llvertexbuffer.cpp

{
	LLMemType mt2(LLMemType::MTYPE_VERTEX_DESTRUCTOR);
	destroyGLBuffer();
	destroyGLIndices();

	if (mGLArray)
	{
#if GL_ARB_vertex_array_object
		releaseVAOName(mGLArray);
#endif
	}

	sCount--;

	if (mFence)
	{
		delete mFence;
	}
	
	mFence = NULL;

	sVertexCount -= mNumVerts;
	sIndexCount -= mNumIndices;

	llassert_always(!mMappedData && !mMappedIndexData);
};

void LLVertexBuffer::placeFence() const
{
	/*if (!mFence && useVBOs())
	{
		if (gGLManager.mHasSync)
		{
			mFence = new LLGLSyncFence();
		}
	}

	if (mFence)
	{
		mFence->placeFence();
	}*/
}

void LLVertexBuffer::waitFence() const
{
	/*if (mFence)
	{
		mFence->wait();
	}*/
}

//----------------------------------------------------------------------------

void LLVertexBuffer::genBuffer(U32 size)
{
	mSize = vbo_block_size(size);

	if (mUsage == GL_STREAM_DRAW_ARB)
	{
		mMappedData = sStreamVBOPool.allocate(mGLBuffer, mSize);
	}
	else
	{
		mMappedData = sDynamicVBOPool.allocate(mGLBuffer, mSize);
	}
	
	sGLCount++;
}

void LLVertexBuffer::genIndices(U32 size)
{
	mIndicesSize = vbo_block_size(size);

	if (mUsage == GL_STREAM_DRAW_ARB)
	{
		mMappedIndexData = sStreamIBOPool.allocate(mGLIndices, mIndicesSize);
	}
	else
	{
		mMappedIndexData = sDynamicIBOPool.allocate(mGLIndices, mIndicesSize);
	}
	
	sGLCount++;
}

void LLVertexBuffer::releaseBuffer()
{
	if (mUsage == GL_STREAM_DRAW_ARB)
	{
		sStreamVBOPool.release(mGLBuffer, mMappedData, mSize);
	}
	else
	{
		sDynamicVBOPool.release(mGLBuffer, mMappedData, mSize);
	}
	
	mGLBuffer = 0;
	mMappedData = NULL;

	sGLCount--;
}

void LLVertexBuffer::releaseIndices()
{
	if (mUsage == GL_STREAM_DRAW_ARB)
	{
		sStreamIBOPool.release(mGLIndices, mMappedIndexData, mIndicesSize);
	}
	else
	{
		sDynamicIBOPool.release(mGLIndices, mMappedIndexData, mIndicesSize);
	}

	mGLIndices = 0;
	mMappedIndexData = NULL;
	
	sGLCount--;
}

void LLVertexBuffer::createGLBuffer(U32 size)
{
	LLMemType mt2(LLMemType::MTYPE_VERTEX_CREATE_VERTICES);
	
	if (mGLBuffer)
	{
		destroyGLBuffer();
	}

	if (size == 0)
	{
		return;
	}

	mEmpty = true;

	mMappedDataUsingVBOs = useVBOs();
	
	if (mMappedDataUsingVBOs)
	{
		genBuffer(size);
	}
	else
	{
		static int gl_buffer_idx = 0;
		mGLBuffer = ++gl_buffer_idx;
		mMappedData = (U8*)ALLOCATE_MEM(sPrivatePoolp, size);
		mSize = size;
	}
}

void LLVertexBuffer::createGLIndices(U32 size)
{
	LLMemType mt2(LLMemType::MTYPE_VERTEX_CREATE_INDICES);
	
	if (mGLIndices)
	{
		destroyGLIndices();
	}
	
	if (size == 0)
	{
		return;
	}

	mEmpty = true;

	//pad by 16 bytes for aligned copies
	size += 16;

	mMappedIndexDataUsingVBOs = useVBOs();

	if (mMappedIndexDataUsingVBOs)
	{
		//pad by another 16 bytes for VBO pointer adjustment
		size += 16;
		genIndices(size);
	}
	else
	{
		mMappedIndexData = (U8*)ALLOCATE_MEM(sPrivatePoolp, size);
		static int gl_buffer_idx = 0;
		mGLIndices = ++gl_buffer_idx;
		mIndicesSize = size;
	}
}

void LLVertexBuffer::destroyGLBuffer()
{
	LLMemType mt2(LLMemType::MTYPE_VERTEX_DESTROY_BUFFER);
	if (mGLBuffer)
	{
		if (mMappedDataUsingVBOs)
		{
			releaseBuffer();
		}
		else
		{
			FREE_MEM(sPrivatePoolp, (void*) mMappedData);
			mMappedData = NULL;
			mEmpty = true;
		}
	}
	
	mGLBuffer = 0;
	//unbind();
}

void LLVertexBuffer::destroyGLIndices()
{
	LLMemType mt2(LLMemType::MTYPE_VERTEX_DESTROY_INDICES);
	if (mGLIndices)
	{
		if (mMappedIndexDataUsingVBOs)
		{
			releaseIndices();
		}
		else
		{
			FREE_MEM(sPrivatePoolp, (void*) mMappedIndexData);
			mMappedIndexData = NULL;
			mEmpty = true;
		}
	}

	mGLIndices = 0;
	//unbind();
}

void LLVertexBuffer::updateNumVerts(S32 nverts)
{
	LLMemType mt2(LLMemType::MTYPE_VERTEX_UPDATE_VERTS);

	llassert(nverts >= 0);

	if (nverts > 65536)
	{
		llwarns << "Vertex buffer overflow!" << llendl;
		nverts = 65536;
	}

	U32 needed_size = calcOffsets(mTypeMask, mOffsets, nverts);

	if (needed_size > mSize || needed_size <= mSize/2)
	{
		createGLBuffer(needed_size);
	}

	sVertexCount -= mNumVerts;
	mNumVerts = nverts;
	sVertexCount += mNumVerts;
}

void LLVertexBuffer::updateNumIndices(S32 nindices)
{
	LLMemType mt2(LLMemType::MTYPE_VERTEX_UPDATE_INDICES);

	llassert(nindices >= 0);

	U32 needed_size = sizeof(U16) * nindices;

	if (needed_size > mIndicesSize || needed_size <= mIndicesSize/2)
	{
		createGLIndices(needed_size);
	}

	sIndexCount -= mNumIndices;
	mNumIndices = nindices;
	sIndexCount += mNumIndices;
}

void LLVertexBuffer::allocateBuffer(S32 nverts, S32 nindices, bool create)
{
	LLMemType mt2(LLMemType::MTYPE_VERTEX_ALLOCATE_BUFFER);
	
	stop_glerror();

	if (nverts < 0 || nindices < 0 ||
		nverts > 65536)
	{
		llerrs << "Bad vertex buffer allocation: " << nverts << " : " << nindices << llendl;
	}

	updateNumVerts(nverts);
	updateNumIndices(nindices);
	
	if (create && (nverts || nindices))
	{
		//actually allocate space for the vertex buffer if using VBO mapping
		flush();

		if (gGLManager.mHasVertexArrayObject && useVBOs() && (LLRender::sGLCoreProfile || sUseVAO))
		{
#if GL_ARB_vertex_array_object
			mGLArray = getVAOName();
#endif
			setupVertexArray();
		}
	}
}

static LLFastTimer::DeclareTimer FTM_SETUP_VERTEX_ARRAY("Setup VAO");

void LLVertexBuffer::setupVertexArray()
{
	if (!mGLArray)
	{
		return;
	}

	LLFastTimer t(FTM_SETUP_VERTEX_ARRAY);
#if GL_ARB_vertex_array_object
	glBindVertexArray(mGLArray);
#endif
	sGLRenderArray = mGLArray;

	U32 attrib_size[] = 
	{
		3, //TYPE_VERTEX,
		3, //TYPE_NORMAL,
		2, //TYPE_TEXCOORD0,
		2, //TYPE_TEXCOORD1,
		2, //TYPE_TEXCOORD2,
		2, //TYPE_TEXCOORD3,
		4, //TYPE_COLOR,
		4, //TYPE_EMISSIVE,
		3, //TYPE_BINORMAL,
		1, //TYPE_WEIGHT,
		4, //TYPE_WEIGHT4,
		4, //TYPE_CLOTHWEIGHT,
		1, //TYPE_TEXTURE_INDEX
	};

	U32 attrib_type[] =
	{
		GL_FLOAT, //TYPE_VERTEX,
		GL_FLOAT, //TYPE_NORMAL,
		GL_FLOAT, //TYPE_TEXCOORD0,
		GL_FLOAT, //TYPE_TEXCOORD1,
		GL_FLOAT, //TYPE_TEXCOORD2,
		GL_FLOAT, //TYPE_TEXCOORD3,
		GL_UNSIGNED_BYTE, //TYPE_COLOR,
		GL_UNSIGNED_BYTE, //TYPE_EMISSIVE,
		GL_FLOAT,   //TYPE_BINORMAL,
		GL_FLOAT, //TYPE_WEIGHT,
		GL_FLOAT, //TYPE_WEIGHT4,
		GL_FLOAT, //TYPE_CLOTHWEIGHT,
		GL_UNSIGNED_INT, //TYPE_TEXTURE_INDEX
	};

	bool attrib_integer[] = 
	{
		false, //TYPE_VERTEX,
		false, //TYPE_NORMAL,
		false, //TYPE_TEXCOORD0,
		false, //TYPE_TEXCOORD1,
		false, //TYPE_TEXCOORD2,
		false, //TYPE_TEXCOORD3,
		false, //TYPE_COLOR,
		false, //TYPE_EMISSIVE,
		false, //TYPE_BINORMAL,
		false, //TYPE_WEIGHT,
		false, //TYPE_WEIGHT4,
		false, //TYPE_CLOTHWEIGHT,
		true, //TYPE_TEXTURE_INDEX
	};

	U32 attrib_normalized[] =
	{
		GL_FALSE, //TYPE_VERTEX,
		GL_FALSE, //TYPE_NORMAL,
		GL_FALSE, //TYPE_TEXCOORD0,
		GL_FALSE, //TYPE_TEXCOORD1,
		GL_FALSE, //TYPE_TEXCOORD2,
		GL_FALSE, //TYPE_TEXCOORD3,
		GL_TRUE, //TYPE_COLOR,
		GL_TRUE, //TYPE_EMISSIVE,
		GL_FALSE,   //TYPE_BINORMAL,
		GL_FALSE, //TYPE_WEIGHT,
		GL_FALSE, //TYPE_WEIGHT4,
		GL_FALSE, //TYPE_CLOTHWEIGHT,
		GL_FALSE, //TYPE_TEXTURE_INDEX
	};

	bindGLBuffer(true);
	bindGLIndices(true);

	for (U32 i = 0; i < TYPE_MAX; ++i)
	{
		if (mTypeMask & (1 << i))
		{
			glEnableVertexAttribArrayARB(i);

			if (attrib_integer[i])
			{
#if !LL_DARWIN
				//glVertexattribIPointer requires GLSL 1.30 or later
				if (gGLManager.mGLSLVersionMajor > 1 || gGLManager.mGLSLVersionMinor >= 30)
				{
					glVertexAttribIPointer(i, attrib_size[i], attrib_type[i], sTypeSize[i], (void*) mOffsets[i]); 
				}
#endif
			}
			else
			{
				glVertexAttribPointerARB(i, attrib_size[i], attrib_type[i], attrib_normalized[i], sTypeSize[i], (void*) mOffsets[i]); 
			}
		}
		else
		{
			glDisableVertexAttribArrayARB(i);
		}
	}

	//draw a dummy triangle to set index array pointer
	//glDrawElements(GL_TRIANGLES, 0, GL_UNSIGNED_SHORT, NULL);

	unbind();
}

void LLVertexBuffer::resizeBuffer(S32 newnverts, S32 newnindices)
{
	llassert(newnverts >= 0);
	llassert(newnindices >= 0);

	LLMemType mt2(LLMemType::MTYPE_VERTEX_RESIZE_BUFFER);
	
	updateNumVerts(newnverts);		
	updateNumIndices(newnindices);
	
	if (useVBOs())
	{
		flush();

		if (mGLArray)
		{ //if size changed, offsets changed
			setupVertexArray();
		}
	}
}

bool LLVertexBuffer::useVBOs() const
{
	//it's generally ineffective to use VBO for things that are streaming on apple
	return (mUsage != 0);
}

//----------------------------------------------------------------------------

bool expand_region(LLVertexBuffer::MappedRegion& region, S32 index, S32 count)
{
	S32 end = index+count;
	S32 region_end = region.mIndex+region.mCount;
	
	if (end < region.mIndex ||
		index > region_end)
	{ //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;
	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
volatile U8* LLVertexBuffer::mapVertexBuffer(S32 type, S32 index, S32 count, bool map_range)
{
	bindGLBuffer(true);
	LLMemType mt2(LLMemType::MTYPE_VERTEX_MAP_BUFFER);
	if (mFinal)
	{
		llerrs << "LLVertexBuffer::mapVeretxBuffer() called on a finalized buffer." << llendl;
	}
	if (!useVBOs() && !mMappedData && !mMappedIndexData)
	{
		llerrs << "LLVertexBuffer::mapVertexBuffer() called on unallocated buffer." << llendl;
	}
		
	if (useVBOs())
	{
		if (!mMappable || gGLManager.mHasMapBufferRange || gGLManager.mHasFlushBufferRange)
		{
			if (count == -1)
			{
				count = mNumVerts-index;
			}

			bool mapped = false;
			//see if range is already 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 (!mapped)
			{
				//not already mapped, map new region
				MappedRegion region(type, mMappable && map_range ? -1 : index, count);
				mMappedVertexRegions.push_back(region);
			}
		}

		if (mVertexLocked && map_range)
		{
			llerrs << "Attempted to map a specific range of a buffer that was already mapped." << llendl;
		}

		if (!mVertexLocked)
		{
			LLMemType mt_v(LLMemType::MTYPE_VERTEX_MAP_BUFFER_VERTICES);
			mVertexLocked = true;
			sMappedCount++;
			stop_glerror();	

			if(!mMappable)
			{
				map_range = false;
			}
			else
			{
				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, 
							GL_MAP_WRITE_BIT | 
							GL_MAP_FLUSH_EXPLICIT_BIT | 
							GL_MAP_INVALIDATE_RANGE_BIT);
#endif
					}
					else
					{
#ifdef GL_ARB_map_buffer_range

						if (gDebugGL)
						{
							GLint size = 0;
							glGetBufferParameterivARB(GL_ARRAY_BUFFER_ARB, GL_BUFFER_SIZE_ARB, &size);

							if (size < mSize)
							{
								llerrs << "Invalid buffer size." << llendl;
							}
						}

						LLFastTimer t(FTM_VBO_MAP_BUFFER);
						src = (U8*) glMapBufferRange(GL_ARRAY_BUFFER_ARB, 0, mSize, 
							GL_MAP_WRITE_BIT | 
							GL_MAP_FLUSH_EXPLICIT_BIT);
#endif
					}
				}
				else if (gGLManager.mHasFlushBufferRange)
				{
					if (map_range)
					{
						glBufferParameteriAPPLE(GL_ARRAY_BUFFER_ARB, GL_BUFFER_SERIALIZED_MODIFY_APPLE, GL_FALSE);
						glBufferParameteriAPPLE(GL_ARRAY_BUFFER_ARB, GL_BUFFER_FLUSHING_UNMAP_APPLE, GL_FALSE);
						src = (U8*) glMapBufferARB(GL_ARRAY_BUFFER_ARB, GL_WRITE_ONLY_ARB);
					}
					else
					{
						src = (U8*) glMapBufferARB(GL_ARRAY_BUFFER_ARB, GL_WRITE_ONLY_ARB);
					}
				}
				else
				{
					map_range = false;
					src = (U8*) glMapBufferARB(GL_ARRAY_BUFFER_ARB, GL_WRITE_ONLY_ARB);
				}

				llassert(src != NULL);

				mMappedData = LL_NEXT_ALIGNED_ADDRESS<volatile U8>(src);
				mAlignedOffset = mMappedData - src;
			
				stop_glerror();
			}
				
			if (!mMappedData)
			{
				log_glerror();

				//check the availability of memory
				LLMemory::logMemoryInfo(true);
			
				if(mMappable)
				{			
					//--------------------
					//print out more debug info before crash
					llinfos << "vertex buffer size: (num verts : num indices) = " << getNumVerts() << " : " << getNumIndices() << llendl;
					GLint size;
					glGetBufferParameterivARB(GL_ARRAY_BUFFER_ARB, GL_BUFFER_SIZE_ARB, &size);
					llinfos << "GL_ARRAY_BUFFER_ARB size is " << size << llendl;
					//--------------------

					GLint buff;
					glGetIntegerv(GL_ARRAY_BUFFER_BINDING_ARB, &buff);
					if ((GLuint)buff != mGLBuffer)
					{
						llerrs << "Invalid GL vertex buffer bound: " << buff << llendl;
					}

							
					llerrs << "glMapBuffer returned NULL (no vertex data)" << llendl;
				}
				else
				{
					llerrs << "memory allocation for vertex data failed." << llendl;
				}
			}
		}
	}
	else
	{
		map_range = false;
	}
	
	if (map_range && gGLManager.mHasMapBufferRange && mMappable)
	{
		return mMappedData;
	}
	else
	{
		return mMappedData+mOffsets[type]+sTypeSize[type]*index;
	}
}


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);
	if (mFinal)
	{
		llerrs << "LLVertexBuffer::mapIndexBuffer() called on a finalized buffer." << llendl;
	}
	if (!useVBOs() && !mMappedData && !mMappedIndexData)
	{
		llerrs << "LLVertexBuffer::mapIndexBuffer() called on unallocated buffer." << llendl;
	}

	if (useVBOs())
	{
		if (!mMappable || gGLManager.mHasMapBufferRange || gGLManager.mHasFlushBufferRange)
		{
			if (count == -1)
			{
				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;
				}
			}

			if (!mapped)
			{
				//not already mapped, map new region
				MappedRegion region(TYPE_INDEX, mMappable && map_range ? -1 : index, count);
				mMappedIndexRegions.push_back(region);
			}
		}

		if (mIndexLocked && map_range)
		{
			llerrs << "Attempted to map a specific range of a buffer that was already mapped." << llendl;
		}

		if (!mIndexLocked)
		{
			LLMemType mt_v(LLMemType::MTYPE_VERTEX_MAP_BUFFER_INDICES);

			mIndexLocked = true;
			sMappedCount++;
			stop_glerror();	

			if (gDebugGL && useVBOs())
			{
				GLint elem = 0;
				glGetIntegerv(GL_ELEMENT_ARRAY_BUFFER_BINDING_ARB, &elem);

				if (elem != mGLIndices)
				{
					llerrs << "Wrong index buffer bound!" << llendl;
				}
			}

			if(!mMappable)
			{
				map_range = false;
			}
			else
			{
				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, 
							GL_MAP_WRITE_BIT | 
							GL_MAP_FLUSH_EXPLICIT_BIT | 
							GL_MAP_INVALIDATE_RANGE_BIT);
#endif
					}
					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);
#endif
					}
				}
				else if (gGLManager.mHasFlushBufferRange)
				{
					if (map_range)
					{
						glBufferParameteriAPPLE(GL_ELEMENT_ARRAY_BUFFER_ARB, GL_BUFFER_SERIALIZED_MODIFY_APPLE, GL_FALSE);
						glBufferParameteriAPPLE(GL_ELEMENT_ARRAY_BUFFER_ARB, GL_BUFFER_FLUSHING_UNMAP_APPLE, GL_FALSE);
						src = (U8*) glMapBufferARB(GL_ELEMENT_ARRAY_BUFFER_ARB, GL_WRITE_ONLY_ARB);
					}
					else
					{
						src = (U8*) glMapBufferARB(GL_ELEMENT_ARRAY_BUFFER_ARB, GL_WRITE_ONLY_ARB);
					}
				}
				else
				{
					LLFastTimer t(FTM_VBO_MAP_INDEX);
					map_range = false;
					src = (U8*) glMapBufferARB(GL_ELEMENT_ARRAY_BUFFER_ARB, GL_WRITE_ONLY_ARB);
				}

				llassert(src != NULL);


				mMappedIndexData = src; //LL_NEXT_ALIGNED_ADDRESS<U8>(src);
				mAlignedIndexOffset = mMappedIndexData - src;
				stop_glerror();
			}
		}

		if (!mMappedIndexData)
		{
			log_glerror();
			LLMemory::logMemoryInfo(true);

			if(mMappable)
			{
				GLint buff;
				glGetIntegerv(GL_ELEMENT_ARRAY_BUFFER_BINDING_ARB, &buff);
				if ((GLuint)buff != mGLIndices)
				{
					llerrs << "Invalid GL index buffer bound: " << buff << llendl;
				}

				llerrs << "glMapBuffer returned NULL (no index data)" << llendl;
			}
			else
			{
				llerrs << "memory allocation for Index data failed. " << llendl;
			}
		}
	}
	else
	{
		map_range = false;
	}

	if (map_range && gGLManager.mHasMapBufferRange && mMappable)
	{
		return mMappedIndexData;
	}
	else
	{
		return mMappedIndexData + sizeof(U16)*index;
	}
}

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);
	if (!useVBOs())
	{
		return; //nothing to unmap
	}

	bool updated_all = false;

	if (mMappedData && mVertexLocked)
	{
		LLFastTimer t(FTM_VBO_UNMAP);
		bindGLBuffer(true);
		updated_all = mIndexLocked; //both vertex and index buffers done updating

		if(!mMappable)
		{
			if (!mMappedVertexRegions.empty())
			{
				stop_glerror();
				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;
					glBufferSubDataARB(GL_ARRAY_BUFFER_ARB, offset, length, (U8*) mMappedData+offset);
					stop_glerror();
				}

				mMappedVertexRegions.clear();
			}
			else
			{
				stop_glerror();
				glBufferSubDataARB(GL_ARRAY_BUFFER_ARB, 0, getSize(), (U8*) mMappedData);
				stop_glerror();
			}
		}
		else
		{
			if (gGLManager.mHasMapBufferRange || gGLManager.mHasFlushBufferRange)
			{
				if (!mMappedVertexRegions.empty())
				{
					stop_glerror();
					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 (gGLManager.mHasMapBufferRange)
						{
							LLFastTimer t(FTM_VBO_FLUSH_RANGE);
#ifdef GL_ARB_map_buffer_range
							glFlushMappedBufferRange(GL_ARRAY_BUFFER_ARB, offset, length);
#endif
						}
						else if (gGLManager.mHasFlushBufferRange)
						{
							glFlushMappedBufferRangeAPPLE(GL_ARRAY_BUFFER_ARB, offset, length);
						}
						stop_glerror();
					}

					mMappedVertexRegions.clear();
				}
			}
			stop_glerror();
			glUnmapBufferARB(GL_ARRAY_BUFFER_ARB);
			stop_glerror();

			mMappedData = NULL;
		}

		mVertexLocked = false;
		sMappedCount--;
	}
	
	if (mMappedIndexData && mIndexLocked)
	{
		LLFastTimer t(FTM_IBO_UNMAP);
		bindGLIndices();
		if(!mMappable)
		{
			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;
					glBufferSubDataARB(GL_ELEMENT_ARRAY_BUFFER_ARB, offset, length, (U8*) mMappedIndexData+offset);
					stop_glerror();
				}

				mMappedIndexRegions.clear();
			}
			else
			{
				stop_glerror();
				glBufferSubDataARB(GL_ELEMENT_ARRAY_BUFFER_ARB, 0, getIndicesSize(), (U8*) mMappedIndexData);
				stop_glerror();
			}
		}
		else
		{
			if (gGLManager.mHasMapBufferRange || gGLManager.mHasFlushBufferRange)
			{
				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 (gGLManager.mHasMapBufferRange)
						{
							LLFastTimer t(FTM_IBO_FLUSH_RANGE);
#ifdef GL_ARB_map_buffer_range
							glFlushMappedBufferRange(GL_ELEMENT_ARRAY_BUFFER_ARB, offset, length);
#endif
						}
						else if (gGLManager.mHasFlushBufferRange)
						{
#ifdef GL_APPLE_flush_buffer_range
							glFlushMappedBufferRangeAPPLE(GL_ELEMENT_ARRAY_BUFFER_ARB, offset, length);
#endif
						}
						stop_glerror();
					}

					mMappedIndexRegions.clear();
				}
			}
			stop_glerror();
			glUnmapBufferARB(GL_ELEMENT_ARRAY_BUFFER_ARB);
			stop_glerror();

			mMappedIndexData = NULL;
		}

		mIndexLocked = false;
		sMappedCount--;
	}

	if(updated_all)
	{
		mEmpty = false;
	}
}

//----------------------------------------------------------------------------

template <class T,S32 type> struct VertexBufferStrider
{
	typedef LLStrider<T> strider_t;
	static bool get(LLVertexBuffer& vbo, 
					strider_t& strider, 
					S32 index, S32 count, bool map_range)
	{
		if (type == LLVertexBuffer::TYPE_INDEX)
		{
			volatile U8* ptr = vbo.mapIndexBuffer(index, count, map_range);

			if (ptr == NULL)
			{
				llwarns << "mapIndexBuffer failed!" << llendl;
				return false;
			}

			strider = (T*)ptr;
			strider.setStride(0);
			return true;
		}
		else if (vbo.hasDataType(type))
		{
			S32 stride = LLVertexBuffer::sTypeSize[type];