merge from viewer2, conflicts resolved...

7 files changed, 204 insertions, 57 deletions

diff --git a/indra/llrender/llfontgl.cpp b/indra/llrender/llfontgl.cpp
index f1f86fd638..32ebbd1e3b 100644
--- a/indra/llrender/llfontgl.cpp
+++ b/indra/llrender/llfontgl.cpp
@@ -151,14 +151,16 @@ S32 LLFontGL::render(const LLWString &wstr, S32 begin_offset, F32 x, F32 y, cons
 		}
 	}
 
-	gGL.pushMatrix();
-	glLoadIdentity();
-	gGL.translatef(floorf(sCurOrigin.mX*sScaleX), floorf(sCurOrigin.mY*sScaleY), sCurOrigin.mZ);
+	gGL.pushUIMatrix();
+
+	gGL.loadUIIdentity();
+	
+	gGL.translateUI(floorf(sCurOrigin.mX*sScaleX), floorf(sCurOrigin.mY*sScaleY), sCurOrigin.mZ);
 
 	// this code snaps the text origin to a pixel grid to start with
 	F32 pixel_offset_x = llround((F32)sCurOrigin.mX) - (sCurOrigin.mX);
 	F32 pixel_offset_y = llround((F32)sCurOrigin.mY) - (sCurOrigin.mY);
-	gGL.translatef(-pixel_offset_x, -pixel_offset_y, 0.f);
+	gGL.translateUI(-pixel_offset_x, -pixel_offset_y, 0.f);
 
 	LLFastTimer t(FTM_RENDER_FONTS);
 
@@ -268,12 +270,8 @@ S32 LLFontGL::render(const LLWString &wstr, S32 begin_offset, F32 x, F32 y, cons
 		}
 		// Per-glyph bitmap texture.
 		LLImageGL *image_gl = mFontFreetype->getFontBitmapCache()->getImageGL(fgi->mBitmapNum);
-		if (last_bound_texture != image_gl)
-		{
-			gGL.getTexUnit(0)->bind(image_gl);
-			last_bound_texture = image_gl;
-		}
-
+		gGL.getTexUnit(0)->bind(image_gl);
+	
 		if ((start_x + scaled_max_pixels) < (cur_x + fgi->mXBearing + fgi->mWidth))
 		{
 			// Not enough room for this character.
@@ -338,10 +336,7 @@ S32 LLFontGL::render(const LLWString &wstr, S32 begin_offset, F32 x, F32 y, cons
 		
 		// recursively render ellipses at end of string
 		// we've already reserved enough room
-		gGL.pushMatrix();
-		//glLoadIdentity();
-		//gGL.translatef(sCurOrigin.mX, sCurOrigin.mY, 0.0f);
-		//glScalef(sScaleX, sScaleY, 1.f);
+		gGL.pushUIMatrix();
 		renderUTF8(std::string("..."), 
 				0,
 				cur_x / sScaleX, (F32)y,
@@ -352,10 +347,10 @@ S32 LLFontGL::render(const LLWString &wstr, S32 begin_offset, F32 x, F32 y, cons
 				S32_MAX, max_pixels,
 				right_x,
 				FALSE); 
-		gGL.popMatrix();
+		gGL.popUIMatrix();
 	}
 
-	gGL.popMatrix();
+	gGL.popUIMatrix();
 
 	return chars_drawn;
 }
diff --git a/indra/llrender/llimagegl.cpp b/indra/llrender/llimagegl.cpp
index 36ac3ff119..4251392546 100644
--- a/indra/llrender/llimagegl.cpp
+++ b/indra/llrender/llimagegl.cpp
@@ -1639,7 +1639,7 @@ void LLImageGL::calcAlphaChannelOffsetAndStride()
 	}
 }
 
-void LLImageGL::analyzeAlpha(const void* data_in, S32 w, S32 h)
+void LLImageGL::analyzeAlpha(const void* data_in, U32 w, U32 h)
 {
 	if(!mNeedsAlphaAndPickMask)
 	{
@@ -1647,24 +1647,64 @@ void LLImageGL::analyzeAlpha(const void* data_in, S32 w, S32 h)
 	}
 
 	U32 length = w * h;
-	const GLubyte* current = ((const GLubyte*) data_in) + mAlphaOffset ;
 	
-	S32 sample[16];
-	memset(sample, 0, sizeof(S32)*16);
-
-	for (U32 i = 0; i < length; i++)
+	U32 sample[16];
+	memset(sample, 0, sizeof(U32)*16);
+
+	// generate histogram of quantized alpha.
+	// also add-in the histogram of a 2x2 box-sampled version.  The idea is
+	// this will mid-skew the data (and thus increase the chances of not
+	// being used as a mask) from high-frequency alpha maps which
+	// suffer the worst from aliasing when used as alpha masks.
+	if (w >= 2 && h >= 2)
+	{
+		llassert(w%2 == 0);
+		llassert(h%2 == 0);
+		const GLubyte* rowstart = ((const GLubyte*) data_in) + mAlphaOffset;
+		for (U32 y = 0; y < h; y+=2)
+		{
+			const GLubyte* current = rowstart;
+			for (U32 x = 0; x < w; x+=2)
+			{
+				U32 s1 = current[0];
+				U32 s2 = current[w * mAlphaStride];
+				current += mAlphaStride;
+				U32 s3 = current[0];
+				U32 s4 = current[w * mAlphaStride];
+				current += mAlphaStride;
+
+				++sample[s1/16];
+				++sample[s2/16];
+				++sample[s3/16];
+				++sample[s4/16];
+
+				sample[(s1+s2+s3+s4)/(16 * 4)] += 4;
+			}
+			
+			rowstart += 2 * w * mAlphaStride;
+		}
+		length += length;
+	}
+	else
 	{
-		++sample[*current/16];
-		current += mAlphaStride ;
+		const GLubyte* current = ((const GLubyte*) data_in) + mAlphaOffset;
+		for (U32 i = 0; i < length; i++)
+		{
+			++sample[*current/16];
+			current += mAlphaStride;
+		}
 	}
+	
+	// if more than 1/16th of alpha samples are mid-range, this
+	// shouldn't be treated as a 1-bit mask
 
-	U32 total = 0;
+	U32 midrangetotal = 0;
 	for (U32 i = 4; i < 11; i++)
 	{
-		total += sample[i];
+		midrangetotal += sample[i];
 	}
 
-	if (total > length/16)
+	if (midrangetotal > length/16)
 	{
 		mIsMask = FALSE;
 	}
diff --git a/indra/llrender/llimagegl.h b/indra/llrender/llimagegl.h
index f0870c3fc4..1b303307f6 100644
--- a/indra/llrender/llimagegl.h
+++ b/indra/llrender/llimagegl.h
@@ -91,7 +91,7 @@ public:
 protected:
 	virtual ~LLImageGL();
 
-	void analyzeAlpha(const void* data_in, S32 w, S32 h);
+	void analyzeAlpha(const void* data_in, U32 w, U32 h);
 	void calcAlphaChannelOffsetAndStride();
 
 public:
diff --git a/indra/llrender/llrender.cpp b/indra/llrender/llrender.cpp
index 595b8577ff..57e7dbe77d 100644
--- a/indra/llrender/llrender.cpp
+++ b/indra/llrender/llrender.cpp
@@ -49,6 +49,9 @@ F64 gGLLastProjection[16];
 F64 gGLProjection[16];
 S32	gGLViewport[4];
 
+U32 LLRender::sUICalls = 0;
+U32 LLRender::sUIVerts = 0;
+
 static const U32 LL_NUM_TEXTURE_LAYERS = 16; 
 
 static GLenum sGLTextureType[] =
@@ -90,7 +93,9 @@ static GLenum sGLBlendFactor[] =
 	GL_DST_ALPHA,
 	GL_SRC_ALPHA,
 	GL_ONE_MINUS_DST_ALPHA,
-	GL_ONE_MINUS_SRC_ALPHA
+	GL_ONE_MINUS_SRC_ALPHA,
+
+	GL_ZERO // 'BF_UNDEF'
 };
 
 LLTexUnit::LLTexUnit(S32 index)
@@ -255,10 +260,9 @@ bool LLTexUnit::bind(LLImageGL* texture, bool for_rendering, bool forceBind)
 		return false ;
 	}
 
-	gGL.flush();
-
 	if ((mCurrTexture != texture->getTexName()) || forceBind)
 	{
+		gGL.flush();
 		activate();
 		enable(texture->getTarget());
 		mCurrTexture = texture->getTexName();
@@ -445,6 +449,8 @@ void LLTexUnit::setTextureBlendType(eTextureBlendType type)
 		return;
 	}
 
+	gGL.flush();
+
 	activate();
 	mCurrBlendType = type;
 	S32 scale_amount = 1;
@@ -759,6 +765,9 @@ LLRender::LLRender()
 
 	mCurrAlphaFunc = CF_DEFAULT;
 	mCurrAlphaFuncVal = 0.01f;
+
+	mCurrBlendSFactor = BF_UNDEF;
+	mCurrBlendDFactor = BF_UNDEF;
 }
 
 LLRender::~LLRender()
@@ -821,6 +830,80 @@ void LLRender::popMatrix()
 	glPopMatrix();
 }
 
+void LLRender::translateUI(F32 x, F32 y, F32 z)
+{
+	if (mUIOffset.empty())
+	{
+		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;
+}
+
+void LLRender::scaleUI(F32 x, F32 y, F32 z)
+{
+	if (mUIScale.empty())
+	{
+		llerrs << "Need to push a UI transformation frame before scaling." << llendl;
+	}
+
+	mUIScale.front().scaleVec(LLVector3(x,y,z));
+}
+
+void LLRender::pushUIMatrix()
+{
+	mUIOffset.push_front(mUIOffset.front());
+	if (mUIScale.empty())
+	{
+		mUIScale.push_front(LLVector3(1,1,1));
+	}
+	else
+	{
+		mUIScale.push_front(mUIScale.front());
+	}
+}
+
+void LLRender::popUIMatrix()
+{
+	if (mUIOffset.empty())
+	{
+		llerrs << "UI offset stack blown." << llendl;
+	}
+	mUIOffset.pop_front();
+	mUIScale.pop_front();
+}
+
+LLVector3 LLRender::getUITranslation()
+{
+	if (mUIOffset.empty())
+	{
+		llerrs << "UI offset stack empty." << llendl;
+	}
+	return mUIOffset.front();
+}
+
+LLVector3 LLRender::getUIScale()
+{
+	if (mUIScale.empty())
+	{
+		llerrs << "UI scale stack empty." << llendl;
+	}
+	return mUIScale.front();
+}
+
+
+void LLRender::loadUIIdentity()
+{
+	if (mUIOffset.empty())
+	{
+		llerrs << "Need to push UI translation frame before clearing offset." << llendl;
+	}
+	mUIOffset.front().setVec(0,0,0);
+	mUIScale.front().setVec(1,1,1);
+}
+
 void LLRender::setColorMask(bool writeColor, bool writeAlpha)
 {
 	setColorMask(writeColor, writeColor, writeColor, writeAlpha);
@@ -843,29 +926,28 @@ void LLRender::setColorMask(bool writeColorR, bool writeColorG, bool writeColorB
 
 void LLRender::setSceneBlendType(eBlendType type)
 {
-	flush();
 	switch (type) 
 	{
 		case BT_ALPHA:
-			glBlendFunc(GL_SRC_ALPHA, GL_ONE_MINUS_SRC_ALPHA);
+			blendFunc(BF_SOURCE_ALPHA, BF_ONE_MINUS_SOURCE_ALPHA);
 			break;
 		case BT_ADD:
-			glBlendFunc(GL_ONE, GL_ONE);
+			blendFunc(BF_ONE, BF_ONE);
 			break;
 		case BT_ADD_WITH_ALPHA:
-			glBlendFunc(GL_SRC_ALPHA, GL_ONE);
+			blendFunc(BF_SOURCE_ALPHA, BF_ONE);
 			break;
 		case BT_MULT:
-			glBlendFunc(GL_DST_COLOR, GL_ZERO);
+			blendFunc(BF_DEST_COLOR, BF_ZERO);
 			break;
 		case BT_MULT_ALPHA:
-			glBlendFunc(GL_DST_ALPHA, GL_ZERO);
+			blendFunc(BF_DEST_ALPHA, BF_ZERO);
 			break;
 		case BT_MULT_X2:
-			glBlendFunc(GL_DST_COLOR, GL_SRC_COLOR);
+			blendFunc(BF_DEST_COLOR, BF_SOURCE_COLOR);
 			break;
 		case BT_REPLACE:
-			glBlendFunc(GL_ONE, GL_ZERO);
+			blendFunc(BF_ONE, BF_ZERO);
 			break;
 		default:
 			llerrs << "Unknown Scene Blend Type: " << type << llendl;
@@ -891,8 +973,15 @@ void LLRender::setAlphaRejectSettings(eCompareFunc func, F32 value)
 
 void LLRender::blendFunc(eBlendFactor sfactor, eBlendFactor dfactor)
 {
-	flush();
-	glBlendFunc(sGLBlendFactor[sfactor], sGLBlendFactor[dfactor]);
+	llassert(sfactor < BF_UNDEF);
+	llassert(dfactor < BF_UNDEF);
+	if (mCurrBlendSFactor != sfactor || mCurrBlendDFactor != dfactor)
+	{
+		mCurrBlendSFactor = sfactor;
+		mCurrBlendDFactor = dfactor;
+		flush();
+		glBlendFunc(sGLBlendFactor[sfactor], sGLBlendFactor[dfactor]);
+	}
 }
 
 LLTexUnit* LLRender::getTexUnit(U32 index)
@@ -1012,6 +1101,12 @@ void LLRender::flush()
 		}
 #endif
 				
+		if (!mUIOffset.empty())
+		{
+			sUICalls++;
+			sUIVerts += mCount;
+		}
+
 		mBuffer->setBuffer(immediate_mask);
 		mBuffer->drawArrays(mMode, 0, mCount);
 		
@@ -1031,7 +1126,16 @@ void LLRender::vertex3f(const GLfloat& x, const GLfloat& y, const GLfloat& z)
 		return;
 	}
 
-	mVerticesp[mCount] = LLVector3(x,y,z);
+	if (mUIOffset.empty())
+	{
+		mVerticesp[mCount] = LLVector3(x,y,z);
+	}
+	else
+	{
+		LLVector3 vert = (LLVector3(x,y,z)+mUIOffset.front()).scaledVec(mUIScale.front());
+		mVerticesp[mCount] = vert;
+	}
+
 	mCount++;
 	if (mCount < 4096)
 	{
diff --git a/indra/llrender/llrender.h b/indra/llrender/llrender.h
index 0121a190ee..df24c985bd 100644
--- a/indra/llrender/llrender.h
+++ b/indra/llrender/llrender.h
@@ -270,7 +270,9 @@ public:
 		BF_DEST_ALPHA,
 		BF_SOURCE_ALPHA,
 		BF_ONE_MINUS_DEST_ALPHA,
-		BF_ONE_MINUS_SOURCE_ALPHA
+		BF_ONE_MINUS_SOURCE_ALPHA,
+
+		BF_UNDEF
 	} eBlendFactor;
 
 	LLRender();
@@ -286,6 +288,14 @@ public:
 	void pushMatrix();
 	void popMatrix();
 
+	void translateUI(F32 x, F32 y, F32 z);
+	void scaleUI(F32 x, F32 y, F32 z);
+	void pushUIMatrix();
+	void popUIMatrix();
+	void loadUIIdentity();
+	LLVector3 getUITranslation();
+	LLVector3 getUIScale();
+
 	void flush();
 
 	void begin(const GLuint& mode);
@@ -333,7 +343,9 @@ public:
 	};
 
 public:
-
+	static U32 sUICalls;
+	static U32 sUIVerts;
+	
 private:
 	bool				mDirty;
 	U32				mCount;
@@ -351,6 +363,13 @@ private:
 	LLTexUnit*			mDummyTexUnit;
 
 	F32				mMaxAnisotropy;
+
+	eBlendFactor mCurrBlendSFactor;
+	eBlendFactor mCurrBlendDFactor;
+
+	std::list<LLVector3> mUIOffset;
+	std::list<LLVector3> mUIScale;
+
 };
 
 extern F64 gGLModelView[16];
diff --git a/indra/llrender/llrendertarget.cpp b/indra/llrender/llrendertarget.cpp
index d9520b3bf6..3f2558f1f5 100644
--- a/indra/llrender/llrendertarget.cpp
+++ b/indra/llrender/llrendertarget.cpp
@@ -390,8 +390,6 @@ void LLRenderTarget::flush(BOOL fetch_depth)
 	}
 	else
 	{
-#if !LL_DARWIN
-
 		stop_glerror();
 
 		glBindFramebufferEXT(GL_FRAMEBUFFER_EXT, 0);
@@ -435,7 +433,6 @@ void LLRenderTarget::flush(BOOL fetch_depth)
 				}
 			}
 		}
-#endif
 
 		glBindFramebufferEXT(GL_FRAMEBUFFER_EXT, 0);
 	}
@@ -444,7 +441,6 @@ void LLRenderTarget::flush(BOOL fetch_depth)
 void LLRenderTarget::copyContents(LLRenderTarget& source, S32 srcX0, S32 srcY0, S32 srcX1, S32 srcY1,
 						S32 dstX0, S32 dstY0, S32 dstX1, S32 dstY1, U32 mask, U32 filter)
 {
-#if !LL_DARWIN
 	gGL.flush();
 	if (!source.mFBO || !mFBO)
 	{
@@ -483,14 +479,12 @@ void LLRenderTarget::copyContents(LLRenderTarget& source, S32 srcX0, S32 srcY0,
 			stop_glerror();
 		}
 	}
-#endif
 }
 
 //static
 void LLRenderTarget::copyContentsToFramebuffer(LLRenderTarget& source, S32 srcX0, S32 srcY0, S32 srcX1, S32 srcY1,
 						S32 dstX0, S32 dstY0, S32 dstX1, S32 dstY1, U32 mask, U32 filter)
 {
-#if !LL_DARWIN
 	if (!source.mFBO)
 	{
 		llerrs << "Cannot copy framebuffer contents for non FBO render targets." << llendl;
@@ -507,7 +501,6 @@ void LLRenderTarget::copyContentsToFramebuffer(LLRenderTarget& source, S32 srcX0
 		glBindFramebufferEXT(GL_FRAMEBUFFER_EXT, 0);
 		stop_glerror();
 	}
-#endif
 }
 
 BOOL LLRenderTarget::isComplete() const
@@ -652,7 +645,6 @@ void LLMultisampleBuffer::allocate(U32 resx, U32 resy, U32 color_fmt, BOOL depth
 
 void LLMultisampleBuffer::addColorAttachment(U32 color_fmt)
 {
-#if !LL_DARWIN
 	if (color_fmt == 0)
 	{
 		return;
@@ -693,12 +685,10 @@ void LLMultisampleBuffer::addColorAttachment(U32 color_fmt)
 	}
 
 	mTex.push_back(tex);
-#endif
 }
 
 void LLMultisampleBuffer::allocateDepth()
 {
-#if !LL_DARWIN
 	glGenRenderbuffersEXT(1, (GLuint* ) &mDepth);
 	glBindRenderbufferEXT(GL_RENDERBUFFER_EXT, mDepth);
 	if (mStencil)
@@ -709,6 +699,5 @@ void LLMultisampleBuffer::allocateDepth()
 	{
 		glRenderbufferStorageMultisampleEXT(GL_RENDERBUFFER_EXT, mSamples, GL_DEPTH_COMPONENT16_ARB, mResX, mResY);	
 	}
-#endif
 }
 
diff --git a/indra/llrender/llvertexbuffer.cpp b/indra/llrender/llvertexbuffer.cpp
index bf5eda21eb..116e265914 100644
--- a/indra/llrender/llvertexbuffer.cpp
+++ b/indra/llrender/llvertexbuffer.cpp
@@ -579,7 +579,7 @@ void LLVertexBuffer::destroyGLBuffer()
 	}
 	
 	mGLBuffer = 0;
-	unbind();
+	//unbind();
 }
 
 void LLVertexBuffer::destroyGLIndices()
@@ -606,7 +606,7 @@ void LLVertexBuffer::destroyGLIndices()
 	}
 
 	mGLIndices = 0;
-	unbind();
+	//unbind();
 }
 
 void LLVertexBuffer::updateNumVerts(S32 nverts)