merge updates from Geenz

11 files changed, 122 insertions, 42 deletions

diff --git a/indra/edit-me-to-trigger-new-build.txt b/indra/edit-me-to-trigger-new-build.txt
index e69de29bb2..8d1c8b69c3 100644
--- a/indra/edit-me-to-trigger-new-build.txt
+++ b/indra/edit-me-to-trigger-new-build.txt
@@ -0,0 +1 @@
+ 
diff --git a/indra/llrender/llglslshader.cpp b/indra/llrender/llglslshader.cpp
index 7cbf39096e..39e81c4bfc 100644
--- a/indra/llrender/llglslshader.cpp
+++ b/indra/llrender/llglslshader.cpp
@@ -150,7 +150,7 @@ BOOL LLGLSLShader::createShader(vector<string> * attributes,
 	vector< pair<string,GLenum> >::iterator fileIter = mShaderFiles.begin();
 	for ( ; fileIter != mShaderFiles.end(); fileIter++ )
 	{
-		GLhandleARB shaderhandle = LLShaderMgr::instance()->loadShaderFile((*fileIter).first, mShaderLevel, (*fileIter).second, mFeatures.mIndexedTextureChannels);
+		GLhandleARB shaderhandle = LLShaderMgr::instance()->loadShaderFile((*fileIter).first, mShaderLevel, (*fileIter).second, mDefines, mFeatures.mIndexedTextureChannels);
 		LL_DEBUGS("ShaderLoading") << "SHADER FILE: " << (*fileIter).first << " mShaderLevel=" << mShaderLevel << LL_ENDL;
 		if (shaderhandle > 0)
 		{
@@ -372,7 +372,17 @@ void LLGLSLShader::mapUniform(GLint index, const vector<string> * uniforms)
 			}
 		}
 	}
- }
+}
+
+void LLGLSLShader::addPermutation(std::string name, std::string value)
+{
+	mDefines[name] = value;
+}
+
+void LLGLSLShader::removePermutation(std::string name)
+{
+	mDefines[name].erase();
+}
 
 GLint LLGLSLShader::mapUniformTextureChannel(GLint location, GLenum type)
 {
@@ -471,6 +481,58 @@ void LLGLSLShader::bindNoShader(void)
 	}
 }
 
+S32 LLGLSLShader::bindTexture(const std::string &uniform, LLTexture *texture, LLTexUnit::eTextureType mode)
+{
+	S32 channel = 0;
+	channel = getUniformLocation(uniform);
+	
+	return bindTexture(channel, texture, mode);
+}
+
+S32 LLGLSLShader::bindTexture(S32 uniform, LLTexture *texture, LLTexUnit::eTextureType mode)
+{
+	if (uniform < 0 || uniform >= (S32)mTexture.size())
+	{
+		UNIFORM_ERRS << "Uniform out of range: " << uniform << LL_ENDL;
+		return -1;
+	}
+	
+	uniform = mTexture[uniform];
+	
+	if (uniform > -1)
+	{
+		gGL.getTexUnit(uniform)->bind(texture, mode);
+	}
+	
+	return uniform;
+}
+
+S32 LLGLSLShader::unbindTexture(const std::string &uniform, LLTexUnit::eTextureType mode)
+{
+	S32 channel = 0;
+	channel = getUniformLocation(uniform);
+	
+	return unbindTexture(channel);
+}
+
+S32 LLGLSLShader::unbindTexture(S32 uniform, LLTexUnit::eTextureType mode)
+{
+	if (uniform < 0 || uniform >= (S32)mTexture.size())
+	{
+		UNIFORM_ERRS << "Uniform out of range: " << uniform << LL_ENDL;
+		return -1;
+	}
+	
+	uniform = mTexture[uniform];
+	
+	if (uniform > -1)
+	{
+		gGL.getTexUnit(uniform)->unbind(mode);
+	}
+	
+	return uniform;
+}
+
 S32 LLGLSLShader::enableTexture(S32 uniform, LLTexUnit::eTextureType mode)
 {
 	if (uniform < 0 || uniform >= (S32)mTexture.size())
diff --git a/indra/llrender/llglslshader.h b/indra/llrender/llglslshader.h
index 5c68cb46eb..2af74c20ff 100644
--- a/indra/llrender/llglslshader.h
+++ b/indra/llrender/llglslshader.h
@@ -123,12 +123,22 @@ public:
 	GLint getAttribLocation(U32 attrib);
 	GLint mapUniformTextureChannel(GLint location, GLenum type);
 	
+	void addPermutation(std::string name, std::string value);
+	void removePermutation(std::string name);
+	
 	//enable/disable texture channel for specified uniform
 	//if given texture uniform is active in the shader, 
 	//the corresponding channel will be active upon return
 	//returns channel texture is enabled in from [0-MAX)
 	S32 enableTexture(S32 uniform, LLTexUnit::eTextureType mode = LLTexUnit::TT_TEXTURE);
-	S32 disableTexture(S32 uniform, LLTexUnit::eTextureType mode = LLTexUnit::TT_TEXTURE); 
+	S32 disableTexture(S32 uniform, LLTexUnit::eTextureType mode = LLTexUnit::TT_TEXTURE);
+	
+	// bindTexture returns the texture unit we've bound the texture to.
+	// You can reuse the return value to unbind a texture when required.
+	S32 bindTexture(const std::string& uniform, LLTexture *texture, LLTexUnit::eTextureType mode = LLTexUnit::TT_TEXTURE);
+	S32 bindTexture(S32 uniform, LLTexture *texture, LLTexUnit::eTextureType mode = LLTexUnit::TT_TEXTURE);
+	S32 unbindTexture(const std::string& uniform, LLTexUnit::eTextureType mode = LLTexUnit::TT_TEXTURE);
+	S32 unbindTexture(S32 uniform, LLTexUnit::eTextureType mode = LLTexUnit::TT_TEXTURE);
 	
     BOOL link(BOOL suppress_errors = FALSE);
 	void bind();
@@ -153,6 +163,7 @@ public:
 	LLShaderFeatures mFeatures;
 	std::vector< std::pair< std::string, GLenum > > mShaderFiles;
 	std::string mName;
+	boost::unordered_map<std::string, std::string> mDefines;
 };
 
 //UI shader (declared here so llui_libtest will link properly)
diff --git a/indra/llrender/llshadermgr.cpp b/indra/llrender/llshadermgr.cpp
index b6a9a6b653..3b1ec281c9 100644
--- a/indra/llrender/llshadermgr.cpp
+++ b/indra/llrender/llshadermgr.cpp
@@ -521,7 +521,7 @@ void LLShaderMgr::dumpObjectLog(GLhandleARB ret, BOOL warns)
 	}
  }
 
-GLhandleARB LLShaderMgr::loadShaderFile(const std::string& filename, S32 & shader_level, GLenum type, S32 texture_index_channels)
+GLhandleARB LLShaderMgr::loadShaderFile(const std::string& filename, S32 & shader_level, GLenum type, boost::unordered_map<std::string, std::string> defines, S32 texture_index_channels)
 {
 	GLenum error = GL_NO_ERROR;
 	if (gDebugGL)
@@ -657,6 +657,12 @@ GLhandleARB LLShaderMgr::loadShaderFile(const std::string& filename, S32 & shade
 		std::string define = "#define " + iter->first + " " + iter->second + "\n";
 		text[count++] = (GLcharARB *) strdup(define.c_str());
 	}
+	
+	for (boost::unordered_map<std::string,std::string>::iterator iter = defines.begin(); iter != defines.end(); ++iter)
+	{
+		std::string define = "#define " + iter->first + " " + iter->second + "\n";
+		text[count++] = (GLcharARB *) strdup(define.c_str());
+	}
 
 	if (texture_index_channels > 0 && type == GL_FRAGMENT_SHADER_ARB)
 	{
@@ -854,7 +860,7 @@ GLhandleARB LLShaderMgr::loadShaderFile(const std::string& filename, S32 & shade
 		if (shader_level > 1)
 		{
 			shader_level--;
-			return loadShaderFile(filename,shader_level,type,texture_index_channels);
+			return loadShaderFile(filename,shader_level,type, defines, texture_index_channels);
 		}
 		LL_WARNS("ShaderLoading") << "Failed to load " << filename << LL_ENDL;	
 	}
diff --git a/indra/llrender/llshadermgr.h b/indra/llrender/llshadermgr.h
index 7a16b7c20f..4b93aae735 100644
--- a/indra/llrender/llshadermgr.h
+++ b/indra/llrender/llshadermgr.h
@@ -176,7 +176,7 @@ public:
 	void dumpObjectLog(GLhandleARB ret, BOOL warns = TRUE);
 	BOOL	linkProgramObject(GLhandleARB obj, BOOL suppress_errors = FALSE);
 	BOOL	validateProgramObject(GLhandleARB obj);
-	GLhandleARB loadShaderFile(const std::string& filename, S32 & shader_level, GLenum type, S32 texture_index_channels = -1);
+	GLhandleARB loadShaderFile(const std::string& filename, S32 & shader_level, GLenum type, boost::unordered_map<std::string, std::string> defines, S32 texture_index_channels = -1);
 
 	// Implemented in the application to actually point to the shader directory.
 	virtual std::string getShaderDirPrefix(void) = 0; // Pure Virtual
diff --git a/indra/newview/app_settings/shaders/class1/deferred/multiPointLightF.glsl b/indra/newview/app_settings/shaders/class1/deferred/multiPointLightF.glsl
index 7e79317543..9e194cdcd1 100644
--- a/indra/newview/app_settings/shaders/class1/deferred/multiPointLightF.glsl
+++ b/indra/newview/app_settings/shaders/class1/deferred/multiPointLightF.glsl
@@ -128,8 +128,7 @@ void main()
 				
 				if (sa > 0.0)
 				{
-					sa = 6 * texture2D(lightFunc, vec2(sa, spec.a)).r * min(dist_atten*4.0, 1.0);
-					sa *= noise;
+					sa = texture2D(lightFunc, vec2(sa, spec.a)).r * min(dist_atten*4.0, 1.0);
 					col += da*sa*light_col[i].rgb*spec.rgb;
 				}
 			}
diff --git a/indra/newview/app_settings/shaders/class1/deferred/pointLightF.glsl b/indra/newview/app_settings/shaders/class1/deferred/pointLightF.glsl
index 75757b26c8..6e5ac8317b 100644
--- a/indra/newview/app_settings/shaders/class1/deferred/pointLightF.glsl
+++ b/indra/newview/app_settings/shaders/class1/deferred/pointLightF.glsl
@@ -110,8 +110,7 @@ void main()
 		float sa = dot(normalize(lv-normalize(pos)),norm);
 		if (sa > 0.0)
 		{
-			sa = 6 * texture2D(lightFunc, vec2(sa, spec.a)).r * min(dist_atten*4.0, 1.0);
-			sa *= noise;
+			sa = texture2D(lightFunc, vec2(sa, spec.a)).r * min(dist_atten*4.0, 1.0);
 			col += da*sa*color.rgb*spec.rgb;
 		}
 	}
diff --git a/indra/newview/app_settings/shaders/class1/deferred/softenLightF.glsl b/indra/newview/app_settings/shaders/class1/deferred/softenLightF.glsl
index 89448e2167..14eaafeb68 100644
--- a/indra/newview/app_settings/shaders/class1/deferred/softenLightF.glsl
+++ b/indra/newview/app_settings/shaders/class1/deferred/softenLightF.glsl
@@ -301,7 +301,7 @@ void main()
 			//
 			vec3 refnormpersp = normalize(reflect(pos.xyz, norm.xyz));
 			float sa = dot(refnormpersp, sun_dir.xyz);
-			vec3 dumbshiny = vary_SunlitColor*(6 * texture2D(lightFunc, vec2(sa, spec.a)).r);
+			vec3 dumbshiny = vary_SunlitColor*(texture2D(lightFunc, vec2(sa, spec.a)).r);
 			
 			// add the two types of shiny together
 			vec3 spec_contrib = dumbshiny * spec.rgb;
diff --git a/indra/newview/app_settings/shaders/class2/deferred/softenLightF.glsl b/indra/newview/app_settings/shaders/class2/deferred/softenLightF.glsl
index 9df9d75905..457189b48e 100644
--- a/indra/newview/app_settings/shaders/class2/deferred/softenLightF.glsl
+++ b/indra/newview/app_settings/shaders/class2/deferred/softenLightF.glsl
@@ -309,7 +309,7 @@ void main()
 			//
 			vec3 refnormpersp = normalize(reflect(pos.xyz, norm.xyz));
 			float sa = dot(refnormpersp, sun_dir.xyz);
-			vec3 dumbshiny = vary_SunlitColor*scol_ambocc.r*(6 * texture2D(lightFunc, vec2(sa, spec.a)).r);
+			vec3 dumbshiny = vary_SunlitColor*scol_ambocc.r*(texture2D(lightFunc, vec2(sa, spec.a)).r);
 
 			// add the two types of shiny together
 			vec3 spec_contrib = dumbshiny * spec.rgb;
diff --git a/indra/newview/llviewershadermgr.cpp b/indra/newview/llviewershadermgr.cpp
index 4b0e0598f6..43de7450c8 100644
--- a/indra/newview/llviewershadermgr.cpp
+++ b/indra/newview/llviewershadermgr.cpp
@@ -826,12 +826,14 @@ BOOL LLViewerShaderMgr::loadBasicShaders()
 		shaders.push_back( make_pair( "objects/indexedTextureV.glsl",			1 ) );
 	}
 	shaders.push_back( make_pair( "objects/nonindexedTextureV.glsl",		1 ) );
-
+	
+	boost::unordered_map<std::string, std::string> attribs;
+	
 	// We no longer have to bind the shaders to global glhandles, they are automatically added to a map now.
 	for (U32 i = 0; i < shaders.size(); i++)
 	{
 		// Note usage of GL_VERTEX_SHADER_ARB
-		if (loadShaderFile(shaders[i].first, shaders[i].second, GL_VERTEX_SHADER_ARB) == 0)
+		if (loadShaderFile(shaders[i].first, shaders[i].second, GL_VERTEX_SHADER_ARB, attribs) == 0)
 		{
 			return FALSE;
 		}
@@ -879,11 +881,11 @@ BOOL LLViewerShaderMgr::loadBasicShaders()
 	index_channels.push_back(ch);	 shaders.push_back( make_pair( "lighting/lightFullbrightShinyF.glsl",	mVertexShaderLevel[SHADER_LIGHTING] ) );
 	index_channels.push_back(ch);	 shaders.push_back( make_pair( "lighting/lightShinyWaterF.glsl",			mVertexShaderLevel[SHADER_LIGHTING] ) );
 	index_channels.push_back(ch);	 shaders.push_back( make_pair( "lighting/lightFullbrightShinyWaterF.glsl", mVertexShaderLevel[SHADER_LIGHTING] ) );
-	
+
 	for (U32 i = 0; i < shaders.size(); i++)
 	{
 		// Note usage of GL_FRAGMENT_SHADER_ARB
-		if (loadShaderFile(shaders[i].first, shaders[i].second, GL_FRAGMENT_SHADER_ARB, index_channels[i]) == 0)
+		if (loadShaderFile(shaders[i].first, shaders[i].second, GL_FRAGMENT_SHADER_ARB, attribs, index_channels[i]) == 0)
 		{
 			return FALSE;
 		}
diff --git a/indra/newview/pipeline.cpp b/indra/newview/pipeline.cpp
index 920bac93a2..28fc70e38a 100644
--- a/indra/newview/pipeline.cpp
+++ b/indra/newview/pipeline.cpp
@@ -397,7 +397,7 @@ void validate_framebuffer_object();
 bool addDeferredAttachments(LLRenderTarget& target)
 {
 	return target.addColorAttachment(GL_RGBA) && //specular
-			target.addColorAttachment(GL_RGBA); //normal+z	
+			target.addColorAttachment(GL_RGB10_A2); //normal+z
 }
 
 LLPipeline::LLPipeline() :
@@ -863,7 +863,7 @@ bool LLPipeline::allocateScreenBuffer(U32 resX, U32 resY, U32 samples)
 		if (!mDeferredScreen.allocate(resX, resY, GL_RGBA, TRUE, TRUE, LLTexUnit::TT_RECT_TEXTURE, FALSE, samples)) return false;
 		if (!mDeferredDepth.allocate(resX, resY, 0, TRUE, FALSE, LLTexUnit::TT_RECT_TEXTURE, FALSE, samples)) return false;
 		if (!addDeferredAttachments(mDeferredScreen)) return false;
-	
+		
 		if (!mScreen.allocate(resX, resY, GL_RGBA, FALSE, FALSE, LLTexUnit::TT_RECT_TEXTURE, FALSE, samples)) return false;
 		if (samples > 0)
 		{
@@ -1206,6 +1206,11 @@ void LLPipeline::createGLBuffers()
 	gBumpImageList.restoreGL();
 }
 
+F32 lerpf(F32 a, F32 b, F32 w)
+{
+	return a + w * (b - a);
+}
+
 void LLPipeline::createLUTBuffers()
 {
 	if (sRenderDeferred)
@@ -1214,45 +1219,40 @@ void LLPipeline::createLUTBuffers()
 		{
 			U32 lightResX = gSavedSettings.getU32("RenderSpecularResX");
 			U32 lightResY = gSavedSettings.getU32("RenderSpecularResY");
-			U8* ls = new U8[lightResX*lightResY];
-			F32 specExp = gSavedSettings.getF32("RenderSpecularExponent");
-            // Calculate the (normalized) Blinn-Phong specular lookup texture.
+			F32* ls = new F32[lightResX*lightResY];
+			//F32 specExp = gSavedSettings.getF32("RenderSpecularExponent"); // Note: only use this when creating new specular lighting functions.
+            // Calculate the (normalized) Gaussian specular lookup texture. (with a few tweaks)
 			for (U32 y = 0; y < lightResY; ++y)
 			{
 				for (U32 x = 0; x < lightResX; ++x)
 				{
 					ls[y*lightResX+x] = 0;
 					F32 sa = (F32) x/(lightResX-1);
-					F32 spec = (F32) y/(lightResY-1);
-					F32 n = spec * spec * specExp;
+					F32 spec = (F32) y/(lightResY);
+					F32 n = spec;
 					
-					// Nothing special here.  Just your typical blinn-phong term.
-					spec = powf(sa, n);
+					float angleNormalHalf = acosf(sa);
+					float exponent = angleNormalHalf / ((1 - n));
+					exponent = -(exponent * exponent);
+					spec = expf(exponent);
 					
 					// Apply our normalization function.
-					// Note: This is the full equation that applies the full normalization curve, not an approximation.
-					// This is fine, given we only need to create our LUT once per buffer initialization.
-					// The only trade off is we have a really low dynamic range.
-					// This means we have to account for things not being able to exceed 0 to 1 in our shaders.
-					spec *= (((n + 2) * (n + 4)) / (8 * F_PI * (powf(2, -n/2) + n)));
+					// This is based around the phong normalization function, trading n+2 for n+1 instead.
+					// Since we're using a gaussian model here, we actually don't really need as large of an exponent as blinn-phong shading.
+					// Instead, we assume that the correct exponent is 8 here.
+					// This was achieved through much tweaking to find a decent "middleground" with our specular highlights with the gaussian term.
+					// Bigger highlights don't look too soft, smaller highlights don't look too bright, and everything in the middle seems to have a well maintained highlight curvature.
+					// There isn't really much theory behind this one.  This was done purely to produce a nice and mostly customizable BRDF.
 					
-					// Always sample at a 1.0/2.2 curve.
-					// This "Gamma corrects" our specular term, boosting our lower exponent reflections.
-					spec = powf(spec, 1.f/2.2f);
+					spec = lerpf(spec, spec * (n * 8 + 1) / 4.5, n);
 					
-					// Easy fix for our dynamic range problem: divide by 6 here, multiply by 6 in our shaders.
-					// This allows for our specular term to exceed a value of 1 in our shaders.
-					// This is something that can be important for energy conserving specular models where higher exponents can result in highlights that exceed a range of 0 to 1.
-					// Technically, we could just use an R16F texture, but driver support for R16F textures can be somewhat spotty at times.
-					// This works remarkably well for higher specular exponents, though banding can sometimes be seen on lower exponents.
-					// Combined with a bit of noise and trilinear filtering, the banding is hardly noticable.
-					ls[y*lightResX+x] = (U8)(llclamp(spec * (1.f / 6), 0.f, 1.f) * 255);
+					ls[y*lightResX+x] = spec;
 				}
 			}
 			
-			LLImageGL::generateTextures(LLTexUnit::TT_TEXTURE, GL_R8, 1, &mLightFunc);
+			LLImageGL::generateTextures(LLTexUnit::TT_TEXTURE, GL_R16F, 1, &mLightFunc);
 			gGL.getTexUnit(0)->bindManual(LLTexUnit::TT_TEXTURE, mLightFunc);
-			LLImageGL::setManualImage(LLTexUnit::getInternalType(LLTexUnit::TT_TEXTURE), 0, GL_R8, lightResX, lightResY, GL_RED, GL_UNSIGNED_BYTE, ls, false);
+			LLImageGL::setManualImage(LLTexUnit::getInternalType(LLTexUnit::TT_TEXTURE), 0, GL_R16F, lightResX, lightResY, GL_RED, GL_FLOAT, ls, false);
 			gGL.getTexUnit(0)->setTextureAddressMode(LLTexUnit::TAM_CLAMP);
 			gGL.getTexUnit(0)->setTextureFilteringOption(LLTexUnit::TFO_TRILINEAR);