w = pop(); + w(); + //mSync = glFenceSync(GL_SYNC_GPU_COMMANDS_COMPLETE, 0); } -void LLVBOPool::release(U32 name, U8* buffer, U32 size) +void GLWorkQueue::runUntilClose() { - llassert(vbo_block_size(size) == size); + while (!isClosed()) + { + runOne(); + } +} - deleteBuffer(name); - ll_aligned_free_fallback((U8*) buffer); +void GLWorkQueue::close() +{ + LL_PROFILE_ZONE_SCOPED_CATEGORY_THREAD; + { + std::lock_guard<std::mutex> lock(mMutex); + mClosed = true; + } - if (mType == GL_ARRAY_BUFFER) - { - LLVertexBuffer::sAllocatedBytes -= size; - } - else - { - LLVertexBuffer::sAllocatedIndexBytes -= size; - } + mCondition.notify_all(); } -void LLVBOPool::seedPool() +bool GLWorkQueue::isClosed() { - LL_PROFILE_ZONE_SCOPED_CATEGORY_VERTEX - if (mMissCountDirty) - { - U32 dummy_name = 0; - U32 size = LL_VBO_BLOCK_SIZE; - - for (U32 i = 0; i < LL_VBO_POOL_SEED_COUNT; i++) - { - if (mMissCount[i] > mFreeList[i].size()) - { - S32 count = mMissCount[i] - mFreeList[i].size(); - for (U32 j = 0; j < count; ++j) - { - allocate(dummy_name, size, true); - } - } - size += LL_VBO_BLOCK_SIZE; - } - mMissCountDirty = false; - } + LL_PROFILE_ZONE_SCOPED_CATEGORY_THREAD; + std::lock_guard<std::mutex> lock(mMutex); + return mClosed; } -void LLVBOPool::cleanup() +#include "llwindow.h" + +class LLGLWorkerThread : public LLThread { - U32 size = LL_VBO_BLOCK_SIZE; +public: + LLGLWorkerThread(const std::string& name, GLWorkQueue* queue, LLWindow* window) + : LLThread(name) + { + mWindow = window; + mContext = mWindow->createSharedContext(); + mQueue = queue; + } - for (U32 i = 0; i < mFreeList.size(); ++i) - { - record_list_t& l = mFreeList[i]; + void run() override + { + mWindow->makeContextCurrent(mContext); + gGL.init(false); + mQueue->runUntilClose(); + gGL.shutdown(); + mWindow->destroySharedContext(mContext); + } - while (!l.empty()) - { - Record& r = l.front(); + GLWorkQueue* mQueue; + LLWindow* mWindow; + void* mContext = nullptr; +}; - deleteBuffer(r.mGLName); - - if (r.mClientData) - { - ll_aligned_free<64>((void*) r.mClientData); - } - l.pop_front(); +static LLGLWorkerThread* sVBOThread[THREAD_COUNT]; +static GLWorkQueue* sQueue = nullptr; - if (mType == GL_ARRAY_BUFFER) - { - sBytesPooled -= size; - LLVertexBuffer::sAllocatedBytes -= size; - } - else - { - sIndexBytesPooled -= size; - LLVertexBuffer::sAllocatedIndexBytes -= size; - } - } +#endif - size += LL_VBO_BLOCK_SIZE; - } +//============================================================================ - //reset miss counts - std::fill(mMissCount.begin(), mMissCount.end(), 0); -} +//static +std::list<U32> LLVertexBuffer::sAvailableVAOName; +U32 LLVertexBuffer::sCurVAOName = 1; + +U32 LLVertexBuffer::sAllocatedIndexBytes = 0; +U32 LLVertexBuffer::sIndexCount = 0; + +U32 LLVertexBuffer::sBindCount = 0; +U32 LLVertexBuffer::sSetCount = 0; +S32 LLVertexBuffer::sCount = 0; +S32 LLVertexBuffer::sGLCount = 0; +S32 LLVertexBuffer::sMappedCount = 0; +bool LLVertexBuffer::sDisableVBOMapping = false; +bool LLVertexBuffer::sEnableVBOs = true; +U32 LLVertexBuffer::sGLRenderBuffer = 0; +U32 LLVertexBuffer::sGLRenderArray = 0; +U32 LLVertexBuffer::sGLRenderIndices = 0; +U32 LLVertexBuffer::sLastMask = 0; +bool LLVertexBuffer::sVBOActive = false; +bool LLVertexBuffer::sIBOActive = false; +U32 LLVertexBuffer::sAllocatedBytes = 0; +U32 LLVertexBuffer::sVertexCount = 0; +bool LLVertexBuffer::sMapped = false; +bool LLVertexBuffer::sUseStreamDraw = true; +bool LLVertexBuffer::sUseVAO = false; +bool LLVertexBuffer::sPreferStreamDraw = false; //NOTE: each component must be AT LEAST 4 bytes in size to avoid a performance penalty on AMD hardware @@ -420,17 +379,6 @@ void LLVertexBuffer::releaseVAOName(U32 name) //static -void LLVertexBuffer::seedPools() -{ - LL_PROFILE_ZONE_SCOPED_CATEGORY_VERTEX - sStreamVBOPool.seedPool(); - sDynamicVBOPool.seedPool(); - sDynamicCopyVBOPool.seedPool(); - sStreamIBOPool.seedPool(); - sDynamicIBOPool.seedPool(); -} - -//static void LLVertexBuffer::setupClientArrays(U32 data_mask) { if (sLastMask != data_mask) @@ -473,7 +421,7 @@ void LLVertexBuffer::drawArrays(U32 mode, const std::vector<LLVector3>& pos) } gGL.end(); gGL.flush(); - } +} //static void LLVertexBuffer::drawElements(U32 mode, const LLVector4a* pos, const LLVector2* tc, S32 num_indices, const U16* indicesp) @@ -612,12 +560,8 @@ void LLVertexBuffer::drawRange(U32 mode, U32 start, U32 end, U32 count, U32 indi U16* idx = ((U16*) getIndicesPointer())+indices_offset; - stop_glerror(); - LLGLSLShader::startProfile(); glDrawRangeElements(sGLMode[mode], start, end, count, GL_UNSIGNED_SHORT, idx); - LLGLSLShader::stopProfile(count, mode); - stop_glerror(); } void LLVertexBuffer::drawRangeFast(U32 mode, U32 start, U32 end, U32 count, U32 indices_offset) const @@ -659,12 +603,8 @@ void LLVertexBuffer::draw(U32 mode, U32 count, U32 indices_offset) const return; } - stop_glerror(); - LLGLSLShader::startProfile(); glDrawElements(sGLMode[mode], count, GL_UNSIGNED_SHORT, ((U16*) getIndicesPointer()) + indices_offset); - LLGLSLShader::stopProfile(count, mode); - stop_glerror(); } @@ -694,20 +634,24 @@ void LLVertexBuffer::drawArrays(U32 mode, U32 first, U32 count) const } #endif - LLGLSLShader::startProfile(); - { - glDrawArrays(sGLMode[mode], first, count); - } - LLGLSLShader::stopProfile(count, mode); - - stop_glerror(); + glDrawArrays(sGLMode[mode], first, count); } //static -void LLVertexBuffer::initClass(bool use_vbo, bool no_vbo_mapping) +void LLVertexBuffer::initClass(LLWindow* window) { - sEnableVBOs = use_vbo; - sDisableVBOMapping = sEnableVBOs && no_vbo_mapping; + sEnableVBOs = true; + sDisableVBOMapping = true; + +#if ENABLE_GL_WORK_QUEUE + sQueue = new GLWorkQueue(); + + for (int i = 0; i < THREAD_COUNT; ++i) + { + sVBOThread[i] = new LLGLWorkerThread("VBO Worker", sQueue, window); + sVBOThread[i]->start(); + } +#endif } //static @@ -743,14 +687,19 @@ void LLVertexBuffer::cleanupClass() { unbind(); - sStreamIBOPool.cleanup(); - sDynamicIBOPool.cleanup(); - sStreamVBOPool.cleanup(); - sDynamicVBOPool.cleanup(); - sDynamicCopyVBOPool.cleanup(); - - llassert(0 == LLVBOPool::sBytesPooled); - llassert(0 == LLVBOPool::sIndexBytesPooled); +#if ENABLE_GL_WORK_QUEUE + sQueue->close(); + for (int i = 0; i < THREAD_COUNT; ++i) + { + sVBOThread[i]->shutdown(); + delete sVBOThread[i]; + sVBOThread[i] = nullptr; + } + + delete sQueue; + sQueue = nullptr; +#endif + //llassert(0 == sAllocatedBytes); //llassert(0 == sAllocatedIndexBytes); } @@ -781,21 +730,6 @@ S32 LLVertexBuffer::determineUsage(S32 usage) ret_usage = GL_STREAM_DRAW; } - if (ret_usage && ret_usage != GL_STREAM_DRAW) - { //only stream_draw and dynamic_draw are supported when using VBOs, dynamic draw is the default - if (ret_usage != GL_DYNAMIC_COPY) - { - if (sDisableVBOMapping) - { //always use stream draw if VBO mapping is disabled - ret_usage = GL_STREAM_DRAW; - } - else - { - ret_usage = GL_DYNAMIC_DRAW; - } - } - } - return ret_usage; } @@ -848,7 +782,7 @@ S32 LLVertexBuffer::calcOffsets(const U32& typemask, S32* offsets, S32 num_verti offsets[TYPE_TEXTURE_INDEX] = offsets[TYPE_VERTEX] + 12; - return offset+16; + return offset; } //static @@ -896,74 +830,103 @@ LLVertexBuffer::~LLVertexBuffer() //---------------------------------------------------------------------------- +// batch glGenBuffers +static GLuint gen_buffer() +{ + LL_PROFILE_ZONE_SCOPED_CATEGORY_VERTEX; + constexpr U32 pool_size = 4096; + + thread_local static GLuint sNamePool[pool_size]; + thread_local static U32 sIndex = 0; + + if (sIndex == 0) + { + LL_PROFILE_ZONE_NAMED_CATEGORY_VERTEX("gen ibo"); + sIndex = pool_size; + glGenBuffers(pool_size, sNamePool); + } + + return sNamePool[--sIndex]; +} + +// batch glDeleteBuffers +static void release_buffer(U32 buff) +{ + LL_PROFILE_ZONE_SCOPED_CATEGORY_VERTEX; +#if 0 + + constexpr U32 pool_size = 4096; + + thread_local static GLuint sNamePool[pool_size]; + thread_local static U32 sIndex = 0; + + if (sIndex == pool_size) + { + LL_PROFILE_ZONE_NAMED_CATEGORY_VERTEX("gen ibo"); + sIndex = 0; + glDeleteBuffers(pool_size, sNamePool); + } + + sNamePool[sIndex++] = buff; +#else + glDeleteBuffers(1, &buff); +#endif +} + void LLVertexBuffer::genBuffer(U32 size) { - mSize = vbo_block_size(size); + LL_PROFILE_ZONE_SCOPED_CATEGORY_VERTEX; - if (mUsage == GL_STREAM_DRAW) - { - mMappedData = sStreamVBOPool.allocate(mGLBuffer, mSize); - } - else if (mUsage == GL_DYNAMIC_DRAW) - { - mMappedData = sDynamicVBOPool.allocate(mGLBuffer, mSize); - } - else - { - mMappedData = sDynamicCopyVBOPool.allocate(mGLBuffer, mSize); - } - - - sGLCount++; + mSize = size; + mMappedData = (U8*) ll_aligned_malloc_16(size); + mGLBuffer = gen_buffer(); + + glBindBuffer(GL_ARRAY_BUFFER, mGLBuffer); + glBufferData(GL_ARRAY_BUFFER, mSize, nullptr, mUsage); + glBindBuffer(GL_ARRAY_BUFFER, 0); + sGLRenderBuffer = 0; + + sGLCount++; } void LLVertexBuffer::genIndices(U32 size) { - mIndicesSize = vbo_block_size(size); + LL_PROFILE_ZONE_SCOPED_CATEGORY_VERTEX; + + mIndicesSize = size; + mMappedIndexData = (U8*) ll_aligned_malloc_16(size); + + mGLIndices = gen_buffer(); + + glBindBuffer(GL_ELEMENT_ARRAY_BUFFER, mGLIndices); + glBufferData(GL_ELEMENT_ARRAY_BUFFER, mIndicesSize, nullptr, mUsage); + glBindBuffer(GL_ELEMENT_ARRAY_BUFFER, 0); + sGLRenderIndices = 0; - if (mUsage == GL_STREAM_DRAW) - { - mMappedIndexData = sStreamIBOPool.allocate(mGLIndices, mIndicesSize); - } - else - { - mMappedIndexData = sDynamicIBOPool.allocate(mGLIndices, mIndicesSize); - } - sGLCount++; } void LLVertexBuffer::releaseBuffer() { - if (mUsage == GL_STREAM_DRAW) - { - sStreamVBOPool.release(mGLBuffer, mMappedData, mSize); - } - else - { - sDynamicVBOPool.release(mGLBuffer, mMappedData, mSize); - } - - mGLBuffer = 0; - mMappedData = NULL; + LL_PROFILE_ZONE_SCOPED_CATEGORY_VERTEX; + release_buffer(mGLBuffer); + mGLBuffer = 0; + ll_aligned_free_16(mMappedData); + mMappedData = nullptr; + sGLCount--; } void LLVertexBuffer::releaseIndices() { - if (mUsage == GL_STREAM_DRAW) - { - sStreamIBOPool.release(mGLIndices, mMappedIndexData, mIndicesSize); - } - else - { - sDynamicIBOPool.release(mGLIndices, mMappedIndexData, mIndicesSize); - } + LL_PROFILE_ZONE_SCOPED_CATEGORY_VERTEX; + release_buffer(mGLIndices); + mGLIndices = 0; + + ll_aligned_free_16(mMappedIndexData); + mMappedIndexData = nullptr; - mGLIndices = 0; - mMappedIndexData = NULL; - sGLCount--; } @@ -1183,21 +1146,20 @@ bool LLVertexBuffer::useVBOs() const //---------------------------------------------------------------------------- -bool expand_region(LLVertexBuffer::MappedRegion& region, S32 index, S32 count) +// if no gap between region and given range exists, expand region to cover given range and return true +// otherwise return false +bool expand_region(LLVertexBuffer::MappedRegion& region, S32 start, S32 end) { - S32 end = index+count; - S32 region_end = region.mIndex+region.mCount; - if (end < region.mIndex || - index > region_end) + if (end < region.mStart || + start > region.mEnd) { //gap exists, do not merge return false; } - S32 new_end = llmax(end, region_end); - S32 new_index = llmin(index, region.mIndex); - region.mIndex = new_index; - region.mCount = new_end-new_index; + region.mStart = llmin(region.mStart, start); + region.mEnd = llmax(region.mEnd, end); + return true; } @@ -1206,7 +1168,6 @@ bool expand_region(LLVertexBuffer::MappedRegion& region, S32 index, S32 count) U8* LLVertexBuffer::mapVertexBuffer(S32 type, S32 index, S32 count, bool map_range) { LL_PROFILE_ZONE_SCOPED_CATEGORY_VERTEX; - bindGLBuffer(true); if (mFinal) { LL_ERRS() << "LLVertexBuffer::mapVeretxBuffer() called on a finalized buffer." << LL_ENDL; @@ -1215,34 +1176,34 @@ U8* LLVertexBuffer::mapVertexBuffer(S32 type, S32 index, S32 count, bool map_ran { LL_ERRS() << "LLVertexBuffer::mapVertexBuffer() called on unallocated buffer." << LL_ENDL; } - - if (useVBOs()) - { - if (count == -1) - { - count = mNumVerts-index; - } - bool mapped = false; - //see if range is already mapped + + if (useVBOs()) + { + if (count == -1) + { + count = mNumVerts - index; + } + + S32 start = mOffsets[type] + sTypeSize[type] * index; + S32 end = start + sTypeSize[type] * count; + + bool flagged = false; + // flag region as mapped for (U32 i = 0; i < mMappedVertexRegions.size(); ++i) { MappedRegion& region = mMappedVertexRegions[i]; - if (region.mType == type) - { - if (expand_region(region, index, count)) - { - mapped = true; - break; - } - } + if (expand_region(region, start, end)) + { + flagged = true; + break; + } } - if (!mapped) + if (!flagged) { - //not already mapped, map new region - MappedRegion region(type, index, count); - mMappedVertexRegions.push_back(region); + //didn't expand an existing region, make a new one + mMappedVertexRegions.push_back({ start, end }); } if (mVertexLocked && map_range) @@ -1282,7 +1243,6 @@ U8* LLVertexBuffer::mapVertexBuffer(S32 type, S32 index, S32 count, bool map_ran U8* LLVertexBuffer::mapIndexBuffer(S32 index, S32 count, bool map_range) { LL_PROFILE_ZONE_SCOPED_CATEGORY_VERTEX; - bindGLIndices(true); if (mFinal) { LL_ERRS() << "LLVertexBuffer::mapIndexBuffer() called on a finalized buffer." << LL_ENDL; @@ -1299,25 +1259,26 @@ U8* LLVertexBuffer::mapIndexBuffer(S32 index, S32 count, bool map_range) count = mNumIndices-index; } - bool mapped = false; - //see if range is already mapped - for (U32 i = 0; i < mMappedIndexRegions.size(); ++i) - { - MappedRegion& region = mMappedIndexRegions[i]; - if (expand_region(region, index, count)) - { - mapped = true; - break; - } - } + S32 start = sizeof(U16) * index; + S32 end = start + sizeof(U16) * count; - if (!mapped) - { - //not already mapped, map new region - MappedRegion region(TYPE_INDEX, index, count); - mMappedIndexRegions.push_back(region); - } - + bool flagged = false; + // flag region as mapped + for (U32 i = 0; i < mMappedIndexRegions.size(); ++i) + { + MappedRegion& region = mMappedIndexRegions[i]; + if (expand_region(region, start, end)) + { + flagged = true; + break; + } + } + + if (!flagged) + { + //didn't expand an existing region, make a new one + mMappedIndexRegions.push_back({ start, end }); + } if (mIndexLocked && map_range) { @@ -1330,17 +1291,6 @@ U8* LLVertexBuffer::mapIndexBuffer(S32 index, S32 count, bool map_range) sMappedCount++; stop_glerror(); - if (gDebugGL && useVBOs()) - { - GLint elem = 0; - glGetIntegerv(GL_ELEMENT_ARRAY_BUFFER_BINDING, &elem); - - if (elem != mGLIndices) - { - LL_ERRS() << "Wrong index buffer bound!" << LL_ENDL; - } - } - map_range = false; } @@ -1360,6 +1310,27 @@ U8* LLVertexBuffer::mapIndexBuffer(S32 index, S32 count, bool map_range) return mMappedIndexData + sizeof(U16)*index; } +static void flush_vbo(GLenum target, S32 start, S32 end, void* data) +{ + if (end != 0) + { + LL_PROFILE_ZONE_NAMED_CATEGORY_VERTEX("glBufferSubData"); + LL_PROFILE_ZONE_NUM(start); + LL_PROFILE_ZONE_NUM(end); + LL_PROFILE_ZONE_NUM(end-start); + + constexpr S32 block_size = 65536; + + for (S32 i = start; i < end; i += block_size) + { + LL_PROFILE_ZONE_NAMED_CATEGORY_VERTEX("glBufferSubData block"); + LL_PROFILE_GPU_ZONE("glBufferSubData"); + S32 tend = llmin(i + block_size, end); + glBufferSubData(target, i, tend - i, (U8*) data + (i-start)); + } + } +} + void LLVertexBuffer::unmapBuffer() { if (!useVBOs()) @@ -1377,37 +1348,31 @@ void LLVertexBuffer::unmapBuffer() if (!mMappedVertexRegions.empty()) { - stop_glerror(); + S32 start = 0; + S32 end = 0; + for (U32 i = 0; i < mMappedVertexRegions.size(); ++i) { const MappedRegion& region = mMappedVertexRegions[i]; - S32 offset = region.mIndex >= 0 ? mOffsets[region.mType]+sTypeSize[region.mType]*region.mIndex : 0; - S32 length = sTypeSize[region.mType]*region.mCount; - if (mSize >= length + offset) - { - glBufferSubData(GL_ARRAY_BUFFER, offset, length, (U8*)mMappedData + offset); - } - else - { - GLint size = 0; - glGetBufferParameteriv(GL_ARRAY_BUFFER, GL_BUFFER_SIZE, &size); - LL_WARNS() << "Attempted to map regions to a buffer that is too small, " - << "mapped size: " << mSize - << ", gl buffer size: " << size - << ", length: " << length - << ", offset: " << offset - << LL_ENDL; - } - stop_glerror(); + if (region.mStart == end + 1) + { + end = region.mEnd; + } + else + { + flush_vbo(GL_ARRAY_BUFFER, start, end, (U8*)mMappedData + start); + start = region.mStart; + end = region.mEnd; + } } + flush_vbo(GL_ARRAY_BUFFER, start, end, (U8*)mMappedData + start); + mMappedVertexRegions.clear(); } else { - stop_glerror(); - glBufferSubData(GL_ARRAY_BUFFER, 0, getSize(), (U8*) mMappedData); - stop_glerror(); + llassert(false); // this shouldn't happen -- a buffer must always be explicitly mapped } mVertexLocked = false; @@ -1421,36 +1386,31 @@ void LLVertexBuffer::unmapBuffer() if (!mMappedIndexRegions.empty()) { - for (U32 i = 0; i < mMappedIndexRegions.size(); ++i) - { - const MappedRegion& region = mMappedIndexRegions[i]; - S32 offset = region.mIndex >= 0 ? sizeof(U16)*region.mIndex : 0; - S32 length = sizeof(U16)*region.mCount; - if (mIndicesSize >= length + offset) - { - glBufferSubData(GL_ELEMENT_ARRAY_BUFFER, offset, length, (U8*) mMappedIndexData+offset); - } - else - { - GLint size = 0; - glGetBufferParameteriv(GL_ELEMENT_ARRAY_BUFFER, GL_BUFFER_SIZE, &size); - LL_WARNS() << "Attempted to map regions to a buffer that is too small, " - << "mapped size: " << mIndicesSize - << ", gl buffer size: " << size - << ", length: " << length - << ", offset: " << offset - << LL_ENDL; - } - stop_glerror(); - } + S32 start = 0; + S32 end = 0; + + for (U32 i = 0; i < mMappedIndexRegions.size(); ++i) + { + const MappedRegion& region = mMappedIndexRegions[i]; + if (region.mStart == end + 1) + { + end = region.mEnd; + } + else + { + flush_vbo(GL_ELEMENT_ARRAY_BUFFER, start, end, (U8*)mMappedIndexData + start); + start = region.mStart; + end = region.mEnd; + } + } + + flush_vbo(GL_ELEMENT_ARRAY_BUFFER, start, end, (U8*)mMappedIndexData + start); mMappedIndexRegions.clear(); } else { - stop_glerror(); - glBufferSubData(GL_ELEMENT_ARRAY_BUFFER, 0, getIndicesSize(), (U8*) mMappedIndexData); - stop_glerror(); + llassert(false); // this shouldn't happen -- a buffer must always be explicitly mapped } mIndexLocked = false; @@ -1640,11 +1600,53 @@ bool LLVertexBuffer::bindGLIndicesFast() return false; } -void LLVertexBuffer::flush() +void LLVertexBuffer::flush(bool discard) { if (useVBOs()) { - unmapBuffer(); + if (discard) + { // discard existing VBO data if the buffer must be updated + + if (!mMappedVertexRegions.empty()) + { + LL_PROFILE_ZONE_NAMED_CATEGORY_VERTEX("flush discard vbo"); + LL_PROFILE_ZONE_NUM(mSize); + release_buffer(mGLBuffer); + mGLBuffer = gen_buffer(); + bindGLBuffer(); + { + LL_PROFILE_GPU_ZONE("glBufferData"); + glBufferData(GL_ARRAY_BUFFER, mSize, nullptr, mUsage); + + for (int i = 0; i < mSize; i += 65536) + { + LL_PROFILE_GPU_ZONE("glBufferSubData"); + S32 end = llmin(i + 65536, mSize); + S32 count = end - i; + glBufferSubData(GL_ARRAY_BUFFER, i, count, mMappedData + i); + } + } + mMappedVertexRegions.clear(); + } + if (!mMappedIndexRegions.empty()) + { + LL_PROFILE_ZONE_NAMED_CATEGORY_VERTEX("flush discard ibo"); + LL_PROFILE_ZONE_NUM(mIndicesSize); + release_buffer(mGLIndices); + mGLIndices = gen_buffer(); + bindGLIndices(); + { + LL_PROFILE_GPU_ZONE("glBufferData (ibo)"); + glBufferData(GL_ELEMENT_ARRAY_BUFFER, mIndicesSize, mMappedIndexData, mUsage); + } + mMappedIndexRegions.clear(); + } + } + else + { + unmapBuffer(); + } + } } @@ -2043,12 +2045,24 @@ void LLVertexBuffer::setupVertexBufferFast(U32 data_mask) void* ptr = (void*)(base + mOffsets[TYPE_VERTEX]); glVertexAttribPointer(loc, 3, GL_FLOAT, GL_FALSE, LLVertexBuffer::sTypeSize[TYPE_VERTEX], ptr); } - } +} + +void LLVertexBuffer::setPositionData(const LLVector4a* data) +{ + bindGLBuffer(); + flush_vbo(GL_ARRAY_BUFFER, 0, sizeof(LLVector4a) * getNumVerts(), (U8*) data); +} -LLVertexBuffer::MappedRegion::MappedRegion(S32 type, S32 index, S32 count) -: mType(type), mIndex(index), mCount(count) -{ - mEnd = mIndex+mCount; -} +void LLVertexBuffer::setTexCoordData(const LLVector2* data) +{ + bindGLBuffer(); + flush_vbo(GL_ARRAY_BUFFER, mOffsets[TYPE_TEXCOORD0], mOffsets[TYPE_TEXCOORD0] + sTypeSize[TYPE_TEXCOORD0] * getNumVerts(), (U8*)data); +} + +void LLVertexBuffer::setColorData(const LLColor4U* data) +{ + bindGLBuffer(); + flush_vbo(GL_ARRAY_BUFFER, mOffsets[TYPE_COLOR], mOffsets[TYPE_COLOR] + sTypeSize[TYPE_COLOR] * getNumVerts(), (U8*) data); +} diff --git a/indra/llrender/llvertexbuffer.h b/indra/llrender/llvertexbuffer.h index bb7460fb2a..926d37b052 100644 --- a/indra/llrender/llvertexbuffer.h +++ b/indra/llrender/llvertexbuffer.h @@ -51,66 +51,15 @@ //============================================================================ -// gl name pools for dynamic and streaming buffers -class LLVBOPool -{ -public: - static U32 sBytesPooled; - static U32 sIndexBytesPooled; - - LLVBOPool(U32 vboUsage, U32 vboType); - - const U32 mUsage; - const U32 mType; - - //size MUST be a power of 2 - U8* allocate(U32& name, U32 size, bool for_seed = false); - - //size MUST be the size provided to allocate that returned the given name - void release(U32 name, U8* buffer, U32 size); - - //batch allocate buffers to be provided to the application on demand - void seedPool(); - - //destroy all records in mFreeList - void cleanup(); - - U32 genBuffer(); - void deleteBuffer(U32 name); - - class Record - { - public: - U32 mGLName; - U8* mClientData; - }; - - typedef std::list<Record> record_list_t; - std::vector<record_list_t> mFreeList; - std::vector<U32> mMissCount; - bool mMissCountDirty; // flag any changes to mFreeList or mMissCount - - //used to avoid calling glGenBuffers for every VBO creation - static U32 sNamePool[1024]; - static U32 sNameIdx; -}; - - -//============================================================================ // base class class LLPrivateMemoryPool; class LLVertexBuffer : public LLRefCount { public: - class MappedRegion + struct MappedRegion { - public: - S32 mType; - S32 mIndex; - S32 mCount; - S32 mEnd; - - MappedRegion(S32 type, S32 index, S32 count); + S32 mStart; + S32 mEnd; }; LLVertexBuffer(const LLVertexBuffer& rhs) @@ -125,12 +74,6 @@ public: return *this; } - static LLVBOPool sStreamVBOPool; - static LLVBOPool sDynamicVBOPool; - static LLVBOPool sDynamicCopyVBOPool; - static LLVBOPool sStreamIBOPool; - static LLVBOPool sDynamicIBOPool; - static std::list<U32> sAvailableVAOName; static U32 sCurVAOName; @@ -138,12 +81,10 @@ public: static bool sUseVAO; static bool sPreferStreamDraw; - static void seedPools(); - static U32 getVAOName(); static void releaseVAOName(U32 name); - static void initClass(bool use_vbo, bool no_vbo_mapping); + static void initClass(LLWindow* window); static void cleanupClass(); static void setupClientArrays(U32 data_mask); static void drawArrays(U32 mode, const std::vector<LLVector3>& pos); @@ -240,7 +181,7 @@ public: virtual void setBuffer(U32 data_mask); // calls setupVertexBuffer() if data_mask is not 0 void setBufferFast(U32 data_mask); // calls setupVertexBufferFast(), assumes data_mask is not 0 among other assumptions - void flush(); //flush pending data to GL memory + void flush(bool discard = false); //flush pending data to GL memory, if discard is true, discard previous VBO // allocate buffer bool allocateBuffer(S32 nverts, S32 nindices, bool create); virtual bool resizeBuffer(S32 newnverts, S32 newnindices); @@ -271,6 +212,10 @@ public: bool getMetallicRoughnessTexcoordStrider(LLStrider<LLVector2>& strider, S32 index=0, S32 count = -1, bool map_range = false); bool getEmissiveTexcoordStrider(LLStrider<LLVector2>& strider, S32 index=0, S32 count = -1, bool map_range = false); + void setPositionData(const LLVector4a* data); + void setTexCoordData(const LLVector2* data); + void setColorData(const LLColor4U* data); + bool useVBOs() const; bool isEmpty() const { return mEmpty; } diff --git a/indra/newview/app_settings/shaders/class1/deferred/attachmentAlphaMaskShadowF.glsl b/indra/newview/app_settings/shaders/class1/deferred/attachmentAlphaMaskShadowF.glsl deleted file mode 100644 index c2372fcbc0..0000000000 --- a/indra/newview/app_settings/shaders/class1/deferred/attachmentAlphaMaskShadowF.glsl +++ /dev/null @@ -1,64 +0,0 @@ -/** - * @file attachmentAlphaMaskShadowF.glsl - * - * $LicenseInfo:firstyear=2005&license=viewerlgpl$ - * Second Life Viewer Source Code - * Copyright (C) 2005, Linden Research, Inc. - * - * This library is free software; you can redistribute it and/or - * modify it under the terms of the GNU Lesser General Public - * License as published by the Free Software Foundation; - * version 2.1 of the License only. - * - * This library 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 - * Lesser General Public License for more details. - * - * You should have received a copy of the GNU Lesser General Public - * License along with this library; if not, write to the Free Software - * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA - * - * Linden Research, Inc., 945 Battery Street, San Francisco, CA 94111 USA - * $/LicenseInfo$ - */ - -#ifdef DEFINE_GL_FRAGCOLOR -out vec4 frag_color; -#else -#define frag_color gl_FragColor -#endif - -uniform float minimum_alpha; - -uniform sampler2D diffuseMap; - -VARYING vec4 post_pos; -VARYING vec2 vary_texcoord0; -VARYING float pos_w; -VARYING float target_pos_x; -VARYING vec4 vertex_color; - -void main() -{ - float alpha = texture2D(diffuseMap, vary_texcoord0.xy).a; - - if (alpha < 0.05) // treat as totally transparent - { - discard; - } - - if (alpha < minimum_alpha) // treat as semi-transparent - { - //if (fract(0.5*floor(target_pos_x / pos_w )) < 0.25) - { - discard; - } - } - - frag_color = vec4(1,1,1,1); - -#if !defined(DEPTH_CLAMP) - gl_FragDepth = max(post_pos.z/post_pos.w*0.5+0.5, 0.0); -#endif -} diff --git a/indra/newview/app_settings/shaders/class1/deferred/attachmentAlphaShadowF.glsl b/indra/newview/app_settings/shaders/class1/deferred/attachmentAlphaShadowF.glsl deleted file mode 100644 index d0a049d372..0000000000 --- a/indra/newview/app_settings/shaders/class1/deferred/attachmentAlphaShadowF.glsl +++ /dev/null @@ -1,68 +0,0 @@ -/** - * @file attachmentAlphaShadowF.glsl - * - * $LicenseInfo:firstyear=2005&license=viewerlgpl$ - * Second Life Viewer Source Code - * Copyright (C) 2005, Linden Research, Inc. - * - * This library is free software; you can redistribute it and/or - * modify it under the terms of the GNU Lesser General Public - * License as published by the Free Software Foundation; - * version 2.1 of the License only. - * - * This library 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 - * Lesser General Public License for more details. - * - * You should have received a copy of the GNU Lesser General Public - * License along with this library; if not, write to the Free Software - * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA - * - * Linden Research, Inc., 945 Battery Street, San Francisco, CA 94111 USA - * $/LicenseInfo$ - */ - -#ifdef DEFINE_GL_FRAGCOLOR -out vec4 frag_color; -#else -#define frag_color gl_FragColor -#endif - -uniform float minimum_alpha; -uniform sampler2D diffuseMap; - -VARYING float pos_w; -VARYING float target_pos_x; - -#if !defined(DEPTH_CLAMP) -VARYING vec4 post_pos; -#endif - -VARYING vec2 vary_texcoord0; -VARYING vec4 vertex_color; - -void main() -{ - float alpha = texture2D(diffuseMap, vary_texcoord0.xy).a * vertex_color.a; - - if (alpha < 0.05) // treat as totally transparent - { - discard; - } - - if (alpha < minimum_alpha) - { - if (fract(0.5*floor(target_pos_x / pos_w )) < 0.25) - { - discard; - } - } - - frag_color = vec4(1,1,1,1); - -#if !defined(DEPTH_CLAMP) - gl_FragDepth = max(post_pos.z/post_pos.w*0.5+0.5, 0.0); -#endif - -} diff --git a/indra/newview/app_settings/shaders/class1/deferred/attachmentAlphaShadowV.glsl b/indra/newview/app_settings/shaders/class1/deferred/attachmentAlphaShadowV.glsl deleted file mode 100644 index a7bf4d7780..0000000000 --- a/indra/newview/app_settings/shaders/class1/deferred/attachmentAlphaShadowV.glsl +++ /dev/null @@ -1,74 +0,0 @@ -/** - * @file attachmentAlphaShadowV.glsl - * $LicenseInfo:firstyear=2007&license=viewerlgpl$ - * Second Life Viewer Source Code - * Copyright (C) 2007, Linden Research, Inc. - * - * This library is free software; you can redistribute it and/or - * modify it under the terms of the GNU Lesser General Public - * License as published by the Free Software Foundation; - * version 2.1 of the License only. - * - * This library 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 - * Lesser General Public License for more details. - * - * You should have received a copy of the GNU Lesser General Public - * License along with this library; if not, write to the Free Software - * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA - * - * Linden Research, Inc., 945 Battery Street, San Francisco, CA 94111 USA - * $/LicenseInfo$ - */ - -uniform mat4 projection_matrix; -uniform mat4 modelview_matrix; -uniform mat4 texture_matrix0; -uniform float shadow_target_width; - -ATTRIBUTE vec4 diffuse_color; -ATTRIBUTE vec3 position; -ATTRIBUTE vec3 normal; -ATTRIBUTE vec2 texcoord0; - -mat4 getObjectSkinnedTransform(); -void passTextureIndex(); - -#if !defined(DEPTH_CLAMP) -VARYING vec4 post_pos; -#endif -VARYING vec2 vary_texcoord0; -VARYING float pos_w; -VARYING float target_pos_x; -VARYING vec4 vertex_color; - -void main() -{ - //transform vertex - mat4 mat = getObjectSkinnedTransform(); - - mat = modelview_matrix * mat; - vec3 pos = (mat*vec4(position.xyz, 1.0)).xyz; - - vec4 p = projection_matrix * vec4(pos, 1.0); - - pos_w = p.w; - - target_pos_x = 0.5 * (shadow_target_width - 1.0) * pos.x; - - vary_texcoord0 = (texture_matrix0 * vec4(texcoord0,0,1)).xy; - - vertex_color = diffuse_color; - -#if !defined(DEPTH_CLAMP) - p.z = max(p.z, -p.w+0.01); - post_pos = p; - gl_Position = p; -#else - gl_Position = p; -#endif - - passTextureIndex(); -} - diff --git a/indra/newview/app_settings/shaders/class1/deferred/avatarAlphaMaskShadowF.glsl b/indra/newview/app_settings/shaders/class1/deferred/avatarAlphaMaskShadowF.glsl index f231213ac8..44f67899e4 100644 --- a/indra/newview/app_settings/shaders/class1/deferred/avatarAlphaMaskShadowF.glsl +++ b/indra/newview/app_settings/shaders/class1/deferred/avatarAlphaMaskShadowF.glsl @@ -32,10 +32,6 @@ out vec4 frag_color; uniform float minimum_alpha; uniform sampler2D diffuseMap; -#if !defined(DEPTH_CLAMP) -VARYING vec4 post_pos; -#endif - VARYING float target_pos_x; VARYING float pos_w; VARYING vec2 vary_texcoord0; @@ -58,8 +54,4 @@ void main() } frag_color = vec4(1,1,1,1); - -#if !defined(DEPTH_CLAMP) - gl_FragDepth = max(post_pos.z/post_pos.w*0.5+0.5, 0.0); -#endif } diff --git a/indra/newview/app_settings/shaders/class1/deferred/avatarAlphaShadowF.glsl b/indra/newview/app_settings/shaders/class1/deferred/avatarAlphaShadowF.glsl index 0e66c722b6..9a0024ede9 100644 --- a/indra/newview/app_settings/shaders/class1/deferred/avatarAlphaShadowF.glsl +++ b/indra/newview/app_settings/shaders/class1/deferred/avatarAlphaShadowF.glsl @@ -33,10 +33,6 @@ uniform float minimum_alpha; uniform sampler2D diffuseMap; -#if !defined(DEPTH_CLAMP) -VARYING vec4 post_pos; -#endif - VARYING float pos_w; VARYING float target_pos_x; VARYING vec2 vary_texcoord0; @@ -60,9 +56,4 @@ void main() } frag_color = vec4(1,1,1,1); - -#if !defined(DEPTH_CLAMP) - gl_FragDepth = max(post_pos.z/post_pos.w*0.5+0.5, 0.0); -#endif - } diff --git a/indra/newview/app_settings/shaders/class1/deferred/avatarAlphaShadowV.glsl b/indra/newview/app_settings/shaders/class1/deferred/avatarAlphaShadowV.glsl index 40ac7b1f95..dfe7077db5 100644 --- a/indra/newview/app_settings/shaders/class1/deferred/avatarAlphaShadowV.glsl +++ b/indra/newview/app_settings/shaders/class1/deferred/avatarAlphaShadowV.glsl @@ -34,9 +34,6 @@ ATTRIBUTE vec3 position; ATTRIBUTE vec3 normal; ATTRIBUTE vec2 texcoord0; -#if !defined(DEPTH_CLAMP) -VARYING vec4 post_pos; -#endif VARYING float pos_w; VARYING float target_pos_x; VARYING vec2 vary_texcoord0; @@ -66,13 +63,7 @@ void main() vary_texcoord0 = (texture_matrix0 * vec4(texcoord0,0,1)).xy; -#if !defined(DEPTH_CLAMP) - post_pos = pos; - - gl_Position = vec4(pos.x, pos.y, pos.w*0.5, pos.w); -#else gl_Position = pos; -#endif passTextureIndex(); } diff --git a/indra/newview/app_settings/shaders/class1/deferred/avatarShadowF.glsl b/indra/newview/app_settings/shaders/class1/deferred/avatarShadowF.glsl index 50020a50d8..0c386ade6c 100644 --- a/indra/newview/app_settings/shaders/class1/deferred/avatarShadowF.glsl +++ b/indra/newview/app_settings/shaders/class1/deferred/avatarShadowF.glsl @@ -25,24 +25,12 @@ /*[EXTRA_CODE_HERE]*/ -#ifdef DEFINE_GL_FRAGCOLOR out vec4 frag_color; -#else -#define frag_color gl_FragColor -#endif uniform sampler2D diffuseMap; -#if !defined(DEPTH_CLAMP) -VARYING vec4 post_pos; -#endif - void main() { frag_color = vec4(1,1,1,1); - -#if !defined(DEPTH_CLAMP) - gl_FragDepth = max(post_pos.z/post_pos.w*0.5+0.5, 0.0); -#endif } diff --git a/indra/newview/app_settings/shaders/class1/deferred/avatarShadowV.glsl b/indra/newview/app_settings/shaders/class1/deferred/avatarShadowV.glsl index 91b25613e0..ac8a483a6a 100644 --- a/indra/newview/app_settings/shaders/class1/deferred/avatarShadowV.glsl +++ b/indra/newview/app_settings/shaders/class1/deferred/avatarShadowV.glsl @@ -31,10 +31,6 @@ ATTRIBUTE vec3 position; ATTRIBUTE vec3 normal; ATTRIBUTE vec2 texcoord0; -#if !defined(DEPTH_CLAMP) -VARYING vec4 post_pos; -#endif - void main() { vec4 pos; @@ -53,13 +49,7 @@ void main() norm = normalize(norm); pos = projection_matrix * pos; -#if !defined(DEPTH_CLAMP) - post_pos = pos; - - gl_Position = vec4(pos.x, pos.y, pos.w*0.5, pos.w); -#else gl_Position = pos; -#endif } diff --git a/indra/newview/app_settings/shaders/class1/deferred/shadowAlphaMaskF.glsl b/indra/newview/app_settings/shaders/class1/deferred/shadowAlphaMaskF.glsl index 06ebeb8281..68e9addc1b 100644 --- a/indra/newview/app_settings/shaders/class1/deferred/shadowAlphaMaskF.glsl +++ b/indra/newview/app_settings/shaders/class1/deferred/shadowAlphaMaskF.glsl @@ -69,8 +69,4 @@ void main() #endif frag_color = vec4(1,1,1,1); - -#if !defined(DEPTH_CLAMP) - gl_FragDepth = max(post_pos.z/post_pos.w*0.5+0.5, 0.0); -#endif } diff --git a/indra/newview/app_settings/shaders/class1/deferred/shadowAlphaMaskSkinnedV.glsl b/indra/newview/app_settings/shaders/class1/deferred/shadowAlphaMaskSkinnedV.glsl index 2b17aea75a..61c9e60744 100644 --- a/indra/newview/app_settings/shaders/class1/deferred/shadowAlphaMaskSkinnedV.glsl +++ b/indra/newview/app_settings/shaders/class1/deferred/shadowAlphaMaskSkinnedV.glsl @@ -57,11 +57,7 @@ void main() post_pos = pos; -#if !defined(DEPTH_CLAMP) - gl_Position = vec4(pos.x, pos.y, pos.w*0.5, pos.w); -#else gl_Position = pos; -#endif passTextureIndex(); diff --git a/indra/newview/app_settings/shaders/class1/deferred/shadowAlphaMaskV.glsl b/indra/newview/app_settings/shaders/class1/deferred/shadowAlphaMaskV.glsl index b6a0f0b165..40f8fc9894 100644 --- a/indra/newview/app_settings/shaders/class1/deferred/shadowAlphaMaskV.glsl +++ b/indra/newview/app_settings/shaders/class1/deferred/shadowAlphaMaskV.glsl @@ -47,11 +47,7 @@ void main() post_pos = pos; -#if !defined(DEPTH_CLAMP) - gl_Position = vec4(pos.x, pos.y, pos.w*0.5, pos.w); -#else gl_Position = pos; -#endif passTextureIndex(); diff --git a/indra/newview/app_settings/shaders/class1/deferred/shadowCubeV.glsl b/indra/newview/app_settings/shaders/class1/deferred/shadowCubeV.glsl index 0e74d2eb8a..9608e89169 100644 --- a/indra/newview/app_settings/shaders/class1/deferred/shadowCubeV.glsl +++ b/indra/newview/app_settings/shaders/class1/deferred/shadowCubeV.glsl @@ -27,10 +27,6 @@ uniform mat4 modelview_projection_matrix; ATTRIBUTE vec3 position; -#if !defined(DEPTH_CLAMP) -VARYING vec4 post_pos; -#endif - uniform vec3 box_center; uniform vec3 box_size; @@ -40,11 +36,5 @@ void main() vec3 p = position*box_size+box_center; vec4 pos = modelview_projection_matrix*vec4(p.xyz, 1.0); -#if !defined(DEPTH_CLAMP) - post_pos = pos; - - gl_Position = vec4(pos.x, pos.y, pos.w*0.5, pos.w); -#else gl_Position = pos; -#endif } diff --git a/indra/newview/app_settings/shaders/class1/deferred/shadowF.glsl b/indra/newview/app_settings/shaders/class1/deferred/shadowF.glsl index 1ea96918bb..406a7ef3b3 100644 --- a/indra/newview/app_settings/shaders/class1/deferred/shadowF.glsl +++ b/indra/newview/app_settings/shaders/class1/deferred/shadowF.glsl @@ -25,20 +25,9 @@ /*[EXTRA_CODE_HERE]*/ -#ifdef DEFINE_GL_FRAGCOLOR out vec4 frag_color; -#else -#define frag_color gl_FragColor -#endif - -VARYING vec4 post_pos; void main() { frag_color = vec4(1,1,1,1); - -#if !defined(DEPTH_CLAMP) - gl_FragDepth = max(post_pos.z/post_pos.w*0.5+0.5, 0.0); -#endif - } diff --git a/indra/newview/app_settings/shaders/class1/deferred/shadowSkinnedV.glsl b/indra/newview/app_settings/shaders/class1/deferred/shadowSkinnedV.glsl index bdf8e0854d..186ba49969 100644 --- a/indra/newview/app_settings/shaders/class1/deferred/shadowSkinnedV.glsl +++ b/indra/newview/app_settings/shaders/class1/deferred/shadowSkinnedV.glsl @@ -28,8 +28,6 @@ uniform mat4 projection_matrix; ATTRIBUTE vec3 position; -VARYING vec4 post_pos; - mat4 getObjectSkinnedTransform(); void main() @@ -41,12 +39,5 @@ void main() vec4 pos = (mat*vec4(position.xyz, 1.0)); pos = projection_matrix*pos; - post_pos = pos; - -#if !defined(DEPTH_CLAMP) - gl_Position = vec4(pos.x, pos.y, pos.w*0.5, pos.w); -#else gl_Position = pos; -#endif - } diff --git a/indra/newview/app_settings/shaders/class1/deferred/shadowV.glsl b/indra/newview/app_settings/shaders/class1/deferred/shadowV.glsl index 72bd0f0f34..bd62954ff8 100644 --- a/indra/newview/app_settings/shaders/class1/deferred/shadowV.glsl +++ b/indra/newview/app_settings/shaders/class1/deferred/shadowV.glsl @@ -25,21 +25,10 @@ uniform mat4 modelview_projection_matrix; -ATTRIBUTE vec3 position; - -VARYING vec4 post_pos; +in vec3 position; void main() { //transform vertex - vec4 pos = modelview_projection_matrix*vec4(position.xyz, 1.0); - - post_pos = pos; - -#if !defined(DEPTH_CLAMP) - gl_Position = vec4(pos.x, pos.y, pos.w*0.5, pos.w); -#else - gl_Position = pos; -#endif - + gl_Position = modelview_projection_matrix*vec4(position.xyz, 1.0); } diff --git a/indra/newview/app_settings/shaders/class1/deferred/treeShadowF.glsl b/indra/newview/app_settings/shaders/class1/deferred/treeShadowF.glsl index e34d75ba1d..c850a39d6f 100644 --- a/indra/newview/app_settings/shaders/class1/deferred/treeShadowF.glsl +++ b/indra/newview/app_settings/shaders/class1/deferred/treeShadowF.glsl @@ -25,18 +25,14 @@ /*[EXTRA_CODE_HERE]*/ -#ifdef DEFINE_GL_FRAGCOLOR + out vec4 frag_color; -#else -#define frag_color gl_FragColor -#endif uniform float minimum_alpha; uniform sampler2D diffuseMap; -VARYING vec4 post_pos; -VARYING vec2 vary_texcoord0; +in vec2 vary_texcoord0; void main() { @@ -48,6 +44,4 @@ void main() } frag_color = vec4(1,1,1,1); - - gl_FragDepth = max(post_pos.z/post_pos.w*0.5+0.5, 0.0); } diff --git a/indra/newview/app_settings/shaders/class1/deferred/treeShadowSkinnedV.glsl b/indra/newview/app_settings/shaders/class1/deferred/treeShadowSkinnedV.glsl index d9ca6d3a46..5a7cf369b5 100644 --- a/indra/newview/app_settings/shaders/class1/deferred/treeShadowSkinnedV.glsl +++ b/indra/newview/app_settings/shaders/class1/deferred/treeShadowSkinnedV.glsl @@ -30,8 +30,7 @@ uniform mat4 projection_matrix; ATTRIBUTE vec3 position; ATTRIBUTE vec2 texcoord0; -VARYING vec4 post_pos; -VARYING vec2 vary_texcoord0; +out vec2 vary_texcoord0; mat4 getObjectSkinnedTransform(); @@ -43,11 +42,7 @@ void main() mat = modelview_matrix * mat; vec4 pos = mat * vec4(position.xyz, 1.0); - pos = projection_matrix * pos; - - post_pos = pos; - - gl_Position = vec4(pos.x, pos.y, pos.w*0.5, pos.w); + gl_Position = projection_matrix * pos; vary_texcoord0 = (texture_matrix0 * vec4(texcoord0,0,1)).xy; } diff --git a/indra/newview/app_settings/shaders/class1/deferred/treeShadowV.glsl b/indra/newview/app_settings/shaders/class1/deferred/treeShadowV.glsl index e472a75304..3102c870fc 100644 --- a/indra/newview/app_settings/shaders/class1/deferred/treeShadowV.glsl +++ b/indra/newview/app_settings/shaders/class1/deferred/treeShadowV.glsl @@ -26,20 +26,15 @@ uniform mat4 texture_matrix0; uniform mat4 modelview_projection_matrix; -ATTRIBUTE vec3 position; -ATTRIBUTE vec2 texcoord0; +in vec3 position; +in vec2 texcoord0; -VARYING vec4 post_pos; -VARYING vec2 vary_texcoord0; +out vec2 vary_texcoord0; void main() { //transform vertex - vec4 pos = modelview_projection_matrix*vec4(position.xyz, 1.0); - - post_pos = pos; - - gl_Position = vec4(pos.x, pos.y, pos.w*0.5, pos.w); + gl_Position = modelview_projection_matrix*vec4(position.xyz, 1.0); vary_texcoord0 = (texture_matrix0 * vec4(texcoord0,0,1)).xy; } diff --git a/indra/newview/app_settings/shaders/class1/deferred/attachmentShadowF.glsl b/indra/newview/app_settings/shaders/class1/interface/copyF.glsl index 8e9a5fcd41..65d3c20091 100644 --- a/indra/newview/app_settings/shaders/class1/deferred/attachmentShadowF.glsl +++ b/indra/newview/app_settings/shaders/class1/interface/copyF.glsl @@ -1,8 +1,9 @@ /** - * @file avatarShadowF.glsl - * $LicenseInfo:firstyear=2007&license=viewerlgpl$ + * @file copyF.glsl + * + * $LicenseInfo:firstyear=2023&license=viewerlgpl$ * Second Life Viewer Source Code - * Copyright (C) 2007, Linden Research, Inc. + * Copyright (C) 2023, Linden Research, Inc. * * This library is free software; you can redistribute it and/or * modify it under the terms of the GNU Lesser General Public @@ -22,20 +23,18 @@ * $/LicenseInfo$ */ -/*[EXTRA_CODE_HERE]*/ - -#ifdef DEFINE_GL_FRAGCOLOR -out vec4 frag_color; -#else -#define frag_color gl_FragColor -#endif +in vec2 tc; +uniform sampler2D depthMap; uniform sampler2D diffuseMap; -VARYING vec2 vary_texcoord0; +out vec4 frag_color; void main() { - frag_color = vec4(1,1,1,1); + frag_color = texture(diffuseMap, tc); +#if defined(COPY_DEPTH) + gl_FragDepth = texture(depthMap, tc).r; +#endif } diff --git a/indra/newview/app_settings/shaders/class1/deferred/attachmentShadowV.glsl b/indra/newview/app_settings/shaders/class1/interface/copyV.glsl index 0fa0edfd67..ace5da6578 100644 --- a/indra/newview/app_settings/shaders/class1/deferred/attachmentShadowV.glsl +++ b/indra/newview/app_settings/shaders/class1/interface/copyV.glsl @@ -1,8 +1,9 @@ /** - * @file attachmentShadowV.glsl - * $LicenseInfo:firstyear=2007&license=viewerlgpl$ + * @file copyV.glsl + * + * $LicenseInfo:firstyear=2023&license=viewerlgpl$ * Second Life Viewer Source Code - * Copyright (C) 2007, Linden Research, Inc. + * Copyright (C) 2023, Linden Research, Inc. * * This library is free software; you can redistribute it and/or * modify it under the terms of the GNU Lesser General Public @@ -22,29 +23,12 @@ * $/LicenseInfo$ */ -uniform mat4 projection_matrix; -uniform mat4 modelview_matrix; -uniform mat4 texture_matrix0; -ATTRIBUTE vec3 position; -ATTRIBUTE vec2 texcoord0; +in vec3 position; +out vec2 tc; -mat4 getObjectSkinnedTransform(); - -void main() +void main() { - //transform vertex - mat4 mat = getObjectSkinnedTransform(); - - mat = modelview_matrix * mat; - vec3 pos = (mat*vec4(position.xyz, 1.0)).xyz; - - - vec4 p = projection_matrix * vec4(pos, 1.0); -#if !defined(DEPTH_CLAMP) - p.z = max(p.z, -p.w+0.01); - gl_Position = p; -#else - gl_Position = p; -#endif + tc = position.xy * 0.5 + 0.5; + gl_Position = vec4(position, 1.0); } diff --git a/indra/newview/lldrawpoolalpha.cpp b/indra/newview/lldrawpoolalpha.cpp index 3ce2ced255..07381acd25 100644 --- a/indra/newview/lldrawpoolalpha.cpp +++ b/indra/newview/lldrawpoolalpha.cpp @@ -600,6 +600,16 @@ void LLDrawPoolAlpha::renderAlpha(U32 mask, bool depth_only, bool rigged) begin = gPipeline.beginAlphaGroups(); end = gPipeline.endAlphaGroups(); } + + LLEnvironment& env = LLEnvironment::instance(); + F32 water_height = env.getWaterHeight(); + + bool above_water = getType() == LLDrawPool::POOL_ALPHA_POST_WATER; + if (LLPipeline::sUnderWaterRender) + { + above_water = !above_water; + } + for (LLCullResult::sg_iterator i = begin; i != end; ++i) { @@ -611,6 +621,25 @@ void LLDrawPoolAlpha::renderAlpha(U32 mask, bool depth_only, bool rigged) if (group->getSpatialPartition()->mRenderByGroup && !group->isDead()) { + + LLSpatialBridge* bridge = group->getSpatialPartition()->asBridge(); + const LLVector4a* ext = bridge ? bridge->getSpatialExtents() : group->getExtents(); + + if (above_water) + { // reject any spatial groups that have no part above water + if (ext[1].getF32ptr()[2] < water_height) + { + continue; + } + } + else + { // reject any spatial groups that he no part below water + if (ext[0].getF32ptr()[2] > water_height) + { + continue; + } + } + static std::vector<LLDrawInfo*> emissives; static std::vector<LLDrawInfo*> rigged_emissives; emissives.resize(0); diff --git a/indra/newview/lldrawpoolwater.cpp b/indra/newview/lldrawpoolwater.cpp index 77da29061d..c2fe52683b 100644 --- a/indra/newview/lldrawpoolwater.cpp +++ b/indra/newview/lldrawpoolwater.cpp @@ -133,11 +133,28 @@ void LLDrawPoolWater::beginPostDeferredPass(S32 pass) // reflections and refractions LLRenderTarget& src = gPipeline.mRT->screen; LLRenderTarget& dst = gPipeline.mWaterDis; + +#if 0 dst.copyContents(src, 0, 0, src.getWidth(), src.getHeight(), 0, 0, dst.getWidth(), dst.getHeight(), GL_COLOR_BUFFER_BIT | GL_DEPTH_BUFFER_BIT, GL_NEAREST); +#else + dst.bindTarget(); + gCopyDepthProgram.bind(); + + S32 diff_map = gCopyDepthProgram.enableTexture(LLShaderMgr::DIFFUSE_MAP); + S32 depth_map = gCopyDepthProgram.enableTexture(LLShaderMgr::DEFERRED_DEPTH); + + gGL.getTexUnit(diff_map)->bind(&src); + gGL.getTexUnit(depth_map)->bind(&src, true); + + gPipeline.mScreenTriangleVB->setBuffer(LLVertexBuffer::MAP_VERTEX); + gPipeline.mScreenTriangleVB->drawArrays(LLRender::TRIANGLES, 0, 3); + + dst.flush(); +#endif } } diff --git a/indra/newview/llgltfmateriallist.cpp b/indra/newview/llgltfmateriallist.cpp index 9539ffc700..9399342a61 100644 --- a/indra/newview/llgltfmateriallist.cpp +++ b/indra/newview/llgltfmateriallist.cpp @@ -297,7 +297,7 @@ public: // object not ready to receive override data, queue for later gGLTFMaterialList.queueOverrideUpdate(object_override.mObjectId, results[i].mSide, results[i].mMaterial); } - else if (obj && obj->getTE(i) && obj->getTE(i)->isSelected()) + else if (obj && obj->getTE(results[i].mSide) && obj->getTE(results[i].mSide)->isSelected()) { doSelectionCallbacks(object_override.mObjectId, results[i].mSide); } @@ -305,7 +305,7 @@ public: else { // unblock material editor - if (obj && obj->getTE(i) && obj->getTE(i)->isSelected()) + if (obj && obj->getTE(results[i].mSide) && obj->getTE(results[i].mSide)->isSelected()) { doSelectionCallbacks(object_override.mObjectId, results[i].mSide); } diff --git a/indra/newview/llselectmgr.cpp b/indra/newview/llselectmgr.cpp index cb1e46068e..2b6578c272 100644 --- a/indra/newview/llselectmgr.cpp +++ b/indra/newview/llselectmgr.cpp @@ -5853,7 +5853,6 @@ void LLSelectMgr::processObjectProperties(LLMessageSystem* msg, void** user_data if (can_copy && can_transfer) { - // this should be the only place that saved textures is called node->saveTextures(texture_ids); } diff --git a/indra/newview/lltooldraganddrop.cpp b/indra/newview/lltooldraganddrop.cpp index cde33a5f96..53bc77daa9 100644 --- a/indra/newview/lltooldraganddrop.cpp +++ b/indra/newview/lltooldraganddrop.cpp @@ -2080,24 +2080,101 @@ EAcceptance LLToolDragAndDrop::dad3dApplyToObject( { if (cargo_type == DAD_TEXTURE) { + LLSelectNode *nodep = nullptr; + if (obj->isSelected()) + { + // update object's saved textures + nodep = LLSelectMgr::getInstance()->getSelection()->findNode(obj); + } + if((mask & MASK_SHIFT)) { dropTextureAllFaces(obj, item, mSource, mSourceID); + + // If user dropped a texture onto face it implies + // applying texture now without cancel, save to selection + if (nodep) + { + uuid_vec_t texture_ids; + S32 num_faces = obj->getNumTEs(); + for (S32 face = 0; face < num_faces; face++) + { + LLViewerTexture *tex = obj->getTEImage(face); + if (tex != nullptr) + { + texture_ids.push_back(tex->getID()); + } + else + { + texture_ids.push_back(LLUUID::null); + } + } + nodep->saveTextures(texture_ids); + } } else { dropTextureOneFace(obj, face, item, mSource, mSourceID); + + // If user dropped a texture onto face it implies + // applying texture now without cancel, save to selection + LLPanelFace* panel_face = gFloaterTools->getPanelFace(); + if (nodep + && gFloaterTools->getVisible() + && panel_face + && panel_face->getTextureDropChannel() == 0 /*texture*/ + && nodep->mSavedGLTFMaterialIds.size() > face) + { + LLViewerTexture *tex = obj->getTEImage(face); + if (tex != nullptr) + { + nodep->mSavedTextures[face] = tex->getID(); + } + else + { + nodep->mSavedTextures[face] = LLUUID::null; + } + } } } else if (cargo_type == DAD_MATERIAL) { + LLSelectNode *nodep = nullptr; + if (obj->isSelected()) + { + // update object's saved materials + nodep = LLSelectMgr::getInstance()->getSelection()->findNode(obj); + } + + // If user dropped a material onto face it implies + // applying texture now without cancel, save to selection if ((mask & MASK_SHIFT)) { dropMaterialAllFaces(obj, item, mSource, mSourceID); + + if (nodep) + { + uuid_vec_t material_ids; + S32 num_faces = obj->getNumTEs(); + for (S32 face = 0; face < num_faces; face++) + { + material_ids.push_back(obj->getRenderMaterialID(face)); + } + nodep->saveGLTFMaterialIds(material_ids); + } } else { dropMaterialOneFace(obj, face, item, mSource, mSourceID); + + // If user dropped a material onto face it implies + // applying texture now without cancel, save to selection + if (nodep + && gFloaterTools->getVisible() + && nodep->mSavedGLTFMaterialIds.size() > face) + { + nodep->mSavedGLTFMaterialIds[face] = obj->getRenderMaterialID(face); + } } } else if (cargo_type == DAD_MESH) diff --git a/indra/newview/llviewerdisplay.cpp b/indra/newview/llviewerdisplay.cpp index 58b1716caa..d1a89a5846 100644 --- a/indra/newview/llviewerdisplay.cpp +++ b/indra/newview/llviewerdisplay.cpp @@ -710,11 +710,11 @@ void display(BOOL rebuild, F32 zoom_factor, int subfield, BOOL for_snapshot) if (!for_snapshot) { - if (gFrameCount > 1) - { //for some reason, ATI 4800 series will error out if you - //try to generate a shadow before the first frame is through - gPipeline.generateSunShadow(*LLViewerCamera::getInstance()); - } + if (gFrameCount > 1 && !for_snapshot) + { //for some reason, ATI 4800 series will error out if you + //try to generate a shadow before the first frame is through + gPipeline.generateSunShadow(*LLViewerCamera::getInstance()); + } LLVertexBuffer::unbind(); @@ -936,8 +936,7 @@ void display(BOOL rebuild, F32 zoom_factor, int subfield, BOOL for_snapshot) else { gPipeline.renderGeom(*LLViewerCamera::getInstance(), TRUE); - } - + } gGL.setColorMask(true, true); //store this frame's modelview matrix for use @@ -967,13 +966,6 @@ void display(BOOL rebuild, F32 zoom_factor, int subfield, BOOL for_snapshot) LLRenderTarget &rt = (gPipeline.sRenderDeferred ? gPipeline.mRT->deferredScreen : gPipeline.mRT->screen); rt.flush(); - /*if (rt.sUseFBO) - { - LLRenderTarget::copyContentsToFramebuffer(rt, 0, 0, rt.getWidth(), rt.getHeight(), 0, 0, rt.getWidth(), - rt.getHeight(), GL_DEPTH_BUFFER_BIT | GL_STENCIL_BUFFER_BIT, - GL_NEAREST); - }*/ - if (LLPipeline::sRenderDeferred) { gPipeline.renderDeferredLighting(); @@ -1364,6 +1356,8 @@ void render_ui(F32 zoom_factor, int subfield) render_hud_elements(); render_hud_attachments(); + LLGLState::checkStates(); + LLGLSDefault gls_default; LLGLSUIDefault gls_ui; { @@ -1378,6 +1372,7 @@ void render_ui(F32 zoom_factor, int subfield) if (!gDisconnected) { LL_PROFILE_ZONE_NAMED_CATEGORY_UI("UI 3D"); //LL_RECORD_BLOCK_TIME(FTM_RENDER_UI_3D); + LLGLState::checkStates(); render_ui_3d(); LLGLState::checkStates(); } diff --git a/indra/newview/llviewershadermgr.cpp b/indra/newview/llviewershadermgr.cpp index 757a946fda..5d79c8e6b2 100644 --- a/indra/newview/llviewershadermgr.cpp +++ b/indra/newview/llviewershadermgr.cpp @@ -93,6 +93,8 @@ LLGLSLShader gDownsampleDepthRectProgram; LLGLSLShader gAlphaMaskProgram; LLGLSLShader gBenchmarkProgram; LLGLSLShader gReflectionProbeDisplayProgram; +LLGLSLShader gCopyProgram; +LLGLSLShader gCopyDepthProgram; //object shaders LLGLSLShader gObjectSimpleProgram; @@ -222,9 +224,6 @@ LLGLSLShader gDeferredSkinnedShadowFullbrightAlphaMaskProgram; LLGLSLShader gDeferredAvatarShadowProgram; LLGLSLShader gDeferredAvatarAlphaShadowProgram; LLGLSLShader gDeferredAvatarAlphaMaskShadowProgram; -LLGLSLShader gDeferredAttachmentShadowProgram; -LLGLSLShader gDeferredAttachmentAlphaShadowProgram; -LLGLSLShader gDeferredAttachmentAlphaMaskShadowProgram; LLGLSLShader gDeferredAlphaProgram; LLGLSLShader gDeferredSkinnedAlphaProgram; LLGLSLShader gDeferredAlphaImpostorProgram; @@ -1271,9 +1270,6 @@ BOOL LLViewerShaderMgr::loadShadersDeferred() gDeferredAvatarShadowProgram.unload(); gDeferredAvatarAlphaShadowProgram.unload(); gDeferredAvatarAlphaMaskShadowProgram.unload(); - gDeferredAttachmentShadowProgram.unload(); - gDeferredAttachmentAlphaShadowProgram.unload(); - gDeferredAttachmentAlphaMaskShadowProgram.unload(); gDeferredAvatarProgram.unload(); gDeferredAvatarAlphaProgram.unload(); gDeferredAlphaProgram.unload(); @@ -2640,7 +2636,6 @@ BOOL LLViewerShaderMgr::loadShadersDeferred() success = gDeferredAvatarAlphaShadowProgram.createShader(NULL, NULL); llassert(success); } - if (success) { gDeferredAvatarAlphaMaskShadowProgram.mName = "Deferred Avatar Alpha Mask Shadow Shader"; @@ -2655,43 +2650,6 @@ BOOL LLViewerShaderMgr::loadShadersDeferred() if (success) { - gDeferredAttachmentShadowProgram.mName = "Deferred Attachment Shadow Shader"; - gDeferredAttachmentShadowProgram.mFeatures.hasObjectSkinning = true; - - gDeferredAttachmentShadowProgram.mShaderFiles.clear(); - gDeferredAttachmentShadowProgram.mShaderFiles.push_back(make_pair("deferred/attachmentShadowV.glsl", GL_VERTEX_SHADER)); - gDeferredAttachmentShadowProgram.mShaderFiles.push_back(make_pair("deferred/attachmentShadowF.glsl", GL_FRAGMENT_SHADER)); - gDeferredAttachmentShadowProgram.mShaderLevel = mShaderLevel[SHADER_DEFERRED]; - success = gDeferredAttachmentShadowProgram.createShader(NULL, NULL); - llassert(success); - } - - if (success) - { - gDeferredAttachmentAlphaShadowProgram.mName = "Deferred Attachment Alpha Shadow Shader"; - gDeferredAttachmentAlphaShadowProgram.mFeatures.hasObjectSkinning = true; - gDeferredAttachmentAlphaShadowProgram.mShaderFiles.clear(); - gDeferredAttachmentAlphaShadowProgram.mShaderFiles.push_back(make_pair("deferred/attachmentAlphaShadowV.glsl", GL_VERTEX_SHADER)); - gDeferredAttachmentAlphaShadowProgram.mShaderFiles.push_back(make_pair("deferred/attachmentAlphaShadowF.glsl", GL_FRAGMENT_SHADER)); - gDeferredAttachmentAlphaShadowProgram.mShaderLevel = mShaderLevel[SHADER_DEFERRED]; - success = gDeferredAttachmentAlphaShadowProgram.createShader(NULL, NULL); - llassert(success); - } - - if (success) - { - gDeferredAttachmentAlphaMaskShadowProgram.mName = "Deferred Attachment Alpha Mask Shadow Shader"; - gDeferredAttachmentAlphaMaskShadowProgram.mFeatures.hasObjectSkinning = true; - gDeferredAttachmentAlphaMaskShadowProgram.mShaderFiles.clear(); - gDeferredAttachmentAlphaMaskShadowProgram.mShaderFiles.push_back(make_pair("deferred/attachmentAlphaShadowV.glsl", GL_VERTEX_SHADER)); - gDeferredAttachmentAlphaMaskShadowProgram.mShaderFiles.push_back(make_pair("deferred/attachmentAlphaMaskShadowF.glsl", GL_FRAGMENT_SHADER)); - gDeferredAttachmentAlphaMaskShadowProgram.mShaderLevel = mShaderLevel[SHADER_DEFERRED]; - success = gDeferredAttachmentAlphaMaskShadowProgram.createShader(NULL, NULL); - llassert(success); - } - - if (success) - { gDeferredTerrainProgram.mName = "Deferred Terrain Shader"; gDeferredTerrainProgram.mFeatures.encodesNormal = true; gDeferredTerrainProgram.mFeatures.hasSrgb = true; @@ -3916,6 +3874,27 @@ BOOL LLViewerShaderMgr::loadShadersInterface() success = gReflectionProbeDisplayProgram.createShader(NULL, NULL); } + if (success) + { + gCopyProgram.mName = "Copy Shader"; + gCopyProgram.mShaderFiles.clear(); + gCopyProgram.mShaderFiles.push_back(make_pair("interface/copyV.glsl", GL_VERTEX_SHADER)); + gCopyProgram.mShaderFiles.push_back(make_pair("interface/copyF.glsl", GL_FRAGMENT_SHADER)); + gCopyProgram.mShaderLevel = mShaderLevel[SHADER_INTERFACE]; + success = gCopyProgram.createShader(NULL, NULL); + } + + if (success) + { + gCopyDepthProgram.mName = "Copy Depth Shader"; + gCopyDepthProgram.mShaderFiles.clear(); + gCopyDepthProgram.mShaderFiles.push_back(make_pair("interface/copyV.glsl", GL_VERTEX_SHADER)); + gCopyDepthProgram.mShaderFiles.push_back(make_pair("interface/copyF.glsl", GL_FRAGMENT_SHADER)); + gCopyDepthProgram.clearPermutations(); + gCopyDepthProgram.addPermutation("COPY_DEPTH", "1"); + gCopyDepthProgram.mShaderLevel = mShaderLevel[SHADER_INTERFACE]; + success = gCopyDepthProgram.createShader(NULL, NULL); + } if (success) { diff --git a/indra/newview/llviewershadermgr.h b/indra/newview/llviewershadermgr.h index 9ded72c6a7..593a8c14f3 100644 --- a/indra/newview/llviewershadermgr.h +++ b/indra/newview/llviewershadermgr.h @@ -164,6 +164,8 @@ extern LLGLSLShader gDownsampleDepthProgram; extern LLGLSLShader gDownsampleDepthRectProgram; extern LLGLSLShader gBenchmarkProgram; extern LLGLSLShader gReflectionProbeDisplayProgram; +extern LLGLSLShader gCopyProgram; +extern LLGLSLShader gCopyDepthProgram; //output tex0[tc0] + tex1[tc1] extern LLGLSLShader gTwoTextureAddProgram; @@ -287,9 +289,6 @@ extern LLGLSLShader gFXAAProgram; extern LLGLSLShader gDeferredPostNoDoFProgram; extern LLGLSLShader gDeferredPostGammaCorrectProgram; extern LLGLSLShader gDeferredAvatarShadowProgram; -extern LLGLSLShader gDeferredAttachmentShadowProgram; -extern LLGLSLShader gDeferredAttachmentAlphaShadowProgram; -extern LLGLSLShader gDeferredAttachmentAlphaMaskShadowProgram; extern LLGLSLShader gDeferredAvatarAlphaShadowProgram; extern LLGLSLShader gDeferredAvatarAlphaMaskShadowProgram; extern LLGLSLShader gDeferredAlphaProgram; diff --git a/indra/newview/llviewerwindow.cpp b/indra/newview/llviewerwindow.cpp index bc4f00bd3f..5848cbfd9d 100644 --- a/indra/newview/llviewerwindow.cpp +++ b/indra/newview/llviewerwindow.cpp @@ -658,12 +658,6 @@ public: } - addText(xpos, ypos, llformat("%d MB Index Data (%d MB Pooled, %d KIndices)", LLVertexBuffer::sAllocatedIndexBytes/(1024*1024), LLVBOPool::sIndexBytesPooled/(1024*1024), LLVertexBuffer::sIndexCount/1024)); - ypos += y_inc; - - addText(xpos, ypos, llformat("%d MB Vertex Data (%d MB Pooled, %d KVerts)", LLVertexBuffer::sAllocatedBytes/(1024*1024), LLVBOPool::sBytesPooled/(1024*1024), LLVertexBuffer::sVertexCount/1024)); - ypos += y_inc; - addText(xpos, ypos, llformat("%d Vertex Buffers", LLVertexBuffer::sGLCount)); ypos += y_inc; @@ -1974,7 +1968,7 @@ LLViewerWindow::LLViewerWindow(const Params& p) LL_DEBUGS("Window") << "Loading feature tables." << LL_ENDL; // Initialize OpenGL Renderer - LLVertexBuffer::initClass(gSavedSettings.getBOOL("RenderVBOEnable"), gSavedSettings.getBOOL("RenderVBOMappingDisable")); + LLVertexBuffer::initClass(mWindow); LL_INFOS("RenderInit") << "LLVertexBuffer initialization done." << LL_ENDL ; gGL.init(true); diff --git a/indra/newview/pipeline.cpp b/indra/newview/pipeline.cpp index 9851d4bc6a..66083470e2 100644 --- a/indra/newview/pipeline.cpp +++ b/indra/newview/pipeline.cpp @@ -855,17 +855,13 @@ bool LLPipeline::allocateScreenBuffer(U32 resX, U32 resY, U32 samples) S32 shadow_detail = RenderShadowDetail; bool ssao = RenderDeferredSSAO; - const U32 occlusion_divisor = 3; - //allocate deferred rendering color buffers - if (!mRT->deferredScreen.allocate(resX, resY, GL_RGBA, TRUE, FALSE, LLTexUnit::TT_TEXTURE, FALSE, samples)) return false; - //if (!mRT->deferredDepth.allocate(resX, resY, 0, TRUE, FALSE, LLTexUnit::TT_TEXTURE, FALSE, samples)) return false; - if (!mRT->occlusionDepth.allocate(resX/occlusion_divisor, resY/occlusion_divisor, 0, TRUE, FALSE, LLTexUnit::TT_TEXTURE, FALSE, samples)) return false; + if (!mRT->deferredScreen.allocate(resX, resY, GL_RGBA, true, true, LLTexUnit::TT_TEXTURE, false, samples)) return false; if (!addDeferredAttachments(mRT->deferredScreen)) return false; GLuint screenFormat = GL_RGBA16; - if (!mRT->screen.allocate(resX, resY, screenFormat, FALSE, FALSE, LLTexUnit::TT_TEXTURE, FALSE, samples)) return false; + if (!mRT->screen.allocate(resX, resY, screenFormat, FALSE, true, LLTexUnit::TT_TEXTURE, FALSE, samples)) return false; mRT->deferredScreen.shareDepthBuffer(mRT->screen); @@ -905,9 +901,7 @@ bool LLPipeline::allocateScreenBuffer(U32 resX, U32 resY, U32 samples) mRT->fxaaBuffer.release(); mRT->screen.release(); mRT->deferredScreen.release(); //make sure to release any render targets that share a depth buffer with mRT->deferredScreen first - //mRT->deferredDepth.release(); - mRT->occlusionDepth.release(); - + if (!mRT->screen.allocate(resX, resY, GL_RGBA, TRUE, TRUE, LLTexUnit::TT_TEXTURE, FALSE)) return false; } @@ -926,8 +920,6 @@ bool LLPipeline::allocateShadowBuffer(U32 resX, U32 resY) LL_PROFILE_ZONE_SCOPED_CATEGORY_DISPLAY; S32 shadow_detail = RenderShadowDetail; - const U32 occlusion_divisor = 3; - F32 scale = llmax(0.f, RenderShadowResolutionScale); U32 sun_shadow_map_width = BlurHappySize(resX, scale); U32 sun_shadow_map_height = BlurHappySize(resY, scale); @@ -936,12 +928,7 @@ bool LLPipeline::allocateShadowBuffer(U32 resX, U32 resY) { //allocate 4 sun shadow maps for (U32 i = 0; i < 4; i++) { - if (!mRT->shadow[i].allocate(sun_shadow_map_width, sun_shadow_map_height, 0, TRUE, FALSE, LLTexUnit::TT_TEXTURE)) - { - return false; - } - - if (!mRT->shadowOcclusion[i].allocate(sun_shadow_map_width / occlusion_divisor, sun_shadow_map_height / occlusion_divisor, 0, TRUE, FALSE, LLTexUnit::TT_TEXTURE)) + if (!mRT->shadow[i].allocate(sun_shadow_map_width, sun_shadow_map_height, 0, true, true, LLTexUnit::TT_TEXTURE)) { return false; } @@ -966,11 +953,7 @@ bool LLPipeline::allocateShadowBuffer(U32 resX, U32 resY) U32 spot_shadow_map_height = height; for (U32 i = 0; i < 2; i++) { - if (!mSpotShadow[i].allocate(spot_shadow_map_width, spot_shadow_map_height, 0, TRUE, FALSE)) - { - return false; - } - if (!mSpotShadowOcclusion[i].allocate(spot_shadow_map_width / occlusion_divisor, height / occlusion_divisor, 0, TRUE, FALSE)) + if (!mSpotShadow[i].allocate(spot_shadow_map_width, spot_shadow_map_height, 0, true, true)) { return false; } @@ -1001,7 +984,7 @@ bool LLPipeline::allocateShadowBuffer(U32 resX, U32 resY) } } - if (shadow_detail > 1) + if (shadow_detail > 1 && !gCubeSnapshot) { for (U32 i = 0; i < 2; i++) { @@ -1198,7 +1181,6 @@ void LLPipeline::releaseScreenBuffers() mRT->deferredScreen.release(); mRT->deferredDepth.release(); mRT->deferredLight.release(); - mRT->occlusionDepth.release(); } @@ -1206,7 +1188,6 @@ void LLPipeline::releaseSunShadowTarget(U32 index) { llassert(index < 4); mRT->shadow[index].release(); - mRT->shadowOcclusion[index].release(); } void LLPipeline::releaseSunShadowTargets() @@ -1224,7 +1205,6 @@ void LLPipeline::releaseSpotShadowTargets() for (U32 i = 0; i < 2; i++) { mSpotShadow[i].release(); - mSpotShadowOcclusion[i].release(); } } } @@ -1239,7 +1219,7 @@ void LLPipeline::createGLBuffers() if (LLPipeline::sRenderTransparentWater) { //water reflection texture U32 res = (U32) llmax(gSavedSettings.getS32("RenderWaterRefResolution"), 512); - mWaterDis.allocate(res,res,GL_RGBA,TRUE,FALSE,LLTexUnit::TT_TEXTURE); + mWaterDis.allocate(res,res,GL_RGBA,true,true,LLTexUnit::TT_TEXTURE); } // Use FBO for bake tex @@ -2353,16 +2333,35 @@ bool LLPipeline::getVisibleExtents(LLCamera& camera, LLVector3& min, LLVector3& static LLTrace::BlockTimerStatHandle FTM_CULL("Object Culling"); -void LLPipeline::updateCull(LLCamera& camera, LLCullResult& result, LLPlane* planep) +void LLPipeline::updateCull(LLCamera& camera, LLCullResult& result) { - static LLCachedControl<bool> use_occlusion(gSavedSettings,"UseOcclusion"); - static bool can_use_occlusion = LLFeatureManager::getInstance()->isFeatureAvailable("UseOcclusion"); - LL_PROFILE_ZONE_SCOPED_CATEGORY_PIPELINE; //LL_RECORD_BLOCK_TIME(FTM_CULL); + LL_PROFILE_GPU_ZONE("updateCull"); // should always be zero GPU time, but drop a timer to flush stuff out - if (planep != nullptr) + bool water_clip = !sRenderTransparentWater; + + if (water_clip) { - camera.setUserClipPlane(*planep); + + LLVector3 pnorm; + + F32 water_height = LLEnvironment::instance().getWaterHeight(); + + if (sUnderWaterRender) + { + //camera is below water, cull above water + pnorm.setVec(0, 0, 1); + } + else + { + //camera is above water, cull below water + pnorm = LLVector3(0, 0, -1); + } + + LLPlane plane; + plane.setVec(LLVector3(0, 0, water_height), pnorm); + + camera.setUserClipPlane(plane); } else { @@ -2373,56 +2372,6 @@ void LLPipeline::updateCull(LLCamera& camera, LLCullResult& result, LLPlane* pla sCull->clear(); - bool to_texture = LLPipeline::sUseOcclusion > 1 && gPipeline.shadersLoaded(); - - if (to_texture) - { - if (LLPipeline::sRenderDeferred && can_use_occlusion) - { - mRT->occlusionDepth.bindTarget(); - } - else - { - mRT->screen.bindTarget(); - } - } - - if (sUseOcclusion > 1) - { - gGL.setColorMask(false, false); - } - - gGL.matrixMode(LLRender::MM_PROJECTION); - gGL.pushMatrix(); - gGL.loadMatrix(gGLLastProjection); - gGL.matrixMode(LLRender::MM_MODELVIEW); - gGL.pushMatrix(); - gGLLastMatrix = NULL; - gGL.loadMatrix(gGLLastModelView); - - LLGLDisable blend(GL_BLEND); - LLGLDisable test(GL_ALPHA_TEST); - gGL.getTexUnit(0)->unbind(LLTexUnit::TT_TEXTURE); - - LLGLDepthTest depth(GL_TRUE, GL_FALSE); - - bool bound_shader = false; - if (gPipeline.shadersLoaded() && LLGLSLShader::sCurBoundShader == 0) - { //if no shader is currently bound, use the occlusion shader instead of fixed function if we can - // (shadow render uses a special shader that clamps to clip planes) - bound_shader = true; - gOcclusionCubeProgram.bind(); - } - - if (sUseOcclusion > 1) - { - if (mCubeVB.isNull()) - { //cube VB will be used for issuing occlusion queries - mCubeVB = ll_create_cube_vb(LLVertexBuffer::MAP_VERTEX, GL_STATIC_DRAW); - } - mCubeVB->setBuffer(LLVertexBuffer::MAP_VERTEX); - } - for (LLWorld::region_list_t::const_iterator iter = LLWorld::getInstance()->getRegionList().begin(); iter != LLWorld::getInstance()->getRegionList().end(); ++iter) { @@ -2444,16 +2393,10 @@ void LLPipeline::updateCull(LLCamera& camera, LLCullResult& result, LLPlane* pla LLVOCachePartition* vo_part = region->getVOCachePartition(); if(vo_part) { - bool do_occlusion_cull = can_use_occlusion && use_occlusion && !gUseWireframe; - vo_part->cull(camera, do_occlusion_cull); + vo_part->cull(camera, sUseOcclusion > 0); } } - if (bound_shader) - { - gOcclusionCubeProgram.unbind(); - } - if (hasRenderType(LLPipeline::RENDER_TYPE_SKY) && gSky.mVOSkyp.notNull() && gSky.mVOSkyp->mDrawable.notNull()) @@ -2479,28 +2422,6 @@ void LLPipeline::updateCull(LLCamera& camera, LLCullResult& result, LLPlane* pla { LLWorld::getInstance()->precullWaterObjects(camera, sCull, render_water); } - - gGL.matrixMode(LLRender::MM_PROJECTION); - gGL.popMatrix(); - gGL.matrixMode(LLRender::MM_MODELVIEW); - gGL.popMatrix(); - - if (sUseOcclusion > 1) - { - gGL.setColorMask(true, false); - } - - if (to_texture) - { - if (LLPipeline::sRenderDeferred && can_use_occlusion) - { - mRT->occlusionDepth.flush(); - } - else - { - mRT->screen.flush(); - } - } } void LLPipeline::markNotCulled(LLSpatialGroup* group, LLCamera& camera) @@ -2563,19 +2484,22 @@ void LLPipeline::markOccluder(LLSpatialGroup* group) void LLPipeline::downsampleDepthBuffer(LLRenderTarget& source, LLRenderTarget& dest, LLRenderTarget* scratch_space) { + LL_PROFILE_ZONE_SCOPED_CATEGORY_PIPELINE; + LL_PROFILE_GPU_ZONE("downsampleDepthBuffer"); + LLGLSLShader* last_shader = LLGLSLShader::sCurBoundShaderPtr; LLGLSLShader* shader = NULL; if (scratch_space) { +#if 0 // TODO -- restore occlusion culling functionality GLint bits = 0; - llassert(!source.hasStencil()); // stencil buffer usage is deprecated - bits |= (source.hasStencil() && dest.hasStencil()) ? GL_STENCIL_BUFFER_BIT : 0; - bits |= GL_DEPTH_BUFFER_BIT; - scratch_space->copyContents(source, + bits = GL_DEPTH_BUFFER_BIT; + scratch_space->copyContents(source, 0, 0, source.getWidth(), source.getHeight(), 0, 0, scratch_space->getWidth(), scratch_space->getHeight(), bits, GL_NEAREST); +#endif } dest.bindTarget(); @@ -2624,23 +2548,6 @@ void LLPipeline::downsampleDepthBuffer(LLRenderTarget& source, LLRenderTarget& d } } -void LLPipeline::doOcclusion(LLCamera& camera, LLRenderTarget& source, LLRenderTarget& dest, LLRenderTarget* scratch_space) -{ - LL_PROFILE_ZONE_SCOPED_CATEGORY_DISPLAY; - llassert(!gCubeSnapshot); -#if 0 - downsampleDepthBuffer(source, dest, scratch_space); - dest.bindTarget(); - doOcclusion(camera); - dest.flush(); -#else - // none of the above shenanigans should matter (enough) because we've preserved hierarchical Z before issuing occlusion queries - //source.bindTarget(); - doOcclusion(camera); - //source.flush(); -#endif -} - void LLPipeline::doOcclusion(LLCamera& camera) { LL_PROFILE_ZONE_SCOPED_CATEGORY_PIPELINE; @@ -2741,10 +2648,6 @@ void LLPipeline::updateGL() LLGLUpdate::sGLQ.pop_front(); } } - - { //seed VBO Pools - LLVertexBuffer::seedPools(); - } } void LLPipeline::clearRebuildGroups() @@ -3321,6 +3224,7 @@ void LLPipeline::markRebuild(LLDrawable *drawablep, LLDrawable::EDrawableFlags f void LLPipeline::stateSort(LLCamera& camera, LLCullResult &result) { LL_PROFILE_ZONE_SCOPED_CATEGORY_PIPELINE; + LL_PROFILE_GPU_ZONE("stateSort"); if (hasAnyRenderType(LLPipeline::RENDER_TYPE_AVATAR, LLPipeline::RENDER_TYPE_CONTROL_AV, @@ -3929,6 +3833,7 @@ void LLPipeline::postSort(LLCamera &camera) // flush particle VB if (LLVOPartGroup::sVB) { + LL_PROFILE_GPU_ZONE("flush particle vb"); LLVOPartGroup::sVB->flush(); } else @@ -3952,9 +3857,12 @@ void LLPipeline::postSort(LLCamera &camera) }*/ // pack vertex buffers for groups that chose to delay their updates - for (LLSpatialGroup::sg_vector_t::iterator iter = mMeshDirtyGroup.begin(); iter != mMeshDirtyGroup.end(); ++iter) { - (*iter)->rebuildMesh(); + LL_PROFILE_GPU_ZONE("rebuildMesh"); + for (LLSpatialGroup::sg_vector_t::iterator iter = mMeshDirtyGroup.begin(); iter != mMeshDirtyGroup.end(); ++iter) + { + (*iter)->rebuildMesh(); + } } /*if (use_transform_feedback) @@ -7351,8 +7259,6 @@ void LLPipeline::doResetVertexBuffers(bool forced) LLVOPartGroup::destroyGL(); gGL.resetVertexBuffer(); - SUBSYSTEM_CLEANUP(LLVertexBuffer); - if (LLVertexBuffer::sGLCount != 0) { LL_WARNS() << "VBO wipe failed -- " << LLVertexBuffer::sGLCount << " buffers remaining." << LL_ENDL; @@ -7372,7 +7278,6 @@ void LLPipeline::doResetVertexBuffers(bool forced) sNoAlpha = gSavedSettings.getBOOL("RenderNoAlpha"); LLPipeline::sTextureBindTest = gSavedSettings.getBOOL("RenderDebugTextureBind"); - LLVertexBuffer::initClass(LLVertexBuffer::sEnableVBOs, LLVertexBuffer::sDisableVBOMapping); gGL.initVertexBuffer(); mDeferredVB = new LLVertexBuffer(DEFERRED_VB_MASK, 0); @@ -7445,30 +7350,28 @@ void LLPipeline::renderAlphaObjects(U32 mask, bool texture, bool batch_texture, for (LLCullResult::drawinfo_iterator i = begin; i != end; ) { LLDrawInfo* pparams = *i; - if (pparams) - { - LLCullResult::increment_iterator(i, end); + LLCullResult::increment_iterator(i, end); - if (rigged) + if (rigged) + { + if (pparams->mAvatar != nullptr) { - if (pparams->mAvatar != nullptr) + if (lastAvatar != pparams->mAvatar || lastMeshId != pparams->mSkinInfo->mHash) { - if (lastAvatar != pparams->mAvatar || lastMeshId != pparams->mSkinInfo->mHash) - { - mSimplePool->uploadMatrixPalette(*pparams); - lastAvatar = pparams->mAvatar; - lastMeshId = pparams->mSkinInfo->mHash; - } - - mSimplePool->pushBatch(*pparams, mask | LLVertexBuffer::MAP_WEIGHT4, texture, batch_texture); + mSimplePool->uploadMatrixPalette(*pparams); + lastAvatar = pparams->mAvatar; + lastMeshId = pparams->mSkinInfo->mHash; } + + mSimplePool->pushBatch(*pparams, mask | LLVertexBuffer::MAP_WEIGHT4, texture, batch_texture); } - else if (pparams->mAvatar == nullptr) - { - mSimplePool->pushBatch(*pparams, mask, texture, batch_texture); - } + } + else if (pparams->mAvatar == nullptr) + { + mSimplePool->pushBatch(*pparams, mask, texture, batch_texture); } } + gGL.loadMatrix(gGLModelView); gGLLastMatrix = NULL; } @@ -9565,20 +9468,17 @@ void LLPipeline::renderShadow(glh::matrix4f& view, glh::matrix4f& proj, LLCamera LLGLEnable cull(GL_CULL_FACE); //enable depth clamping if available - //LLGLEnable depth_clamp(GL_DEPTH_CLAMP); + LLGLEnable depth_clamp(GL_DEPTH_CLAMP); + + LLGLDepthTest depth_test(GL_TRUE, GL_TRUE, GL_LESS); if (use_shader) { gDeferredShadowCubeProgram.bind(); } - LLRenderTarget& occlusion_target = LLViewerCamera::sCurCameraID >= LLViewerCamera::CAMERA_SPOT_SHADOW0 ? - mSpotShadowOcclusion[LLViewerCamera::sCurCameraID - LLViewerCamera::CAMERA_SPOT_SHADOW0] : - mRT->shadowOcclusion[LLViewerCamera::sCurCameraID - LLViewerCamera::CAMERA_SUN_SHADOW0]; - - occlusion_target.bindTarget(); + updateCull(shadow_cam, result); - occlusion_target.flush(); stateSort(shadow_cam, result); @@ -9613,15 +9513,8 @@ void LLPipeline::renderShadow(glh::matrix4f& view, glh::matrix4f& proj, LLCamera for (int j = 0; j < 2; ++j) // 0 -- static, 1 -- rigged { bool rigged = j == 1; - if (!use_shader) - { //occlusion program is general purpose depth-only no-textures - gOcclusionProgram.bind(rigged); - } - else - { - gDeferredShadowProgram.bind(rigged); - LLGLSLShader::sCurBoundShaderPtr->uniform1i(LLShaderMgr::SUN_UP_FACTOR, environment.getIsSunUp() ? 1 : 0); - } + gDeferredShadowProgram.bind(rigged); + LLGLSLShader::sCurBoundShaderPtr->uniform1i(LLShaderMgr::SUN_UP_FACTOR, environment.getIsSunUp() ? 1 : 0); gGL.diffuseColor4f(1, 1, 1, 1); @@ -9655,12 +9548,14 @@ void LLPipeline::renderShadow(glh::matrix4f& view, glh::matrix4f& proj, LLCamera } gGL.getTexUnit(0)->enable(LLTexUnit::TT_TEXTURE); - if (!use_shader) - { - gOcclusionProgram.unbind(); - } } + if (occlude > 1) + { // do occlusion culling against non-masked only to take advantage of hierarchical Z + doOcclusion(shadow_cam); + } + + if (use_shader) { LL_PROFILE_ZONE_NAMED_CATEGORY_PIPELINE("shadow geom"); @@ -9760,15 +9655,6 @@ void LLPipeline::renderShadow(glh::matrix4f& view, glh::matrix4f& proj, LLCamera gGLLastMatrix = NULL; gGL.loadMatrix(gGLModelView); - LLRenderTarget& occlusion_source = LLViewerCamera::sCurCameraID >= LLViewerCamera::CAMERA_SPOT_SHADOW0 ? - mSpotShadow[LLViewerCamera::sCurCameraID - LLViewerCamera::CAMERA_SPOT_SHADOW0] : - mRT->shadow[LLViewerCamera::sCurCameraID - LLViewerCamera::CAMERA_SUN_SHADOW0]; - - if (occlude > 1) - { - doOcclusion(shadow_cam, occlusion_source, occlusion_target); - } - if (use_shader) { gDeferredShadowProgram.unbind(); diff --git a/indra/newview/pipeline.h b/indra/newview/pipeline.h index c698374c8b..d3793da347 100644 --- a/indra/newview/pipeline.h +++ b/indra/newview/pipeline.h @@ -173,7 +173,6 @@ public: // if source's depth buffer cannot be bound for reading, a scratch space depth buffer must be provided void downsampleDepthBuffer(LLRenderTarget& source, LLRenderTarget& dest, LLRenderTarget* scratch_space = NULL); - void doOcclusion(LLCamera& camera, LLRenderTarget& source, LLRenderTarget& dest, LLRenderTarget* scratch_space = NULL); void doOcclusion(LLCamera& camera); void markNotCulled(LLSpatialGroup* group, LLCamera &camera); void markMoved(LLDrawable *drawablep, bool damped_motion = false); @@ -244,7 +243,9 @@ public: bool visibleObjectsInFrustum(LLCamera& camera); bool getVisibleExtents(LLCamera& camera, LLVector3 &min, LLVector3& max); bool getVisiblePointCloud(LLCamera& camera, LLVector3 &min, LLVector3& max, std::vector<LLVector3>& fp, LLVector3 light_dir = LLVector3(0,0,0)); - void updateCull(LLCamera& camera, LLCullResult& result, LLPlane* plane = NULL); //if water_clip is 0, ignore water plane, 1, cull to above plane, -1, cull to below plane + + // Populate given LLCullResult with results of a frustum cull of the entire scene against the given LLCamera + void updateCull(LLCamera& camera, LLCullResult& result); void createObjects(F32 max_dtime); void createObject(LLViewerObject* vobj); void processPartitionQ(); @@ -673,12 +674,10 @@ public: LLRenderTarget fxaaBuffer; LLRenderTarget edgeMap; LLRenderTarget deferredDepth; - LLRenderTarget occlusionDepth; LLRenderTarget deferredLight; //sun shadow map LLRenderTarget shadow[4]; - LLRenderTarget shadowOcclusion[4]; }; // main full resoltuion render target @@ -691,7 +690,6 @@ public: RenderTargetPack* mRT; LLRenderTarget mSpotShadow[2]; - LLRenderTarget mSpotShadowOcclusion[2]; LLRenderTarget mPbrBrdfLut; |