Automated merge with http://hg.secondlife.com/viewer-release

11 files changed, 394 insertions, 107 deletions

diff --git a/indra/llrender/llgl.cpp b/indra/llrender/llgl.cpp
index 946e602fee..628a8d6131 100644
--- a/indra/llrender/llgl.cpp
+++ b/indra/llrender/llgl.cpp
@@ -94,9 +94,15 @@ void APIENTRY gl_debug_callback(GLenum source,
 	llwarns << "Severity: " << std::hex << severity << llendl;
 	llwarns << "Message: " << message << llendl;
 	llwarns << "-----------------------" << llendl;
+	if (severity == GL_DEBUG_SEVERITY_HIGH_ARB)
+	{
+		llerrs << "Halting on GL Error" << llendl;
+	}
 }
 #endif
 
+void parse_glsl_version(S32& major, S32& minor);
+
 void ll_init_fail_log(std::string filename)
 {
 	gFailLog.open(filename.c_str());
@@ -295,6 +301,7 @@ PFNGLGETACTIVEUNIFORMARBPROC glGetActiveUniformARB = NULL;
 PFNGLGETUNIFORMFVARBPROC glGetUniformfvARB = NULL;
 PFNGLGETUNIFORMIVARBPROC glGetUniformivARB = NULL;
 PFNGLGETSHADERSOURCEARBPROC glGetShaderSourceARB = NULL;
+PFNGLVERTEXATTRIBIPOINTERPROC glVertexAttribIPointer = NULL;
 
 #if LL_WINDOWS
 PFNWGLCREATECONTEXTATTRIBSARBPROC wglCreateContextAttribsARB = NULL;
@@ -443,7 +450,8 @@ LLGLManager::LLGLManager() :
 	mDriverVersionMinor(0),
 	mDriverVersionRelease(0),
 	mGLVersion(1.0f),
-		
+	mGLSLVersionMajor(0),
+	mGLSLVersionMinor(0),		
 	mVRAM(0),
 	mGLMaxVertexRange(0),
 	mGLMaxIndexRange(0)
@@ -554,6 +562,29 @@ bool LLGLManager::initGL()
 
 	mGLVersion = mDriverVersionMajor + mDriverVersionMinor * .1f;
 
+	if (mGLVersion >= 2.f)
+	{
+		parse_glsl_version(mGLSLVersionMajor, mGLSLVersionMinor);
+
+#if LL_DARWIN
+		//never use GLSL greater than 1.20 on OSX
+		if (mGLSLVersionMajor > 1 || mGLSLVersionMinor >= 30)
+		{
+			mGLSLVersionMajor = 1;
+			mGLSLVersionMinor = 20;
+		}
+#endif
+	}
+
+	if (mGLVersion >= 2.1f && LLImageGL::sCompressTextures)
+	{ //use texture compression
+		glHint(GL_TEXTURE_COMPRESSION_HINT, GL_NICEST);
+	}
+	else
+	{ //GL version is < 3.0, always disable texture compression
+		LLImageGL::sCompressTextures = false;
+	}
+	
 	// Trailing space necessary to keep "nVidia Corpor_ati_on" cards
 	// from being recognized as ATI.
 	if (mGLVendor.substr(0,4) == "ATI ")
@@ -574,11 +605,8 @@ bool LLGLManager::initGL()
 #endif // LL_WINDOWS
 
 #if (LL_WINDOWS || LL_LINUX) && !LL_MESA_HEADLESS
-		// release 7277 is a point at which we verify that ATI OpenGL
-		// drivers get pretty stable with SL, ~Catalyst 8.2,
-		// for both Win32 and Linux.
-		if (mDriverVersionRelease < 7277 &&
-		    mDriverVersionRelease != 0) // 0 == Undetectable driver version - these get to pretend to be new ATI drivers, though that decision may be revisited.
+		// count any pre OpenGL 3.0 implementation as an old driver
+		if (mGLVersion < 3.f) 
 		{
 			mATIOldDriver = TRUE;
 		}
@@ -717,6 +745,11 @@ bool LLGLManager::initGL()
 	}
 #endif
 
+	if (mIsIntel && mGLVersion <= 3.f)
+	{ //never try to use framebuffer objects on older intel drivers (crashy)
+		mHasFramebufferObject = FALSE;
+	}
+
 	if (mHasFramebufferObject)
 	{
 		glGetIntegerv(GL_MAX_SAMPLES, &mMaxSamples);
@@ -1300,6 +1333,7 @@ void LLGLManager::initExtensions()
 		glVertexAttrib4uivARB = (PFNGLVERTEXATTRIB4UIVARBPROC) GLH_EXT_GET_PROC_ADDRESS("glVertexAttrib4uivARB");
 		glVertexAttrib4usvARB = (PFNGLVERTEXATTRIB4USVARBPROC) GLH_EXT_GET_PROC_ADDRESS("glVertexAttrib4usvARB");
 		glVertexAttribPointerARB = (PFNGLVERTEXATTRIBPOINTERARBPROC) GLH_EXT_GET_PROC_ADDRESS("glVertexAttribPointerARB");
+		glVertexAttribIPointer = (PFNGLVERTEXATTRIBIPOINTERPROC) GLH_EXT_GET_PROC_ADDRESS("glVertexAttribIPointer");
 		glEnableVertexAttribArrayARB = (PFNGLENABLEVERTEXATTRIBARRAYARBPROC) GLH_EXT_GET_PROC_ADDRESS("glEnableVertexAttribArrayARB");
 		glDisableVertexAttribArrayARB = (PFNGLDISABLEVERTEXATTRIBARRAYARBPROC) GLH_EXT_GET_PROC_ADDRESS("glDisableVertexAttribArrayARB");
 		glProgramStringARB = (PFNGLPROGRAMSTRINGARBPROC) GLH_EXT_GET_PROC_ADDRESS("glProgramStringARB");
@@ -1878,7 +1912,7 @@ void LLGLState::checkClientArrays(const std::string& msg, U32 data_mask)
 	glClientActiveTextureARB(GL_TEXTURE0_ARB);
 	gGL.getTexUnit(0)->activate();
 
-	if (gGLManager.mHasVertexShader)
+	if (gGLManager.mHasVertexShader && LLGLSLShader::sNoFixedFunction)
 	{	//make sure vertex attribs are all disabled
 		GLint count;
 		glGetIntegerv(GL_MAX_VERTEX_ATTRIBS_ARB, &count);
@@ -2098,6 +2132,55 @@ void parse_gl_version( S32* major, S32* minor, S32* release, std::string* vendor
 	}
 }
 
+
+void parse_glsl_version(S32& major, S32& minor)
+{
+	// GL_SHADING_LANGUAGE_VERSION returns a null-terminated string with the format: 
+	// <major>.<minor>[.<release>] [<vendor specific>]
+
+	const char* version = (const char*) glGetString(GL_SHADING_LANGUAGE_VERSION);
+	major = 0;
+	minor = 0;
+	
+	if( !version )
+	{
+		return;
+	}
+
+	std::string ver_copy( version );
+	S32 len = (S32)strlen( version );	/* Flawfinder: ignore */
+	S32 i = 0;
+	S32 start;
+	// Find the major version
+	start = i;
+	for( ; i < len; i++ )
+	{
+		if( '.' == version[i] )
+		{
+			break;
+		}
+	}
+	std::string major_str = ver_copy.substr(start,i-start);
+	LLStringUtil::convertToS32(major_str, major);
+
+	if( '.' == version[i] )
+	{
+		i++;
+	}
+
+	// Find the minor version
+	start = i;
+	for( ; i < len; i++ )
+	{
+		if( ('.' == version[i]) || isspace(version[i]) )
+		{
+			break;
+		}
+	}
+	std::string minor_str = ver_copy.substr(start,i-start);
+	LLStringUtil::convertToS32(minor_str, minor);
+}
+
 LLGLUserClipPlane::LLGLUserClipPlane(const LLPlane& p, const glh::matrix4f& modelview, const glh::matrix4f& projection, bool apply)
 {
 	mApply = apply;
diff --git a/indra/llrender/llgl.h b/indra/llrender/llgl.h
index 6a147b8e19..5a33c98708 100644
--- a/indra/llrender/llgl.h
+++ b/indra/llrender/llgl.h
@@ -138,6 +138,8 @@ public:
 	S32 mDriverVersionMinor;
 	S32 mDriverVersionRelease;
 	F32 mGLVersion; // e.g = 1.4
+	S32 mGLSLVersionMajor;
+	S32 mGLSLVersionMinor;
 	std::string mDriverVersionVendorString;
 
 	S32 mVRAM; // VRAM in MB
diff --git a/indra/llrender/llglheaders.h b/indra/llrender/llglheaders.h
index 10aad202e1..d61ec707f0 100644
--- a/indra/llrender/llglheaders.h
+++ b/indra/llrender/llglheaders.h
@@ -199,6 +199,7 @@ extern PFNGLVERTEXATTRIB4UBVARBPROC glVertexAttrib4ubvARB;
 extern PFNGLVERTEXATTRIB4UIVARBPROC glVertexAttrib4uivARB;
 extern PFNGLVERTEXATTRIB4USVARBPROC glVertexAttrib4usvARB;
 extern PFNGLVERTEXATTRIBPOINTERARBPROC glVertexAttribPointerARB;
+extern PFNGLVERTEXATTRIBIPOINTERPROC glVertexAttribIPointer;
 extern PFNGLENABLEVERTEXATTRIBARRAYARBPROC glEnableVertexAttribArrayARB;
 extern PFNGLDISABLEVERTEXATTRIBARRAYARBPROC glDisableVertexAttribArrayARB;
 extern PFNGLPROGRAMSTRINGARBPROC glProgramStringARB;
@@ -460,6 +461,7 @@ extern PFNGLVERTEXATTRIB4UBVARBPROC glVertexAttrib4ubvARB;
 extern PFNGLVERTEXATTRIB4UIVARBPROC glVertexAttrib4uivARB;
 extern PFNGLVERTEXATTRIB4USVARBPROC glVertexAttrib4usvARB;
 extern PFNGLVERTEXATTRIBPOINTERARBPROC glVertexAttribPointerARB;
+extern PFNGLVERTEXATTRIBIPOINTERPROC glVertexAttribIPointer;
 extern PFNGLENABLEVERTEXATTRIBARRAYARBPROC glEnableVertexAttribArrayARB;
 extern PFNGLDISABLEVERTEXATTRIBARRAYARBPROC glDisableVertexAttribArrayARB;
 extern PFNGLPROGRAMSTRINGARBPROC glProgramStringARB;
@@ -693,6 +695,7 @@ extern PFNGLVERTEXATTRIB4UBVARBPROC glVertexAttrib4ubvARB;
 extern PFNGLVERTEXATTRIB4UIVARBPROC glVertexAttrib4uivARB;
 extern PFNGLVERTEXATTRIB4USVARBPROC glVertexAttrib4usvARB;
 extern PFNGLVERTEXATTRIBPOINTERARBPROC glVertexAttribPointerARB;
+extern PFNGLVERTEXATTRIBIPOINTERPROC glVertexAttribIPointer;
 extern PFNGLENABLEVERTEXATTRIBARRAYARBPROC glEnableVertexAttribArrayARB;
 extern PFNGLDISABLEVERTEXATTRIBARRAYARBPROC glDisableVertexAttribArrayARB;
 extern PFNGLPROGRAMSTRINGARBPROC glProgramStringARB;
diff --git a/indra/llrender/llglslshader.cpp b/indra/llrender/llglslshader.cpp
index 3773568ad8..4b7e639aed 100644
--- a/indra/llrender/llglslshader.cpp
+++ b/indra/llrender/llglslshader.cpp
@@ -109,7 +109,12 @@ void LLGLSLShader::unload()
 		glGetAttachedObjectsARB(mProgramObject, 1024, &count, obj);
 		for (GLsizei i = 0; i < count; i++)
 		{
-			glDeleteObjectARB(obj[i]);
+#if !LL_DARWIN
+			if (glIsProgramARB(obj[i]))
+#endif
+			{
+				glDeleteObjectARB(obj[i]);
+			}
 		}
 
 		glDeleteObjectARB(mProgramObject);
@@ -161,8 +166,9 @@ BOOL LLGLSLShader::createShader(vector<string> * attributes,
 		return FALSE;
 	}
 
-	if (gGLManager.mGLVersion < 3.1f)
-	{ //attachShaderFeatures may have set the number of indexed texture channels, so set to 1 again
+	if (gGLManager.mGLSLVersionMajor < 2 && gGLManager.mGLSLVersionMinor < 3)
+	{ //indexed texture rendering requires GLSL 1.3 or later
+		//attachShaderFeatures may have set the number of indexed texture channels, so set to 1 again
 		mFeatures.mIndexedTextureChannels = llmin(mFeatures.mIndexedTextureChannels, 1);
 	}
 
diff --git a/indra/llrender/llimagegl.cpp b/indra/llrender/llimagegl.cpp
index 78591ddd38..17131c9d8a 100644
--- a/indra/llrender/llimagegl.cpp
+++ b/indra/llrender/llimagegl.cpp
@@ -57,6 +57,7 @@ BOOL LLImageGL::sGlobalUseAnisotropic	= FALSE;
 F32 LLImageGL::sLastFrameTime			= 0.f;
 BOOL LLImageGL::sAllowReadBackRaw       = FALSE ;
 LLImageGL* LLImageGL::sDefaultGLTexture = NULL ;
+bool LLImageGL::sCompressTextures = false;
 
 std::set<LLImageGL*> LLImageGL::sImageList;
 
@@ -477,6 +478,8 @@ void LLImageGL::init(BOOL usemipmaps)
 	mDiscardLevelInAtlas = -1 ;
 	mTexelsInAtlas = 0 ;
 	mTexelsInGLTexture = 0 ;
+
+	mAllowCompression = true;
 	
 	mTarget = GL_TEXTURE_2D;
 	mBindTarget = LLTexUnit::TT_TEXTURE;
@@ -705,7 +708,7 @@ void LLImageGL::setImage(const U8* data_in, BOOL data_hasmips)
 						stop_glerror();
 					}
 						
-					LLImageGL::setManualImage(mTarget, gl_level, mFormatInternal, w, h, mFormatPrimary, GL_UNSIGNED_BYTE, (GLvoid*)data_in);
+					LLImageGL::setManualImage(mTarget, gl_level, mFormatInternal, w, h, mFormatPrimary, GL_UNSIGNED_BYTE, (GLvoid*)data_in, mAllowCompression);
 					if (gl_level == 0)
 					{
 						analyzeAlpha(data_in, w, h);
@@ -747,7 +750,7 @@ void LLImageGL::setImage(const U8* data_in, BOOL data_hasmips)
 					LLImageGL::setManualImage(mTarget, 0, mFormatInternal,
 								 w, h, 
 								 mFormatPrimary, mFormatType,
-								 data_in);
+								 data_in, mAllowCompression);
 					analyzeAlpha(data_in, w, h);
 					stop_glerror();
 
@@ -805,7 +808,7 @@ void LLImageGL::setImage(const U8* data_in, BOOL data_hasmips)
 							stop_glerror();
 						}
 
-						LLImageGL::setManualImage(mTarget, m, mFormatInternal, w, h, mFormatPrimary, mFormatType, cur_mip_data);
+						LLImageGL::setManualImage(mTarget, m, mFormatInternal, w, h, mFormatPrimary, mFormatType, cur_mip_data, mAllowCompression);
 						if (m == 0)
 						{
 							analyzeAlpha(data_in, w, h);
@@ -863,7 +866,7 @@ void LLImageGL::setImage(const U8* data_in, BOOL data_hasmips)
 			}
 
 			LLImageGL::setManualImage(mTarget, 0, mFormatInternal, w, h,
-						 mFormatPrimary, mFormatType, (GLvoid *)data_in);
+						 mFormatPrimary, mFormatType, (GLvoid *)data_in, mAllowCompression);
 			analyzeAlpha(data_in, w, h);
 			
 			updatePickMask(w, h, data_in);
@@ -1110,7 +1113,7 @@ void LLImageGL::deleteTextures(S32 numTextures, U32 *textures, bool immediate)
 }
 
 // static
-void LLImageGL::setManualImage(U32 target, S32 miplevel, S32 intformat, S32 width, S32 height, U32 pixformat, U32 pixtype, const void *pixels)
+void LLImageGL::setManualImage(U32 target, S32 miplevel, S32 intformat, S32 width, S32 height, U32 pixformat, U32 pixtype, const void *pixels, bool allow_compression)
 {
 	bool use_scratch = false;
 	U32* scratch = NULL;
@@ -1173,6 +1176,36 @@ void LLImageGL::setManualImage(U32 target, S32 miplevel, S32 intformat, S32 widt
 		}
 	}
 
+	if (LLImageGL::sCompressTextures && allow_compression)
+	{
+		switch (intformat)
+		{
+			case GL_RGB: 
+			case GL_RGB8:
+				intformat = GL_COMPRESSED_RGB; 
+				break;
+			case GL_RGBA:
+			case GL_RGBA8:
+				intformat = GL_COMPRESSED_RGBA; 
+				break;
+			case GL_LUMINANCE:
+			case GL_LUMINANCE8:
+				intformat = GL_COMPRESSED_LUMINANCE;
+				break;
+			case GL_LUMINANCE_ALPHA:
+			case GL_LUMINANCE8_ALPHA8:
+				intformat = GL_COMPRESSED_LUMINANCE_ALPHA;
+				break;
+			case GL_ALPHA:
+			case GL_ALPHA8:
+				intformat = GL_COMPRESSED_ALPHA;
+				break;
+			default:
+				llwarns << "Could not compress format: " << std::hex << intformat << llendl;
+				break;
+		}
+	}
+
 	stop_glerror();
 	glTexImage2D(target, miplevel, intformat, width, height, 0, pixformat, pixtype, use_scratch ? scratch : pixels);
 	stop_glerror();
diff --git a/indra/llrender/llimagegl.h b/indra/llrender/llimagegl.h
index 2cfb15b0d9..e23005fe29 100644
--- a/indra/llrender/llimagegl.h
+++ b/indra/llrender/llimagegl.h
@@ -94,12 +94,13 @@ public:
 	
 	void setSize(S32 width, S32 height, S32 ncomponents);
 	void setComponents(S32 ncomponents) { mComponents = (S8)ncomponents ;}
+	void setAllowCompression(bool allow) { mAllowCompression = allow; }
 
 	// These 3 functions currently wrap glGenTextures(), glDeleteTextures(), and glTexImage2D() 
 	// for tracking purposes and will be deprecated in the future
 	static void generateTextures(S32 numTextures, U32 *textures);
 	static void deleteTextures(S32 numTextures, U32 *textures, bool immediate = false);
-	static void setManualImage(U32 target, S32 miplevel, S32 intformat, S32 width, S32 height, U32 pixformat, U32 pixtype, const void *pixels);
+	static void setManualImage(U32 target, S32 miplevel, S32 intformat, S32 width, S32 height, U32 pixformat, U32 pixtype, const void *pixels, bool allow_compression = true);
 
 	BOOL createGLTexture() ;
 	BOOL createGLTexture(S32 discard_level, const LLImageRaw* imageraw, S32 usename = 0, BOOL to_create = TRUE, 
@@ -209,6 +210,8 @@ private:
 	U32      mTexelsInAtlas ;
 	U32      mTexelsInGLTexture;
 
+	bool mAllowCompression;
+
 protected:
 	LLGLenum mTarget;		// Normally GL_TEXTURE2D, sometimes something else (ex. cube maps)
 	LLTexUnit::eTextureType mBindTarget;	// Normally TT_TEXTURE, sometimes something else (ex. cube maps)
@@ -246,7 +249,7 @@ public:
 	static BOOL sGlobalUseAnisotropic;
 	static LLImageGL* sDefaultGLTexture ;	
 	static BOOL sAutomatedTest;
-
+	static bool sCompressTextures;			//use GL texture compression
 #if DEBUG_MISS
 	BOOL mMissed; // Missed on last bind?
 	BOOL getMissed() const { return mMissed; };
diff --git a/indra/llrender/llrendertarget.cpp b/indra/llrender/llrendertarget.cpp
index ef2a7395da..780f1dc484 100644
--- a/indra/llrender/llrendertarget.cpp
+++ b/indra/llrender/llrendertarget.cpp
@@ -143,7 +143,7 @@ bool LLRenderTarget::addColorAttachment(U32 color_fmt)
 
 	{
 		clear_glerror();
-		LLImageGL::setManualImage(LLTexUnit::getInternalType(mUsage), 0, color_fmt, mResX, mResY, GL_RGBA, GL_UNSIGNED_BYTE, NULL);
+		LLImageGL::setManualImage(LLTexUnit::getInternalType(mUsage), 0, color_fmt, mResX, mResY, GL_RGBA, GL_UNSIGNED_BYTE, NULL, false);
 		if (glGetError() != GL_NO_ERROR)
 		{
 			llwarns << "Could not allocate color buffer for render target." << llendl;
@@ -223,7 +223,7 @@ bool LLRenderTarget::allocateDepth()
 		U32 internal_type = LLTexUnit::getInternalType(mUsage);
 		stop_glerror();
 		clear_glerror();
-		LLImageGL::setManualImage(internal_type, 0, GL_DEPTH_COMPONENT24, mResX, mResY, GL_DEPTH_COMPONENT, GL_UNSIGNED_INT, NULL);
+		LLImageGL::setManualImage(internal_type, 0, GL_DEPTH_COMPONENT24, mResX, mResY, GL_DEPTH_COMPONENT, GL_UNSIGNED_INT, NULL, false);
 		gGL.getTexUnit(0)->setTextureFilteringOption(LLTexUnit::TFO_POINT);
 	}
 
diff --git a/indra/llrender/llshadermgr.cpp b/indra/llrender/llshadermgr.cpp
index 908443e8cf..7d384450e6 100644
--- a/indra/llrender/llshadermgr.cpp
+++ b/indra/llrender/llshadermgr.cpp
@@ -575,34 +575,46 @@ GLhandleARB LLShaderMgr::loadShaderFile(const std::string& filename, S32 & shade
 	GLcharARB* text[4096];
 	GLuint count = 0;
 
-	F32 version = gGLManager.mGLVersion;
-
-//hack to never use GLSL > 1.20 on OSX
-#if LL_DARWIN
-	version = llmin(version, 2.9f);
-#endif
-
-	if (version < 2.1f)
-	{
-		text[count++] = strdup("#version 110\n");
-		text[count++] = strdup("#define ATTRIBUTE attribute\n");
-		text[count++] = strdup("#define VARYING varying\n");
-	}
-	else if (version < 3.3f)
+	S32 major_version = gGLManager.mGLSLVersionMajor;
+	S32 minor_version = gGLManager.mGLSLVersionMinor;
+	
+	if (major_version == 1 && minor_version < 30)
 	{
-		//set version to 1.20
-		text[count++] = strdup("#version 120\n");
-		text[count++] = strdup("#define FXAA_GLSL_120 1\n");
-		text[count++] = strdup("#define FXAA_FAST_PIXEL_OFFSET 0\n");
-		text[count++] = strdup("#define ATTRIBUTE attribute\n");
-		text[count++] = strdup("#define VARYING varying\n");
+		if (minor_version < 10)
+		{
+			//should NEVER get here -- if major version is 1 and minor version is less than 10, 
+			// viewer should never attempt to use shaders, continuing will result in undefined behavior
+			llerrs << "Unsupported GLSL Version." << llendl;
+		}
+
+		if (minor_version <= 19)
+		{
+			text[count++] = strdup("#version 110\n");
+			text[count++] = strdup("#define ATTRIBUTE attribute\n");
+			text[count++] = strdup("#define VARYING varying\n");
+			text[count++] = strdup("#define VARYING_FLAT varying\n");
+		}
+		else if (minor_version <= 29)
+		{
+			//set version to 1.20
+			text[count++] = strdup("#version 120\n");
+			text[count++] = strdup("#define FXAA_GLSL_120 1\n");
+			text[count++] = strdup("#define FXAA_FAST_PIXEL_OFFSET 0\n");
+			text[count++] = strdup("#define ATTRIBUTE attribute\n");
+			text[count++] = strdup("#define VARYING varying\n");
+			text[count++] = strdup("#define VARYING_FLAT varying\n");
+		}
 	}
 	else
 	{  
-		if (version < 4.f)
+		if (major_version < 4)
 		{
 			//set version to 1.30
 			text[count++] = strdup("#version 130\n");
+
+			//some implementations of GLSL 1.30 require integer precision be explicitly declared
+			text[count++] = strdup("precision mediump int;\n");
+			text[count++] = strdup("precision highp float;\n");
 		}
 		else
 		{ //set version to 400
@@ -618,16 +630,25 @@ GLhandleARB LLShaderMgr::loadShaderFile(const std::string& filename, S32 & shade
 		{ //"varying" state is "out" in a vertex program, "in" in a fragment program 
 			// ("varying" is deprecated after version 1.20)
 			text[count++] = strdup("#define VARYING out\n");
+			text[count++] = strdup("#define VARYING_FLAT flat out\n");
 		}
 		else
 		{
 			text[count++] = strdup("#define VARYING in\n");
+			text[count++] = strdup("#define VARYING_FLAT flat in\n");
 		}
 
 		//backwards compatibility with legacy texture lookup syntax
+		text[count++] = strdup("#define texture2D texture\n");
 		text[count++] = strdup("#define textureCube texture\n");
 		text[count++] = strdup("#define texture2DLod textureLod\n");
 		text[count++] = strdup("#define	shadow2D(a,b) vec2(texture(a,b))\n");
+
+		if (major_version > 1 || minor_version >= 40)
+		{ //GLSL 1.40 replaces texture2DRect et al with texture
+			text[count++] = strdup("#define texture2DRect texture\n");
+			text[count++] = strdup("#define shadow2DRect(a,b) vec2(texture(a,b))\n");
+		}
 	}
 
 	//copy preprocessor definitions into buffer
@@ -651,22 +672,24 @@ GLhandleARB LLShaderMgr::loadShaderFile(const std::string& filename, S32 & shade
 		.
 		uniform sampler2D texN;
 		
-		VARYING float vary_texture_index;
+		VARYING_FLAT ivec4 vary_texture_index;
+
+		vec4 ret = vec4(1,0,1,1);
 
 		vec4 diffuseLookup(vec2 texcoord)
 		{
-			switch (int(vary_texture_index+0.25))
+			switch (vary_texture_index.r))
 			{
-				case 0: return texture2D(tex0, texcoord);
-				case 1: return texture2D(tex1, texcoord);
-				case 2: return texture2D(tex2, texcoord);
+				case 0: ret = texture2D(tex0, texcoord); break;
+				case 1: ret = texture2D(tex1, texcoord); break;
+				case 2: ret = texture2D(tex2, texcoord); break;
 				.
 				.
 				.
-				case N: return texture2D(texN, texcoord);
+				case N: return texture2D(texN, texcoord); break;
 			}
 
-			return vec4(0,0,0,0);
+			return ret;
 		}
 		*/
 
@@ -679,7 +702,7 @@ GLhandleARB LLShaderMgr::loadShaderFile(const std::string& filename, S32 & shade
 
 		if (texture_index_channels > 1)
 		{
-			text[count++] = strdup("VARYING float vary_texture_index;\n");
+			text[count++] = strdup("VARYING_FLAT ivec4 vary_texture_index;\n");
 		}
 
 		text[count++] = strdup("vec4 diffuseLookup(vec2 texcoord)\n");
@@ -691,45 +714,28 @@ GLhandleARB LLShaderMgr::loadShaderFile(const std::string& filename, S32 & shade
 			text[count++] = strdup("return texture2D(tex0, texcoord);\n");
 			text[count++] = strdup("}\n");
 		}
-		else if (gGLManager.mGLVersion >= 3.f)
-		{ 
-			text[count++] = strdup("\tswitch (int(vary_texture_index+0.25))\n");
+		else if (major_version > 1 || minor_version >= 30)
+		{  //switches are supported in GLSL 1.30 and later
+			text[count++] = strdup("\tvec4 ret = vec4(1,0,1,1);\n");
+			text[count++] = strdup("\tswitch (vary_texture_index.r)\n");
 			text[count++] = strdup("\t{\n");
 		
 			//switch body
 			for (S32 i = 0; i < texture_index_channels; ++i)
 			{
-				std::string case_str = llformat("\t\tcase %d: return texture2D(tex%d, texcoord);\n", i, i);
+				std::string case_str = llformat("\t\tcase %d: ret = texture2D(tex%d, texcoord); break;\n", i, i);
 				text[count++] = strdup(case_str.c_str());
 			}
 
 			text[count++] = strdup("\t}\n");
-			text[count++] = strdup("\treturn vec4(1,0,1,1);\n");
+			text[count++] = strdup("\treturn ret;\n");
 			text[count++] = strdup("}\n");
 		}
 		else
-		{
-			//switches aren't supported, make block that looks like:
-			/*
-				int ti = int(vary_texture_index+0.25);
-				if (ti == 0) return texture2D(tex0, texcoord);
-				if (ti == 1) return texture2D(tex1, texcoord);
-				.
-				.
-				.
-				if (ti == N) return texture2D(texN, texcoord);
-			*/
-				
-			text[count++] = strdup("int ti = int(vary_texture_index+0.25);\n");
-			for (S32 i = 0; i < texture_index_channels; ++i)
-			{
-				std::string if_str = llformat("if (ti == %d) return texture2D(tex%d, texcoord);\n", i, i);
-				text[count++] = strdup(if_str.c_str());
-			}
-
-			text[count++] = strdup("\treturn vec4(1,0,1,1);\n");
-			text[count++] = strdup("}\n");
-		}			
+		{ //should never get here.  Indexed texture rendering requires GLSL 1.30 or later 
+			// (for passing integers between vertex and fragment shaders)
+			llerrs << "Indexed texture rendering requires GLSL 1.30 or later." << llendl;
+		}
 	}
 
 	//copy file into memory
@@ -1070,6 +1076,8 @@ void LLShaderMgr::initAttribsAndUniforms()
 	mReservedUniforms.push_back("magnification");
 	mReservedUniforms.push_back("max_cof");
 	mReservedUniforms.push_back("res_scale");
+	mReservedUniforms.push_back("dof_width");
+	mReservedUniforms.push_back("dof_height");
 
 	mReservedUniforms.push_back("depthMap");
 	mReservedUniforms.push_back("shadowMap0");
diff --git a/indra/llrender/llshadermgr.h b/indra/llrender/llshadermgr.h
index 950e6c9c2f..e28bda6de2 100644
--- a/indra/llrender/llshadermgr.h
+++ b/indra/llrender/llshadermgr.h
@@ -142,6 +142,8 @@ public:
 		DOF_MAGNIFICATION,
 		DOF_MAX_COF,
 		DOF_RES_SCALE,
+		DOF_WIDTH,
+		DOF_HEIGHT,
 
 		DEFERRED_DEPTH,
 		DEFERRED_SHADOW0,
diff --git a/indra/llrender/llvertexbuffer.cpp b/indra/llrender/llvertexbuffer.cpp
index e4a5cd0299..b4899209c4 100644
--- a/indra/llrender/llvertexbuffer.cpp
+++ b/indra/llrender/llvertexbuffer.cpp
@@ -38,6 +38,8 @@
 #include "llglslshader.h"
 #include "llmemory.h"
 
+#define LL_VBO_POOLING 0
+
 //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)
@@ -49,6 +51,35 @@ U32 nhpo2(U32 v)
 	return r;
 }
 
+//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;
+}
+
+
+const U32 LL_VBO_BLOCK_SIZE = 2048;
+
+U32 vbo_block_size(U32 size)
+{ //what block size will fit size?
+	U32 mod = size % LL_VBO_BLOCK_SIZE;
+	return mod == 0 ? size : size + (LL_VBO_BLOCK_SIZE-mod);
+}
+
+U32 vbo_block_index(U32 size)
+{
+	return vbo_block_size(size)/LL_VBO_BLOCK_SIZE;
+}
+
+
 
 //============================================================================
 
@@ -57,7 +88,11 @@ LLVBOPool LLVertexBuffer::sStreamVBOPool(GL_STREAM_DRAW_ARB, GL_ARRAY_BUFFER_ARB
 LLVBOPool LLVertexBuffer::sDynamicVBOPool(GL_DYNAMIC_DRAW_ARB, GL_ARRAY_BUFFER_ARB);
 LLVBOPool LLVertexBuffer::sStreamIBOPool(GL_STREAM_DRAW_ARB, GL_ELEMENT_ARRAY_BUFFER_ARB);
 LLVBOPool LLVertexBuffer::sDynamicIBOPool(GL_DYNAMIC_DRAW_ARB, GL_ELEMENT_ARRAY_BUFFER_ARB);
+
 U32 LLVBOPool::sBytesPooled = 0;
+U32 LLVBOPool::sIndexBytesPooled = 0;
+U32 LLVertexBuffer::sAllocatedIndexBytes = 0;
+U32 LLVertexBuffer::sIndexCount = 0;
 
 LLPrivateMemoryPool* LLVertexBuffer::sPrivatePoolp = NULL;
 U32 LLVertexBuffer::sBindCount = 0;
@@ -74,6 +109,7 @@ U32 LLVertexBuffer::sLastMask = 0;
 bool LLVertexBuffer::sVBOActive = false;
 bool LLVertexBuffer::sIBOActive = false;
 U32 LLVertexBuffer::sAllocatedBytes = 0;
+U32 LLVertexBuffer::sVertexCount = 0;
 bool LLVertexBuffer::sMapped = false;
 bool LLVertexBuffer::sUseStreamDraw = true;
 bool LLVertexBuffer::sUseVAO = false;
@@ -134,39 +170,35 @@ public:
 };
 
 
-//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);
+	llassert(vbo_block_size(size) == size);
+	
+	volatile U8* ret = NULL;
+
+#if LL_VBO_POOLING
 
-	U32 i = wpo2(size);
+	U32 i = vbo_block_index(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);
-		LLVertexBuffer::sAllocatedBytes += size;
+
+		if (mType == GL_ARRAY_BUFFER_ARB)
+		{
+			LLVertexBuffer::sAllocatedBytes += size;
+		}
+		else
+		{
+			LLVertexBuffer::sAllocatedIndexBytes += size;
+		}
 
 		if (LLVertexBuffer::sDisableVBOMapping || mUsage != GL_DYNAMIC_DRAW_ARB)
 		{
@@ -185,19 +217,55 @@ volatile U8* LLVBOPool::allocate(U32& name, U32 size)
 		name = mFreeList[i].front().mGLName;
 		ret = mFreeList[i].front().mClientData;
 
-		sBytesPooled -= size;
+		if (mType == GL_ARRAY_BUFFER_ARB)
+		{
+			sBytesPooled -= size;
+		}
+		else
+		{
+			sIndexBytesPooled -= size;
+		}
 
 		mFreeList[i].pop_front();
 	}
+#else //no pooling
+
+	glGenBuffersARB(1, &name);
+	glBindBufferARB(mType, name);
+
+	if (mType == GL_ARRAY_BUFFER_ARB)
+	{
+		LLVertexBuffer::sAllocatedBytes += size;
+	}
+	else
+	{
+		LLVertexBuffer::sAllocatedIndexBytes += size;
+	}
+
+	if (LLVertexBuffer::sDisableVBOMapping || mUsage != GL_DYNAMIC_DRAW_ARB)
+	{
+		glBufferDataARB(mType, size, 0, mUsage);
+		ret = (U8*) ll_aligned_malloc_16(size);
+	}
+	else
+	{ //always use a true hint of static draw when allocating non-client-backed buffers
+		glBufferDataARB(mType, size, 0, GL_STATIC_DRAW_ARB);
+	}
+
+	glBindBufferARB(mType, 0);
+
+#endif
 
 	return ret;
 }
 
 void LLVBOPool::release(U32 name, volatile U8* buffer, U32 size)
 {
-	llassert(nhpo2(size) == size);
+	llassert(vbo_block_size(size) == size);
 
-	U32 i = wpo2(size);
+#if LL_VBO_POOLING
+
+	U32 i = vbo_block_index(size);
 
 	llassert(mFreeList.size() > i);
 
@@ -211,9 +279,29 @@ void LLVBOPool::release(U32 name, volatile U8* buffer, U32 size)
 	}
 	else
 	{
-		sBytesPooled += size;
+		if (mType == GL_ARRAY_BUFFER_ARB)
+		{
+			sBytesPooled += size;
+		}
+		else
+		{
+			sIndexBytesPooled += size;
+		}
 		mFreeList[i].push_back(rec);
 	}
+#else //no pooling
+	glDeleteBuffersARB(1, &name);
+	ll_aligned_free_16((U8*) buffer);
+
+	if (mType == GL_ARRAY_BUFFER_ARB)
+	{
+		LLVertexBuffer::sAllocatedBytes -= size;
+	}
+	else
+	{
+		LLVertexBuffer::sAllocatedIndexBytes -= size;
+	}
+#endif
 }
 
 void LLVBOPool::cleanup()
@@ -237,8 +325,16 @@ void LLVBOPool::cleanup()
 
 			l.pop_front();
 
-			LLVertexBuffer::sAllocatedBytes -= size;
-			sBytesPooled -= size;
+			if (mType == GL_ARRAY_BUFFER_ARB)
+			{
+				sBytesPooled -= size;
+				LLVertexBuffer::sAllocatedBytes -= size;
+			}
+			else
+			{
+				sIndexBytesPooled -= size;
+				LLVertexBuffer::sAllocatedIndexBytes -= size;
+			}
 		}
 
 		size *= 2;
@@ -284,6 +380,12 @@ void LLVertexBuffer::setupClientArrays(U32 data_mask)
 	{
 		bool error = false;
 
+		if (gGLManager.mGLSLVersionMajor < 2 && gGLManager.mGLSLVersionMinor < 30)
+		{
+			//make sure texture index is disabled
+			data_mask = data_mask & ~MAP_TEXTURE_INDEX;
+		}
+
 		if (LLGLSLShader::sNoFixedFunction)
 		{
 			for (U32 i = 0; i < TYPE_MAX; ++i)
@@ -892,6 +994,9 @@ LLVertexBuffer::~LLVertexBuffer()
 	
 	mFence = NULL;
 
+	sVertexCount -= mNumVerts;
+	sIndexCount -= mNumIndices;
+
 	llassert_always(!mMappedData && !mMappedIndexData);
 };
 
@@ -923,7 +1028,7 @@ void LLVertexBuffer::waitFence() const
 
 void LLVertexBuffer::genBuffer(U32 size)
 {
-	mSize = nhpo2(size);
+	mSize = vbo_block_size(size);
 
 	if (mUsage == GL_STREAM_DRAW_ARB)
 	{
@@ -939,7 +1044,7 @@ void LLVertexBuffer::genBuffer(U32 size)
 
 void LLVertexBuffer::genIndices(U32 size)
 {
-	mIndicesSize = nhpo2(size);
+	mIndicesSize = vbo_block_size(size);
 
 	if (mUsage == GL_STREAM_DRAW_ARB)
 	{
@@ -1115,7 +1220,9 @@ void LLVertexBuffer::updateNumVerts(S32 nverts)
 		createGLBuffer(needed_size);
 	}
 
+	sVertexCount -= mNumVerts;
 	mNumVerts = nverts;
+	sVertexCount += mNumVerts;
 }
 
 void LLVertexBuffer::updateNumIndices(S32 nindices)
@@ -1131,7 +1238,9 @@ void LLVertexBuffer::updateNumIndices(S32 nindices)
 		createGLIndices(needed_size);
 	}
 
+	sIndexCount -= mNumIndices;
 	mNumIndices = nindices;
+	sIndexCount += mNumIndices;
 }
 
 void LLVertexBuffer::allocateBuffer(S32 nverts, S32 nindices, bool create)
@@ -1193,7 +1302,7 @@ void LLVertexBuffer::setupVertexArray()
 		1, //TYPE_WEIGHT,
 		4, //TYPE_WEIGHT4,
 		4, //TYPE_CLOTHWEIGHT,
-		1, //TYPE_TEXTURE_INDEX
+		4, //TYPE_TEXTURE_INDEX
 	};
 
 	U32 attrib_type[] =
@@ -1210,7 +1319,24 @@ void LLVertexBuffer::setupVertexArray()
 		GL_FLOAT, //TYPE_WEIGHT,
 		GL_FLOAT, //TYPE_WEIGHT4,
 		GL_FLOAT, //TYPE_CLOTHWEIGHT,
-		GL_FLOAT, //TYPE_TEXTURE_INDEX
+		GL_UNSIGNED_BYTE, //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[] =
@@ -1238,7 +1364,21 @@ void LLVertexBuffer::setupVertexArray()
 		if (mTypeMask & (1 << i))
 		{
 			glEnableVertexAttribArrayARB(i);
-			glVertexAttribPointerARB(i, attrib_size[i], attrib_type[i], attrib_normalized[i], sTypeSize[i], (void*) mOffsets[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
 		{
@@ -2220,11 +2360,14 @@ void LLVertexBuffer::setupVertexBuffer(U32 data_mask)
 			void* ptr = (void*)(base + mOffsets[TYPE_CLOTHWEIGHT]);
 			glVertexAttribPointerARB(loc, 4, GL_FLOAT, GL_TRUE,  LLVertexBuffer::sTypeSize[TYPE_CLOTHWEIGHT], ptr);
 		}
-		if (data_mask & MAP_TEXTURE_INDEX)
+		if (data_mask & MAP_TEXTURE_INDEX && 
+				(gGLManager.mGLSLVersionMajor >= 2 || gGLManager.mGLSLVersionMinor >= 30)) //indexed texture rendering requires GLSL 1.30 or later
 		{
+#if !LL_DARWIN
 			S32 loc = TYPE_TEXTURE_INDEX;
 			void *ptr = (void*) (base + mOffsets[TYPE_VERTEX] + 12);
-			glVertexAttribPointerARB(loc, 1, GL_FLOAT, GL_FALSE, LLVertexBuffer::sTypeSize[TYPE_VERTEX], ptr);
+			glVertexAttribIPointer(loc, 4, GL_UNSIGNED_BYTE, LLVertexBuffer::sTypeSize[TYPE_VERTEX], ptr);
+#endif
 		}
 		if (data_mask & MAP_VERTEX)
 		{
diff --git a/indra/llrender/llvertexbuffer.h b/indra/llrender/llvertexbuffer.h
index d859199663..7477dec3ad 100644
--- a/indra/llrender/llvertexbuffer.h
+++ b/indra/llrender/llvertexbuffer.h
@@ -55,6 +55,7 @@ class LLVBOPool
 {
 public:
 	static U32 sBytesPooled;
+	static U32 sIndexBytesPooled;
 	
 	LLVBOPool(U32 vboUsage, U32 vboType)
 		: mUsage(vboUsage)
@@ -332,6 +333,9 @@ public:
 	static bool sIBOActive;
 	static U32 sLastMask;
 	static U32 sAllocatedBytes;
+	static U32 sAllocatedIndexBytes;
+	static U32 sVertexCount;
+	static U32 sIndexCount;
 	static U32 sBindCount;
 	static U32 sSetCount;
 };