diff options
Diffstat (limited to 'indra/llrender')
| -rw-r--r-- | indra/llrender/llfontfreetype.cpp | 7 | ||||
| -rw-r--r-- | indra/llrender/llfontgl.cpp | 12 | ||||
| -rw-r--r-- | indra/llrender/llglslshader.cpp | 2 | ||||
| -rw-r--r-- | indra/llrender/llglslshader.h | 1 | ||||
| -rw-r--r-- | indra/llrender/llrender.cpp | 9 | ||||
| -rw-r--r-- | indra/llrender/llrender.h | 1 | ||||
| -rw-r--r-- | indra/llrender/llrendertarget.cpp | 3 | ||||
| -rw-r--r-- | indra/llrender/llshadermgr.cpp | 74 | ||||
| -rw-r--r-- | indra/llrender/llvertexbuffer.cpp | 132 | ||||
| -rw-r--r-- | indra/llrender/llvertexbuffer.h | 24 |
10 files changed, 182 insertions, 83 deletions
diff --git a/indra/llrender/llfontfreetype.cpp b/indra/llrender/llfontfreetype.cpp index 91c8a37022..66d4ad2d87 100644 --- a/indra/llrender/llfontfreetype.cpp +++ b/indra/llrender/llfontfreetype.cpp @@ -55,7 +55,10 @@ FT_Library gFTLibrary = NULL; //static void LLFontManager::initClass() { - gFontManagerp = new LLFontManager; + if (!gFontManagerp) + { + gFontManagerp = new LLFontManager; + } } //static @@ -136,7 +139,7 @@ BOOL LLFontFreetype::loadFace(const std::string& filename, F32 point_size, F32 v FT_Done_Face(mFTFace); mFTFace = NULL; } - + int error; error = FT_New_Face( gFTLibrary, diff --git a/indra/llrender/llfontgl.cpp b/indra/llrender/llfontgl.cpp index 607473d416..82e8227ffe 100644 --- a/indra/llrender/llfontgl.cpp +++ b/indra/llrender/llfontgl.cpp @@ -304,6 +304,18 @@ S32 LLFontGL::render(const LLWString &wstr, S32 begin_offset, F32 x, F32 y, cons S32 next_bitmap_num = fgi->mBitmapNum; if (next_bitmap_num != bitmap_num) { + // Actually draw the queued glyphs before switching their texture; + // otherwise the queued glyphs will be taken from wrong textures. + if (glyph_count > 0) + { + gGL.begin(LLRender::QUADS); + { + gGL.vertexBatchPreTransformed(vertices, uvs, colors, glyph_count * 4); + } + gGL.end(); + glyph_count = 0; + } + bitmap_num = next_bitmap_num; LLImageGL *font_image = font_bitmap_cache->getImageGL(bitmap_num); gGL.getTexUnit(0)->bind(font_image); diff --git a/indra/llrender/llglslshader.cpp b/indra/llrender/llglslshader.cpp index 5a6f3d8292..6b2852670a 100644 --- a/indra/llrender/llglslshader.cpp +++ b/indra/llrender/llglslshader.cpp @@ -66,7 +66,7 @@ BOOL shouldChange(const LLVector4& v1, const LLVector4& v2) LLShaderFeatures::LLShaderFeatures() : calculatesLighting(false), isShiny(false), isFullbright(false), hasWaterFog(false), hasTransport(false), hasSkinning(false), hasObjectSkinning(false), hasAtmospherics(false), isSpecular(false), -hasGamma(false), hasLighting(false), calculatesAtmospherics(false), mIndexedTextureChannels(0), disableTextureIndex(false), +hasGamma(false), hasLighting(false), isAlphaLighting(false), calculatesAtmospherics(false), mIndexedTextureChannels(0), disableTextureIndex(false), hasAlphaMask(false) { } diff --git a/indra/llrender/llglslshader.h b/indra/llrender/llglslshader.h index 2a6c050eac..00b4b0dbd4 100644 --- a/indra/llrender/llglslshader.h +++ b/indra/llrender/llglslshader.h @@ -36,6 +36,7 @@ public: 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; diff --git a/indra/llrender/llrender.cpp b/indra/llrender/llrender.cpp index 812fa7024b..b0ddacbb05 100644 --- a/indra/llrender/llrender.cpp +++ b/indra/llrender/llrender.cpp @@ -997,7 +997,7 @@ void LLLightState::setSpotDirection(const LLVector3& direction) const glh::matrix4f& mat = gGL.getModelviewMatrix(); mat.mult_matrix_dir(dir); - mSpotDirection.set(direction); + mSpotDirection.set(dir.v); } } @@ -1434,6 +1434,8 @@ void LLRender::loadIdentity() flush(); { + llassert_always(mMatrixMode < NUM_MATRIX_MODES) ; + mMatrix[mMatrixMode][mMatIdx[mMatrixMode]].make_identity(); mMatHash[mMatrixMode]++; } @@ -1444,6 +1446,11 @@ const glh::matrix4f& LLRender::getModelviewMatrix() return mMatrix[MM_MODELVIEW][mMatIdx[MM_MODELVIEW]]; } +const glh::matrix4f& LLRender::getProjectionMatrix() +{ + return mMatrix[MM_PROJECTION][mMatIdx[MM_PROJECTION]]; +} + void LLRender::translateUI(F32 x, F32 y, F32 z) { if (mUIOffset.empty()) diff --git a/indra/llrender/llrender.h b/indra/llrender/llrender.h index 7581b9f908..fa5f7f311d 100644 --- a/indra/llrender/llrender.h +++ b/indra/llrender/llrender.h @@ -348,6 +348,7 @@ public: void matrixMode(U32 mode); const glh::matrix4f& getModelviewMatrix(); + const glh::matrix4f& getProjectionMatrix(); void syncMatrices(); void syncLightState(); diff --git a/indra/llrender/llrendertarget.cpp b/indra/llrender/llrendertarget.cpp index 1aa12614ea..ef2a7395da 100644 --- a/indra/llrender/llrendertarget.cpp +++ b/indra/llrender/llrendertarget.cpp @@ -457,7 +457,8 @@ void LLRenderTarget::copyContents(LLRenderTarget& source, S32 srcX0, S32 srcY0, gGL.flush(); if (!source.mFBO || !mFBO) { - llerrs << "Cannot copy framebuffer contents for non FBO render targets." << llendl; + llwarns << "Cannot copy framebuffer contents for non FBO render targets." << llendl; + return; } diff --git a/indra/llrender/llshadermgr.cpp b/indra/llrender/llshadermgr.cpp index eea768a3ea..1a03aeebb7 100644 --- a/indra/llrender/llshadermgr.cpp +++ b/indra/llrender/llshadermgr.cpp @@ -107,10 +107,13 @@ BOOL LLShaderMgr::attachShaderFeatures(LLGLSLShader * shader) { return FALSE; } - - if (!shader->attachObject("lighting/sumLightsSpecularV.glsl")) + + if (!features->isAlphaLighting) { - return FALSE; + if (!shader->attachObject("lighting/sumLightsSpecularV.glsl")) + { + return FALSE; + } } if (!shader->attachObject("lighting/lightSpecularV.glsl")) @@ -125,9 +128,12 @@ BOOL LLShaderMgr::attachShaderFeatures(LLGLSLShader * shader) return FALSE; } - if (!shader->attachObject("lighting/sumLightsV.glsl")) + if (!features->isAlphaLighting) { - return FALSE; + if (!shader->attachObject("lighting/sumLightsV.glsl")) + { + return FALSE; + } } if (!shader->attachObject("lighting/lightV.glsl")) @@ -296,7 +302,7 @@ BOOL LLShaderMgr::attachShaderFeatures(LLGLSLShader * shader) } shader->mFeatures.mIndexedTextureChannels = llmax(LLGLSLShader::sIndexedTextureChannels-1, 1); } - } + } } // NOTE order of shader object attaching is VERY IMPORTANT!!! @@ -579,7 +585,7 @@ GLhandleARB LLShaderMgr::loadShaderFile(const std::string& filename, S32 & shade text[count++] = strdup("#define ATTRIBUTE attribute\n"); text[count++] = strdup("#define VARYING varying\n"); } - else if (version < 3.f) + else if (version < 3.3f) { //set version to 1.20 text[count++] = strdup("#version 120\n"); @@ -845,28 +851,42 @@ BOOL LLShaderMgr::linkProgramObject(GLhandleARB obj, BOOL suppress_errors) LL_WARNS("ShaderLoading") << "GLSL Linker Error:" << LL_ENDL; } -// NOTE: Removing LL_DARWIN block as it doesn't seem to actually give the correct answer, -// but want it for reference once I move it. -#if 0 - // Force an evaluation of the gl state so the driver can tell if the shader will run in hardware or software - // per Apple's suggestion - glBegin(gGL.mMode); - glEnd(); - - // Query whether the shader can or cannot run in hardware - // http://developer.apple.com/qa/qa2007/qa1502.html - long vertexGPUProcessing; - CGLContextObj ctx = CGLGetCurrentContext(); - CGLGetParameter (ctx, kCGLCPGPUVertexProcessing, &vertexGPUProcessing); - long fragmentGPUProcessing; - CGLGetParameter (ctx, kCGLCPGPUFragmentProcessing, &fragmentGPUProcessing); - if (!fragmentGPUProcessing || !vertexGPUProcessing) +#if LL_DARWIN + + // For some reason this absolutely kills the frame rate when VBO's are enabled + if (0) { - LL_WARNS("ShaderLoading") << "GLSL Linker: Running in Software:" << LL_ENDL; - success = GL_FALSE; - suppress_errors = FALSE; + // Force an evaluation of the gl state so the driver can tell if the shader will run in hardware or software + // per Apple's suggestion + LLGLSLShader::sNoFixedFunction = false; + + glUseProgramObjectARB(obj); + + gGL.begin(LLRender::TRIANGLES); + gGL.vertex3f(0.0f, 0.0f, 0.0f); + gGL.vertex3f(0.0f, 0.0f, 0.0f); + gGL.vertex3f(0.0f, 0.0f, 0.0f); + gGL.end(); + gGL.flush(); + + glUseProgramObjectARB(0); + + LLGLSLShader::sNoFixedFunction = true; + + // Query whether the shader can or cannot run in hardware + // http://developer.apple.com/qa/qa2007/qa1502.html + GLint vertexGPUProcessing, fragmentGPUProcessing; + CGLContextObj ctx = CGLGetCurrentContext(); + CGLGetParameter(ctx, kCGLCPGPUVertexProcessing, &vertexGPUProcessing); + CGLGetParameter(ctx, kCGLCPGPUFragmentProcessing, &fragmentGPUProcessing); + if (!fragmentGPUProcessing || !vertexGPUProcessing) + { + LL_WARNS("ShaderLoading") << "GLSL Linker: Running in Software:" << LL_ENDL; + success = GL_FALSE; + suppress_errors = FALSE; + } } - + #else std::string log = get_object_log(obj); LLStringUtil::toLower(log); diff --git a/indra/llrender/llvertexbuffer.cpp b/indra/llrender/llvertexbuffer.cpp index 3948145580..eb302392bb 100644 --- a/indra/llrender/llvertexbuffer.cpp +++ b/indra/llrender/llvertexbuffer.cpp @@ -95,7 +95,7 @@ public: #endif } - ~LLGLSyncFence() + virtual ~LLGLSyncFence() { #ifdef GL_ARB_sync if (mSync) @@ -148,7 +148,7 @@ U32 wpo2(U32 i) return r; } -U8* LLVBOPool::allocate(U32& name, U32 size) +volatile U8* LLVBOPool::allocate(U32& name, U32 size) { llassert(nhpo2(size) == size); @@ -159,20 +159,25 @@ U8* LLVBOPool::allocate(U32& name, U32 size) mFreeList.resize(i+1); } - U8* ret = NULL; + volatile U8* ret = NULL; if (mFreeList[i].empty()) { //make a new buffer glGenBuffersARB(1, &name); glBindBufferARB(mType, name); - glBufferDataARB(mType, size, 0, mUsage); LLVertexBuffer::sAllocatedBytes += size; - if (LLVertexBuffer::sDisableVBOMapping) + if (LLVertexBuffer::sDisableVBOMapping || mUsage != GL_DYNAMIC_DRAW_ARB) { + glBufferDataARB(mType, size, 0, mUsage); ret = (U8*) ll_aligned_malloc_16(size); } + else + { //always use a true hint of static draw when allocating non-client-backed buffers + glBufferDataARB(mType, size, 0, GL_STATIC_DRAW_ARB); + } + glBindBufferARB(mType, 0); } else @@ -188,7 +193,7 @@ U8* LLVBOPool::allocate(U32& name, U32 size) return ret; } -void LLVBOPool::release(U32 name, U8* buffer, U32 size) +void LLVBOPool::release(U32 name, volatile U8* buffer, U32 size) { llassert(nhpo2(size) == size); @@ -201,8 +206,15 @@ void LLVBOPool::release(U32 name, U8* buffer, U32 size) rec.mClientData = buffer; sBytesPooled += size; - - mFreeList[i].push_back(rec); + + if (!LLVertexBuffer::sDisableVBOMapping && mUsage == GL_DYNAMIC_DRAW_ARB) + { + glDeleteBuffersARB(1, &rec.mGLName); + } + else + { + mFreeList[i].push_back(rec); + } } void LLVBOPool::cleanup() @@ -221,7 +233,7 @@ void LLVBOPool::cleanup() if (r.mClientData) { - ll_aligned_free_16(r.mClientData); + ll_aligned_free_16((void*) r.mClientData); } l.pop_front(); @@ -536,7 +548,7 @@ void LLVertexBuffer::validateRange(U32 start, U32 end, U32 count, U32 indices_of void LLVertexBuffer::drawRange(U32 mode, U32 start, U32 end, U32 count, U32 indices_offset) const { validateRange(start, end, count, indices_offset); - + mMappable = FALSE; gGL.syncMatrices(); llassert(mNumVerts >= 0); @@ -591,7 +603,7 @@ void LLVertexBuffer::drawRange(U32 mode, U32 start, U32 end, U32 count, U32 indi void LLVertexBuffer::draw(U32 mode, U32 count, U32 indices_offset) const { llassert(!LLGLSLShader::sNoFixedFunction || LLGLSLShader::sCurBoundShaderPtr != NULL); - + mMappable = FALSE; gGL.syncMatrices(); llassert(mNumIndices >= 0); @@ -637,7 +649,7 @@ void LLVertexBuffer::draw(U32 mode, U32 count, U32 indices_offset) const void LLVertexBuffer::drawArrays(U32 mode, U32 first, U32 count) const { llassert(!LLGLSLShader::sNoFixedFunction || LLGLSLShader::sCurBoundShaderPtr != NULL); - + mMappable = FALSE; gGL.syncMatrices(); llassert(mNumVerts >= 0); @@ -787,9 +799,26 @@ LLVertexBuffer::LLVertexBuffer(U32 typemask, S32 usage) : if (mUsage && mUsage != GL_STREAM_DRAW_ARB) { //only stream_draw and dynamic_draw are supported when using VBOs, dynamic draw is the default - mUsage = GL_DYNAMIC_DRAW_ARB; + if (sDisableVBOMapping) + { //always use stream draw if VBO mapping is disabled + mUsage = GL_STREAM_DRAW_ARB; + } + else + { + mUsage = GL_DYNAMIC_DRAW_ARB; + } } - + + + if (mUsage == GL_DYNAMIC_DRAW_ARB && !sDisableVBOMapping) + { + mMappable = TRUE; + } + else + { + mMappable = FALSE; + } + //zero out offsets for (U32 i = 0; i < TYPE_MAX; i++) { @@ -1042,7 +1071,7 @@ void LLVertexBuffer::destroyGLBuffer() } else { - FREE_MEM(sPrivatePoolp, mMappedData) ; + FREE_MEM(sPrivatePoolp, (void*) mMappedData) ; mMappedData = NULL; mEmpty = TRUE; } @@ -1063,7 +1092,7 @@ void LLVertexBuffer::destroyGLIndices() } else { - FREE_MEM(sPrivatePoolp, mMappedIndexData) ; + FREE_MEM(sPrivatePoolp, (void*) mMappedIndexData) ; mMappedIndexData = NULL; mEmpty = TRUE; } @@ -1282,8 +1311,11 @@ bool expand_region(LLVertexBuffer::MappedRegion& region, S32 index, S32 count) return true; } +static LLFastTimer::DeclareTimer FTM_VBO_MAP_BUFFER_RANGE("VBO Map Range"); +static LLFastTimer::DeclareTimer FTM_VBO_MAP_BUFFER("VBO Map"); + // Map for data access -U8* LLVertexBuffer::mapVertexBuffer(S32 type, S32 index, S32 count, bool map_range) +volatile U8* LLVertexBuffer::mapVertexBuffer(S32 type, S32 index, S32 count, bool map_range) { bindGLBuffer(true); LLMemType mt2(LLMemType::MTYPE_VERTEX_MAP_BUFFER); @@ -1298,7 +1330,7 @@ U8* LLVertexBuffer::mapVertexBuffer(S32 type, S32 index, S32 count, bool map_ran if (useVBOs()) { - if (sDisableVBOMapping || gGLManager.mHasMapBufferRange || gGLManager.mHasFlushBufferRange) + if (!mMappable || gGLManager.mHasMapBufferRange || gGLManager.mHasFlushBufferRange) { if (count == -1) { @@ -1323,7 +1355,7 @@ U8* LLVertexBuffer::mapVertexBuffer(S32 type, S32 index, S32 count, bool map_ran if (!mapped) { //not already mapped, map new region - MappedRegion region(type, !sDisableVBOMapping && map_range ? -1 : index, count); + MappedRegion region(type, mMappable && map_range ? -1 : index, count); mMappedVertexRegions.push_back(region); } } @@ -1340,19 +1372,20 @@ U8* LLVertexBuffer::mapVertexBuffer(S32 type, S32 index, S32 count, bool map_ran sMappedCount++; stop_glerror(); - if(sDisableVBOMapping) + if(!mMappable) { map_range = false; } else { - U8* src = NULL; + volatile U8* src = NULL; waitFence(); if (gGLManager.mHasMapBufferRange) { if (map_range) { #ifdef GL_ARB_map_buffer_range + LLFastTimer t(FTM_VBO_MAP_BUFFER_RANGE); S32 offset = mOffsets[type] + sTypeSize[type]*index; S32 length = (sTypeSize[type]*count+0xF) & ~0xF; src = (U8*) glMapBufferRange(GL_ARRAY_BUFFER_ARB, offset, length, @@ -1376,6 +1409,7 @@ U8* LLVertexBuffer::mapVertexBuffer(S32 type, S32 index, S32 count, bool map_ran } } + LLFastTimer t(FTM_VBO_MAP_BUFFER); src = (U8*) glMapBufferRange(GL_ARRAY_BUFFER_ARB, 0, mSize, GL_MAP_WRITE_BIT | GL_MAP_FLUSH_EXPLICIT_BIT); @@ -1403,7 +1437,7 @@ U8* LLVertexBuffer::mapVertexBuffer(S32 type, S32 index, S32 count, bool map_ran llassert(src != NULL); - mMappedData = LL_NEXT_ALIGNED_ADDRESS<U8>(src); + mMappedData = LL_NEXT_ALIGNED_ADDRESS<volatile U8>(src); mAlignedOffset = mMappedData - src; stop_glerror(); @@ -1416,7 +1450,7 @@ U8* LLVertexBuffer::mapVertexBuffer(S32 type, S32 index, S32 count, bool map_ran //check the availability of memory LLMemory::logMemoryInfo(TRUE) ; - if(!sDisableVBOMapping) + if(mMappable) { //-------------------- //print out more debug info before crash @@ -1448,7 +1482,7 @@ U8* LLVertexBuffer::mapVertexBuffer(S32 type, S32 index, S32 count, bool map_ran map_range = false; } - if (map_range && gGLManager.mHasMapBufferRange && !sDisableVBOMapping) + if (map_range && gGLManager.mHasMapBufferRange && mMappable) { return mMappedData; } @@ -1458,7 +1492,11 @@ U8* LLVertexBuffer::mapVertexBuffer(S32 type, S32 index, S32 count, bool map_ran } } -U8* LLVertexBuffer::mapIndexBuffer(S32 index, S32 count, bool map_range) + +static LLFastTimer::DeclareTimer FTM_VBO_MAP_INDEX_RANGE("IBO Map Range"); +static LLFastTimer::DeclareTimer FTM_VBO_MAP_INDEX("IBO Map"); + +volatile U8* LLVertexBuffer::mapIndexBuffer(S32 index, S32 count, bool map_range) { LLMemType mt2(LLMemType::MTYPE_VERTEX_MAP_BUFFER); bindGLIndices(true); @@ -1473,7 +1511,7 @@ U8* LLVertexBuffer::mapIndexBuffer(S32 index, S32 count, bool map_range) if (useVBOs()) { - if (sDisableVBOMapping || gGLManager.mHasMapBufferRange || gGLManager.mHasFlushBufferRange) + if (!mMappable || gGLManager.mHasMapBufferRange || gGLManager.mHasFlushBufferRange) { if (count == -1) { @@ -1495,7 +1533,7 @@ U8* LLVertexBuffer::mapIndexBuffer(S32 index, S32 count, bool map_range) if (!mapped) { //not already mapped, map new region - MappedRegion region(TYPE_INDEX, !sDisableVBOMapping && map_range ? -1 : index, count); + MappedRegion region(TYPE_INDEX, mMappable && map_range ? -1 : index, count); mMappedIndexRegions.push_back(region); } } @@ -1524,19 +1562,20 @@ U8* LLVertexBuffer::mapIndexBuffer(S32 index, S32 count, bool map_range) } } - if(sDisableVBOMapping) + if(!mMappable) { map_range = false; } else { - U8* src = NULL; + volatile U8* src = NULL; waitFence(); if (gGLManager.mHasMapBufferRange) { if (map_range) { #ifdef GL_ARB_map_buffer_range + LLFastTimer t(FTM_VBO_MAP_INDEX_RANGE); S32 offset = sizeof(U16)*index; S32 length = sizeof(U16)*count; src = (U8*) glMapBufferRange(GL_ELEMENT_ARRAY_BUFFER_ARB, offset, length, @@ -1548,6 +1587,7 @@ U8* LLVertexBuffer::mapIndexBuffer(S32 index, S32 count, bool map_range) else { #ifdef GL_ARB_map_buffer_range + LLFastTimer t(FTM_VBO_MAP_INDEX); src = (U8*) glMapBufferRange(GL_ELEMENT_ARRAY_BUFFER_ARB, 0, sizeof(U16)*mNumIndices, GL_MAP_WRITE_BIT | GL_MAP_FLUSH_EXPLICIT_BIT); @@ -1569,6 +1609,7 @@ U8* LLVertexBuffer::mapIndexBuffer(S32 index, S32 count, bool map_range) } else { + LLFastTimer t(FTM_VBO_MAP_INDEX); map_range = false; src = (U8*) glMapBufferARB(GL_ELEMENT_ARRAY_BUFFER_ARB, GL_WRITE_ONLY_ARB); } @@ -1587,7 +1628,7 @@ U8* LLVertexBuffer::mapIndexBuffer(S32 index, S32 count, bool map_range) log_glerror(); LLMemory::logMemoryInfo(TRUE) ; - if(!sDisableVBOMapping) + if(mMappable) { GLint buff; glGetIntegerv(GL_ELEMENT_ARRAY_BUFFER_BINDING_ARB, &buff); @@ -1609,7 +1650,7 @@ U8* LLVertexBuffer::mapIndexBuffer(S32 index, S32 count, bool map_range) map_range = false; } - if (map_range && gGLManager.mHasMapBufferRange && !sDisableVBOMapping) + if (map_range && gGLManager.mHasMapBufferRange && mMappable) { return mMappedIndexData; } @@ -1619,6 +1660,13 @@ U8* LLVertexBuffer::mapIndexBuffer(S32 index, S32 count, bool map_range) } } +static LLFastTimer::DeclareTimer FTM_VBO_UNMAP("VBO Unmap"); +static LLFastTimer::DeclareTimer FTM_VBO_FLUSH_RANGE("Flush VBO Range"); + + +static LLFastTimer::DeclareTimer FTM_IBO_UNMAP("IBO Unmap"); +static LLFastTimer::DeclareTimer FTM_IBO_FLUSH_RANGE("Flush IBO Range"); + void LLVertexBuffer::unmapBuffer() { LLMemType mt2(LLMemType::MTYPE_VERTEX_UNMAP_BUFFER); @@ -1631,10 +1679,11 @@ void LLVertexBuffer::unmapBuffer() if (mMappedData && mVertexLocked) { + LLFastTimer t(FTM_VBO_UNMAP); bindGLBuffer(true); updated_all = mIndexLocked; //both vertex and index buffers done updating - if(sDisableVBOMapping) + if(!mMappable) { if (!mMappedVertexRegions.empty()) { @@ -1644,7 +1693,7 @@ void LLVertexBuffer::unmapBuffer() 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; - glBufferSubDataARB(GL_ARRAY_BUFFER_ARB, offset, length, mMappedData+offset); + glBufferSubDataARB(GL_ARRAY_BUFFER_ARB, offset, length, (U8*) mMappedData+offset); stop_glerror(); } @@ -1653,7 +1702,7 @@ void LLVertexBuffer::unmapBuffer() else { stop_glerror(); - glBufferSubDataARB(GL_ARRAY_BUFFER_ARB, 0, getSize(), mMappedData); + glBufferSubDataARB(GL_ARRAY_BUFFER_ARB, 0, getSize(), (U8*) mMappedData); stop_glerror(); } } @@ -1671,6 +1720,7 @@ void LLVertexBuffer::unmapBuffer() S32 length = sTypeSize[region.mType]*region.mCount; if (gGLManager.mHasMapBufferRange) { + LLFastTimer t(FTM_VBO_FLUSH_RANGE); #ifdef GL_ARB_map_buffer_range glFlushMappedBufferRange(GL_ARRAY_BUFFER_ARB, offset, length); #endif @@ -1698,8 +1748,9 @@ void LLVertexBuffer::unmapBuffer() if (mMappedIndexData && mIndexLocked) { + LLFastTimer t(FTM_IBO_UNMAP); bindGLIndices(); - if(sDisableVBOMapping) + if(!mMappable) { if (!mMappedIndexRegions.empty()) { @@ -1708,7 +1759,7 @@ void LLVertexBuffer::unmapBuffer() const MappedRegion& region = mMappedIndexRegions[i]; S32 offset = region.mIndex >= 0 ? sizeof(U16)*region.mIndex : 0; S32 length = sizeof(U16)*region.mCount; - glBufferSubDataARB(GL_ELEMENT_ARRAY_BUFFER_ARB, offset, length, mMappedIndexData+offset); + glBufferSubDataARB(GL_ELEMENT_ARRAY_BUFFER_ARB, offset, length, (U8*) mMappedIndexData+offset); stop_glerror(); } @@ -1717,7 +1768,7 @@ void LLVertexBuffer::unmapBuffer() else { stop_glerror(); - glBufferSubDataARB(GL_ELEMENT_ARRAY_BUFFER_ARB, 0, getIndicesSize(), mMappedIndexData); + glBufferSubDataARB(GL_ELEMENT_ARRAY_BUFFER_ARB, 0, getIndicesSize(), (U8*) mMappedIndexData); stop_glerror(); } } @@ -1734,6 +1785,7 @@ void LLVertexBuffer::unmapBuffer() S32 length = sizeof(U16)*region.mCount; if (gGLManager.mHasMapBufferRange) { + LLFastTimer t(FTM_IBO_FLUSH_RANGE); #ifdef GL_ARB_map_buffer_range glFlushMappedBufferRange(GL_ELEMENT_ARRAY_BUFFER_ARB, offset, length); #endif @@ -1778,7 +1830,7 @@ template <class T,S32 type> struct VertexBufferStrider { if (type == LLVertexBuffer::TYPE_INDEX) { - U8* ptr = vbo.mapIndexBuffer(index, count, map_range); + volatile U8* ptr = vbo.mapIndexBuffer(index, count, map_range); if (ptr == NULL) { @@ -1794,7 +1846,7 @@ template <class T,S32 type> struct VertexBufferStrider { S32 stride = LLVertexBuffer::sTypeSize[type]; - U8* ptr = vbo.mapVertexBuffer(type, index, count, map_range); + volatile U8* ptr = vbo.mapVertexBuffer(type, index, count, map_range); if (ptr == NULL) { @@ -2109,7 +2161,7 @@ void LLVertexBuffer::setupVertexBuffer(U32 data_mask) { LLMemType mt2(LLMemType::MTYPE_VERTEX_SETUP_VERTEX_BUFFER); stop_glerror(); - U8* base = useVBOs() ? (U8*) mAlignedOffset : mMappedData; + volatile U8* base = useVBOs() ? (U8*) mAlignedOffset : mMappedData; /*if ((data_mask & mTypeMask) != data_mask) { diff --git a/indra/llrender/llvertexbuffer.h b/indra/llrender/llvertexbuffer.h index 3e6f6a959a..e1cbfd3b61 100644 --- a/indra/llrender/llvertexbuffer.h +++ b/indra/llrender/llvertexbuffer.h @@ -60,10 +60,10 @@ public: U32 mType; //size MUST be a power of 2 - U8* allocate(U32& name, U32 size); + volatile U8* allocate(U32& name, U32 size); //size MUST be the size provided to allocate that returned the given name - void release(U32 name, U8* buffer, U32 size); + void release(U32 name, volatile U8* buffer, U32 size); //destroy all records in mFreeList void cleanup(); @@ -72,7 +72,7 @@ public: { public: U32 mGLName; - U8* mClientData; + volatile U8* mClientData; }; typedef std::list<Record> record_list_t; @@ -208,8 +208,8 @@ public: LLVertexBuffer(U32 typemask, S32 usage); // map for data access - U8* mapVertexBuffer(S32 type, S32 index, S32 count, bool map_range); - U8* mapIndexBuffer(S32 index, S32 count, bool map_range); + volatile U8* mapVertexBuffer(S32 type, S32 index, S32 count, bool map_range); + volatile U8* mapIndexBuffer(S32 index, S32 count, bool map_range); // set for rendering virtual void setBuffer(U32 data_mask); // calls setupVertexBuffer() if data_mask is not 0 @@ -244,16 +244,17 @@ public: S32 getNumVerts() const { return mNumVerts; } S32 getNumIndices() const { return mNumIndices; } - U8* getIndicesPointer() const { return useVBOs() ? (U8*) mAlignedIndexOffset : mMappedIndexData; } - U8* getVerticesPointer() const { return useVBOs() ? (U8*) mAlignedOffset : mMappedData; } + volatile U8* getIndicesPointer() const { return useVBOs() ? (U8*) mAlignedIndexOffset : mMappedIndexData; } + volatile U8* getVerticesPointer() const { return useVBOs() ? (U8*) mAlignedOffset : mMappedData; } U32 getTypeMask() const { return mTypeMask; } bool hasDataType(S32 type) const { return ((1 << type) & getTypeMask()); } S32 getSize() const; S32 getIndicesSize() const { return mIndicesSize; } - U8* getMappedData() const { return mMappedData; } - U8* getMappedIndices() const { return mMappedIndexData; } + volatile U8* getMappedData() const { return mMappedData; } + volatile U8* getMappedIndices() const { return mMappedIndexData; } S32 getOffset(S32 type) const { return mOffsets[type]; } S32 getUsage() const { return mUsage; } + BOOL isWriteable() const { return (mMappable || mUsage == GL_STREAM_DRAW_ARB) ? TRUE : FALSE; } void draw(U32 mode, U32 count, U32 indices_offset) const; void drawArrays(U32 mode, U32 offset, U32 count) const; @@ -278,12 +279,13 @@ protected: U32 mGLIndices; // GL IBO handle U32 mGLArray; // GL VAO handle - U8* mMappedData; // pointer to currently mapped data (NULL if unmapped) - U8* mMappedIndexData; // pointer to currently mapped indices (NULL if unmapped) + volatile U8* mMappedData; // pointer to currently mapped data (NULL if unmapped) + volatile U8* mMappedIndexData; // pointer to currently mapped indices (NULL if unmapped) BOOL mVertexLocked; // if TRUE, vertex buffer is being or has been written to in client memory BOOL mIndexLocked; // if TRUE, index buffer is being or has been written to in client memory BOOL mFinal; // if TRUE, buffer can not be mapped again BOOL mEmpty; // if TRUE, client buffer is empty (or NULL). Old values have been discarded. + mutable BOOL mMappable; // if TRUE, use memory mapping to upload data (otherwise doublebuffer and use glBufferSubData) S32 mOffsets[TYPE_MAX]; std::vector<MappedRegion> mMappedVertexRegions; |