Fix merge issues from genBBoxes optimizations fighting with genBBoxes optimizations and missing statics

6 files changed, 119 insertions, 147 deletions

diff --git a/indra/llmath/llvolume.cpp b/indra/llmath/llvolume.cpp
index ad0ca618a0..a030d889af 100755
--- a/indra/llmath/llvolume.cpp
+++ b/indra/llmath/llvolume.cpp
@@ -136,6 +136,82 @@ BOOL LLLineSegmentBoxIntersect(const F32* start, const F32* end, const F32* cent
 	return true;
 }
 
+// Finds tangent vec based on three vertices with texture coordinates.
+// Fills in dummy values if the triangle has degenerate texture coordinates.
+void calc_tangent_from_triangle(
+	LLVector4a&			normal,
+	LLVector4a&			tangent_out,
+	const LLVector4a& v1,
+	const LLVector2&  w1,
+	const LLVector4a& v2,
+	const LLVector2&  w2,
+	const LLVector4a& v3,
+	const LLVector2&  w3)
+{
+	const F32* v1ptr = v1.getF32ptr();
+	const F32* v2ptr = v2.getF32ptr();
+	const F32* v3ptr = v3.getF32ptr();
+
+	float x1 = v2ptr[0] - v1ptr[0];
+	float x2 = v3ptr[0] - v1ptr[0];
+	float y1 = v2ptr[1] - v1ptr[1];
+	float y2 = v3ptr[1] - v1ptr[1];
+	float z1 = v2ptr[2] - v1ptr[2];
+	float z2 = v3ptr[2] - v1ptr[2];
+
+	float s1 = w2.mV[0] - w1.mV[0];
+	float s2 = w3.mV[0] - w1.mV[0];
+	float t1 = w2.mV[1] - w1.mV[1];
+	float t2 = w3.mV[1] - w1.mV[1];
+
+	F32 rd = s1*t2-s2*t1;
+
+	float r = ((rd*rd) > FLT_EPSILON) ? 1.0F / rd : 1024.f; //some made up large ratio for division by zero
+
+	llassert(llfinite(r));
+	llassert(!llisnan(r));
+
+	LLVector4a sdir(
+		(t2 * x1 - t1 * x2) * r,
+		(t2 * y1 - t1 * y2) * r,
+		(t2 * z1 - t1 * z2) * r);
+
+	LLVector4a tdir(
+		(s1 * x2 - s2 * x1) * r,
+		(s1 * y2 - s2 * y1) * r,
+		(s1 * z2 - s2 * z1) * r);
+
+	LLVector4a	n = normal;
+	LLVector4a	t = sdir;
+
+	LLVector4a ncrosst;
+	ncrosst.setCross3(n,t);
+
+	// Gram-Schmidt orthogonalize
+	n.mul(n.dot3(t).getF32());
+
+	LLVector4a tsubn;
+	tsubn.setSub(t,n);
+
+	if (tsubn.dot3(tsubn).getF32() > F_APPROXIMATELY_ZERO)
+	{
+		tsubn.normalize3fast_checked();
+
+		// Calculate handedness
+		F32 handedness = ncrosst.dot3(tdir).getF32() < 0.f ? -1.f : 1.f;
+
+		tsubn.getF32ptr()[3] = handedness;
+
+		tangent_out = tsubn;
+	}
+	else
+	{
+		// degenerate, make up a value
+		//
+		tangent_out.set(0,0,1,1);
+	}
+
+}
 
 
 // intersect test between triangle vert0, vert1, vert2 and a ray from orig in direction dir.
@@ -5908,10 +5984,10 @@ void LLVolumeFace::cacheOptimize()
 		wght = (LLVector4a*) ll_aligned_malloc_16(sizeof(LLVector4a)*num_verts);
 	}
 
-	LLVector4a* binorm = NULL;
+	LLVector4a* tangent = NULL;
 	if (mTangents)
 	{
-		binorm = (LLVector4a*) ll_aligned_malloc_16(sizeof(LLVector4a)*num_verts);
+		tangent = (LLVector4a*) ll_aligned_malloc_16(sizeof(LLVector4a)*num_verts);
 	}
 
 	//allocate mapping of old indices to new indices
@@ -7277,83 +7353,6 @@ void calc_binormal_from_triangle(LLVector4a& binormal,
 	}
 }
 
-// Finds binormal based on three vertices with texture coordinates.
-// Fills in dummy values if the triangle has degenerate texture coordinates.
-void calc_tangent_from_triangle(
-	LLVector4a&			normal,
-	LLVector4a&			tangent_out,
-	const LLVector4a& v1,
-	const LLVector2&  w1,
-	const LLVector4a& v2,
-	const LLVector2&  w2,
-	const LLVector4a& v3,
-	const LLVector2&  w3)
-{
-	const F32* v1ptr = v1.getF32ptr();
-	const F32* v2ptr = v2.getF32ptr();
-	const F32* v3ptr = v3.getF32ptr();
-
-	float x1 = v2ptr[0] - v1ptr[0];
-	float x2 = v3ptr[0] - v1ptr[0];
-	float y1 = v2ptr[1] - v1ptr[1];
-	float y2 = v3ptr[1] - v1ptr[1];
-	float z1 = v2ptr[2] - v1ptr[2];
-	float z2 = v3ptr[2] - v1ptr[2];
-
-	float s1 = w2.mV[0] - w1.mV[0];
-	float s2 = w3.mV[0] - w1.mV[0];
-	float t1 = w2.mV[1] - w1.mV[1];
-	float t2 = w3.mV[1] - w1.mV[1];
-
-	F32 rd = s1*t2-s2*t1;
-
-	float r = ((rd*rd) > FLT_EPSILON) ? 1.0F / rd : 1024.f; //some made up large ratio for division by zero
-
-	llassert(llfinite(r));
-	llassert(!llisnan(r));
-
-	LLVector4a sdir(
-		(t2 * x1 - t1 * x2) * r,
-		(t2 * y1 - t1 * y2) * r,
-		(t2 * z1 - t1 * z2) * r);
-
-	LLVector4a tdir(
-		(s1 * x2 - s2 * x1) * r,
-		(s1 * y2 - s2 * y1) * r,
-		(s1 * z2 - s2 * z1) * r);
-
-	LLVector4a	n = normal;
-	LLVector4a	t = sdir;
-
-	LLVector4a ncrosst;
-	ncrosst.setCross3(n,t);
-
-	// Gram-Schmidt orthogonalize
-	n.mul(n.dot3(t).getF32());
-
-	LLVector4a tsubn;
-	tsubn.setSub(t,n);
-
-	if (tsubn.dot3(tsubn).getF32() > F_APPROXIMATELY_ZERO)
-	{
-		tsubn.normalize3fast_checked();
-
-		// Calculate handedness
-		F32 handedness = ncrosst.dot3(tdir).getF32() < 0.f ? -1.f : 1.f;
-
-		tsubn.getF32ptr()[3] = handedness;
-
-		tangent_out = tsubn;
-	}
-	else
-	{
-		// degenerate, make up a value
-		//
-		tangent_out.set(0,0,1,1);
-	}
-
-}
-
 //adapted from Lengyel, Eric. “Computing Tangent Space Basis Vectors for an Arbitrary Mesh”. Terathon Software 3D Graphics Library, 2001. http://www.terathon.com/code/tangent.html
 void CalculateTangentArray(U32 vertexCount, const LLVector4a *vertex, const LLVector4a *normal,
         const LLVector2 *texcoord, U32 triangleCount, const U16* index_array, LLVector4a *tangent)
diff --git a/indra/newview/llface.cpp b/indra/newview/llface.cpp
index 700b31f8d3..53e5b55b89 100755
--- a/indra/newview/llface.cpp
+++ b/indra/newview/llface.cpp
@@ -768,7 +768,7 @@ bool less_than_max_mag(const LLVector4a& vec)
 }
 
 BOOL LLFace::genVolumeBBoxes(const LLVolume &volume, S32 f,
-								const LLMatrix4& mat_vert_in, const LLMatrix3& mat_normal_in, BOOL global_volume)
+								const LLMatrix4& mat_vert_in, BOOL global_volume)
 {
 	//get bounding box
 	if (mDrawablep->isState(LLDrawable::REBUILD_VOLUME | LLDrawable::REBUILD_POSITION | LLDrawable::REBUILD_RIGGED))
@@ -777,10 +777,6 @@ BOOL LLFace::genVolumeBBoxes(const LLVolume &volume, S32 f,
 		LLMatrix4a mat_vert;
 		mat_vert.loadu(mat_vert_in);
 
-		LLMatrix4a mat_normal;
-		mat_normal.loadu(mat_normal_in);
-
-		//VECTORIZE THIS
 		LLVector4a min,max;
 	
 		if (f >= volume.getNumVolumeFaces())
@@ -797,95 +793,69 @@ BOOL LLFace::genVolumeBBoxes(const LLVolume &volume, S32 f,
 		llassert(less_than_max_mag(max));
 
 		//min, max are in volume space, convert to drawable render space
-		LLVector4a center;
-		LLVector4a t;
-		t.setAdd(min, max);
-		t.mul(0.5f);
-		mat_vert.affineTransform(t, center);
-		LLVector4a size;
-		size.setSub(max, min);
-		size.mul(0.5f);
 
-		llassert(less_than_max_mag(min));
-		llassert(less_than_max_mag(max));
+		//get 8 corners of bounding box
+		LLVector4Logical mask[6];
 
-		if (!global_volume)
+		for (U32 i = 0; i < 6; ++i)
 		{
-			//VECTORIZE THIS
-			LLVector4a scale;
-			scale.load3(mDrawablep->getVObj()->getScale().mV);
-			size.mul(scale);
+			mask[i].clear();
 		}
 
-		mat_normal.mMatrix[0].normalize3fast();
-		mat_normal.mMatrix[1].normalize3fast();
-		mat_normal.mMatrix[2].normalize3fast();
+		mask[0].setElement<2>(); //001
+		mask[1].setElement<1>(); //010
+		mask[2].setElement<1>(); //011
+		mask[2].setElement<2>();
+		mask[3].setElement<0>(); //100
+		mask[4].setElement<0>(); //101
+		mask[4].setElement<2>();
+		mask[5].setElement<0>(); //110
+		mask[5].setElement<1>();
 		
-		LLVector4a v[4];
+		LLVector4a v[8];
 
-		//get 4 corners of bounding box
-		mat_normal.rotate(size,v[0]);
+		v[6] = min;
+		v[7] = max;
 
-		//VECTORIZE THIS
-		LLVector4a scale;
-		
-		scale.set(-1.f, -1.f, 1.f);
-		scale.mul(size);
-		mat_normal.rotate(scale, v[1]);
+		for (U32 i = 0; i < 6; ++i)
+		{
+			v[i].setSelectWithMask(mask[i], min, max);
+		}
 		
-		scale.set(1.f, -1.f, -1.f);
-		scale.mul(size);
-		mat_normal.rotate(scale, v[2]);
+		LLVector4a tv[8];
 		
-		scale.set(-1.f, 1.f, -1.f);
-		scale.mul(size);
-		mat_normal.rotate(scale, v[3]);
+		//transform bounding box into drawable space
+		for (U32 i = 0; i < 8; ++i)
+		{
+			mat_vert.affineTransform(v[i], tv[i]);
+		}
 
+		//find bounding box
 		LLVector4a& newMin = mExtents[0];
 		LLVector4a& newMax = mExtents[1];
 		
-		newMin = newMax = center;
-		
-		llassert(less_than_max_mag(center));
+		newMin = newMax = tv[0];
 		
-		for (U32 i = 0; i < 4; i++)
+		for (U32 i = 1; i < 8; ++i)
 		{
-			LLVector4a delta;
-			delta.setAbs(v[i]);
-			LLVector4a min;
-			min.setSub(center, delta);
-			LLVector4a max;
-			max.setAdd(center, delta);
-
-			newMin.setMin(newMin,min);
-			newMax.setMax(newMax,max);
-
-			llassert(less_than_max_mag(newMin));
-			llassert(less_than_max_mag(newMax));
+			newMin.setMin(newMin, tv[i]);
+			newMax.setMax(newMax, tv[i]);
 		}
 
 		if (!mDrawablep->isActive())
-		{
+		{	// Shift position for region
 			LLVector4a offset;
 			offset.load3(mDrawablep->getRegion()->getOriginAgent().mV);
 			newMin.add(offset);
 			newMax.add(offset);
-			
-			llassert(less_than_max_mag(newMin));
-			llassert(less_than_max_mag(newMax));
 		}
 
+		LLVector4a t;
 		t.setAdd(newMin, newMax);
 		t.mul(0.5f);
 
-		llassert(less_than_max_mag(t));
-		
-		//VECTORIZE THIS
 		mCenterLocal.set(t.getF32ptr());
 		
-		llassert(less_than_max_mag(newMin));
-		llassert(less_than_max_mag(newMax));
-
 		t.setSub(newMax,newMin);
 		mBoundingSphereRadius = t.getLength3().getF32()*0.5f;
 
diff --git a/indra/newview/llface.h b/indra/newview/llface.h
index 763634a3ab..66b5f13740 100755
--- a/indra/newview/llface.h
+++ b/indra/newview/llface.h
@@ -194,8 +194,7 @@ public:
 
 	void		setSize(S32 numVertices, S32 num_indices = 0, bool align = false);
 	
-	BOOL		genVolumeBBoxes(const LLVolume &volume, S32 f,
-								   const LLMatrix4& mat, BOOL global_volume = FALSE);
+	BOOL		genVolumeBBoxes(const LLVolume &volume, S32 f,const LLMatrix4& mat, BOOL global_volume = FALSE);
 	
 	void		init(LLDrawable* drawablep, LLViewerObject* objp);
 	void		destroy();
diff --git a/indra/newview/llvovolume.cpp b/indra/newview/llvovolume.cpp
index 113bdd2ce3..33b26494de 100755
--- a/indra/newview/llvovolume.cpp
+++ b/indra/newview/llvovolume.cpp
@@ -1458,7 +1458,7 @@ BOOL LLVOVolume::genBBoxes(BOOL force_global)
 			continue;
 		}
 		res &= face->genVolumeBBoxes(*volume, i,
-										mRelativeXform, mRelativeXformInvTrans,
+										mRelativeXform, /*mRelativeXformInvTrans,*/
 										(mVolumeImpl && mVolumeImpl->isVolumeGlobal()) || force_global);
 		
 		if (rebuild)
diff --git a/indra/newview/pipeline.cpp b/indra/newview/pipeline.cpp
index c4286e73eb..8be7b2ab53 100755
--- a/indra/newview/pipeline.cpp
+++ b/indra/newview/pipeline.cpp
@@ -169,6 +169,7 @@ S32 LLPipeline::RenderGlowIterations;
 F32 LLPipeline::RenderGlowWidth;
 F32 LLPipeline::RenderGlowStrength;
 BOOL LLPipeline::RenderDepthOfField;
+BOOL LLPipeline::RenderDepthOfFieldInEditMode;
 F32 LLPipeline::CameraFocusTransitionTime;
 F32 LLPipeline::CameraFNumber;
 F32 LLPipeline::CameraFocalLength;
@@ -379,6 +380,7 @@ BOOL	LLPipeline::sWaterReflections = FALSE;
 BOOL	LLPipeline::sRenderGlow = FALSE;
 BOOL	LLPipeline::sReflectionRender = FALSE;
 BOOL	LLPipeline::sImpostorRender = FALSE;
+BOOL	LLPipeline::sImpostorRenderAlphaDepthPass = FALSE;
 BOOL	LLPipeline::sUnderWaterRender = FALSE;
 BOOL	LLPipeline::sTextureBindTest = FALSE;
 BOOL	LLPipeline::sRenderFrameTest = FALSE;
diff --git a/indra/newview/pipeline.h b/indra/newview/pipeline.h
index f0bebbe20d..fbbcf8f06d 100755
--- a/indra/newview/pipeline.h
+++ b/indra/newview/pipeline.h
@@ -295,7 +295,8 @@ public:
 
 	void unbindDeferredShader(LLGLSLShader& shader);
 	void renderDeferredLighting();
-	
+	void renderDeferredLightingToRT(LLRenderTarget* target);
+
 	void generateWaterReflection(LLCamera& camera);
 	void generateSunShadow(LLCamera& camera);
 	void generateHighlight(LLCamera& camera);
@@ -594,6 +595,7 @@ public:
 	static BOOL				sPickAvatar;
 	static BOOL				sReflectionRender;
 	static BOOL				sImpostorRender;
+	static BOOL				sImpostorRenderAlphaDepthPass;
 	static BOOL				sUnderWaterRender;
 	static BOOL				sRenderGlow;
 	static BOOL				sTextureBindTest;