llvertexbuffer.cpp

/** 
 * @file llvertexbuffer.cpp
 * @brief LLVertexBuffer implementation
 *
 * $LicenseInfo:firstyear=2003&license=viewerlgpl$
 * 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
 * $/LicenseInfo$
 */

#include "linden_common.h"

#include <boost/static_assert.hpp>
#include "llsys.h"
#include "llvertexbuffer.h"
// #include "llrender.h"
#include "llglheaders.h"
#include "llmemtype.h"
#include "llrender.h"
#include "llvector4a.h"
#include "llshadermgr.h"
#include "llglslshader.h"
#include "llmemory.h"

//Next Highest Power Of Two
//helper function, returns first number > v that is a power of 2, or v if v is already a power of 2
U32 nhpo2(U32 v)
{
	U32 r = 1;
	while (r < v) {
		r *= 2;
	}
	return r;
}


//============================================================================

//static
LLVBOPool LLVertexBuffer::sStreamVBOPool;
LLVBOPool LLVertexBuffer::sDynamicVBOPool;
LLVBOPool LLVertexBuffer::sStreamIBOPool;
LLVBOPool LLVertexBuffer::sDynamicIBOPool;
U32 LLVBOPool::sBytesPooled = 0;

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;
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;
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

class LLGLSyncFence : public LLGLFence
{
public:
#ifdef GL_ARB_sync
	GLsync mSync;
#endif
	
	LLGLSyncFence()
	{
#ifdef GL_ARB_sync
		mSync = 0;
#endif
	}

	virtual ~LLGLSyncFence()
	{
#ifdef GL_ARB_sync
		if (mSync)
		{
			glDeleteSync(mSync);
		}
#endif
	}

	void placeFence()
	{
#ifdef GL_ARB_sync
		if (mSync)
		{
			glDeleteSync(mSync);
		}
		mSync = glFenceSync(GL_SYNC_GPU_COMMANDS_COMPLETE, 0);
#endif
	}

	void wait()
	{
#ifdef GL_ARB_sync
		if (mSync)
		{
			while (glClientWaitSync(mSync, 0, FENCE_WAIT_TIME_NANOSECONDS) == GL_TIMEOUT_EXPIRED)
			{ //track the number of times we've waited here
				static S32 waits = 0;
				waits++;
			}
		}
#endif
	}


};


//which power of 2 is i?
//assumes i is a power of 2 > 0
U32 wpo2(U32 i)
{
	llassert(i > 0);
	llassert(nhpo2(i) == i);

	U32 r = 0;

	while (i >>= 1) ++r;

	return r;
}

volatile U8* LLVBOPool::allocate(U32& name, U32 size)
{
	llassert(nhpo2(size) == size);

	U32 i = wpo2(size);

	if (mFreeList.size() <= i)
	{
		mFreeList.resize(i+1);
	}

	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 || mUsage != GL_DYNAMIC_DRAW_ARB)
		{
			ret = (U8*) ll_aligned_malloc_16(size);
		}
		glBindBufferARB(mType, 0);
	}
	else
	{
		name = mFreeList[i].front().mGLName;
		ret = mFreeList[i].front().mClientData;

		sBytesPooled -= size;

		mFreeList[i].pop_front();
	}

	return ret;
}

void LLVBOPool::release(U32 name, volatile U8* buffer, U32 size)
{
	llassert(nhpo2(size) == size);

	U32 i = wpo2(size);

	llassert(mFreeList.size() > i);

	Record rec;
	rec.mGLName = name;
	rec.mClientData = buffer;

	sBytesPooled += size;
	
	if (!LLVertexBuffer::sDisableVBOMapping && mUsage == GL_DYNAMIC_DRAW_ARB)
	{
		glDeleteBuffersARB(1, &rec.mGLName);
	}
	else
	{
		mFreeList[i].push_back(rec);
	}
}

void LLVBOPool::cleanup()
{
	U32 size = 1;

	for (U32 i = 0; i < mFreeList.size(); ++i)
	{
		record_list_t& l = mFreeList[i];

		while (!l.empty())
		{
			Record& r = l.front();

			glDeleteBuffersARB(1, &r.mGLName);

			if (r.mClientData)
			{
				ll_aligned_free_16((void*) r.mClientData);
			}

			l.pop_front();

			LLVertexBuffer::sAllocatedBytes -= size;
			sBytesPooled -= size;
		}

		size *= 2;
	}
}


//NOTE: each component must be AT LEAST 4 bytes in size to avoid a performance penalty on AMD hardware
S32 LLVertexBuffer::sTypeSize[LLVertexBuffer::TYPE_MAX] =
{
	sizeof(LLVector4), // TYPE_VERTEX,
	sizeof(LLVector4), // TYPE_NORMAL,
	sizeof(LLVector2), // TYPE_TEXCOORD0,
	sizeof(LLVector2), // TYPE_TEXCOORD1,
	sizeof(LLVector2), // TYPE_TEXCOORD2,
	sizeof(LLVector2), // TYPE_TEXCOORD3,
	sizeof(LLColor4U), // TYPE_COLOR,
	sizeof(LLColor4U), // TYPE_EMISSIVE, only alpha is used currently
	sizeof(LLVector4), // TYPE_BINORMAL,
	sizeof(F32),	   // TYPE_WEIGHT,
	sizeof(LLVector4), // TYPE_WEIGHT4,
	sizeof(LLVector4), // TYPE_CLOTHWEIGHT,
	sizeof(LLVector4), // TYPE_TEXTURE_INDEX (actually exists as position.w), no extra data, but stride is 16 bytes
};

U32 LLVertexBuffer::sGLMode[LLRender::NUM_MODES] = 
{
	GL_TRIANGLES,
	GL_TRIANGLE_STRIP,
	GL_TRIANGLE_FAN,
	GL_POINTS,
	GL_LINES,
	GL_LINE_STRIP,
	GL_QUADS,
	GL_LINE_LOOP,
};


//static
void LLVertexBuffer::setupClientArrays(U32 data_mask)
{
	if (sLastMask != data_mask)
	{
		BOOL error = FALSE;

		if (LLGLSLShader::sNoFixedFunction)
		{
			for (U32 i = 0; i < TYPE_MAX; ++i)
			{
				S32 loc = i;
										
				U32 mask = 1 << i;

				if (sLastMask & (1 << i))
				{ //was enabled
					if (!(data_mask & mask))
					{ //needs to be disabled
						glDisableVertexAttribArrayARB(loc);
					}
				}
				else 
				{	//was disabled
					if (data_mask & mask)
					{ //needs to be enabled
						glEnableVertexAttribArrayARB(loc);
					}
				}
			}
		}
		else
		{

			GLenum array[] =
			{
				GL_VERTEX_ARRAY,
				GL_NORMAL_ARRAY,
				GL_TEXTURE_COORD_ARRAY,
				GL_COLOR_ARRAY,
			};

			GLenum mask[] = 
			{
				MAP_VERTEX,
				MAP_NORMAL,
				MAP_TEXCOORD0,
				MAP_COLOR
			};



			for (U32 i = 0; i < 4; ++i)
			{
				if (sLastMask & mask[i])
				{ //was enabled
					if (!(data_mask & mask[i]))
					{ //needs to be disabled
						glDisableClientState(array[i]);
					}
					else if (gDebugGL)
					{ //needs to be enabled, make sure it was (DEBUG)
						if (!glIsEnabled(array[i]))
						{
							if (gDebugSession)
							{
								error = TRUE;
								gFailLog << "Bad client state! " << array[i] << " disabled." << std::endl;
							}
							else
							{
								llerrs << "Bad client state! " << array[i] << " disabled." << llendl;
							}
						}
					}
				}
				else 
				{	//was disabled
					if (data_mask & mask[i])
					{ //needs to be enabled
						glEnableClientState(array[i]);
					}
					else if (gDebugGL && glIsEnabled(array[i]))
					{ //needs to be disabled, make sure it was (DEBUG TEMPORARY)
						if (gDebugSession)
						{
							error = TRUE;
							gFailLog << "Bad client state! " << array[i] << " enabled." << std::endl;
						}
						else
						{
							llerrs << "Bad client state! " << array[i] << " enabled." << llendl;
						}
					}
				}
			}
		
			U32 map_tc[] = 
			{
				MAP_TEXCOORD1,
				MAP_TEXCOORD2,
				MAP_TEXCOORD3
			};

			for (U32 i = 0; i < 3; i++)
			{
				if (sLastMask & map_tc[i])
				{
					if (!(data_mask & map_tc[i]))
					{ //disable
						glClientActiveTextureARB(GL_TEXTURE1_ARB+i);
						glDisableClientState(GL_TEXTURE_COORD_ARRAY);
						glClientActiveTextureARB(GL_TEXTURE0_ARB);
					}
				}
				else if (data_mask & map_tc[i])
				{
					glClientActiveTextureARB(GL_TEXTURE1_ARB+i);
					glEnableClientState(GL_TEXTURE_COORD_ARRAY);
					glClientActiveTextureARB(GL_TEXTURE0_ARB);
				}
			}

			if (sLastMask & MAP_BINORMAL)
			{
				if (!(data_mask & MAP_BINORMAL))
				{
					glClientActiveTextureARB(GL_TEXTURE2_ARB);
					glDisableClientState(GL_TEXTURE_COORD_ARRAY);
					glClientActiveTextureARB(GL_TEXTURE0_ARB);
				}
			}
			else if (data_mask & MAP_BINORMAL)
			{
				glClientActiveTextureARB(GL_TEXTURE2_ARB);
				glEnableClientState(GL_TEXTURE_COORD_ARRAY);
				glClientActiveTextureARB(GL_TEXTURE0_ARB);
			}
		}
				
		sLastMask = data_mask;
	}
}

//static
void LLVertexBuffer::drawArrays(U32 mode, const std::vector<LLVector3>& pos, const std::vector<LLVector3>& norm)
{
	llassert(!LLGLSLShader::sNoFixedFunction || LLGLSLShader::sCurBoundShaderPtr != NULL);
	gGL.syncMatrices();

	U32 count = pos.size();
	llassert_always(norm.size() >= pos.size());
	llassert_always(count > 0) ;

	unbind();
	
	setupClientArrays(MAP_VERTEX | MAP_NORMAL);

	LLGLSLShader* shader = LLGLSLShader::sCurBoundShaderPtr;

	if (shader)
	{
		S32 loc = LLVertexBuffer::TYPE_VERTEX;
		if (loc > -1)
		{
			glVertexAttribPointerARB(loc, 3, GL_FLOAT, GL_FALSE, 0, pos[0].mV);
		}
		loc = LLVertexBuffer::TYPE_NORMAL;
		if (loc > -1)
		{
			glVertexAttribPointerARB(loc, 3, GL_FLOAT, GL_FALSE, 0, norm[0].mV);
		}
	}
	else
	{
		glVertexPointer(3, GL_FLOAT, 0, pos[0].mV);
		glNormalPointer(GL_FLOAT, 0, norm[0].mV);
	}

	glDrawArrays(sGLMode[mode], 0, count);
}

//static
void LLVertexBuffer::drawElements(U32 mode, const LLVector4a* pos, const LLVector2* tc, S32 num_indices, const U16* indicesp)
{
	llassert(!LLGLSLShader::sNoFixedFunction || LLGLSLShader::sCurBoundShaderPtr != NULL);

	gGL.syncMatrices();

	U32 mask = LLVertexBuffer::MAP_VERTEX;
	if (tc)
	{
		mask = mask | LLVertexBuffer::MAP_TEXCOORD0;
	}

	unbind();
	
	setupClientArrays(mask);

	if (LLGLSLShader::sNoFixedFunction)
	{
		S32 loc = LLVertexBuffer::TYPE_VERTEX;
		glVertexAttribPointerARB(loc, 3, GL_FLOAT, GL_FALSE, 16, pos);

		if (tc)
		{
			loc = LLVertexBuffer::TYPE_TEXCOORD0;
			glVertexAttribPointerARB(loc, 2, GL_FLOAT, GL_FALSE, 0, tc);
		}
	}
	else
	{
		glTexCoordPointer(2, GL_FLOAT, 0, tc);
		glVertexPointer(3, GL_FLOAT, 16, pos);
	}

	glDrawElements(sGLMode[mode], num_indices, GL_UNSIGNED_SHORT, indicesp);
}

void LLVertexBuffer::validateRange(U32 start, U32 end, U32 count, U32 indices_offset) const
{
	if (start >= (U32) mNumVerts ||
	    end >= (U32) mNumVerts)
	{
		llerrs << "Bad vertex buffer draw range: [" << start << ", " << end << "] vs " << mNumVerts << llendl;
	}

	llassert(mNumIndices >= 0);

	if (indices_offset >= (U32) mNumIndices ||
	    indices_offset + count > (U32) mNumIndices)
	{
		llerrs << "Bad index buffer draw range: [" << indices_offset << ", " << indices_offset+count << "]" << llendl;
	}

	if (gDebugGL && !useVBOs())
	{
		U16* idx = ((U16*) getIndicesPointer())+indices_offset;
		for (U32 i = 0; i < count; ++i)
		{
			if (idx[i] < start || idx[i] > end)
			{
				llerrs << "Index out of range: " << idx[i] << " not in [" << start << ", " << end << "]" << llendl;
			}
		}

		LLGLSLShader* shader = LLGLSLShader::sCurBoundShaderPtr;

		if (shader && shader->mFeatures.mIndexedTextureChannels > 1)
		{
			LLStrider<LLVector4a> v;
			//hack to get non-const reference
			LLVertexBuffer* vb = (LLVertexBuffer*) this;
			vb->getVertexStrider(v);

			for (U32 i = start; i < end; i++)
			{
				S32 idx = (S32) (v[i][3]+0.25f);
				if (idx < 0 || idx >= shader->mFeatures.mIndexedTextureChannels)
				{
					llerrs << "Bad texture index found in vertex data stream." << llendl;
				}
			}
		}
	}
}

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);
	llassert(!LLGLSLShader::sNoFixedFunction || LLGLSLShader::sCurBoundShaderPtr != NULL);

	if (mGLArray)
	{
		if (mGLArray != sGLRenderArray)
		{
			llerrs << "Wrong vertex array bound." << llendl;
		}
	}
	else
	{
		if (mGLIndices != sGLRenderIndices)
		{
			llerrs << "Wrong index buffer bound." << llendl;
		}

		if (mGLBuffer != sGLRenderBuffer)
		{
			llerrs << "Wrong vertex buffer bound." << llendl;
		}
	}

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

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

	if (mode >= LLRender::NUM_MODES)
	{
		llerrs << "Invalid draw mode: " << mode << llendl;
		return;
	}

	U16* idx = ((U16*) getIndicesPointer())+indices_offset;

	stop_glerror();
	glDrawRangeElements(sGLMode[mode], start, end, count, GL_UNSIGNED_SHORT, 
		idx);
	stop_glerror();
	placeFence();
}

void LLVertexBuffer::draw(U32 mode, U32 count, U32 indices_offset) const
{
	llassert(!LLGLSLShader::sNoFixedFunction || LLGLSLShader::sCurBoundShaderPtr != NULL);
	mMappable = FALSE;
	gGL.syncMatrices();

	llassert(mNumIndices >= 0);
	if (indices_offset >= (U32) mNumIndices ||
	    indices_offset + count > (U32) mNumIndices)
	{
		llerrs << "Bad index buffer draw range: [" << indices_offset << ", " << indices_offset+count << "]" << llendl;
	}

	if (mGLArray)
	{
		if (mGLArray != sGLRenderArray)
		{
			llerrs << "Wrong vertex array bound." << llendl;
		}
	}
	else
	{
		if (mGLIndices != sGLRenderIndices)
		{
			llerrs << "Wrong index buffer bound." << llendl;
		}

		if (mGLBuffer != sGLRenderBuffer)
		{
			llerrs << "Wrong vertex buffer bound." << llendl;
		}
	}

	if (mode >= LLRender::NUM_MODES)
	{
		llerrs << "Invalid draw mode: " << mode << llendl;
		return;
	}

	stop_glerror();
	glDrawElements(sGLMode[mode], count, GL_UNSIGNED_SHORT,
		((U16*) getIndicesPointer()) + indices_offset);
	stop_glerror();
	placeFence();
}

void LLVertexBuffer::drawArrays(U32 mode, U32 first, U32 count) const
{
	llassert(!LLGLSLShader::sNoFixedFunction || LLGLSLShader::sCurBoundShaderPtr != NULL);
	mMappable = FALSE;
	gGL.syncMatrices();
	
	llassert(mNumVerts >= 0);
	if (first >= (U32) mNumVerts ||
	    first + count > (U32) mNumVerts)
	{
		llerrs << "Bad vertex buffer draw range: [" << first << ", " << first+count << "]" << llendl;
	}

	if (mGLArray)
	{
		if (mGLArray != sGLRenderArray)
		{
			llerrs << "Wrong vertex array bound." << llendl;
		}
	}
	else
	{
		if (mGLBuffer != sGLRenderBuffer || useVBOs() != sVBOActive)
		{
			llerrs << "Wrong vertex buffer bound." << llendl;
		}
	}

	if (mode >= LLRender::NUM_MODES)
	{
		llerrs << "Invalid draw mode: " << mode << llendl;
		return;
	}

	stop_glerror();
	glDrawArrays(sGLMode[mode], first, count);
	stop_glerror();
	placeFence();
}

//static
void LLVertexBuffer::initClass(bool use_vbo, bool no_vbo_mapping)
{
	sEnableVBOs = use_vbo && gGLManager.mHasVertexBufferObject ;
	sDisableVBOMapping = sEnableVBOs && no_vbo_mapping ;

	if(!sPrivatePoolp)
	{ 
		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 
void LLVertexBuffer::unbind()
{
	if (sGLRenderArray)
	{
#if GL_ARB_vertex_array_object
		glBindVertexArray(0);
#endif
		sGLRenderArray = 0;
		sGLRenderIndices = 0;
		sIBOActive = FALSE;
	}

	if (sVBOActive)
	{
		glBindBufferARB(GL_ARRAY_BUFFER_ARB, 0);
		sVBOActive = FALSE;
	}
	if (sIBOActive)
	{
		glBindBufferARB(GL_ELEMENT_ARRAY_BUFFER_ARB, 0);
		sIBOActive = FALSE;
	}

	sGLRenderBuffer = 0;
	sGLRenderIndices = 0;

	setupClientArrays(0);
}

//static
void LLVertexBuffer::cleanupClass()
{
	LLMemType mt2(LLMemType::MTYPE_VERTEX_CLEANUP_CLASS);
	unbind();
	
	sStreamIBOPool.cleanup();
	sDynamicIBOPool.cleanup();
	sStreamVBOPool.cleanup();
	sDynamicVBOPool.cleanup();

	if(sPrivatePoolp)
	{
		LLPrivateMemoryPoolManager::getInstance()->deletePool(sPrivatePoolp) ;
		sPrivatePoolp = NULL ;
	}
}

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

LLVertexBuffer::LLVertexBuffer(U32 typemask, S32 usage) :
	LLRefCount(),

	mNumVerts(0),
	mNumIndices(0),
	mUsage(usage),
	mGLBuffer(0),
	mGLArray(0),
	mGLIndices(0), 
	mMappedData(NULL),
	mMappedIndexData(NULL), 
	mVertexLocked(FALSE),
	mIndexLocked(FALSE),
	mFinal(FALSE),
	mEmpty(TRUE),
	mFence(NULL)
{
	LLMemType mt2(LLMemType::MTYPE_VERTEX_CONSTRUCTOR);
	mFence = NULL;
	if (!sEnableVBOs)
	{
		mUsage = 0 ; 
	}

	if (mUsage == GL_STREAM_DRAW_ARB && !sUseStreamDraw)
	{
		mUsage = 0;
	}
	
	if (mUsage == GL_DYNAMIC_DRAW_ARB && sPreferStreamDraw)
	{
		mUsage = GL_STREAM_DRAW_ARB;
	}

	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;
	}
		
	if (mUsage == GL_DYNAMIC_DRAW_ARB && !sDisableVBOMapping)
	{
		mMappable = TRUE;
	}
	else
	{
		mMappable = FALSE;
	}

	//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++;
}

//static
S32 LLVertexBuffer::calcOffsets(const U32& typemask, S32* offsets, S32 num_vertices)
{
	S32 offset = 0;
	for (S32 i=0; i<TYPE_TEXTURE_INDEX; i++)
	{
		U32 mask = 1<<i;
		if (typemask & mask)
		{
			if (offsets && LLVertexBuffer::sTypeSize[i])
			{
				offsets[i] = offset;
				offset += LLVertexBuffer::sTypeSize[i]*num_vertices;
				offset = (offset + 0xF) & ~0xF;
			}
		}
	}

	offsets[TYPE_TEXTURE_INDEX] = offsets[TYPE_VERTEX] + 12;
	
	return offset+16;
}

//static 
S32 LLVertexBuffer::calcVertexSize(const U32& typemask)
{
	S32 size = 0;
	for (S32 i = 0; i < TYPE_TEXTURE_INDEX; i++)
	{
		U32 mask = 1<<i;
		if (typemask & mask)
		{
			size += LLVertexBuffer::sTypeSize[i];
		}
	}

	return size;
}

S32 LLVertexBuffer::getSize() const
{
	return mSize;
}

// protected, use unref()
//virtual
LLVertexBuffer::~LLVertexBuffer()
{
	LLMemType mt2(LLMemType::MTYPE_VERTEX_DESTRUCTOR);
	destroyGLBuffer();
	destroyGLIndices();

	if (mGLArray)
	{
#if GL_ARB_vertex_array_object
		glDeleteVertexArrays(1, &mGLArray);
#endif
	}

	sCount--;

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

	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 = nhpo2(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 = nhpo2(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;
	}