diff options
Diffstat (limited to 'indra/llrender')
-rw-r--r-- | indra/llrender/llglslshader.cpp | 21 | ||||
-rw-r--r-- | indra/llrender/llglslshader.h | 1 | ||||
-rw-r--r-- | indra/llrender/llrender.cpp | 2 | ||||
-rw-r--r-- | indra/llrender/llshadermgr.cpp | 5 | ||||
-rw-r--r-- | indra/llrender/llvertexbuffer.cpp | 357 | ||||
-rw-r--r-- | indra/llrender/llvertexbuffer.h | 73 |
6 files changed, 267 insertions, 192 deletions
diff --git a/indra/llrender/llglslshader.cpp b/indra/llrender/llglslshader.cpp index 6b2852670a..3773568ad8 100644 --- a/indra/llrender/llglslshader.cpp +++ b/indra/llrender/llglslshader.cpp @@ -64,10 +64,23 @@ 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), isAlphaLighting(false), calculatesAtmospherics(false), mIndexedTextureChannels(0), disableTextureIndex(false), -hasAlphaMask(false) + : atmosphericHelpers(false) + , calculatesLighting(false) + , calculatesAtmospherics(false) + , hasLighting(false) + , isAlphaLighting(false) + , isShiny(false) + , isFullbright(false) + , isSpecular(false) + , hasWaterFog(false) + , hasTransport(false) + , hasSkinning(false) + , hasObjectSkinning(false) + , hasAtmospherics(false) + , hasGamma(false) + , mIndexedTextureChannels(0) + , disableTextureIndex(false) + , hasAlphaMask(false) { } diff --git a/indra/llrender/llglslshader.h b/indra/llrender/llglslshader.h index 00b4b0dbd4..7873fe3c4e 100644 --- a/indra/llrender/llglslshader.h +++ b/indra/llrender/llglslshader.h @@ -33,6 +33,7 @@ class LLShaderFeatures { public: + bool atmosphericHelpers; bool calculatesLighting; bool calculatesAtmospherics; bool hasLighting; // implies no transport (it's possible to have neither though) diff --git a/indra/llrender/llrender.cpp b/indra/llrender/llrender.cpp index cd827f5091..f26764cc42 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); } } diff --git a/indra/llrender/llshadermgr.cpp b/indra/llrender/llshadermgr.cpp index 1a03aeebb7..908443e8cf 100644 --- a/indra/llrender/llshadermgr.cpp +++ b/indra/llrender/llshadermgr.cpp @@ -94,13 +94,16 @@ BOOL LLShaderMgr::attachShaderFeatures(LLGLSLShader * shader) } } - if (features->calculatesLighting) + if (features->calculatesLighting || features->atmosphericHelpers) { if (!shader->attachObject("windlight/atmosphericsHelpersV.glsl")) { return FALSE; } + } + if (features->calculatesLighting) + { if (features->isSpecular) { if (!shader->attachObject("lighting/lightFuncSpecularV.glsl")) diff --git a/indra/llrender/llvertexbuffer.cpp b/indra/llrender/llvertexbuffer.cpp index 20a450fbfb..a7151afeb1 100644 --- a/indra/llrender/llvertexbuffer.cpp +++ b/indra/llrender/llvertexbuffer.cpp @@ -53,31 +53,31 @@ U32 nhpo2(U32 v) //============================================================================ //static -LLVBOPool LLVertexBuffer::sStreamVBOPool; -LLVBOPool LLVertexBuffer::sDynamicVBOPool; -LLVBOPool LLVertexBuffer::sStreamIBOPool; -LLVBOPool LLVertexBuffer::sDynamicIBOPool; +LLVBOPool LLVertexBuffer::sStreamVBOPool(GL_STREAM_DRAW_ARB, GL_ARRAY_BUFFER_ARB); +LLVBOPool LLVertexBuffer::sDynamicVBOPool(GL_DYNAMIC_DRAW_ARB, GL_ARRAY_BUFFER_ARB); +LLVBOPool LLVertexBuffer::sStreamIBOPool(GL_STREAM_DRAW_ARB, GL_ELEMENT_ARRAY_BUFFER_ARB); +LLVBOPool LLVertexBuffer::sDynamicIBOPool(GL_DYNAMIC_DRAW_ARB, GL_ELEMENT_ARRAY_BUFFER_ARB); U32 LLVBOPool::sBytesPooled = 0; -LLPrivateMemoryPool* LLVertexBuffer::sPrivatePoolp = NULL ; +LLPrivateMemoryPool* LLVertexBuffer::sPrivatePoolp = NULL; 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; +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; +bool LLVertexBuffer::sVBOActive = false; +bool LLVertexBuffer::sIBOActive = false; U32 LLVertexBuffer::sAllocatedBytes = 0; -BOOL LLVertexBuffer::sMapped = FALSE; -BOOL LLVertexBuffer::sUseStreamDraw = TRUE; -BOOL LLVertexBuffer::sUseVAO = FALSE; -BOOL LLVertexBuffer::sPreferStreamDraw = FALSE; +bool LLVertexBuffer::sMapped = false; +bool LLVertexBuffer::sUseStreamDraw = true; +bool LLVertexBuffer::sUseVAO = false; +bool LLVertexBuffer::sPreferStreamDraw = false; const U32 FENCE_WAIT_TIME_NANOSECONDS = 10000; //1 ms @@ -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); @@ -199,10 +204,16 @@ void LLVBOPool::release(U32 name, U8* buffer, U32 size) Record rec; rec.mGLName = name; rec.mClientData = buffer; - - sBytesPooled += size; - - mFreeList[i].push_back(rec); + + if (buffer == NULL) + { + glDeleteBuffersARB(1, &rec.mGLName); + } + else + { + sBytesPooled += size; + mFreeList[i].push_back(rec); + } } void LLVBOPool::cleanup() @@ -221,7 +232,7 @@ void LLVBOPool::cleanup() if (r.mClientData) { - ll_aligned_free_16(r.mClientData); + ll_aligned_free_16((void*) r.mClientData); } l.pop_front(); @@ -271,7 +282,7 @@ void LLVertexBuffer::setupClientArrays(U32 data_mask) { if (sLastMask != data_mask) { - BOOL error = FALSE; + bool error = false; if (LLGLSLShader::sNoFixedFunction) { @@ -332,7 +343,7 @@ void LLVertexBuffer::setupClientArrays(U32 data_mask) { if (gDebugSession) { - error = TRUE; + error = true; gFailLog << "Bad client state! " << array[i] << " disabled." << std::endl; } else @@ -352,7 +363,7 @@ void LLVertexBuffer::setupClientArrays(U32 data_mask) { //needs to be disabled, make sure it was (DEBUG TEMPORARY) if (gDebugSession) { - error = TRUE; + error = true; gFailLog << "Bad client state! " << array[i] << " enabled." << std::endl; } else @@ -536,7 +547,7 @@ 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 +602,7 @@ 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 +648,7 @@ 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); @@ -677,23 +688,13 @@ void LLVertexBuffer::drawArrays(U32 mode, U32 first, U32 count) const //static void LLVertexBuffer::initClass(bool use_vbo, bool no_vbo_mapping) { - sEnableVBOs = use_vbo && gGLManager.mHasVertexBufferObject ; - sDisableVBOMapping = sEnableVBOs && no_vbo_mapping ; + sEnableVBOs = use_vbo && gGLManager.mHasVertexBufferObject; + sDisableVBOMapping = sEnableVBOs && no_vbo_mapping; - if(!sPrivatePoolp) + if (!sPrivatePoolp) { - sPrivatePoolp = LLPrivateMemoryPoolManager::getInstance()->newPool(LLPrivateMemoryPool::STATIC) ; + sPrivatePoolp = LLPrivateMemoryPoolManager::getInstance()->newPool(LLPrivateMemoryPool::STATIC); } - - sStreamVBOPool.mType = GL_ARRAY_BUFFER_ARB; - sStreamVBOPool.mUsage= GL_STREAM_DRAW_ARB; - sStreamIBOPool.mType = GL_ELEMENT_ARRAY_BUFFER_ARB; - sStreamIBOPool.mUsage= GL_STREAM_DRAW_ARB; - - sDynamicVBOPool.mType = GL_ARRAY_BUFFER_ARB; - sDynamicVBOPool.mUsage= GL_DYNAMIC_DRAW_ARB; - sDynamicIBOPool.mType = GL_ELEMENT_ARRAY_BUFFER_ARB; - sDynamicIBOPool.mUsage= GL_DYNAMIC_DRAW_ARB; } //static @@ -706,18 +707,18 @@ void LLVertexBuffer::unbind() #endif sGLRenderArray = 0; sGLRenderIndices = 0; - sIBOActive = FALSE; + sIBOActive = false; } if (sVBOActive) { glBindBufferARB(GL_ARRAY_BUFFER_ARB, 0); - sVBOActive = FALSE; + sVBOActive = false; } if (sIBOActive) { glBindBufferARB(GL_ELEMENT_ARRAY_BUFFER_ARB, 0); - sIBOActive = FALSE; + sIBOActive = false; } sGLRenderBuffer = 0; @@ -746,62 +747,87 @@ void LLVertexBuffer::cleanupClass() //---------------------------------------------------------------------------- +S32 LLVertexBuffer::determineUsage(S32 usage) +{ + S32 ret_usage = usage; + + if (!sEnableVBOs) + { + ret_usage = 0; + } + + if (usage == GL_STREAM_DRAW_ARB && !sUseStreamDraw) + { + ret_usage = 0; + } + + if (usage == GL_DYNAMIC_DRAW_ARB && sPreferStreamDraw) + { + ret_usage = GL_STREAM_DRAW_ARB; + } + + if (usage == 0 && LLRender::sGLCoreProfile) + { //MUST use VBOs for all rendering + ret_usage = GL_STREAM_DRAW_ARB; + } + + if (usage && usage != GL_STREAM_DRAW_ARB) + { //only stream_draw and dynamic_draw are supported when using VBOs, dynamic draw is the default + if (sDisableVBOMapping) + { //always use stream draw if VBO mapping is disabled + ret_usage = GL_STREAM_DRAW_ARB; + } + else + { + ret_usage = GL_DYNAMIC_DRAW_ARB; + } + } + + return ret_usage; +} + LLVertexBuffer::LLVertexBuffer(U32 typemask, S32 usage) : LLRefCount(), mNumVerts(0), mNumIndices(0), - mUsage(usage), + mAlignedOffset(0), + mAlignedIndexOffset(0), + mSize(0), + mIndicesSize(0), + mTypeMask(typemask), + mUsage(LLVertexBuffer::determineUsage(usage)), mGLBuffer(0), + mGLIndices(0), mGLArray(0), - mGLIndices(0), mMappedData(NULL), - mMappedIndexData(NULL), - mVertexLocked(FALSE), - mIndexLocked(FALSE), - mFinal(FALSE), - mEmpty(TRUE), + mMappedIndexData(NULL), + mMappedDataUsingVBOs(false), + mMappedIndexDataUsingVBOs(false), + mVertexLocked(false), + mIndexLocked(false), + mFinal(false), + mEmpty(true), + mMappable(false), mFence(NULL) { LLMemType mt2(LLMemType::MTYPE_VERTEX_CONSTRUCTOR); - mFence = NULL; - if (!sEnableVBOs) - { - mUsage = 0 ; - } - if (mUsage == GL_STREAM_DRAW_ARB && !sUseStreamDraw) + if (mUsage == GL_DYNAMIC_DRAW_ARB && !sDisableVBOMapping) { - mUsage = 0; + mMappable = true; } - - if (mUsage == GL_DYNAMIC_DRAW_ARB && sPreferStreamDraw) + else { - mUsage = GL_STREAM_DRAW_ARB; + mMappable = false; } - if (mUsage == 0 && LLRender::sGLCoreProfile) - { //MUST use VBOs for all rendering - mUsage = GL_STREAM_DRAW_ARB; - } - - 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; - } - //zero out offsets for (U32 i = 0; i < TYPE_MAX; i++) { mOffsets[i] = 0; } - mTypeMask = typemask; - mSize = 0; - mIndicesSize = 0; - mAlignedOffset = 0; - mAlignedIndexOffset = 0; - sCount++; } @@ -982,9 +1008,11 @@ void LLVertexBuffer::createGLBuffer(U32 size) return; } - mEmpty = TRUE; + mEmpty = true; - if (useVBOs()) + mMappedDataUsingVBOs = useVBOs(); + + if (mMappedDataUsingVBOs) { genBuffer(size); } @@ -1011,12 +1039,14 @@ void LLVertexBuffer::createGLIndices(U32 size) return; } - mEmpty = TRUE; + mEmpty = true; //pad by 16 bytes for aligned copies size += 16; - if (useVBOs()) + mMappedIndexDataUsingVBOs = useVBOs(); + + if (mMappedIndexDataUsingVBOs) { //pad by another 16 bytes for VBO pointer adjustment size += 16; @@ -1036,15 +1066,15 @@ void LLVertexBuffer::destroyGLBuffer() LLMemType mt2(LLMemType::MTYPE_VERTEX_DESTROY_BUFFER); if (mGLBuffer) { - if (useVBOs()) + if (mMappedDataUsingVBOs) { releaseBuffer(); } else { - FREE_MEM(sPrivatePoolp, mMappedData) ; + FREE_MEM(sPrivatePoolp, (void*) mMappedData); mMappedData = NULL; - mEmpty = TRUE; + mEmpty = true; } } @@ -1057,15 +1087,15 @@ void LLVertexBuffer::destroyGLIndices() LLMemType mt2(LLMemType::MTYPE_VERTEX_DESTROY_INDICES); if (mGLIndices) { - if (useVBOs()) + if (mMappedIndexDataUsingVBOs) { releaseIndices(); } else { - FREE_MEM(sPrivatePoolp, mMappedIndexData) ; + FREE_MEM(sPrivatePoolp, (void*) mMappedIndexData); mMappedIndexData = NULL; - mEmpty = TRUE; + mEmpty = true; } } @@ -1250,16 +1280,10 @@ void LLVertexBuffer::resizeBuffer(S32 newnverts, S32 newnindices) } } -BOOL LLVertexBuffer::useVBOs() const +bool LLVertexBuffer::useVBOs() const { //it's generally ineffective to use VBO for things that are streaming on apple - - if (!mUsage) - { - return FALSE; - } - - return TRUE; + return (mUsage != 0); } //---------------------------------------------------------------------------- @@ -1282,8 +1306,11 @@ 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 +1325,7 @@ 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 +1350,7 @@ 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); } } @@ -1336,23 +1363,24 @@ U8* LLVertexBuffer::mapVertexBuffer(S32 type, S32 index, S32 count, bool map_ran if (!mVertexLocked) { LLMemType mt_v(LLMemType::MTYPE_VERTEX_MAP_BUFFER_VERTICES); - mVertexLocked = TRUE; + mVertexLocked = true; 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 +1404,7 @@ 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 +1432,7 @@ 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(); @@ -1413,10 +1442,10 @@ U8* LLVertexBuffer::mapVertexBuffer(S32 type, S32 index, S32 count, bool map_ran { log_glerror(); - //check the availability of memory - LLMemory::logMemoryInfo(TRUE) ; + //check the availability of memory + LLMemory::logMemoryInfo(true) ; - if(!sDisableVBOMapping) + if(mMappable) { //-------------------- //print out more debug info before crash @@ -1448,7 +1477,7 @@ 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 +1487,11 @@ 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 +1506,7 @@ 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 +1528,7 @@ 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); } } @@ -1509,7 +1542,7 @@ U8* LLVertexBuffer::mapIndexBuffer(S32 index, S32 count, bool map_range) { LLMemType mt_v(LLMemType::MTYPE_VERTEX_MAP_BUFFER_INDICES); - mIndexLocked = TRUE; + mIndexLocked = true; sMappedCount++; stop_glerror(); @@ -1524,19 +1557,20 @@ 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 +1582,7 @@ 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 +1604,7 @@ 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); } @@ -1585,9 +1621,9 @@ U8* LLVertexBuffer::mapIndexBuffer(S32 index, S32 count, bool map_range) if (!mMappedIndexData) { log_glerror(); - LLMemory::logMemoryInfo(TRUE) ; + LLMemory::logMemoryInfo(true) ; - if(!sDisableVBOMapping) + if(mMappable) { GLint buff; glGetIntegerv(GL_ELEMENT_ARRAY_BUFFER_BINDING_ARB, &buff); @@ -1609,7 +1645,7 @@ 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 +1655,13 @@ 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 +1674,11 @@ 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 +1688,7 @@ 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 +1697,7 @@ 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 +1715,7 @@ 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 @@ -1692,14 +1737,15 @@ void LLVertexBuffer::unmapBuffer() mMappedData = NULL; } - mVertexLocked = FALSE ; + mVertexLocked = false; sMappedCount--; } if (mMappedIndexData && mIndexLocked) { + LLFastTimer t(FTM_IBO_UNMAP); bindGLIndices(); - if(sDisableVBOMapping) + if(!mMappable) { if (!mMappedIndexRegions.empty()) { @@ -1708,7 +1754,7 @@ 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 +1763,7 @@ 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 +1780,7 @@ 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 @@ -1757,13 +1804,13 @@ void LLVertexBuffer::unmapBuffer() mMappedIndexData = NULL ; } - mIndexLocked = FALSE ; + mIndexLocked = false ; sMappedCount--; } if(updated_all) { - mEmpty = FALSE; + mEmpty = false; } } @@ -1778,39 +1825,39 @@ 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) { llwarns << "mapIndexBuffer failed!" << llendl; - return FALSE; + return false; } strider = (T*)ptr; strider.setStride(0); - return TRUE; + return true; } else if (vbo.hasDataType(type)) { 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) { llwarns << "mapVertexBuffer failed!" << llendl; - return FALSE; + return false; } strider = (T*)ptr; strider.setStride(stride); - return TRUE; + return true; } else { llerrs << "VertexBufferStrider could not find valid vertex data." << llendl; } - return FALSE; + return false; } }; @@ -1909,7 +1956,7 @@ bool LLVertexBuffer::bindGLBuffer(bool force_bind) glBindBufferARB(GL_ARRAY_BUFFER_ARB, mGLBuffer); sGLRenderBuffer = mGLBuffer; sBindCount++; - sVBOActive = TRUE; + sVBOActive = true; if (mGLArray) { @@ -1941,7 +1988,7 @@ bool LLVertexBuffer::bindGLIndices(bool force_bind) sGLRenderIndices = mGLIndices; stop_glerror(); sBindCount++; - sIBOActive = TRUE; + sIBOActive = true; ret = true; } @@ -1963,7 +2010,7 @@ void LLVertexBuffer::setBuffer(U32 data_mask) LLMemType mt2(LLMemType::MTYPE_VERTEX_SET_BUFFER); //set up pointers if the data mask is different ... - BOOL setup = (sLastMask != data_mask); + bool setup = (sLastMask != data_mask); if (gDebugGL && data_mask != 0) { //make sure data requirements are fulfilled @@ -1997,21 +2044,17 @@ void LLVertexBuffer::setBuffer(U32 data_mask) if (mGLArray) { bindGLArray(); - setup = FALSE; //do NOT perform pointer setup if using VAO + setup = false; //do NOT perform pointer setup if using VAO } else { - if (bindGLBuffer()) - { - setup = TRUE; - } - if (bindGLIndices()) - { - setup = TRUE; - } + const bool bindBuffer = bindGLBuffer(); + const bool bindIndices = bindGLIndices(); + + setup = setup || bindBuffer || bindIndices; } - BOOL error = FALSE; + bool error = false; if (gDebugGL && !mGLArray) { GLint buff; @@ -2020,7 +2063,7 @@ void LLVertexBuffer::setBuffer(U32 data_mask) { if (gDebugSession) { - error = TRUE; + error = true; gFailLog << "Invalid GL vertex buffer bound: " << buff << std::endl; } else @@ -2036,7 +2079,7 @@ void LLVertexBuffer::setBuffer(U32 data_mask) { if (gDebugSession) { - error = TRUE; + error = true; gFailLog << "Invalid GL index buffer bound: " << buff << std::endl; } else @@ -2058,7 +2101,7 @@ void LLVertexBuffer::setBuffer(U32 data_mask) #endif sGLRenderArray = 0; sGLRenderIndices = 0; - sIBOActive = FALSE; + sIBOActive = false; } if (mGLBuffer) @@ -2067,13 +2110,13 @@ void LLVertexBuffer::setBuffer(U32 data_mask) { glBindBufferARB(GL_ARRAY_BUFFER_ARB, 0); sBindCount++; - sVBOActive = FALSE; - setup = TRUE; // ... or a VBO is deactivated + sVBOActive = false; + setup = true; // ... or a VBO is deactivated } if (sGLRenderBuffer != mGLBuffer) { sGLRenderBuffer = mGLBuffer; - setup = TRUE; // ... or a client memory pointer changed + setup = true; // ... or a client memory pointer changed } } if (mGLIndices) @@ -2082,7 +2125,7 @@ void LLVertexBuffer::setBuffer(U32 data_mask) { glBindBufferARB(GL_ELEMENT_ARRAY_BUFFER_ARB, 0); sBindCount++; - sIBOActive = FALSE; + sIBOActive = false; } sGLRenderIndices = mGLIndices; @@ -2109,7 +2152,7 @@ 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) { @@ -2170,19 +2213,19 @@ void LLVertexBuffer::setupVertexBuffer(U32 data_mask) { S32 loc = TYPE_WEIGHT; void* ptr = (void*)(base + mOffsets[TYPE_WEIGHT]); - glVertexAttribPointerARB(loc, 1, GL_FLOAT, FALSE, LLVertexBuffer::sTypeSize[TYPE_WEIGHT], ptr); + glVertexAttribPointerARB(loc, 1, GL_FLOAT, GL_FALSE, LLVertexBuffer::sTypeSize[TYPE_WEIGHT], ptr); } if (data_mask & MAP_WEIGHT4) { S32 loc = TYPE_WEIGHT4; void* ptr = (void*)(base+mOffsets[TYPE_WEIGHT4]); - glVertexAttribPointerARB(loc, 4, GL_FLOAT, FALSE, LLVertexBuffer::sTypeSize[TYPE_WEIGHT4], ptr); + glVertexAttribPointerARB(loc, 4, GL_FLOAT, GL_FALSE, LLVertexBuffer::sTypeSize[TYPE_WEIGHT4], ptr); } if (data_mask & MAP_CLOTHWEIGHT) { S32 loc = TYPE_CLOTHWEIGHT; void* ptr = (void*)(base + mOffsets[TYPE_CLOTHWEIGHT]); - glVertexAttribPointerARB(loc, 4, GL_FLOAT, TRUE, LLVertexBuffer::sTypeSize[TYPE_CLOTHWEIGHT], ptr); + glVertexAttribPointerARB(loc, 4, GL_FLOAT, GL_TRUE, LLVertexBuffer::sTypeSize[TYPE_CLOTHWEIGHT], ptr); } if (data_mask & MAP_TEXTURE_INDEX) { diff --git a/indra/llrender/llvertexbuffer.h b/indra/llrender/llvertexbuffer.h index 3e6f6a959a..0bf689bfad 100644 --- a/indra/llrender/llvertexbuffer.h +++ b/indra/llrender/llvertexbuffer.h @@ -55,15 +55,20 @@ class LLVBOPool { public: static U32 sBytesPooled; + + LLVBOPool(U32 vboUsage, U32 vboType) + : mUsage(vboUsage) + , mType(vboType) + {} - U32 mUsage; - U32 mType; + const U32 mUsage; + const 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 +77,7 @@ public: { public: U32 mGLName; - U8* mClientData; + volatile U8* mClientData; }; typedef std::list<Record> record_list_t; @@ -103,6 +108,7 @@ public: }; LLVertexBuffer(const LLVertexBuffer& rhs) + : mUsage(rhs.mUsage) { *this = rhs; } @@ -118,9 +124,9 @@ public: static LLVBOPool sStreamIBOPool; static LLVBOPool sDynamicIBOPool; - static BOOL sUseStreamDraw; - static BOOL sUseVAO; - static BOOL sPreferStreamDraw; + static bool sUseStreamDraw; + static bool sUseVAO; + static bool sPreferStreamDraw; static void initClass(bool use_vbo, bool no_vbo_mapping); static void cleanupClass(); @@ -201,15 +207,15 @@ protected: void destroyGLIndices(); void updateNumVerts(S32 nverts); void updateNumIndices(S32 nindices); - virtual BOOL useVBOs() const; + bool useVBOs() const; void unmapBuffer(); 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 @@ -239,21 +245,22 @@ public: bool getClothWeightStrider(LLStrider<LLVector4>& strider, S32 index=0, S32 count = -1, bool map_range = false); - BOOL isEmpty() const { return mEmpty; } - BOOL isLocked() const { return mVertexLocked || mIndexLocked; } + bool isEmpty() const { return mEmpty; } + bool isLocked() const { return mVertexLocked || mIndexLocked; } 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; @@ -273,17 +280,24 @@ protected: S32 mSize; S32 mIndicesSize; U32 mTypeMask; - S32 mUsage; // GL usage + + const S32 mUsage; // GL usage + U32 mGLBuffer; // GL VBO handle 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) - 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. + volatile U8* mMappedData; // pointer to currently mapped data (NULL if unmapped) + volatile U8* mMappedIndexData; // pointer to currently mapped indices (NULL if unmapped) + + U32 mMappedDataUsingVBOs : 1; + U32 mMappedIndexDataUsingVBOs : 1; + U32 mVertexLocked : 1; // if true, vertex buffer is being or has been written to in client memory + U32 mIndexLocked : 1; // if true, index buffer is being or has been written to in client memory + U32 mFinal : 1; // if true, buffer can not be mapped again + U32 mEmpty : 1; // 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; @@ -294,6 +308,7 @@ protected: void placeFence() const; void waitFence() const; + static S32 determineUsage(S32 usage); private: static LLPrivateMemoryPool* sPrivatePoolp ; @@ -302,18 +317,18 @@ public: static S32 sCount; static S32 sGLCount; static S32 sMappedCount; - static BOOL sMapped; + static bool sMapped; typedef std::list<LLVertexBuffer*> buffer_list_t; - static BOOL sDisableVBOMapping; //disable glMapBufferARB - static BOOL sEnableVBOs; + static bool sDisableVBOMapping; //disable glMapBufferARB + static bool sEnableVBOs; static S32 sTypeSize[TYPE_MAX]; static U32 sGLMode[LLRender::NUM_MODES]; static U32 sGLRenderBuffer; static U32 sGLRenderArray; static U32 sGLRenderIndices; - static BOOL sVBOActive; - static BOOL sIBOActive; + static bool sVBOActive; + static bool sIBOActive; static U32 sLastMask; static U32 sAllocatedBytes; static U32 sBindCount; |