merge

7 files changed, 370 insertions, 155 deletions

diff --git a/indra/llrender/llimagegl.cpp b/indra/llrender/llimagegl.cpp
index 2f02ccf30b..2579bad0b6 100644
--- a/indra/llrender/llimagegl.cpp
+++ b/indra/llrender/llimagegl.cpp
@@ -1813,7 +1813,7 @@ BOOL LLImageGL::getMask(const LLVector2 &tc)
 		{
 			LL_WARNS_ONCE("render") << "Ugh, non-finite u/v in mask pick" << LL_ENDL;
 			u = v = 0.f;
-			llassert(false);
+			//llassert(false);
 		}
 
 		if (LL_UNLIKELY(u < 0.f || u > 1.f ||
@@ -1821,7 +1821,7 @@ BOOL LLImageGL::getMask(const LLVector2 &tc)
 		{
 			LL_WARNS_ONCE("render") << "Ugh, u/v out of range in image mask pick" << LL_ENDL;
 			u = v = 0.f;
-			llassert(false);
+			//llassert(false);
 		}
 
 		S32 x = llfloor(u * mPickMaskWidth);
diff --git a/indra/llrender/llrender.cpp b/indra/llrender/llrender.cpp
index 2472339ec4..70601663e6 100644
--- a/indra/llrender/llrender.cpp
+++ b/indra/llrender/llrender.cpp
@@ -39,6 +39,7 @@
 #include "llimagegl.h"
 #include "llrendertarget.h"
 #include "lltexture.h"
+#include "llvector4a.h"
 
 LLRender gGL;
 
@@ -53,6 +54,7 @@ U32 LLRender::sUICalls = 0;
 U32 LLRender::sUIVerts = 0;
 
 static const U32 LL_NUM_TEXTURE_LAYERS = 16; 
+static const U32 LL_MAX_UI_STACK_DEPTH = 32;
 
 static GLenum sGLTextureType[] =
 {
@@ -757,14 +759,27 @@ LLRender::LLRender()
     mCount(0),
     mMode(LLRender::TRIANGLES),
     mCurrTextureUnitIndex(0),
-    mMaxAnisotropy(0.f) 
+    mMaxAnisotropy(0.f),
+	mUIStackDepth(0)
 {
 	mBuffer = new LLVertexBuffer(immediate_mask, 0);
 	mBuffer->allocateBuffer(4096, 0, TRUE);
-	mBuffer->getVertexStrider(mVerticesp);
-	mBuffer->getTexCoord0Strider(mTexcoordsp);
-	mBuffer->getColorStrider(mColorsp);
-	
+
+	LLStrider<LLVector3> vert;
+	LLStrider<LLVector2> tc;
+	LLStrider<LLColor4U> color;
+
+	mBuffer->getVertexStrider(vert);
+	mBuffer->getTexCoord0Strider(tc);
+	mBuffer->getColorStrider(color);
+
+	mVerticesp = (LLVector4a*) vert.get();
+	mTexcoordsp = tc.get();
+	mColorsp = color.get();
+
+	mUIOffset = (LLVector4a*) ll_aligned_malloc_16(LL_MAX_UI_STACK_DEPTH*sizeof(LLVector4a));
+	mUIScale = (LLVector4a*) ll_aligned_malloc_16(LL_MAX_UI_STACK_DEPTH*sizeof(LLVector4a));
+		
 	mTexUnits.reserve(LL_NUM_TEXTURE_LAYERS);
 	for (U32 i = 0; i < LL_NUM_TEXTURE_LAYERS; i++)
 	{
@@ -800,6 +815,11 @@ void LLRender::shutdown()
 	mTexUnits.clear();
 	delete mDummyTexUnit;
 	mDummyTexUnit = NULL;
+
+	ll_aligned_free_16(mUIOffset);
+	mUIOffset = NULL;
+	ll_aligned_free_16(mUIScale);
+	mUIScale = NULL;
 }
 
 void LLRender::refreshState(void)
@@ -848,84 +868,83 @@ void LLRender::popMatrix()
 
 void LLRender::translateUI(F32 x, F32 y, F32 z)
 {
-	if (mUIOffset.empty())
+	if (mUIStackDepth == 0)
 	{
 		llerrs << "Need to push a UI translation frame before offsetting" << llendl;
 	}
 
-	mUIOffset.front().mV[0] += x;
-	mUIOffset.front().mV[1] += y;
-	mUIOffset.front().mV[2] += z;
+	LLVector4a trans(x,y,z);
+	mUIOffset[mUIStackDepth-1].add(trans);
 }
 
 void LLRender::scaleUI(F32 x, F32 y, F32 z)
 {
-	if (mUIScale.empty())
+	if (mUIStackDepth == 0)
 	{
 		llerrs << "Need to push a UI transformation frame before scaling." << llendl;
 	}
 
-	mUIScale.front().scaleVec(LLVector3(x,y,z));
+	LLVector4a scale(x,y,z);
+	mUIScale[mUIStackDepth-1].mul(scale);
 }
 
 void LLRender::pushUIMatrix()
 {
-	if (mUIOffset.empty())
+	if (mUIStackDepth == 0)
 	{
-		mUIOffset.push_front(LLVector3(0,0,0));
+		mUIOffset[0].clear();
+		mUIScale[0].splat(1.f);
 	}
-	else
+	else if (mUIStackDepth < LL_MAX_UI_STACK_DEPTH)
 	{
-		mUIOffset.push_front(mUIOffset.front());
-	}
-	
-	if (mUIScale.empty())
-	{
-		mUIScale.push_front(LLVector3(1,1,1));
+		mUIOffset[mUIStackDepth] = mUIOffset[mUIStackDepth-1];
+		mUIScale[mUIStackDepth] = mUIScale[mUIStackDepth-1];
 	}
 	else
 	{
-		mUIScale.push_front(mUIScale.front());
+		llerrs << "Blown UI matrix stack." << llendl;
 	}
+	
+	++mUIStackDepth;
+	
 }
 
 void LLRender::popUIMatrix()
 {
-	if (mUIOffset.empty())
+	if (mUIStackDepth == 0)
 	{
 		llerrs << "UI offset stack blown." << llendl;
 	}
-	mUIOffset.pop_front();
-	mUIScale.pop_front();
+	--mUIStackDepth;
 }
 
 LLVector3 LLRender::getUITranslation()
 {
-	if (mUIOffset.empty())
+	if (mUIStackDepth == 0)
 	{
 		llerrs << "UI offset stack empty." << llendl;
 	}
-	return mUIOffset.front();
+	return LLVector3(mUIOffset[mUIStackDepth-1].getF32());
 }
 
 LLVector3 LLRender::getUIScale()
 {
-	if (mUIScale.empty())
+	if (mUIStackDepth == 0)
 	{
 		llerrs << "UI scale stack empty." << llendl;
 	}
-	return mUIScale.front();
+	return LLVector3(mUIScale[mUIStackDepth-1].getF32());
 }
 
 
 void LLRender::loadUIIdentity()
 {
-	if (mUIOffset.empty())
+	if (mUIStackDepth == 0)
 	{
 		llerrs << "Need to push UI translation frame before clearing offset." << llendl;
 	}
-	mUIOffset.front().setVec(0,0,0);
-	mUIScale.front().setVec(1,1,1);
+	mUIOffset[mUIStackDepth-1].clear();
+	mUIScale[mUIStackDepth-1].splat(1.f);
 }
 
 void LLRender::setColorMask(bool writeColor, bool writeAlpha)
@@ -1154,7 +1173,7 @@ void LLRender::flush()
 		}
 #endif
 				
-		if (!mUIOffset.empty())
+		if (mUIStackDepth > 0)
 		{
 			sUICalls++;
 			sUIVerts += mCount;
@@ -1206,24 +1225,22 @@ void LLRender::vertex3f(const GLfloat& x, const GLfloat& y, const GLfloat& z)
 		return;
 	}
 
-	if (mUIOffset.empty())
+	LLVector3& v = reinterpret_cast<LLVector3&>(mVerticesp[mCount]);
+	v.set(x,y,z);
+	if (mUIStackDepth != 0)
 	{
-		mVerticesp[mCount] = LLVector3(x,y,z);
-	}
-	else
-	{
-		LLVector3 vert = (LLVector3(x,y,z)+mUIOffset.front()).scaledVec(mUIScale.front());
-		mVerticesp[mCount] = vert;
+		v += reinterpret_cast<LLVector3&>(mUIOffset[mUIStackDepth-1]);
+		v.scaleVec(reinterpret_cast<LLVector3&>(mUIScale[mUIStackDepth-1]));
 	}
 
 	mCount++;
 	if (mCount < 4096)
 	{
-		mVerticesp[mCount] = mVerticesp[mCount-1];
 		mColorsp[mCount] = mColorsp[mCount-1];
 		mTexcoordsp[mCount] = mTexcoordsp[mCount-1];
 	}
 }
+
 void LLRender::vertex2i(const GLint& x, const GLint& y)
 {
 	vertex3f((GLfloat) x, (GLfloat) y, 0);	
diff --git a/indra/llrender/llrender.h b/indra/llrender/llrender.h
index 3cda4b5770..11cd95646f 100644
--- a/indra/llrender/llrender.h
+++ b/indra/llrender/llrender.h
@@ -45,6 +45,7 @@
 #include "v4coloru.h"
 #include "llstrider.h"
 #include "llpointer.h"
+#include "llmemory.h"
 #include "llglheaders.h"
 
 class LLVertexBuffer;
@@ -52,6 +53,7 @@ class LLCubeMap;
 class LLImageGL;
 class LLRenderTarget;
 class LLTexture ;
+class LLVector4a;
 
 class LLTexUnit
 {
@@ -360,9 +362,10 @@ private:
 	F32				mCurrAlphaFuncVal;
 
 	LLPointer<LLVertexBuffer>	mBuffer;
-	LLStrider<LLVector3>		mVerticesp;
-	LLStrider<LLVector2>		mTexcoordsp;
-	LLStrider<LLColor4U>		mColorsp;
+	LLVector4a*					mVerticesp;
+	LLVector2*					mTexcoordsp;
+	LLColor4U*					mColorsp;
+
 	std::vector<LLTexUnit*>		mTexUnits;
 	LLTexUnit*			mDummyTexUnit;
 
@@ -372,9 +375,10 @@ private:
 	eBlendFactor mCurrBlendAlphaDFactor;
 	F32				mMaxAnisotropy;
 
-	std::list<LLVector3> mUIOffset;
-	std::list<LLVector3> mUIScale;
+	LLVector4a* mUIOffset;
+	LLVector4a* mUIScale;
 
+	U32 mUIStackDepth;
 };
 
 extern F64 gGLModelView[16];
diff --git a/indra/llrender/llrendersphere.cpp b/indra/llrender/llrendersphere.cpp
index e22b753923..212963f270 100644
--- a/indra/llrender/llrendersphere.cpp
+++ b/indra/llrender/llrendersphere.cpp
@@ -68,45 +68,6 @@ void drawSolidSphere(GLdouble radius, GLint slices, GLint stacks)
 }
 
 
-// lat = 0 is Z-axis
-// lon = 0, lat = 90 at X-axis
-void lat2xyz(LLVector3 * result, F32 lat, F32 lon)
-{
-	// Convert a latitude and longitude to x,y,z on a normal sphere and return it in result
-	F32 r;
-	result->mV[VX] = (F32) (cos(lon * DEG_TO_RAD) * sin(lat * DEG_TO_RAD));
-	result->mV[VY] = (F32) (sin(lon * DEG_TO_RAD) * sin(lat * DEG_TO_RAD));
-	r = (F32) pow(result->mV[VX] * result->mV[VX] + result->mV[VY] * result->mV[VY], 0.5f);
-	if (r == 1.0f) 
-	{
-		result->mV[VZ] = 0.0f;
-	}
-	else
-	{
-		result->mV[VZ] = (F32) pow(1 - r*r, 0.5f);
-		if (lat > 90.01)
-		{
-			result->mV[VZ] *= -1.0;
-		}
-	}
-}
-
-void lat2xyz_rad(LLVector3 * result, F32 lat, F32 lon)
-{
-	// Convert a latitude and longitude to x,y,z on a normal sphere and return it in result
-	F32 r;
-	result->mV[VX] = (F32) (cos(lon) * sin(lat));
-	result->mV[VY] = (F32) (sin(lon) * sin(lat));
-	r = (F32) pow(result->mV[VX] * result->mV[VX] + result->mV[VY] * result->mV[VY], 0.5f);
-	if (r == 1.0f) 
-		result->mV[VZ] = 0.0f;
-	else
-	{
-		result->mV[VZ] = (F32) pow(1 - r*r, 0.5f);
-		if (lat > F_PI_BY_TWO) result->mV[VZ] *= -1.0;
-	}
-}
-
 // A couple thoughts on sphere drawing:
 // 1) You need more slices than stacks, but little less than 2:1
 // 2) At low LOD, setting stacks to an odd number avoids a "band" around the equator, making things look smoother
@@ -181,3 +142,50 @@ void LLRenderSphere::render()
 {
 	glCallList(mDList[0]);
 }
+
+inline LLVector3 polar_to_cart(F32 latitude, F32 longitude)
+{
+	return LLVector3(sin(F_TWO_PI * latitude) * cos(F_TWO_PI * longitude),
+					 sin(F_TWO_PI * latitude) * sin(F_TWO_PI * longitude),
+					 cos(F_TWO_PI * latitude));
+}
+
+
+void LLRenderSphere::renderGGL()
+{
+	S32 const LATITUDE_SLICES = 20;
+	S32 const LONGITUDE_SLICES = 30;
+
+	if (mSpherePoints.empty())
+	{
+		mSpherePoints.resize(LATITUDE_SLICES + 1);
+		for (S32 lat_i = 0; lat_i < LATITUDE_SLICES + 1; lat_i++)
+		{
+			mSpherePoints[lat_i].resize(LONGITUDE_SLICES + 1);
+			for (S32 lon_i = 0; lon_i < LONGITUDE_SLICES + 1; lon_i++)
+			{
+				F32 lat = (F32)lat_i / LATITUDE_SLICES;
+				F32 lon = (F32)lon_i / LONGITUDE_SLICES;
+
+				mSpherePoints[lat_i][lon_i] = polar_to_cart(lat, lon);
+			}
+		}
+	}
+	
+	gGL.begin(LLRender::TRIANGLES);
+
+	for (S32 lat_i = 0; lat_i < LATITUDE_SLICES; lat_i++)
+	{
+		for (S32 lon_i = 0; lon_i < LONGITUDE_SLICES; lon_i++)
+		{
+			gGL.vertex3fv(mSpherePoints[lat_i][lon_i].mV);
+			gGL.vertex3fv(mSpherePoints[lat_i][lon_i+1].mV);
+			gGL.vertex3fv(mSpherePoints[lat_i+1][lon_i].mV);
+
+			gGL.vertex3fv(mSpherePoints[lat_i+1][lon_i].mV);
+			gGL.vertex3fv(mSpherePoints[lat_i][lon_i+1].mV);
+			gGL.vertex3fv(mSpherePoints[lat_i+1][lon_i+1].mV);
+		}
+	}
+	gGL.end();
+}
diff --git a/indra/llrender/llrendersphere.h b/indra/llrender/llrendersphere.h
index 617ee3e12e..ebc71b6146 100644
--- a/indra/llrender/llrendersphere.h
+++ b/indra/llrender/llrendersphere.h
@@ -52,6 +52,10 @@ public:
 	void cleanupGL();
 	void render(F32 pixel_area);		// of a box of size 1.0 at that position
 	void render();						// render at highest LOD
+	void renderGGL();                   // render using LLRender
+
+private:
+	std::vector< std::vector<LLVector3> > mSpherePoints;
 };
 
 extern LLRenderSphere gSphere;
diff --git a/indra/llrender/llvertexbuffer.cpp b/indra/llrender/llvertexbuffer.cpp
index 7fa47cd171..514ca25aa0 100644
--- a/indra/llrender/llvertexbuffer.cpp
+++ b/indra/llrender/llvertexbuffer.cpp
@@ -31,6 +31,7 @@
  */
 
 #include "linden_common.h"
+#include "llmemory.h"
 
 #include <boost/static_assert.hpp>
 
@@ -39,6 +40,7 @@
 #include "llglheaders.h"
 #include "llmemtype.h"
 #include "llrender.h"
+#include "llvector4a.h"
 
 //============================================================================
 
@@ -66,6 +68,27 @@ S32	LLVertexBuffer::sWeight4Loc = -1;
 
 std::vector<U32> LLVertexBuffer::sDeleteList;
 
+#define LL_ALIGNED_VB 1
+
+#if LL_ALIGNED_VB
+
+S32 LLVertexBuffer::sTypeOffsets[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(LLVector4), // TYPE_BINORMAL,
+	sizeof(F32),	   // TYPE_WEIGHT,
+	sizeof(LLVector4), // TYPE_WEIGHT4,
+	sizeof(LLVector4), // TYPE_CLOTHWEIGHT,
+};
+
+#else
+
 S32 LLVertexBuffer::sTypeOffsets[LLVertexBuffer::TYPE_MAX] =
 {
 	sizeof(LLVector3), // TYPE_VERTEX,
@@ -81,6 +104,8 @@ S32 LLVertexBuffer::sTypeOffsets[LLVertexBuffer::TYPE_MAX] =
 	sizeof(LLVector4), // TYPE_CLOTHWEIGHT,
 };
 
+#endif
+
 U32 LLVertexBuffer::sGLMode[LLRender::NUM_MODES] = 
 {
 	GL_TRIANGLES,
@@ -237,10 +262,8 @@ void LLVertexBuffer::setupClientArrays(U32 data_mask)
 	}
 }
 
-void LLVertexBuffer::drawRange(U32 mode, U32 start, U32 end, U32 count, U32 indices_offset) const
+void LLVertexBuffer::validateRange(U32 start, U32 end, U32 count, U32 indices_offset) const
 {
-	llassert(mRequestedNumVerts >= 0);
-
 	if (start >= (U32) mRequestedNumVerts ||
 	    end >= (U32) mRequestedNumVerts)
 	{
@@ -255,6 +278,25 @@ void LLVertexBuffer::drawRange(U32 mode, U32 start, U32 end, U32 count, U32 indi
 		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;
+			}
+		}
+	}
+}
+
+void LLVertexBuffer::drawRange(U32 mode, U32 start, U32 end, U32 count, U32 indices_offset) const
+{
+	validateRange(start, end, count, indices_offset);
+
+	llassert(mRequestedNumVerts >= 0);
+
 	if (mGLIndices != sGLRenderIndices)
 	{
 		llerrs << "Wrong index buffer bound." << llendl;
@@ -273,17 +315,6 @@ void LLVertexBuffer::drawRange(U32 mode, U32 start, U32 end, U32 count, U32 indi
 
 	U16* idx = ((U16*) getIndicesPointer())+indices_offset;
 
-	if (gDebugGL && !useVBOs())
-	{
-		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;
-			}
-		}
-	}
-
 	stop_glerror();
 	glDrawRangeElements(sGLMode[mode], start, end, count, GL_UNSIGNED_SHORT, 
 		idx);
@@ -428,11 +459,42 @@ LLVertexBuffer::LLVertexBuffer(U32 typemask, S32 usage) :
 
 	mTypeMask = typemask;
 	mStride = stride;
+	mAlignedOffset = 0;
+	mAlignedIndexOffset = 0;
+
 	sCount++;
 }
 
+#if LL_ALIGNED_VB
+//static
+S32 LLVertexBuffer::calcStride(const U32& typemask, S32* offsets, S32 num_vertices)
+{
+	S32 offset = 0;
+	for (S32 i=0; i<TYPE_MAX; i++)
+	{
+		U32 mask = 1<<i;
+		if (typemask & mask)
+		{
+			if (offsets)
+			{
+				offsets[i] = offset;
+				offset += LLVertexBuffer::sTypeOffsets[i]*num_vertices;
+				offset = (offset + 0xF) & ~0xF;
+			}
+		}
+	}
+
+	return offset+16;
+}
+
+S32 LLVertexBuffer::getSize() const
+{
+	return mStride;
+}
+
+#else
 //static
-S32 LLVertexBuffer::calcStride(const U32& typemask, S32* offsets)
+S32 LLVertexBuffer::calcStride(const U32& typemask, S32* offsets, S32 num_vertices)
 {
 	S32 stride = 0;
 	for (S32 i=0; i<TYPE_MAX; i++)
@@ -451,6 +513,12 @@ S32 LLVertexBuffer::calcStride(const U32& typemask, S32* offsets)
 	return stride;
 }
 
+S32 LLVertexBuffer::getSize() const
+{ 
+	return mNumVerts*mStride; 
+}
+
+#endif
 // protected, use unref()
 //virtual
 LLVertexBuffer::~LLVertexBuffer()
@@ -560,7 +628,7 @@ void LLVertexBuffer::createGLBuffer()
 	{
 		static int gl_buffer_idx = 0;
 		mGLBuffer = ++gl_buffer_idx;
-		mMappedData = new U8[size];
+		mMappedData = (U8*) ll_aligned_malloc_16(size);
 		memset(mMappedData, 0, size);
 	}
 }
@@ -582,16 +650,20 @@ void LLVertexBuffer::createGLIndices()
 
 	mEmpty = TRUE;
 
+	//pad by 16 bytes for aligned copies
+	size += 16;
+
 	if (useVBOs())
 	{
+		//pad by another 16 bytes for VBO pointer adjustment
+		size += 16;
 		mMappedIndexData = NULL;
 		genIndices();
 		mResized = TRUE;
 	}
 	else
 	{
-		mMappedIndexData = new U8[size];
-		memset(mMappedIndexData, 0, size);
+		mMappedIndexData = (U8*) ll_aligned_malloc_16(size);
 		static int gl_buffer_idx = 0;
 		mGLIndices = ++gl_buffer_idx;
 	}
@@ -612,7 +684,7 @@ void LLVertexBuffer::destroyGLBuffer()
 		}
 		else
 		{
-			delete [] mMappedData;
+			ll_aligned_free_16(mMappedData);
 			mMappedData = NULL;
 			mEmpty = TRUE;
 		}
@@ -639,7 +711,7 @@ void LLVertexBuffer::destroyGLIndices()
 		}
 		else
 		{
-			delete [] mMappedIndexData;
+			ll_aligned_free_16(mMappedIndexData);
 			mMappedIndexData = NULL;
 			mEmpty = TRUE;
 		}
@@ -664,7 +736,7 @@ void LLVertexBuffer::updateNumVerts(S32 nverts)
 	}
 
 	mRequestedNumVerts = nverts;
-	
+
 	if (!mDynamicSize)
 	{
 		mNumVerts = nverts;
@@ -679,6 +751,9 @@ void LLVertexBuffer::updateNumVerts(S32 nverts)
 		}
 		mNumVerts = nverts;
 	}
+#if LL_ALIGNED_VB
+	mStride = calcStride(mTypeMask, mOffsets, mNumVerts);
+#endif
 
 }
 
@@ -754,9 +829,6 @@ void LLVertexBuffer::resizeBuffer(S32 newnverts, S32 newnindices)
 	{
 		sAllocatedBytes -= getSize() + getIndicesSize();
 		
-		S32 oldsize = getSize();
-		S32 old_index_size = getIndicesSize();
-
 		updateNumVerts(newnverts);		
 		updateNumIndices(newnindices);
 		
@@ -773,26 +845,10 @@ void LLVertexBuffer::resizeBuffer(S32 newnverts, S32 newnindices)
 			}
 			else
 			{
-				//delete old buffer, keep GL buffer for now
 				if (!useVBOs())
 				{
-					U8* old = mMappedData;
-					mMappedData = new U8[newsize];
-					if (old)
-					{	
-						memcpy(mMappedData, old, llmin(newsize, oldsize));
-						if (newsize > oldsize)
-						{
-							memset(mMappedData+oldsize, 0, newsize-oldsize);
-						}
-
-						delete [] old;
-					}
-					else
-					{
-						memset(mMappedData, 0, newsize);
-						mEmpty = TRUE;
-					}
+					ll_aligned_free_16(mMappedData);
+					mMappedData = (U8*) ll_aligned_malloc_16(newsize);
 				}
 				mResized = TRUE;
 			}
@@ -812,24 +868,8 @@ void LLVertexBuffer::resizeBuffer(S32 newnverts, S32 newnindices)
 			{
 				if (!useVBOs())
 				{
-					//delete old buffer, keep GL buffer for now
-					U8* old = mMappedIndexData;
-					mMappedIndexData = new U8[new_index_size];
-					
-					if (old)
-					{	
-						memcpy(mMappedIndexData, old, llmin(new_index_size, old_index_size));
-						if (new_index_size > old_index_size)
-						{
-							memset(mMappedIndexData+old_index_size, 0, new_index_size - old_index_size);
-						}
-						delete [] old;
-					}
-					else
-					{
-						memset(mMappedIndexData, 0, new_index_size);
-						mEmpty = TRUE;
-					}
+					ll_aligned_free_16(mMappedIndexData);
+					mMappedIndexData = (U8*) ll_aligned_malloc_16(new_index_size);
 				}
 				mResized = TRUE;
 			}
@@ -861,7 +901,7 @@ BOOL LLVertexBuffer::useVBOs() const
 		return FALSE;
 	}
 #endif
-	return sEnableVBOs;
+	return TRUE;
 }
 
 //----------------------------------------------------------------------------
@@ -886,12 +926,19 @@ U8* LLVertexBuffer::mapBuffer(S32 access)
 			setBuffer(0);
 			mLocked = TRUE;
 			stop_glerror();	
-			mMappedData = (U8*) glMapBufferARB(GL_ARRAY_BUFFER_ARB, GL_WRITE_ONLY_ARB);
+
+			U8* src = (U8*) glMapBufferARB(GL_ARRAY_BUFFER_ARB, GL_WRITE_ONLY_ARB);
+			mMappedData = LL_NEXT_ALIGNED_ADDRESS<U8>(src);
+			mAlignedOffset = mMappedData - src;
+			
 			stop_glerror();
 		}
 		{
 			LLMemType mt_v(LLMemType::MTYPE_VERTEX_MAP_BUFFER_INDICES);
-			mMappedIndexData = (U8*) glMapBufferARB(GL_ELEMENT_ARRAY_BUFFER_ARB, GL_WRITE_ONLY_ARB);
+			U8* src = (U8*) glMapBufferARB(GL_ELEMENT_ARRAY_BUFFER_ARB, GL_WRITE_ONLY_ARB);
+			mMappedIndexData = LL_NEXT_ALIGNED_ADDRESS<U8>(src);
+			mAlignedIndexOffset = mMappedIndexData - src;
+
 			stop_glerror();
 		}
 
@@ -975,6 +1022,45 @@ void LLVertexBuffer::unmapBuffer()
 
 //----------------------------------------------------------------------------
 
+#if LL_ALIGNED_VB
+
+template <class T,S32 type> struct VertexBufferStrider
+{
+	typedef LLStrider<T> strider_t;
+	static bool get(LLVertexBuffer& vbo, 
+					strider_t& strider, 
+					S32 index)
+	{
+		if (vbo.mapBuffer() == NULL)
+		{
+			llwarns << "mapBuffer failed!" << llendl;
+			return FALSE;
+		}
+
+		if (type == LLVertexBuffer::TYPE_INDEX)
+		{
+			S32 stride = sizeof(T);
+			strider = (T*)(vbo.getMappedIndices() + index*stride);
+			strider.setStride(0);
+			return TRUE;
+		}
+		else if (vbo.hasDataType(type))
+		{
+			S32 stride = LLVertexBuffer::sTypeOffsets[type];
+			strider = (T*)(vbo.getMappedData() + vbo.getOffset(type)+index*stride);
+			strider.setStride(stride);
+			return TRUE;
+		}
+		else
+		{
+			llerrs << "VertexBufferStrider could not find valid vertex data." << llendl;
+		}
+		return FALSE;
+	}
+};
+
+#else
+
 template <class T,S32 type> struct VertexBufferStrider
 {
 	typedef LLStrider<T> strider_t;
@@ -1010,6 +1096,7 @@ template <class T,S32 type> struct VertexBufferStrider
 	}
 };
 
+#endif
 
 bool LLVertexBuffer::getVertexStrider(LLStrider<LLVector3>& strider, S32 index)
 {
@@ -1231,7 +1318,7 @@ void LLVertexBuffer::setBuffer(U32 data_mask)
 	{		
 		if (mGLBuffer)
 		{
-			if (sEnableVBOs && sVBOActive)
+			if (sVBOActive)
 			{
 				glBindBufferARB(GL_ARRAY_BUFFER_ARB, 0);
 				sBindCount++;
@@ -1243,7 +1330,7 @@ void LLVertexBuffer::setBuffer(U32 data_mask)
 				setup = TRUE; // ... or a client memory pointer changed
 			}
 		}
-		if (sEnableVBOs && mGLIndices && sIBOActive)
+		if (mGLIndices && sIBOActive)
 		{
 			/*if (sMapped)
 			{
@@ -1272,6 +1359,81 @@ void LLVertexBuffer::setBuffer(U32 data_mask)
 	}
 }
 
+#if LL_ALIGNED_VB
+
+// virtual (default)
+void LLVertexBuffer::setupVertexBuffer(U32 data_mask) const
+{
+	LLMemType mt2(LLMemType::MTYPE_VERTEX_SETUP_VERTEX_BUFFER);
+	stop_glerror();
+	U8* base = useVBOs() ? (U8*) mAlignedOffset : mMappedData;
+
+	if ((data_mask & mTypeMask) != data_mask)
+	{
+		llerrs << "LLVertexBuffer::setupVertexBuffer missing required components for supplied data mask." << llendl;
+	}
+
+	if (data_mask & MAP_NORMAL)
+	{
+		glNormalPointer(GL_FLOAT, LLVertexBuffer::sTypeOffsets[TYPE_NORMAL], (void*)(base + mOffsets[TYPE_NORMAL]));
+	}
+	if (data_mask & MAP_TEXCOORD3)
+	{
+		glClientActiveTextureARB(GL_TEXTURE3_ARB);
+		glTexCoordPointer(2,GL_FLOAT, LLVertexBuffer::sTypeOffsets[TYPE_TEXCOORD3], (void*)(base + mOffsets[TYPE_TEXCOORD3]));
+		glClientActiveTextureARB(GL_TEXTURE0_ARB);
+	}
+	if (data_mask & MAP_TEXCOORD2)
+	{
+		glClientActiveTextureARB(GL_TEXTURE2_ARB);
+		glTexCoordPointer(2,GL_FLOAT, LLVertexBuffer::sTypeOffsets[TYPE_TEXCOORD2], (void*)(base + mOffsets[TYPE_TEXCOORD2]));
+		glClientActiveTextureARB(GL_TEXTURE0_ARB);
+	}
+	if (data_mask & MAP_TEXCOORD1)
+	{
+		glClientActiveTextureARB(GL_TEXTURE1_ARB);
+		glTexCoordPointer(2,GL_FLOAT, LLVertexBuffer::sTypeOffsets[TYPE_TEXCOORD1], (void*)(base + mOffsets[TYPE_TEXCOORD1]));
+		glClientActiveTextureARB(GL_TEXTURE0_ARB);
+	}
+	if (data_mask & MAP_BINORMAL)
+	{
+		glClientActiveTextureARB(GL_TEXTURE2_ARB);
+		glTexCoordPointer(3,GL_FLOAT, LLVertexBuffer::sTypeOffsets[TYPE_BINORMAL], (void*)(base + mOffsets[TYPE_BINORMAL]));
+		glClientActiveTextureARB(GL_TEXTURE0_ARB);
+	}
+	if (data_mask & MAP_TEXCOORD0)
+	{
+		glTexCoordPointer(2,GL_FLOAT, LLVertexBuffer::sTypeOffsets[TYPE_TEXCOORD0], (void*)(base + mOffsets[TYPE_TEXCOORD0]));
+	}
+	if (data_mask & MAP_COLOR)
+	{
+		glColorPointer(4, GL_UNSIGNED_BYTE, LLVertexBuffer::sTypeOffsets[TYPE_COLOR], (void*)(base + mOffsets[TYPE_COLOR]));
+	}
+	
+	if (data_mask & MAP_WEIGHT)
+	{
+		glVertexAttribPointerARB(1, 1, GL_FLOAT, FALSE, LLVertexBuffer::sTypeOffsets[TYPE_WEIGHT], (void*)(base + mOffsets[TYPE_WEIGHT]));
+	}
+
+	if (data_mask & MAP_WEIGHT4 && sWeight4Loc != -1)
+	{
+		glVertexAttribPointerARB(sWeight4Loc, 4, GL_FLOAT, FALSE, LLVertexBuffer::sTypeOffsets[TYPE_WEIGHT4], (void*)(base+mOffsets[TYPE_WEIGHT4]));
+	}
+
+	if (data_mask & MAP_CLOTHWEIGHT)
+	{
+		glVertexAttribPointerARB(4, 4, GL_FLOAT, TRUE,  LLVertexBuffer::sTypeOffsets[TYPE_CLOTHWEIGHT], (void*)(base + mOffsets[TYPE_CLOTHWEIGHT]));
+	}
+	if (data_mask & MAP_VERTEX)
+	{
+		glVertexPointer(3,GL_FLOAT, LLVertexBuffer::sTypeOffsets[TYPE_VERTEX], (void*)(base + 0));
+	}
+
+	llglassertok();
+}
+
+#else
+
 // virtual (default)
 void LLVertexBuffer::setupVertexBuffer(U32 data_mask) const
 {
@@ -1344,6 +1506,8 @@ void LLVertexBuffer::setupVertexBuffer(U32 data_mask) const
 	llglassertok();
 }
 
+#endif
+
 void LLVertexBuffer::markDirty(U32 vert_index, U32 vert_count, U32 indices_index, U32 indices_count)
 {
 	// TODO: use GL_APPLE_flush_buffer_range here
diff --git a/indra/llrender/llvertexbuffer.h b/indra/llrender/llvertexbuffer.h
index d1700aa54a..715309b64a 100644
--- a/indra/llrender/llvertexbuffer.h
+++ b/indra/llrender/llvertexbuffer.h
@@ -79,6 +79,17 @@ protected:
 class LLVertexBuffer : public LLRefCount
 {
 public:
+	LLVertexBuffer(const LLVertexBuffer& rhs)
+	{
+		*this = rhs;
+	}
+
+	const LLVertexBuffer& operator=(const LLVertexBuffer& rhs)
+	{
+		llerrs << "Illegal operation!" << llendl;
+		return *this;
+	}
+
 	static LLVBOPool sStreamVBOPool;
 	static LLVBOPool sDynamicVBOPool;
 	static LLVBOPool sStreamIBOPool;
@@ -98,7 +109,7 @@ public:
 	//if offsets is not NULL, its contents will be filled
 	//with the offset of each vertex component in the buffer, 
 	// indexed by the following enum
-	static S32 calcStride(const U32& typemask, S32* offsets = NULL); 										
+	static S32 calcStride(const U32& typemask, S32* offsets = NULL, S32 num_vertices = 0); 										
 
 	enum {
 		TYPE_VERTEX,
@@ -187,12 +198,12 @@ public:
 	S32 getRequestedVerts() const			{ return mRequestedNumVerts; }
 	S32 getRequestedIndices() const			{ return mRequestedNumIndices; }
 
-	U8* getIndicesPointer() const			{ return useVBOs() ? NULL : mMappedIndexData; }
+	U8* getIndicesPointer() const			{ return useVBOs() ? (U8*) mAlignedIndexOffset : mMappedIndexData; }
 	U8* getVerticesPointer() const			{ return useVBOs() ? NULL : mMappedData; }
 	S32 getStride() const					{ return mStride; }
 	U32 getTypeMask() const					{ return mTypeMask; }
 	BOOL hasDataType(S32 type) const		{ return ((1 << type) & getTypeMask()) ? TRUE : FALSE; }
-	S32 getSize() const						{ return mNumVerts*mStride; }
+	S32 getSize() const;
 	S32 getIndicesSize() const				{ return mNumIndices * sizeof(U16); }
 	U8* getMappedData() const				{ return mMappedData; }
 	U8* getMappedIndices() const			{ return mMappedIndexData; }
@@ -207,12 +218,19 @@ public:
 	void drawArrays(U32 mode, U32 offset, U32 count) const;
 	void drawRange(U32 mode, U32 start, U32 end, U32 count, U32 indices_offset) const;
 
+	//for debugging, validate data in given range is valid
+	void validateRange(U32 start, U32 end, U32 count, U32 offset) const;
+
+	
+
 protected:	
 	S32		mNumVerts;		// Number of vertices allocated
 	S32		mNumIndices;	// Number of indices allocated
 	S32		mRequestedNumVerts;  // Number of vertices requested
 	S32		mRequestedNumIndices;  // Number of indices requested
 
+	ptrdiff_t mAlignedOffset;
+	ptrdiff_t mAlignedIndexOffset;
 	S32		mStride;
 	U32		mTypeMask;
 	S32		mUsage;			// GL usage
@@ -227,7 +245,7 @@ protected:
 	S32		mOffsets[TYPE_MAX];
 	BOOL	mResized;		// if TRUE, client buffer has been resized and GL buffer has not
 	BOOL	mDynamicSize;	// if TRUE, buffer has been resized at least once (and should be padded)
-
+	
 	class DirtyRegion
 	{
 	public: