Merge remote-tracking branch 'singu/master'

LICENSES/LEGAL-intel_matrixlib.txt

@@ -0,0 +1,29 @@
+INTEL LICENSE AGREEMENT
+
+IMPORTANT - READ BEFORE COPYING OR USING. 
+Do not use or load this library and any associated materials (collectively, 
+the "Software") until you have read the following terms and conditions. By 
+loading or using the Software, you agree to the terms of this Agreement. If 
+you do not wish to so agree, do not use the Software.
+
+LICENSE:  Subject to the restrictions below, Intel Corporation ("Intel") 
+grants to you the permission to use, copy, distribute and prepare derivative 
+works of this Software for any purpose and without fee, provided, that 
+Intel's copyright notice appear in all copies of the Software files.
+The distribution of derivative works of the Software is also subject to the 
+following limitations:  you (i) are solely responsible to your customers for 
+any liability which may arise from the distribution, (ii) do not make any 
+statement that your product is "certified", or that its performance is 
+guaranteed, by Intel, and (iii) do not use Intel's name or trademarks to 
+market your product without written permission.
+
+EXCLUSION OF ALL WARRANTIES. The Software is provided "AS IS" without any 
+express or implies warranty of any kind including warranties of 
+merchantability, noninfringement, or fitness for a particular purpose.  
+Intel does not warrant or assume responsibility for the accuracy or 
+completeness of any information contained within the Software.
+As this Software is given free of charge, in no event shall Intel be liable 
+for any damages whatsoever arising out of the use of or inability to use the 
+Software, even if Intel has been adviced of the possibility of such damages. 
+Intel does not assume any responsibility for any errors which may appear in 
+this Software nor any responsibility to update it.

indra/aistatemachine/aistatemachine.cpp

@@ -293,7 +293,36 @@ void AIEngine::add(AIStateMachine* state_machine)
   }
 }
 
-extern void print_statemachine_diagnostics(U64 total_clocks, U64 max_delta, AIEngine::queued_type::const_reference slowest_state_machine);
+#if STATE_MACHINE_PROFILING
+// Called from AIStateMachine::mainloop
+void print_statemachine_diagnostics(U64 total_clocks, AIStateMachine::StateTimerBase::TimeData& slowest_timer, AIEngine::queued_type::const_reference slowest_element)
+{
+	AIStateMachine const& slowest_state_machine = slowest_element.statemachine();
+	F64 const tfactor = 1000 / calc_clock_frequency();
+	std::ostringstream msg;
+
+	U64 max_delta = slowest_timer.GetDuration();
+
+	if (total_clocks > max_delta)
+	{
+		msg << "AIStateMachine::mainloop did run for " << (total_clocks * tfactor) << " ms. The slowest ";
+	}
+	else
+	{
+		msg << "AIStateMachine::mainloop: A ";
+	}
+	msg << "state machine " << "(" << slowest_state_machine.getName() << ") " << "ran for " << (max_delta * tfactor) << " ms";
+	if (slowest_state_machine.getRuntime() > max_delta)
+	{
+		msg << " (" << (slowest_state_machine.getRuntime() * tfactor) << " ms in total now)";
+	}
+	msg << ".\n";
+
+	AIStateMachine::StateTimerBase::DumpTimers(msg);
+
+	llwarns << msg.str() << llendl;
+}
+#endif
 
 // MAIN-THREAD
 void AIEngine::mainloop(void)
@@ -305,28 +334,33 @@ void AIEngine::mainloop(void)
 	queued_element = engine_state_w->list.begin();
   }
   U64 total_clocks = 0;
-#ifndef LL_RELEASE_FOR_DOWNLOAD
+#if STATE_MACHINE_PROFILING
-  U64 max_delta = 0;
   queued_type::value_type slowest_element(NULL);
+  AIStateMachine::StateTimerRoot::TimeData slowest_timer;
 #endif
   while (queued_element != end)
   {
 	AIStateMachine& state_machine(queued_element->statemachine());
-	U64 start = get_clock_count();
+	AIStateMachine::StateTimerBase::TimeData time_data;
-	if (!state_machine.sleep(start))
+	if (!state_machine.sleep(get_clock_count()))
 	{
-	  state_machine.multiplex(AIStateMachine::normal_run);
+		AIStateMachine::StateTimerRoot timer(state_machine.getName());
+		state_machine.multiplex(AIStateMachine::normal_run);
+		time_data = timer.GetTimerData();
 	}
-	U64 delta = get_clock_count() - start;
+	if (U64 delta = time_data.GetDuration())
-	state_machine.add(delta);
-	total_clocks += delta;
-#ifndef LL_RELEASE_FOR_DOWNLOAD
-	if (delta > max_delta)
 	{
-	  max_delta = delta;
+		state_machine.add(delta);
-	  slowest_element = *queued_element;
+		total_clocks += delta;
-	}
+#if STATE_MACHINE_PROFILING
+		if (delta > slowest_timer.GetDuration())
+		{
+			slowest_element = *queued_element;
+			slowest_timer = time_data;
+		}
 #endif
+	}
+
 	bool active = state_machine.active(this);		// This locks mState shortly, so it must be called before locking mEngineState because add() locks mEngineState while holding mState.
 	engine_state_type_wat engine_state_w(mEngineState);
 	if (!active)
@@ -340,8 +374,8 @@ void AIEngine::mainloop(void)
 	}
 	if (total_clocks >= sMaxCount)
 	{
-#ifndef LL_RELEASE_FOR_DOWNLOAD
+#if STATE_MACHINE_PROFILING
-	  print_statemachine_diagnostics(total_clocks, max_delta, slowest_element);
+		print_statemachine_diagnostics(total_clocks, slowest_timer, slowest_element);
 #endif
 	  Dout(dc::statemachine, "Sorting " << engine_state_w->list.size() << " state machines.");
 	  engine_state_w->list.sort(QueueElementComp());
@@ -752,6 +786,22 @@ void AIStateMachine::multiplex(event_type event)
   }
 }
 
+#if STATE_MACHINE_PROFILING
+std::vector<AIStateMachine::StateTimerBase*> AIStateMachine::StateTimerBase::mTimerStack;
+AIStateMachine::StateTimerBase::TimeData AIStateMachine::StateTimerBase::TimeData::sRoot("");
+void AIStateMachine::StateTimer::TimeData::DumpTimer(std::ostringstream& msg, std::string prefix)
+{
+	F64 const tfactor = 1000 / calc_clock_frequency();
+	msg << prefix << mName << " " << (mEnd - mStart)*tfactor << "ms" << std::endl;
+	prefix.push_back(' ');
+	std::vector<TimeData>::iterator it;
+	for (it = mChildren.begin(); it != mChildren.end(); ++it)
+	{
+		it->DumpTimer(msg, prefix);
+	}
+}
+#endif
+
 AIStateMachine::state_type AIStateMachine::begin_loop(base_state_type base_state)
 {
   DoutEntering(dc::statemachine(mSMDebug), "AIStateMachine::begin_loop(" << state_str(base_state) << ") [" << (void*)this << "]");

indra/aistatemachine/aistatemachine.h

@@ -36,6 +36,7 @@
 
 #include "aithreadsafe.h"
 #include <llpointer.h>
+#include "lltimer.h"
 #include <list>
 #include <boost/signals2.hpp>
 
@@ -98,11 +99,143 @@ class AIEngine
 extern AIEngine gMainThreadEngine;
 extern AIEngine gStateMachineThreadEngine;
 
+#ifndef STATE_MACHINE_PROFILING
+#ifndef LL_RELEASE_FOR_DOWNLOAD
+#define STATE_MACHINE_PROFILING 1
+#endif
+#endif
+
 class AIStateMachine : public LLThreadSafeRefCount
 {
   public:
 	typedef U32 state_type;		//!< The type of run_state
 
+	// A simple timer class that will calculate time delta between ctor and GetTimerData call.
+	// Time data is stored as a nested TimeData object.
+	// If STATE_MACHINE_PROFILING is defined then a stack of all StateTimers from root is maintained for debug output.
+	class StateTimerBase
+	{
+	public:
+		class TimeData
+		{
+			friend class StateTimerBase;
+		public:
+			TimeData() : mStart(-1), mEnd(-1) {}
+			U64 GetDuration() {	return mEnd - mStart; }
+		private:
+			U64 mStart, mEnd;
+
+#if !STATE_MACHINE_PROFILING
+			TimeData(const std::string& name) : mStart(get_clock_count()), mEnd(get_clock_count()) {}
+#else
+			TimeData(const std::string& name) : mName(name), mStart(get_clock_count()), mEnd(get_clock_count()) {}
+			void DumpTimer(std::ostringstream& msg, std::string prefix);
+			std::vector<TimeData> mChildren;
+			std::string mName;	
+			static TimeData sRoot;
+#endif
+		};
+#if !STATE_MACHINE_PROFILING
+		StateTimerBase(const std::string& name) : mData(name) {}
+		~StateTimerBase() {}
+	protected:
+		TimeData mData;
+		// Return a copy of the underlying timer data.
+		// This allows the data live beyond the scope of the state timer.
+	public:
+		const TimeData GetTimerData()
+		{
+			mData.mEnd = get_clock_count();	//set mEnd to current time, since GetTimerData() will always be called before the dtor, obv.
+			return mData;
+		}
+#else
+	protected:
+		// Ctors/dtors are hidden. Only StateTimerRoot and StateTimer are permitted to access them.
+		StateTimerBase() : mData(NULL) {}
+		~StateTimerBase()
+		{
+			// If mData is null then the timer was not registered due to being in the wrong thread or the root timer wasn't in the expected state.
+			if (!mData)
+				return;
+			mData->mEnd = get_clock_count();
+			mTimerStack.pop_back();
+		}
+
+		// Also hide internals from everything except StateTimerRoot and StateTimer
+		bool AddAsRoot(const std::string& name)
+		{
+			
+			if (!is_main_thread())
+				return true;	//Ignoring this timer, but pretending it was added.
+			if (!mTimerStack.empty())
+				return false;
+			TimeData::sRoot = TimeData(name);
+			mData = &TimeData::sRoot;
+			mData->mChildren.clear();	
+			mTimerStack.push_back(this);
+			return true;
+		}
+		bool AddAsChild(const std::string& name)
+		{
+			if (!is_main_thread())
+				return true;	//Ignoring this timer, but pretending it was added.
+			if (mTimerStack.empty())
+				return false;
+			mTimerStack.back()->mData->mChildren.push_back(TimeData(name));
+			mData = &mTimerStack.back()->mData->mChildren.back();
+			mTimerStack.push_back(this);
+			return true;
+		}
+
+		TimeData* mData;
+		static std::vector<StateTimerBase*> mTimerStack;
+
+	public:
+		// Debug spew
+		static void DumpTimers(std::ostringstream& msg)
+		{
+			TimeData::sRoot.DumpTimer(msg, "");
+		}
+
+		// Return a copy of the underlying timer data.
+		// This allows the data live beyond the scope of the state timer.
+		const TimeData GetTimerData() const
+		{
+			if (mData)
+			{
+				TimeData ret = *mData;
+				ret.mEnd = get_clock_count();	//set mEnd to current time, since GetTimerData() will always be called before the dtor, obv.
+				return ret;
+			}
+			return TimeData();
+		}
+#endif
+	};
+public:
+#if !STATE_MACHINE_PROFILING
+	typedef StateTimerBase StateTimerRoot;
+	typedef StateTimerBase StateTimer;
+#else
+	class StateTimerRoot : public StateTimerBase
+	{ //A StateTimerRoot can become a child if a root already exists.
+	public:
+		StateTimerRoot(const std::string& name)
+		{
+			if(!AddAsRoot(name))
+				AddAsChild(name);	
+		}
+	};
+	class StateTimer : public StateTimerBase
+	{ //A StateTimer can never become a root
+	public:
+		StateTimer(const std::string& name)
+		{
+			AddAsChild(name);
+		}
+	};
+#endif
+
+
   protected:
 	// The type of event that causes multiplex() to be called.
 	enum event_type {
@@ -302,6 +435,9 @@ class AIStateMachine : public LLThreadSafeRefCount
 	void add(U64 count) { mRuntime += count; }
 	U64 getRuntime(void) const { return mRuntime; }
 
+	// For diagnostics. Every derived class must override this.
+	virtual const char* getName() const = 0;
+
   protected:
 	virtual void initialize_impl(void) = 0;
 	virtual void multiplex_impl(state_type run_state) = 0;

indra/aistatemachine/aistatemachinethread.h

@@ -232,6 +232,13 @@ class AIStateMachineThread : public AIStateMachineThreadBase {
 	// Accessor.
 	THREAD_IMPL& thread_impl(void) { return mThreadImpl; }
 
+	/*virtual*/ const char* getName() const
+	{
+#define STRIZE(arg) #arg
+		return "AIStateMachineThread<"STRIZE(THREAD_IMPL)">";
+#undef STRIZE
+	}
+
   protected:
 	/*virtual*/ AIThreadImpl& impl(void) { return mThreadImpl; }
 };

indra/aistatemachine/aitimer.h

@@ -98,6 +98,8 @@ class AITimer : public AIStateMachine {
 	 */
 	F64 getInterval(void) const { return mInterval; }
 
+	/*virtual*/ const char* getName() const { return "AITimer"; }
+
   protected:
 	// Call finish() (or abort()), not delete.
 	/*virtual*/ ~AITimer() { DoutEntering(dc::statemachine(mSMDebug), "~AITimer() [" << (void*)this << "]"); mFrameTimer.cancel(); }

indra/develop.py

@@ -221,6 +221,7 @@ class UnixSetup(PlatformSetup):
     exe_suffixes = ('',)
 
     def __init__(self):
+        PlatformSetup.__init__(self)
         super(UnixSetup, self).__init__()
         self.generator = 'Unix Makefiles'
 
@@ -263,6 +264,7 @@ class UnixSetup(PlatformSetup):
 
 class LinuxSetup(UnixSetup):
     def __init__(self):
+        UnixSetup.__init__(self)
         super(LinuxSetup, self).__init__()
         try:
             self.debian_sarge = open('/etc/debian_version').read().strip() == '3.1'
@@ -384,6 +386,7 @@ class LinuxSetup(UnixSetup):
         
 class DarwinSetup(UnixSetup):
     def __init__(self):
+        UnixSetup.__init__(self)
         super(DarwinSetup, self).__init__()
         self.generator = 'Xcode'
 
@@ -457,6 +460,7 @@ class WindowsSetup(PlatformSetup):
     exe_suffixes = ('.exe', '.bat', '.com')
 
     def __init__(self):
+        PlatformSetup.__init__(self)
         super(WindowsSetup, self).__init__()
         self._generator = None
         self.incredibuild = False
@@ -497,7 +501,10 @@ class WindowsSetup(PlatformSetup):
         return 'win32'
 
     def build_dirs(self):
-        return ['build-' + self.generator]
+        if self.word_size == 64:
+            return ['build-' + self.generator + '-Win64']
+        else:
+            return ['build-' + self.generator]
 
     def cmake_commandline(self, src_dir, build_dir, opts, simple):
         args = dict(

indra/llappearance/llavatarappearance.cpp

@@ -983,19 +983,19 @@ BOOL LLAvatarAppearance::loadSkeletonNode ()
 	mRoot->addChild(mMeshLOD[MESH_ID_SKIRT]);
 	mRoot->addChild(mMeshLOD[MESH_ID_HEAD]);
 
-	LLAvatarJoint *skull = (LLAvatarJoint*)mRoot->findJoint("mSkull");
+	LLJoint *skull = mRoot->findJoint("mSkull");
 	if (skull)
 	{
 		skull->addChild(mMeshLOD[MESH_ID_HAIR] );
 	}
 
-	LLAvatarJoint *eyeL = (LLAvatarJoint*)mRoot->findJoint("mEyeLeft");
+	LLJoint *eyeL = mRoot->findJoint("mEyeLeft");
 	if (eyeL)
 	{
 		eyeL->addChild( mMeshLOD[MESH_ID_EYEBALL_LEFT] );
 	}
 
-	LLAvatarJoint *eyeR = (LLAvatarJoint*)mRoot->findJoint("mEyeRight");
+	LLJoint *eyeR = mRoot->findJoint("mEyeRight");
 	if (eyeR)
 	{
 		eyeR->addChild( mMeshLOD[MESH_ID_EYEBALL_RIGHT] );

indra/llappearance/llavatarjoint.cpp

@@ -105,8 +105,9 @@ void LLAvatarJoint::setValid( BOOL valid, BOOL recursive )
 		for (child_list_t::iterator iter = mChildren.begin();
 			 iter != mChildren.end(); ++iter)
 		{
-			LLAvatarJoint* joint = (LLAvatarJoint*)(*iter);
+			LLAvatarJoint* joint = dynamic_cast<LLAvatarJoint*>(*iter);
-			joint->setValid(valid, TRUE);
+			if (joint)
+				joint->setValid(valid, TRUE);
 		}
 	}
 
@@ -124,7 +125,8 @@ void LLAvatarJoint::setSkeletonComponents( U32 comp, BOOL recursive )
 			 iter != mChildren.end(); ++iter)
 		{
 			LLAvatarJoint* joint = dynamic_cast<LLAvatarJoint*>(*iter);
-			joint->setSkeletonComponents(comp, recursive);
+			if (joint)
+				joint->setSkeletonComponents(comp, recursive);
 		}
 	}
 }
@@ -138,8 +140,9 @@ void LLAvatarJoint::setVisible(BOOL visible, BOOL recursive)
 		for (child_list_t::iterator iter = mChildren.begin();
 			 iter != mChildren.end(); ++iter)
 		{
-			LLAvatarJoint* joint = (LLAvatarJoint*)(*iter);
+			LLAvatarJoint* joint = dynamic_cast<LLAvatarJoint*>(*iter);
-			joint->setVisible(visible, recursive);
+			if(joint)
+				joint->setVisible(visible, recursive);
 		}
 	}
 }
@@ -150,7 +153,8 @@ void LLAvatarJoint::updateFaceSizes(U32 &num_vertices, U32& num_indices, F32 pix
 		 iter != mChildren.end(); ++iter)
 	{
 		LLAvatarJoint* joint = dynamic_cast<LLAvatarJoint*>(*iter);
-		joint->updateFaceSizes(num_vertices, num_indices, pixel_area);
+		if (joint)
+			joint->updateFaceSizes(num_vertices, num_indices, pixel_area);
 	}
 }
 
@@ -160,7 +164,8 @@ void LLAvatarJoint::updateFaceData(LLFace *face, F32 pixel_area, BOOL damp_wind,
 		 iter != mChildren.end(); ++iter)
 	{
 		LLAvatarJoint* joint = dynamic_cast<LLAvatarJoint*>(*iter);
-		joint->updateFaceData(face, pixel_area, damp_wind, terse_update);
+		if (joint)
+			joint->updateFaceData(face, pixel_area, damp_wind, terse_update);
 	}
 }
 
@@ -170,7 +175,8 @@ void LLAvatarJoint::updateJointGeometry()
 		 iter != mChildren.end(); ++iter)
 	{
 		LLAvatarJoint* joint = dynamic_cast<LLAvatarJoint*>(*iter);
-		joint->updateJointGeometry();
+		if (joint)
+			joint->updateJointGeometry();
 	}
 }
 
@@ -184,6 +190,9 @@ BOOL LLAvatarJoint::updateLOD(F32 pixel_area, BOOL activate)
 		 iter != mChildren.end(); ++iter)
 	{
 		LLAvatarJoint* joint = dynamic_cast<LLAvatarJoint*>(*iter);
+		if (!joint)
+			continue;
+
 		F32 jointLOD = joint->getLOD();
 		
 		if (found_lod || jointLOD == DEFAULT_AVATAR_JOINT_LOD)
@@ -213,7 +222,8 @@ void LLAvatarJoint::dump()
 		 iter != mChildren.end(); ++iter)
 	{
 		LLAvatarJoint* joint = dynamic_cast<LLAvatarJoint*>(*iter);
-		joint->dump();
+		if (joint)
+			joint->dump();
 	}
 }
 
@@ -260,7 +270,7 @@ void LLAvatarJointCollisionVolume::renderCollision()
 	updateWorldMatrix();
 	
 	gGL.pushMatrix();
-	gGL.multMatrix( &mXform.getWorldMatrix().mMatrix[0][0] );
+	gGL.multMatrix( mXform.getWorldMatrix() );
 
 	gGL.diffuseColor3f( 0.f, 0.f, 1.f );
 	

indra/llappearance/llavatarjointmesh.cpp

@@ -83,30 +83,28 @@ LLSkinJoint::~LLSkinJoint()
 //-----------------------------------------------------------------------------
 // LLSkinJoint::setupSkinJoint()
 //-----------------------------------------------------------------------------
-BOOL LLSkinJoint::setupSkinJoint( LLAvatarJoint *joint)
+void LLSkinJoint::setupSkinJoint( LLJoint *joint)
 {
+	mRootToJointSkinOffset.clearVec();
+	mRootToParentJointSkinOffset.clearVec();
+	
 	// find the named joint
-	mJoint = joint;
+	if (!(mJoint = joint))
-	if ( !mJoint )
 	{
 		llinfos << "Can't find joint" << llendl;
+		return;
 	}
 
 	// compute the inverse root skin matrix
-	mRootToJointSkinOffset.clearVec();
+	do
-
-	LLVector3 rootSkinOffset;
-	while (joint)
 	{
-		rootSkinOffset += joint->getSkinOffset();
+		mRootToJointSkinOffset -= joint->getSkinOffset();
-		joint = (LLAvatarJoint*)joint->getParent();
+	} while (joint = joint->getParent());
-	}
 
-	mRootToJointSkinOffset = -rootSkinOffset;
 	mRootToParentJointSkinOffset = mRootToJointSkinOffset;
 	mRootToParentJointSkinOffset += mJoint->getSkinOffset();
 
-	return TRUE;
+	return;
 }
 
 
@@ -307,15 +305,14 @@ void LLAvatarJointMesh::setMesh( LLPolyMesh *mesh )
 		for (jn = 0; jn < numJointNames; jn++)
 		{
 			//llinfos << "Setting up joint " << jointNames[jn] << llendl;
-			LLAvatarJoint* joint = (LLAvatarJoint*)(getRoot()->findJoint(jointNames[jn]) );
+			mSkinJoints[jn].setupSkinJoint( getRoot()->findJoint(jointNames[jn]) );
-			mSkinJoints[jn].setupSkinJoint( joint );
 		}
 	}
 
 	// setup joint array
 	if (!mMesh->isLOD())
 	{
-		setupJoint((LLAvatarJoint*)getRoot());
+		setupJoint(getRoot());
 	}
 
 //	llinfos << "joint render entries: " << mMesh->mJointRenderData.count() << llendl;
@@ -324,7 +321,7 @@ void LLAvatarJointMesh::setMesh( LLPolyMesh *mesh )
 //-----------------------------------------------------------------------------
 // setupJoint()
 //-----------------------------------------------------------------------------
-void LLAvatarJointMesh::setupJoint(LLAvatarJoint* current_joint)
+void LLAvatarJointMesh::setupJoint(LLJoint* current_joint)
 {
 //	llinfos << "Mesh: " << getName() << llendl;
 
@@ -345,7 +342,7 @@ void LLAvatarJointMesh::setupJoint(LLAvatarJoint* current_joint)
 		if(mMesh->mJointRenderData.count() && mMesh->mJointRenderData[mMesh->mJointRenderData.count() - 1]->mWorldMatrix == &current_joint->getParent()->getWorldMatrix())
 		{
 			// ...then just add ourselves
-			LLAvatarJoint* jointp = js.mJoint;
+			LLJoint* jointp = js.mJoint;
 			mMesh->mJointRenderData.put(new LLJointRenderData(&jointp->getWorldMatrix(), &js));
 //			llinfos << "joint " << joint_count << js.mJoint->getName() << llendl;
 //			joint_count++;
@@ -366,8 +363,9 @@ void LLAvatarJointMesh::setupJoint(LLAvatarJoint* current_joint)
 	for (LLJoint::child_list_t::iterator iter = current_joint->mChildren.begin();
 		 iter != current_joint->mChildren.end(); ++iter)
 	{
-		LLAvatarJoint* child_joint = (LLAvatarJoint*)(*iter);
+		LLAvatarJoint* child_joint = dynamic_cast<LLAvatarJoint*>(*iter);
-		setupJoint(child_joint);
+		if(child_joint)
+			setupJoint(child_joint);
 	}
 }
 

indra/llappearance/llavatarjointmesh.h

@@ -49,9 +49,9 @@ class LLSkinJoint
 public:
 	LLSkinJoint();
 	~LLSkinJoint();
-	BOOL setupSkinJoint( LLAvatarJoint *joint);
+	void setupSkinJoint( LLJoint *joint);
 
-	LLAvatarJoint	*mJoint;
+	LLJoint*		mJoint;
 	LLVector3		mRootToJointSkinOffset;
 	LLVector3		mRootToParentJointSkinOffset;
 };
@@ -122,7 +122,7 @@ public:
 	void setMesh( LLPolyMesh *mesh );
 
 	// Sets up joint matrix data for rendering
-	void setupJoint(LLAvatarJoint* current_joint);
+	void setupJoint(LLJoint* current_joint);
 
 	// Sets ID for picking
 	void setMeshID( S32 id ) {mMeshID = id;}

indra/llappearance/llpolymesh.cpp

@@ -231,9 +231,9 @@ BOOL LLPolyMeshSharedData::allocateVertexData( U32 numVertices )
         mBaseCoords = (LLVector4a*) ll_aligned_malloc_16(numVertices*sizeof(LLVector4a));
         mBaseNormals = (LLVector4a*) ll_aligned_malloc_16(numVertices*sizeof(LLVector4a));
         mBaseBinormals = (LLVector4a*) ll_aligned_malloc_16(numVertices*sizeof(LLVector4a));
-        mTexCoords = (LLVector2*) ll_aligned_malloc_16(numVertices*sizeof(LLVector2));
+        mTexCoords = (LLVector2*) ll_aligned_malloc_16((numVertices+numVertices%2)*sizeof(LLVector2));
-        mDetailTexCoords = (LLVector2*) ll_aligned_malloc_16(numVertices*sizeof(LLVector2));
+        mDetailTexCoords = (LLVector2*) ll_aligned_malloc_16((numVertices+numVertices%2)*sizeof(LLVector2));
-        mWeights = (F32*) ll_aligned_malloc_16(numVertices*sizeof(F32));
+        mWeights = (F32*) ll_aligned_malloc_16(((numVertices)*sizeof(F32)+0xF) & ~0xF);
         for (i = 0; i < numVertices; i++)
 	{
 		mBaseCoords[i].clear();

indra/llappearance/llpolymesh.h

@@ -146,10 +146,10 @@ public:
 class LLJointRenderData
 {
 public:
-	LLJointRenderData(const LLMatrix4* world_matrix, LLSkinJoint* skin_joint) : mWorldMatrix(world_matrix), mSkinJoint(skin_joint) {}
+	LLJointRenderData(const LLMatrix4a* world_matrix, LLSkinJoint* skin_joint) : mWorldMatrix(world_matrix), mSkinJoint(skin_joint) {}
 	~LLJointRenderData(){}
 
-	const LLMatrix4*		mWorldMatrix;
+	const LLMatrix4a*		mWorldMatrix;
 	LLSkinJoint*			mSkinJoint;
 };
 

indra/llappearance/llpolyskeletaldistortion.cpp

@@ -154,13 +154,13 @@ BOOL LLPolySkeletalDistortion::setInfo(LLPolySkeletalDistortionInfo *info)
                 for (LLJoint::child_list_t::iterator iter = joint->mChildren.begin();
                      iter != joint->mChildren.end(); ++iter)
                 {
-                        LLAvatarJoint* child_joint = (LLAvatarJoint*)(*iter);
+					LLAvatarJoint* child_joint = dynamic_cast<LLAvatarJoint*>(*iter);
-                        if (child_joint->inheritScale())
+					if (child_joint && child_joint->inheritScale())
-                        {
+                    {
-                                LLVector3 childDeformation = LLVector3(child_joint->getScale());
+                            LLVector3 childDeformation = LLVector3(child_joint->getScale());
-                                childDeformation.scaleVec(bone_info->mScaleDeformation);
+                            childDeformation.scaleVec(bone_info->mScaleDeformation);
-                                mJointScales[child_joint] = childDeformation;
+                            mJointScales[child_joint] = childDeformation;
-                        }
+                    }
                 }
 
                 if (bone_info->mHasPositionDeformation)

indra/llappearance/llpolyskeletaldistortion.h

@@ -68,7 +68,16 @@ class LLPolySkeletalDistortionInfo : public LLViewerVisualParamInfo
 {
 	friend class LLPolySkeletalDistortion;
 public:
-	
+	void* operator new(size_t size)
+	{
+		return ll_aligned_malloc_16(size);
+	}
+
+	void operator delete(void* ptr)
+	{
+		ll_aligned_free_16(ptr);
+	}
+
 	LLPolySkeletalDistortionInfo();
 	/*virtual*/ ~LLPolySkeletalDistortionInfo() {};
 	
@@ -77,12 +86,12 @@ public:
 protected:
 	typedef std::vector<LLPolySkeletalBoneInfo> bone_info_list_t;
 	bone_info_list_t mBoneInfoList;
-};
+} LL_ALIGN_POSTFIX(16);
-
 //-----------------------------------------------------------------------------
 // LLPolySkeletalDeformation
 // A set of joint scale data for deforming the avatar mesh
 //-----------------------------------------------------------------------------
+LL_ALIGN_PREFIX(16)
 class LLPolySkeletalDistortion : public LLViewerVisualParam
 {
 public:

indra/llcharacter/llcharacter.cpp

@@ -72,21 +72,11 @@ LLCharacter::~LLCharacter()
 		delete param;
 	}
 
-	U32 i ;
+	bool erased = vector_replace_with_last(sInstances,this);
-	U32 size = sInstances.size() ;
-	for(i = 0 ; i < size ; i++)
-	{
-		if(sInstances[i] == this)
-		{
-			break ;
-		}
-	}
 
-	llassert_always(i < size) ;
+	llassert_always(erased) ;
 
 	llassert_always(sAllowInstancesChange) ;
-	sInstances[i] = sInstances[size - 1] ;
-	sInstances.pop_back() ;
 }
 
 

indra/llcharacter/lljoint.cpp

@@ -312,19 +312,15 @@ void LLJoint::setWorldPosition( const LLVector3& pos )
 		return;
 	}
 
-	LLMatrix4 temp_matrix = getWorldMatrix();
+	LLMatrix4a temp_matrix = getWorldMatrix();
-	temp_matrix.mMatrix[VW][VX] = pos.mV[VX];
+	temp_matrix.setTranslate_affine(pos);
-	temp_matrix.mMatrix[VW][VY] = pos.mV[VY];
-	temp_matrix.mMatrix[VW][VZ] = pos.mV[VZ];
 
-	LLMatrix4 parentWorldMatrix = mParent->getWorldMatrix();
+	LLMatrix4a invParentWorldMatrix = mParent->getWorldMatrix();
-	LLMatrix4 invParentWorldMatrix = parentWorldMatrix.invert();
+	invParentWorldMatrix.invert();
 
-	temp_matrix *= invParentWorldMatrix;
+	invParentWorldMatrix.mul(temp_matrix);
 
-	LLVector3 localPos(	temp_matrix.mMatrix[VW][VX],
+	LLVector3 localPos(	invParentWorldMatrix.getRow<LLMatrix4a::ROW_TRANS>().getF32ptr() );
-						temp_matrix.mMatrix[VW][VY],
-						temp_matrix.mMatrix[VW][VZ] );
 
 	setPosition( localPos );
 }
@@ -383,19 +379,19 @@ void LLJoint::setWorldRotation( const LLQuaternion& rot )
 		this->setRotation( rot );
 		return;
 	}
+	
+	LLMatrix4a parentWorldMatrix = mParent->getWorldMatrix();
+	LLQuaternion2 rota(rot);
+	LLMatrix4a temp_mat(rota);
 
-	LLMatrix4 temp_mat(rot);
+	LLMatrix4a invParentWorldMatrix = mParent->getWorldMatrix();
+	invParentWorldMatrix.setTranslate_affine(LLVector3(0.f));
 
-	LLMatrix4 parentWorldMatrix = mParent->getWorldMatrix();
+	invParentWorldMatrix.invert();
-	parentWorldMatrix.mMatrix[VW][VX] = 0;
-	parentWorldMatrix.mMatrix[VW][VY] = 0;
-	parentWorldMatrix.mMatrix[VW][VZ] = 0;
 
-	LLMatrix4 invParentWorldMatrix = parentWorldMatrix.invert();
+	invParentWorldMatrix.mul(temp_mat);
 
-	temp_mat *= invParentWorldMatrix;
+	setRotation(LLQuaternion(LLMatrix4(invParentWorldMatrix.getF32ptr())));
-
-	setRotation(LLQuaternion(temp_mat));
 }
 
 
@@ -425,7 +421,7 @@ void LLJoint::setScale( const LLVector3& scale )
 //--------------------------------------------------------------------
 // getWorldMatrix()
 //--------------------------------------------------------------------
-const LLMatrix4 &LLJoint::getWorldMatrix()
+const LLMatrix4a &LLJoint::getWorldMatrix()
 {
 	updateWorldMatrixParent();
 

indra/llcharacter/lljoint.h

@@ -162,7 +162,7 @@ public:
 	void setScale( const LLVector3& scale );
 
 	// get/set world matrix
-	const LLMatrix4 &getWorldMatrix();
+	const LLMatrix4a &getWorldMatrix();
 	void setWorldMatrix( const LLMatrix4& mat );
 
 	void updateWorldMatrixChildren();

indra/llcharacter/lljointsolverrp3.cpp

@@ -171,12 +171,14 @@ void LLJointSolverRP3::solve()
 	//-------------------------------------------------------------------------
 	// get the poleVector in world space
 	//-------------------------------------------------------------------------
-	LLMatrix4 worldJointAParentMat;
+	LLVector3 poleVec = mPoleVector;
 	if ( mJointA->getParent() )
 	{
-		worldJointAParentMat = mJointA->getParent()->getWorldMatrix();
+		LLVector4a pole_veca;
+		pole_veca.load3(mPoleVector.mV);
+		mJointA->getParent()->getWorldMatrix().rotate(pole_veca,pole_veca);
+		poleVec.set(pole_veca.getF32ptr());
 	}
-	LLVector3 poleVec = rotate_vector( mPoleVector, worldJointAParentMat );
 
 	//-------------------------------------------------------------------------
 	// compute the following:

indra/llcharacter/llkeyframestandmotion.cpp

@@ -286,40 +286,38 @@ BOOL LLKeyframeStandMotion::onUpdate(F32 time, U8* joint_mask)
 	//-------------------------------------------------------------------------
 	if ( mTrackAnkles )
 	{
-		LLVector4 dirLeft4 = mAnkleLeftJoint.getWorldMatrix().getFwdRow4();
+		const LLVector4a& dirLeft4 = mAnkleLeftJoint.getWorldMatrix().getRow<LLMatrix4a::ROW_FWD>();
-		LLVector4 dirRight4 = mAnkleRightJoint.getWorldMatrix().getFwdRow4();
+		const LLVector4a& dirRight4 = mAnkleRightJoint.getWorldMatrix().getRow<LLMatrix4a::ROW_FWD>();
-		LLVector3 dirLeft = vec4to3( dirLeft4 );
-		LLVector3 dirRight = vec4to3( dirRight4 );
 
-		LLVector3 up;
+		LLVector4a up;
-		LLVector3 dir;
+		LLVector4a dir;
-		LLVector3 left;
+		LLVector4a left;
 
-		up = mNormalLeft;
+		up.load3(mNormalLeft.mV);
-		up.normVec();
+		up.normalize3fast();
 		if (mFlipFeet)
 		{
-			up *= -1.0f;
+			up.negate();
 		}
-		dir = dirLeft;
+		dir = dirLeft4;
-		dir.normVec();
+		dir.normalize3fast();
-		left = up % dir;
+		left.setCross3(up,dir);
-		left.normVec();
+		left.normalize3fast();
-		dir = left % up;
+		dir.setCross3(left,up);
-		mRotationLeft = LLQuaternion( dir, left, up );
+		mRotationLeft = LLQuaternion( LLVector3(dir.getF32ptr()), LLVector3(left.getF32ptr()), LLVector3(up.getF32ptr()));
 
-		up = mNormalRight;
+		up.load3(mNormalRight.mV);
-		up.normVec();
+		up.normalize3fast();
 		if (mFlipFeet)
 		{
-			up *= -1.0f;
+			up.negate();
 		}
-		dir = dirRight;
+		dir = dirRight4;
-		dir.normVec();
+		dir.normalize3fast();
-		left = up % dir;
+		left.setCross3(up,dir);
-		left.normVec();
+		left.normalize3fast();
-		dir = left % up;
+		dir.setCross3(left,up);
-		mRotationRight = LLQuaternion( dir, left, up );
+		mRotationRight = LLQuaternion( LLVector3(dir.getF32ptr()), LLVector3(left.getF32ptr()), LLVector3(up.getF32ptr()));
 	}
 	mAnkleLeftJoint.setWorldRotation( mRotationLeft );
 	mAnkleRightJoint.setWorldRotation( mRotationRight );

indra/llcommon/CMakeLists.txt

@@ -254,7 +254,6 @@ set(llcommon_HEADER_FILES
     metapropertyt.h
     reflective.h
     reflectivet.h
-    roles_constants.h
     stdenums.h
     stdtypes.h
     string_table.h

indra/llcommon/llstl.h

@@ -239,14 +239,14 @@ inline typename T::mapped_type get_ptr_in_map(const T& inmap, typename T::key_ty
 //
 //Singu note: This has been generalized to support a broader range of sequence containers
 template <typename T>
-inline typename T::iterator vector_replace_with_last(T& invec, typename T::iterator& iter)
+inline typename T::iterator vector_replace_with_last(T& invec, typename T::iterator iter)
 {
-	typename T::iterator last = invec.end(); --last;
+	typename T::iterator last = invec.end();
 	if (iter == invec.end())
 	{
 		return iter;
 	}
-	else if (iter == last)
+	else if (iter == --last)
 	{
 		invec.pop_back();
 		return invec.end();

indra/llcommon/stdenums.h

@@ -118,16 +118,6 @@ enum EAddPosition
 	ADD_BOTTOM
 };
 
-enum LLGroupChange
-{
-	GC_PROPERTIES,
-	GC_MEMBER_DATA,
-	GC_ROLE_DATA,
-	GC_ROLE_MEMBER_DATA,
-	GC_TITLES,
-	GC_ALL
-};
-
 //----------------------------------------------------------------------------
 // DEPRECATED - create new, more specific files for shared enums/constants
 //----------------------------------------------------------------------------

indra/llinventory/llinventory.cpp

@@ -178,7 +178,7 @@ BOOL LLInventoryObject::importLegacyStream(std::istream& input_stream)
 	while(input_stream.good())
 	{
 		input_stream.getline(buffer, MAX_STRING);
-		sscanf(buffer, " %254s %254s", keyword, valuestr);	/* Flawfinder: ignore */
+		if (sscanf(buffer, " %254s %254s", keyword, valuestr) < 1) continue;
 		if(0 == strcmp("{",keyword))
 		{
 			continue;
@@ -610,7 +610,7 @@ BOOL LLInventoryItem::importFile(LLFILE* fp)
 			buffer[0] = '\0';
 		}
 		
-		sscanf(buffer, " %254s %254s", keyword, valuestr);	/* Flawfinder: ignore */
+		if (sscanf(buffer, " %254s %254s", keyword, valuestr) < 1) continue;
 		if(0 == strcmp("{",keyword))
 		{
 			continue;
@@ -813,10 +813,10 @@ BOOL LLInventoryItem::importLegacyStream(std::istream& input_stream)
 	while(success && input_stream.good())
 	{
 		input_stream.getline(buffer, MAX_STRING);
-		sscanf(	/* Flawfinder: ignore */
+		if (sscanf(
 			buffer,
 			" %254s %254s",
-			keyword, valuestr);
+			keyword, valuestr) < 1) continue;
 		if(0 == strcmp("{",keyword))
 		{
 			continue;
@@ -1489,10 +1489,10 @@ BOOL LLInventoryCategory::importFile(LLFILE* fp)
 			buffer[0] = '\0';
 		}
 		
-		sscanf(	/* Flawfinder: ignore */
+		if (sscanf(
 			buffer,
 			" %254s %254s",
-			keyword, valuestr);
+			keyword, valuestr) < 1) continue;
 		if(0 == strcmp("{",keyword))
 		{
 			continue;
@@ -1568,10 +1568,10 @@ BOOL LLInventoryCategory::importLegacyStream(std::istream& input_stream)
 	while(input_stream.good())
 	{
 		input_stream.getline(buffer, MAX_STRING);
-		sscanf(	/* Flawfinder: ignore */
+		if (sscanf(
 			buffer,
 			" %254s %254s",
-			keyword, valuestr);
+			keyword, valuestr) < 1) continue;
 		if(0 == strcmp("{",keyword))
 		{
 			continue;

indra/llmath/llcoordframe.cpp

@@ -2,31 +2,25 @@
  * @file llcoordframe.cpp
  * @brief LLCoordFrame class implementation.
  *
- * $LicenseInfo:firstyear=2000&license=viewergpl$
+ * $LicenseInfo:firstyear=2000&license=viewerlgpl$
- * 
- * Copyright (c) 2000-2009, Linden Research, Inc.
- * 
  * Second Life Viewer Source Code
- * The source code in this file ("Source Code") is provided by Linden Lab
+ * Copyright (C) 2010, Linden Research, Inc.
- * to you under the terms of the GNU General Public License, version 2.0
+ *
- * ("GPL"), unless you have obtained a separate licensing agreement
+ * This library is free software; you can redistribute it and/or
- * ("Other License"), formally executed by you and Linden Lab.  Terms of
+ * modify it under the terms of the GNU Lesser General Public
- * the GPL can be found in doc/GPL-license.txt in this distribution, or
+ * License as published by the Free Software Foundation;
- * online at http://secondlifegrid.net/programs/open_source/licensing/gplv2
+ * version 2.1 of the License only.
  * 
- * There are special exceptions to the terms and conditions of the GPL as
+ * This library is distributed in the hope that it will be useful,
- * it is applied to this Source Code. View the full text of the exception
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
- * in the file doc/FLOSS-exception.txt in this software distribution, or
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
- * online at
+ * Lesser General Public License for more details.
- * http://secondlifegrid.net/programs/open_source/licensing/flossexception
  * 
- * By copying, modifying or distributing this software, you acknowledge
+ * You should have received a copy of the GNU Lesser General Public
- * that you have read and understood your obligations described above,
+ * License along with this library; if not, write to the Free Software
- * and agree to abide by those obligations.
+ * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA  02110-1301  USA
  * 
- * ALL LINDEN LAB SOURCE CODE IS PROVIDED "AS IS." LINDEN LAB MAKES NO
+ * Linden Research, Inc., 945 Battery Street, San Francisco, CA  94111  USA
- * WARRANTIES, EXPRESS, IMPLIED OR OTHERWISE, REGARDING ITS ACCURACY,
- * COMPLETENESS OR PERFORMANCE.
  * $/LicenseInfo$
  */
 

indra/llmath/llcoordframe.h

@@ -2,31 +2,25 @@
  * @file llcoordframe.h
  * @brief LLCoordFrame class header file.
  *
- * $LicenseInfo:firstyear=2000&license=viewergpl$
+ * $LicenseInfo:firstyear=2000&license=viewerlgpl$
- * 
- * Copyright (c) 2000-2009, Linden Research, Inc.
- * 
  * Second Life Viewer Source Code
- * The source code in this file ("Source Code") is provided by Linden Lab
+ * Copyright (C) 2010, Linden Research, Inc.
- * to you under the terms of the GNU General Public License, version 2.0
- * ("GPL"), unless you have obtained a separate licensing agreement
- * ("Other License"), formally executed by you and Linden Lab.  Terms of
- * the GPL can be found in doc/GPL-license.txt in this distribution, or
- * online at http://secondlifegrid.net/programs/open_source/licensing/gplv2
  * 
- * There are special exceptions to the terms and conditions of the GPL as
+ * This library is free software; you can redistribute it and/or
- * it is applied to this Source Code. View the full text of the exception
+ * modify it under the terms of the GNU Lesser General Public
- * in the file doc/FLOSS-exception.txt in this software distribution, or
+ * License as published by the Free Software Foundation;
- * online at
+ * version 2.1 of the License only.
- * http://secondlifegrid.net/programs/open_source/licensing/flossexception
  * 
- * By copying, modifying or distributing this software, you acknowledge
+ * This library is distributed in the hope that it will be useful,
- * that you have read and understood your obligations described above,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
- * and agree to abide by those obligations.
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * Lesser General Public License for more details.
  * 
- * ALL LINDEN LAB SOURCE CODE IS PROVIDED "AS IS." LINDEN LAB MAKES NO
+ * You should have received a copy of the GNU Lesser General Public
- * WARRANTIES, EXPRESS, IMPLIED OR OTHERWISE, REGARDING ITS ACCURACY,
+ * License along with this library; if not, write to the Free Software
- * COMPLETENESS OR PERFORMANCE.
+ * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA  02110-1301  USA
+ *
+ * Linden Research, Inc., 945 Battery Street, San Francisco, CA  94111  USA
  * $/LicenseInfo$
  */
 

indra/llmath/llinterp.h

@@ -1,31 +1,25 @@
 /** 
  * @file llinterp.h
  *
- * $LicenseInfo:firstyear=2001&license=viewergpl$
+ * $LicenseInfo:firstyear=2001&license=viewerlgpl$
- * 
- * Copyright (c) 2001-2009, Linden Research, Inc.
- * 
  * Second Life Viewer Source Code
- * The source code in this file ("Source Code") is provided by Linden Lab
+ * Copyright (C) 2010, Linden Research, Inc.
- * to you under the terms of the GNU General Public License, version 2.0
- * ("GPL"), unless you have obtained a separate licensing agreement
- * ("Other License"), formally executed by you and Linden Lab.  Terms of
- * the GPL can be found in doc/GPL-license.txt in this distribution, or
- * online at http://secondlifegrid.net/programs/open_source/licensing/gplv2
  * 
- * There are special exceptions to the terms and conditions of the GPL as
+ * This library is free software; you can redistribute it and/or
- * it is applied to this Source Code. View the full text of the exception
+ * modify it under the terms of the GNU Lesser General Public
- * in the file doc/FLOSS-exception.txt in this software distribution, or
+ * License as published by the Free Software Foundation;
- * online at
+ * version 2.1 of the License only.
- * http://secondlifegrid.net/programs/open_source/licensing/flossexception
  * 
- * By copying, modifying or distributing this software, you acknowledge
+ * This library is distributed in the hope that it will be useful,
- * that you have read and understood your obligations described above,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
- * and agree to abide by those obligations.
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * Lesser General Public License for more details.
+ *
+ * You should have received a copy of the GNU Lesser General Public
+ * License along with this library; if not, write to the Free Software
+ * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA  02110-1301  USA
  * 
- * ALL LINDEN LAB SOURCE CODE IS PROVIDED "AS IS." LINDEN LAB MAKES NO
+ * Linden Research, Inc., 945 Battery Street, San Francisco, CA  94111  USA
- * WARRANTIES, EXPRESS, IMPLIED OR OTHERWISE, REGARDING ITS ACCURACY,
- * COMPLETENESS OR PERFORMANCE.
  * $/LicenseInfo$
  */
 

indra/llmath/llline.cpp

@@ -3,31 +3,25 @@
  * @author Andrew Meadows
  * @brief Simple line class that can compute nearest approach between two lines
  *
- * $LicenseInfo:firstyear=2006&license=viewergpl$
+ * $LicenseInfo:firstyear=2006&license=viewerlgpl$
- * 
- * Copyright (c) 2006-2009, Linden Research, Inc.
- * 
  * Second Life Viewer Source Code
- * The source code in this file ("Source Code") is provided by Linden Lab
+ * Copyright (C) 2010, Linden Research, Inc.
- * to you under the terms of the GNU General Public License, version 2.0
+ *
- * ("GPL"), unless you have obtained a separate licensing agreement
+ * This library is free software; you can redistribute it and/or
- * ("Other License"), formally executed by you and Linden Lab.  Terms of
+ * modify it under the terms of the GNU Lesser General Public
- * the GPL can be found in doc/GPL-license.txt in this distribution, or
+ * License as published by the Free Software Foundation;
- * online at http://secondlifegrid.net/programs/open_source/licensing/gplv2
+ * version 2.1 of the License only.
  * 
- * There are special exceptions to the terms and conditions of the GPL as
+ * This library is distributed in the hope that it will be useful,
- * it is applied to this Source Code. View the full text of the exception
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
- * in the file doc/FLOSS-exception.txt in this software distribution, or
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
- * online at
+ * Lesser General Public License for more details.
- * http://secondlifegrid.net/programs/open_source/licensing/flossexception
  * 
- * By copying, modifying or distributing this software, you acknowledge
+ * You should have received a copy of the GNU Lesser General Public
- * that you have read and understood your obligations described above,
+ * License along with this library; if not, write to the Free Software
- * and agree to abide by those obligations.
+ * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA  02110-1301  USA
  * 
- * ALL LINDEN LAB SOURCE CODE IS PROVIDED "AS IS." LINDEN LAB MAKES NO
+ * Linden Research, Inc., 945 Battery Street, San Francisco, CA  94111  USA
- * WARRANTIES, EXPRESS, IMPLIED OR OTHERWISE, REGARDING ITS ACCURACY,
- * COMPLETENESS OR PERFORMANCE.
  * $/LicenseInfo$
  */
 

indra/llmath/llline.h

@@ -4,31 +4,25 @@
  * @author Andrew Meadows
  * @brief Simple line for computing nearest approach between two infinite lines
  *
- * $LicenseInfo:firstyear=2006&license=viewergpl$
+ * $LicenseInfo:firstyear=2006&license=viewerlgpl$
- * 
- * Copyright (c) 2006-2009, Linden Research, Inc.
- * 
  * Second Life Viewer Source Code
- * The source code in this file ("Source Code") is provided by Linden Lab
+ * Copyright (C) 2010, Linden Research, Inc.
- * to you under the terms of the GNU General Public License, version 2.0
- * ("GPL"), unless you have obtained a separate licensing agreement
- * ("Other License"), formally executed by you and Linden Lab.  Terms of
- * the GPL can be found in doc/GPL-license.txt in this distribution, or
- * online at http://secondlifegrid.net/programs/open_source/licensing/gplv2
  * 
- * There are special exceptions to the terms and conditions of the GPL as
+ * This library is free software; you can redistribute it and/or
- * it is applied to this Source Code. View the full text of the exception
+ * modify it under the terms of the GNU Lesser General Public
- * in the file doc/FLOSS-exception.txt in this software distribution, or
+ * License as published by the Free Software Foundation;
- * online at
+ * version 2.1 of the License only.
- * http://secondlifegrid.net/programs/open_source/licensing/flossexception
  * 
- * By copying, modifying or distributing this software, you acknowledge
+ * This library is distributed in the hope that it will be useful,
- * that you have read and understood your obligations described above,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
- * and agree to abide by those obligations.
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * Lesser General Public License for more details.
  * 
- * ALL LINDEN LAB SOURCE CODE IS PROVIDED "AS IS." LINDEN LAB MAKES NO
+ * You should have received a copy of the GNU Lesser General Public
- * WARRANTIES, EXPRESS, IMPLIED OR OTHERWISE, REGARDING ITS ACCURACY,
+ * License along with this library; if not, write to the Free Software
- * COMPLETENESS OR PERFORMANCE.
+ * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA  02110-1301  USA
+ *
+ * Linden Research, Inc., 945 Battery Street, San Francisco, CA  94111  USA
  * $/LicenseInfo$
  */
 

indra/llmath/llmath.h

@@ -30,6 +30,7 @@
 #include <cmath>
 #include <cstdlib>
 #include <vector>
+#include <limits>
 #include "lldefs.h"
 //#include "llstl.h" // *TODO: Remove when LLString is gone
 //#include "llstring.h" // *TODO: Remove when LLString is gone
@@ -72,15 +73,21 @@ const F32	F_E			= 2.71828182845904523536f;
 const F32	F_SQRT2		= 1.4142135623730950488016887242097f;
 const F32	F_SQRT3		= 1.73205080756888288657986402541f;
 const F32	OO_SQRT2	= 0.7071067811865475244008443621049f;
+const F32	OO_SQRT3	= 0.577350269189625764509f;
 const F32	DEG_TO_RAD	= 0.017453292519943295769236907684886f;
 const F32	RAD_TO_DEG	= 57.295779513082320876798154814105f;
 const F32	F_APPROXIMATELY_ZERO = 0.00001f;
+const F32	F_LN10		= 2.3025850929940456840179914546844f;
+const F32	OO_LN10		= 0.43429448190325182765112891891661f;
 const F32	F_LN2		= 0.69314718056f;
 const F32	OO_LN2		= 1.4426950408889634073599246810019f;
 
 const F32	F_ALMOST_ZERO	= 0.0001f;
 const F32	F_ALMOST_ONE	= 1.0f - F_ALMOST_ZERO;
 
+const F32	GIMBAL_THRESHOLD = 0.000436f; // sets the gimballock threshold 0.025 away from +/-90 degrees
+// formula: GIMBAL_THRESHOLD = sin(DEG_TO_RAD * gimbal_threshold_angle);
+
 // BUG: Eliminate in favor of F_APPROXIMATELY_ZERO above?
 const F32 FP_MAG_THRESHOLD = 0.0000001f;
 
@@ -111,6 +118,12 @@ inline bool is_approx_zero( F32 f ) { return (-F_APPROXIMATELY_ZERO < f) && (f <
 // WARNING: Infinity is comparable with F32_MAX and negative 
 // infinity is comparable with F32_MIN
 
+// handles negative and positive zeros
+inline bool is_zero(F32 x)
+{
+	return (*(U32*)(&x) & 0x7fffffff) == 0;
+}
+
 inline bool is_approx_equal(F32 x, F32 y)
 {
 	const S32 COMPARE_MANTISSA_UP_TO_BIT = 0x02;

indra/llmath/llmatrix3a.cpp

@@ -24,7 +24,6 @@
  * $/LicenseInfo$
  */
 
-#include "sys.h"
 #include "llmath.h"
 
 static LL_ALIGN_16(const F32 M_IDENT_3A[12]) = 

indra/llmath/llmatrix3a.h

@@ -40,6 +40,7 @@
 
 // LLMatrix3a is the base class for LLRotation, which should be used instead any time you're dealing with a 
 // rotation matrix.
+LL_ALIGN_PREFIX(16)
 class LLMatrix3a
 {
 public:
@@ -113,8 +114,9 @@ protected:
 
 	LL_ALIGN_16(LLVector4a mColumns[3]);
 
-};
+} LL_ALIGN_POSTFIX(16);
 
+LL_ALIGN_PREFIX(16)
 class LLRotation : public LLMatrix3a
 {
 public:
@@ -123,6 +125,6 @@ public:
 	
 	// Returns true if this rotation is orthonormal with det ~= 1
 	inline bool isOkRotation() const;		
-};
+} LL_ALIGN_POSTFIX(16);
 
 #endif

indra/llmath/llmatrix4a.h

@@ -31,10 +31,72 @@
 #include "m4math.h"
 #include "m3math.h"
 
+LL_ALIGN_PREFIX(16)
 class LLMatrix4a
 {
-public:
+private:
 	LL_ALIGN_16(LLVector4a mMatrix[4]);
+public:
+	enum
+	{
+		ROW_FWD = 0,
+		ROW_LEFT,
+		ROW_UP,
+		ROW_TRANS
+	};
+
+	void* operator new(size_t size)
+	{
+		return ll_aligned_malloc_16(size);
+	}
+
+	void operator delete(void* ptr)
+	{
+		ll_aligned_free_16(ptr);
+	}
+
+	LLMatrix4a()
+	{}
+	LLMatrix4a(const LLQuad& q1,const LLQuad& q2,const LLQuad& q3,const LLQuad& q4)
+	{
+		mMatrix[0] = q1;
+		mMatrix[1] = q2;
+		mMatrix[2] = q3;
+		mMatrix[3] = q4;
+	}
+	LLMatrix4a(const LLQuaternion2& quat)
+	{
+		const LLVector4a& xyzw = quat.getVector4a(); 
+		LLVector4a nyxwz = _mm_shuffle_ps(xyzw, xyzw, _MM_SHUFFLE(2,3,0,1));
+		nyxwz.negate();
+	
+		const LLVector4a xnyynx = _mm_unpacklo_ps(xyzw,nyxwz);
+		const LLVector4a znwwnz = _mm_unpackhi_ps(xyzw,nyxwz);
+
+		LLMatrix4a mata;
+		mata.setRow<0>(_mm_shuffle_ps(xyzw, xnyynx, _MM_SHUFFLE(0,1,2,3)));
+		mata.setRow<1>(_mm_shuffle_ps(znwwnz, xyzw, _MM_SHUFFLE(1,0,2,3)));
+		mata.setRow<2>(_mm_shuffle_ps(xnyynx, xyzw, _MM_SHUFFLE(2,3,3,2)));
+		mata.setRow<3>(_mm_shuffle_ps(xnyynx, znwwnz, _MM_SHUFFLE(2,3,1,3)));
+
+		LLMatrix4a matb;
+		matb.setRow<0>(_mm_shuffle_ps(xyzw, xnyynx, _MM_SHUFFLE(3,1,2,3)));
+		matb.setRow<1>(_mm_shuffle_ps(znwwnz, xnyynx, _MM_SHUFFLE(1,0,2,3)));
+		matb.setRow<2>(_mm_shuffle_ps(xnyynx, znwwnz, _MM_SHUFFLE(3,2,3,2)));
+		matb.setRow<3>(xyzw);
+
+		setMul(matb,mata);
+	}
+
+	inline F32* getF32ptr()
+	{
+		return mMatrix[0].getF32ptr();
+	}
+	
+	inline const F32* getF32ptr() const
+	{
+		return mMatrix[0].getF32ptr();
+	}
 
 	inline void clear()
 	{
@@ -44,13 +106,21 @@ public:
 		mMatrix[3].clear();
 	}
 
+	inline void setIdentity()
+	{
+		static __m128 ones = _mm_set_ps(1.f,0.f,0.f,1.f);
+		mMatrix[0] = _mm_movelh_ps(ones,_mm_setzero_ps());
+		mMatrix[1] = _mm_movehl_ps(_mm_setzero_ps(),ones);
+		mMatrix[2] = _mm_movelh_ps(_mm_setzero_ps(),ones);
+		mMatrix[3] = _mm_movehl_ps(ones,_mm_setzero_ps());
+	}
+
 	inline void loadu(const LLMatrix4& src)
 	{
-		mMatrix[0] = _mm_loadu_ps(src.mMatrix[0]);
+		mMatrix[0].loadua(src.mMatrix[0]);
-		mMatrix[1] = _mm_loadu_ps(src.mMatrix[1]);
+		mMatrix[1].loadua(src.mMatrix[1]);
-		mMatrix[2] = _mm_loadu_ps(src.mMatrix[2]);
+		mMatrix[2].loadua(src.mMatrix[2]);
-		mMatrix[3] = _mm_loadu_ps(src.mMatrix[3]);
+		mMatrix[3].loadua(src.mMatrix[3]);
-		
 	}
 
 	inline void loadu(const LLMatrix3& src)
@@ -61,6 +131,14 @@ public:
 		mMatrix[3].set(0,0,0,1.f);
 	}
 
+	inline void loadu(const F32* src)
+	{
+		mMatrix[0].loadua(src+0);
+		mMatrix[1].loadua(src+4);
+		mMatrix[2].loadua(src+8);
+		mMatrix[3].loadua(src+12);
+	}
+
 	inline void add(const LLMatrix4a& rhs)
 	{
 		mMatrix[0].add(rhs.mMatrix[0]);
@@ -69,6 +147,75 @@ public:
 		mMatrix[3].add(rhs.mMatrix[3]);
 	}
 
+	inline void mul(const LLMatrix4a& rhs)
+	{
+		//Not using rotate4 to avoid extra copy of *this.
+		LLVector4a x0,y0,z0,w0;
+		LLVector4a x1,y1,z1,w1;
+		LLVector4a x2,y2,z2,w2;
+		LLVector4a x3,y3,z3,w3;
+
+		//16 shuffles
+		x0.splat<0>(rhs.mMatrix[0]);
+		x1.splat<0>(rhs.mMatrix[1]);
+		x2.splat<0>(rhs.mMatrix[2]);
+		x3.splat<0>(rhs.mMatrix[3]);
+
+		y0.splat<1>(rhs.mMatrix[0]);
+		y1.splat<1>(rhs.mMatrix[1]);
+		y2.splat<1>(rhs.mMatrix[2]);
+		y3.splat<1>(rhs.mMatrix[3]);
+
+		z0.splat<2>(rhs.mMatrix[0]);
+		z1.splat<2>(rhs.mMatrix[1]);
+		z2.splat<2>(rhs.mMatrix[2]);
+		z3.splat<2>(rhs.mMatrix[3]);
+
+		w0.splat<3>(rhs.mMatrix[0]);
+		w1.splat<3>(rhs.mMatrix[1]);
+		w2.splat<3>(rhs.mMatrix[2]);
+		w3.splat<3>(rhs.mMatrix[3]);
+
+		//16 muls
+		x0.mul(mMatrix[0]);
+		x1.mul(mMatrix[0]);
+		x2.mul(mMatrix[0]);
+		x3.mul(mMatrix[0]);
+
+		y0.mul(mMatrix[1]);
+		y1.mul(mMatrix[1]);
+		y2.mul(mMatrix[1]);
+		y3.mul(mMatrix[1]);
+
+		z0.mul(mMatrix[2]);
+		z1.mul(mMatrix[2]);
+		z2.mul(mMatrix[2]);
+		z3.mul(mMatrix[2]);
+
+		w0.mul(mMatrix[3]);
+		w1.mul(mMatrix[3]);
+		w2.mul(mMatrix[3]);
+		w3.mul(mMatrix[3]);
+
+		//12 adds
+		x0.add(y0);
+		z0.add(w0);
+
+		x1.add(y1);
+		z1.add(w1);
+
+		x2.add(y2);
+		z2.add(w2);
+
+		x3.add(y3);
+		z3.add(w3);
+
+		mMatrix[0].setAdd(x0,z0);
+		mMatrix[1].setAdd(x1,z1);
+		mMatrix[2].setAdd(x2,z2);
+		mMatrix[3].setAdd(x3,z3);
+	}
+
 	inline void setRows(const LLVector4a& r0, const LLVector4a& r1, const LLVector4a& r2)
 	{
 		mMatrix[0] = r0;
@@ -76,6 +223,44 @@ public:
 		mMatrix[2] = r2;
 	}
 
+	template<int N>
+	inline void setRow(const LLVector4a& row)
+	{
+		mMatrix[N] = row;
+	}
+
+	template<int N>
+	inline const LLVector4a& getRow() const
+	{
+		return mMatrix[N];
+	}
+
+	template<int N>
+	inline LLVector4a& getRow()
+	{
+		return mMatrix[N];
+	}
+
+	template<int N>
+	inline void setColumn(const LLVector4a& col)
+	{
+		mMatrix[0].copyComponent<N>(col.getScalarAt<0>());
+		mMatrix[1].copyComponent<N>(col.getScalarAt<1>());
+		mMatrix[2].copyComponent<N>(col.getScalarAt<2>());
+		mMatrix[3].copyComponent<N>(col.getScalarAt<3>());
+	}
+
+	template<int N>
+	inline LLVector4a getColumn()
+	{
+		LLVector4a v;
+		v.copyComponent<0>(mMatrix[0].getScalarAt<N>());
+		v.copyComponent<1>(mMatrix[1].getScalarAt<N>());
+		v.copyComponent<2>(mMatrix[2].getScalarAt<N>());
+		v.copyComponent<3>(mMatrix[3].getScalarAt<N>());
+		return v;
+	}
+
 	inline void setMul(const LLMatrix4a& m, const F32 s)
 	{
 		mMatrix[0].setMul(m.mMatrix[0], s);
@@ -84,6 +269,14 @@ public:
 		mMatrix[3].setMul(m.mMatrix[3], s);
 	}
 
+	inline void setMul(const LLMatrix4a& m0, const LLMatrix4a& m1)
+	{
+		m0.rotate4(m1.mMatrix[0],mMatrix[0]);
+		m0.rotate4(m1.mMatrix[1],mMatrix[1]);
+		m0.rotate4(m1.mMatrix[2],mMatrix[2]);
+		m0.rotate4(m1.mMatrix[3],mMatrix[3]);
+	}
+
 	inline void setLerp(const LLMatrix4a& a, const LLMatrix4a& b, F32 w)
 	{
 		LLVector4a d0,d1,d2,d3;
@@ -107,13 +300,14 @@ public:
 	
 	//Singu Note: Don't mess with this. It's intentionally different from LL's. 
 	// Note how res isn't manipulated until the very end.
+	//Fast(er). Treats v[VW] as 0.f
 	inline void rotate(const LLVector4a& v, LLVector4a& res) const
 	{
 		LLVector4a x,y,z;
 
-		x = _mm_shuffle_ps(v, v, _MM_SHUFFLE(0, 0, 0, 0));
+		x.splat<0>(v);
-		y = _mm_shuffle_ps(v, v, _MM_SHUFFLE(1, 1, 1, 1));
+		y.splat<1>(v);
-		z = _mm_shuffle_ps(v, v, _MM_SHUFFLE(2, 2, 2, 2));
+		z.splat<2>(v);
 
 		x.mul(mMatrix[0]);
 		y.mul(mMatrix[1]);
@@ -123,14 +317,15 @@ public:
 		res.setAdd(x,z);
 	}
 	
+	//Proper. v[VW] as v[VW]
 	inline void rotate4(const LLVector4a& v, LLVector4a& res) const
 	{
 		LLVector4a x,y,z,w;
 
-		x = _mm_shuffle_ps(v, v, _MM_SHUFFLE(0, 0, 0, 0));
+		x.splat<0>(v);
-		y = _mm_shuffle_ps(v, v, _MM_SHUFFLE(1, 1, 1, 1));
+		y.splat<1>(v);
-		z = _mm_shuffle_ps(v, v, _MM_SHUFFLE(2, 2, 2, 2));
+		z.splat<2>(v);
-		w = _mm_shuffle_ps(v, v, _MM_SHUFFLE(3, 3, 3, 3));
+		w.splat<3>(v);
 
 		x.mul(mMatrix[0]);
 		y.mul(mMatrix[1]);
@@ -142,14 +337,15 @@ public:
 		res.setAdd(x,z);
 	}
 
+	//Fast(er). Treats v[VW] as 1.f
 	inline void affineTransform(const LLVector4a& v, LLVector4a& res) const
 	{
 		LLVector4a x,y,z;
 
-		x = _mm_shuffle_ps(v, v, _MM_SHUFFLE(0, 0, 0, 0));
+		x.splat<0>(v);
-		y = _mm_shuffle_ps(v, v, _MM_SHUFFLE(1, 1, 1, 1));
+		y.splat<1>(v);
-		z = _mm_shuffle_ps(v, v, _MM_SHUFFLE(2, 2, 2, 2));
+		z.splat<2>(v);
-		
+
 		x.mul(mMatrix[0]);
 		y.mul(mMatrix[1]);
 		z.mul(mMatrix[2]);
@@ -158,6 +354,348 @@ public:
 		z.add(mMatrix[3]);
 		res.setAdd(x,z);
 	}
-};
+
+	inline void perspectiveTransform(const LLVector4a& v, LLVector4a& res) const
+	{
+		LLVector4a x,y,z,s,t,p,q;
+
+		x.splat<0>(v);
+		y.splat<1>(v);
+		z.splat<2>(v);
+
+		s.splat<3>(mMatrix[0]);
+		t.splat<3>(mMatrix[1]);
+		p.splat<3>(mMatrix[2]);
+		q.splat<3>(mMatrix[3]);
+
+		s.mul(x);
+		t.mul(y);
+		p.mul(z);
+		q.add(s);
+		t.add(p);
+		q.add(t);
+
+		x.mul(mMatrix[0]);
+		y.mul(mMatrix[1]);
+		z.mul(mMatrix[2]);
+
+		x.add(y);
+		z.add(mMatrix[3]);
+		res.setAdd(x,z);
+		res.div(q);
+	}
+
+	inline void transpose()
+	{
+		__m128 q1 = _mm_unpackhi_ps(mMatrix[0],mMatrix[1]);
+		__m128 q2 = _mm_unpacklo_ps(mMatrix[0],mMatrix[1]);
+		__m128 q3 = _mm_unpacklo_ps(mMatrix[2],mMatrix[3]);
+		__m128 q4 = _mm_unpackhi_ps(mMatrix[2],mMatrix[3]);
+
+		mMatrix[0] = _mm_movelh_ps(q2,q3);
+		mMatrix[1] = _mm_movehl_ps(q3,q2);
+		mMatrix[2] = _mm_movelh_ps(q1,q4);
+		mMatrix[3] = _mm_movehl_ps(q4,q1);
+	}
+
+//  Following procedure adapted from:
+//		http://software.intel.com/en-us/articles/optimized-matrix-library-for-use-with-the-intel-pentiumr-4-processors-sse2-instructions/
+//
+//  License/Copyright Statement:
+//		
+//			Copyright (c) 2001 Intel Corporation.
+//
+//		Permition is granted to use, copy, distribute and prepare derivative works 
+//		of this library for any purpose and without fee, provided, that the above 
+//		copyright notice and this statement appear in all copies.  
+//		Intel makes no representations about the suitability of this library for 
+//		any purpose, and specifically disclaims all warranties. 
+//		See LEGAL-intel_matrixlib.TXT for all the legal information.
+	inline float invert()
+	{
+		LL_ALIGN_16(const unsigned int Sign_PNNP[4]) = { 0x00000000, 0x80000000, 0x80000000, 0x00000000 };
+
+		// The inverse is calculated using "Divide and Conquer" technique. The 
+		// original matrix is divide into four 2x2 sub-matrices. Since each 
+		// register holds four matrix element, the smaller matrices are 
+		// represented as a registers. Hence we get a better locality of the 
+		// calculations.
+
+		LLVector4a A = _mm_movelh_ps(mMatrix[0], mMatrix[1]),    // the four sub-matrices 
+				B = _mm_movehl_ps(mMatrix[1], mMatrix[0]),
+				C = _mm_movelh_ps(mMatrix[2], mMatrix[3]),
+				D = _mm_movehl_ps(mMatrix[3], mMatrix[2]);
+		LLVector4a iA, iB, iC, iD,					// partial inverse of the sub-matrices
+				DC, AB;
+		LLSimdScalar dA, dB, dC, dD;                 // determinant of the sub-matrices
+		LLSimdScalar det, d, d1, d2;
+		LLVector4a rd;
+
+		//  AB = A# * B
+		AB.setMul(_mm_shuffle_ps(A,A,0x0F), B);
+		AB.sub(_mm_mul_ps(_mm_shuffle_ps(A,A,0xA5), _mm_shuffle_ps(B,B,0x4E)));
+		//  DC = D# * C
+		DC.setMul(_mm_shuffle_ps(D,D,0x0F), C);
+		DC.sub(_mm_mul_ps(_mm_shuffle_ps(D,D,0xA5), _mm_shuffle_ps(C,C,0x4E)));
+
+		//  dA = |A|
+		dA = _mm_mul_ps(_mm_shuffle_ps(A, A, 0x5F),A);
+		dA -= _mm_movehl_ps(dA,dA);
+		//  dB = |B|
+		dB = _mm_mul_ps(_mm_shuffle_ps(B, B, 0x5F),B);
+		dB -= _mm_movehl_ps(dB,dB);
+
+		//  dC = |C|
+		dC = _mm_mul_ps(_mm_shuffle_ps(C, C, 0x5F),C);
+		dC -= _mm_movehl_ps(dC,dC);
+		//  dD = |D|
+		dD = _mm_mul_ps(_mm_shuffle_ps(D, D, 0x5F),D);
+		dD -= _mm_movehl_ps(dD,dD);
+
+		//  d = trace(AB*DC) = trace(A#*B*D#*C)
+		d = _mm_mul_ps(_mm_shuffle_ps(DC,DC,0xD8),AB);
+
+		//  iD = C*A#*B
+		iD.setMul(_mm_shuffle_ps(C,C,0xA0), _mm_movelh_ps(AB,AB));
+		iD.add(_mm_mul_ps(_mm_shuffle_ps(C,C,0xF5), _mm_movehl_ps(AB,AB)));
+		//  iA = B*D#*C
+		iA.setMul(_mm_shuffle_ps(B,B,0xA0), _mm_movelh_ps(DC,DC));
+		iA.add(_mm_mul_ps(_mm_shuffle_ps(B,B,0xF5), _mm_movehl_ps(DC,DC)));
+
+		//  d = trace(AB*DC) = trace(A#*B*D#*C) [continue]
+		d = _mm_add_ps(d, _mm_movehl_ps(d, d));
+		d += _mm_shuffle_ps(d, d, 1);
+		d1 = dA*dD;
+		d2 = dB*dC;
+
+		//  iD = D*|A| - C*A#*B
+		iD.setSub(_mm_mul_ps(D,_mm_shuffle_ps(dA,dA,0)), iD);
+
+		//  iA = A*|D| - B*D#*C;
+		iA.setSub(_mm_mul_ps(A,_mm_shuffle_ps(dD,dD,0)), iA);
+
+		//  det = |A|*|D| + |B|*|C| - trace(A#*B*D#*C)
+		det = d1+d2-d;
+
+		__m128 is_zero_mask = _mm_cmpeq_ps(det,_mm_setzero_ps());
+		rd = _mm_div_ss(_mm_set_ss(1.f),_mm_or_ps(_mm_andnot_ps(is_zero_mask, det), _mm_and_ps(is_zero_mask, _mm_set_ss(1.f))));
+#ifdef ZERO_SINGULAR
+		rd = _mm_and_ps(_mm_cmpneq_ss(det,_mm_setzero_ps()), rd);
+#endif
+
+		//  iB = D * (A#B)# = D*B#*A
+		iB.setMul(D, _mm_shuffle_ps(AB,AB,0x33));
+		iB.sub(_mm_mul_ps(_mm_shuffle_ps(D,D,0xB1), _mm_shuffle_ps(AB,AB,0x66)));
+		//  iC = A * (D#C)# = A*C#*D
+		iC.setMul(A, _mm_shuffle_ps(DC,DC,0x33));
+		iC.sub(_mm_mul_ps(_mm_shuffle_ps(A,A,0xB1), _mm_shuffle_ps(DC,DC,0x66)));
+
+		rd = _mm_shuffle_ps(rd,rd,0);
+		rd = _mm_xor_ps(rd, _mm_load_ps((const float*)Sign_PNNP));
+
+		//  iB = C*|B| - D*B#*A
+		iB.setSub(_mm_mul_ps(C,_mm_shuffle_ps(dB,dB,0)), iB);
+
+		//  iC = B*|C| - A*C#*D;
+		iC.setSub(_mm_mul_ps(B,_mm_shuffle_ps(dC,dC,0)), iC);
+
+
+		//  iX = iX / det
+		iA.mul(rd);
+		iB.mul(rd);
+		iC.mul(rd);
+		iD.mul(rd);
+
+		mMatrix[0] = _mm_shuffle_ps(iA,iB,0x77);
+		mMatrix[1] = _mm_shuffle_ps(iA,iB,0x22);
+		mMatrix[2] = _mm_shuffle_ps(iC,iD,0x77);
+		mMatrix[3] = _mm_shuffle_ps(iC,iD,0x22);
+		
+		F32 ret;
+		_mm_store_ss(&ret,det);
+		return ret;
+	}
+
+	//=============Affine transformation matrix only=========================
+
+	//Multiply matrix with a pure translation matrix.
+	inline void applyTranslation_affine(const F32& x, const F32& y, const F32& z)
+	{
+		const LLVector4a xyz0(x,y,z,0);	//load
+		LLVector4a xxxx;
+		xxxx.splat<0>(xyz0);
+		LLVector4a yyyy;
+		yyyy.splat<1>(xyz0);
+		LLVector4a zzzz;
+		zzzz.splat<2>(xyz0);
+
+		LLVector4a sum1;
+		LLVector4a sum2;
+		LLVector4a sum3;
+
+		sum1.setMul(xxxx,mMatrix[0]);
+		sum2.setMul(yyyy,mMatrix[1]);
+		sum3.setMul(zzzz,mMatrix[2]);
+
+		mMatrix[3].add(sum1);
+		mMatrix[3].add(sum2);
+		mMatrix[3].add(sum3);
+	}
+
+	//Multiply matrix with a pure translation matrix.
+	inline void applyTranslation_affine(const LLVector3& trans)
+	{
+		applyTranslation_affine(trans.mV[VX],trans.mV[VY],trans.mV[VZ]);
+	}
+
+	//Multiply matrix with a pure scale matrix.
+	inline void applyScale_affine(const F32& x, const F32& y, const F32& z)
+	{
+		const LLVector4a xyz0(x,y,z,0);	//load
+		LLVector4a xxxx;
+		xxxx.splat<0>(xyz0);
+		LLVector4a yyyy;
+		yyyy.splat<1>(xyz0);
+		LLVector4a zzzz;
+		zzzz.splat<2>(xyz0);
+
+		mMatrix[0].mul(xxxx);
+		mMatrix[1].mul(yyyy);
+		mMatrix[2].mul(zzzz);
+	}
+
+	//Multiply matrix with a pure scale matrix.
+	inline void applyScale_affine(const LLVector3& scale)
+	{
+		applyScale_affine(scale.mV[VX],scale.mV[VY],scale.mV[VZ]);
+	}
+
+	//Multiply matrix with a pure scale matrix.
+	inline void applyScale_affine(const F32& s)
+	{
+		const LLVector4a scale(s);	//load
+		mMatrix[0].mul(scale);
+		mMatrix[1].mul(scale);
+		mMatrix[2].mul(scale);
+	}
+
+	//Direct addition to row3.
+	inline void translate_affine(const LLVector3& trans)
+	{
+		LLVector4a translation;
+		translation.load3(trans.mV);
+		mMatrix[3].add(translation);
+	}
+
+	//Direct assignment of row3.
+	inline void setTranslate_affine(const LLVector3& trans)
+	{
+		static const LLVector4Logical mask = _mm_load_ps((F32*)&S_V4LOGICAL_MASK_TABLE[3*4]);
+
+		LLVector4a translation;
+		translation.load3(trans.mV);
+		
+		mMatrix[3].setSelectWithMask(mask,mMatrix[3],translation);
+	}
+
+	inline void mul_affine(const LLMatrix4a& rhs)
+	{
+		LLVector4a x0,y0,z0;
+		LLVector4a x1,y1,z1;
+		LLVector4a x2,y2,z2;
+		LLVector4a x3,y3,z3;
+
+		//12 shuffles
+		x0.splat<0>(rhs.mMatrix[0]);
+		x1.splat<0>(rhs.mMatrix[1]);
+		x2.splat<0>(rhs.mMatrix[2]);
+		x3.splat<0>(rhs.mMatrix[3]);
+
+		y0.splat<1>(rhs.mMatrix[0]);
+		y1.splat<1>(rhs.mMatrix[1]);
+		y2.splat<1>(rhs.mMatrix[2]);
+		y3.splat<1>(rhs.mMatrix[3]);
+
+		z0.splat<2>(rhs.mMatrix[0]);
+		z1.splat<2>(rhs.mMatrix[1]);
+		z2.splat<2>(rhs.mMatrix[2]);
+		z3.splat<2>(rhs.mMatrix[3]);
+
+		//12 muls
+		x0.mul(mMatrix[0]);
+		x1.mul(mMatrix[0]);
+		x2.mul(mMatrix[0]);
+		x3.mul(mMatrix[0]);
+
+		y0.mul(mMatrix[1]);
+		y1.mul(mMatrix[1]);
+		y2.mul(mMatrix[1]);
+		y3.mul(mMatrix[1]);
+
+		z0.mul(mMatrix[2]);
+		z1.mul(mMatrix[2]);
+		z2.mul(mMatrix[2]);
+		z3.mul(mMatrix[2]);
+
+		//9 adds
+		x0.add(y0);
+
+		x1.add(y1);
+
+		x2.add(y2);
+
+		x3.add(y3);
+		z3.add(mMatrix[3]);
+
+		mMatrix[0].setAdd(x0,z0);
+		mMatrix[1].setAdd(x1,z1);
+		mMatrix[2].setAdd(x2,z2);
+		mMatrix[3].setAdd(x3,z3);
+	}
+
+	inline void extractRotation_affine()
+	{
+		static const LLVector4Logical mask = _mm_load_ps((F32*)&S_V4LOGICAL_MASK_TABLE[3*4]);
+		mMatrix[0].setSelectWithMask(mask,_mm_setzero_ps(),mMatrix[0]);
+		mMatrix[1].setSelectWithMask(mask,_mm_setzero_ps(),mMatrix[1]);
+		mMatrix[2].setSelectWithMask(mask,_mm_setzero_ps(),mMatrix[2]);
+		mMatrix[3].setSelectWithMask(mask,LLVector4a(1.f),_mm_setzero_ps());
+	}
+
+	//======================Logic====================
+private:
+	template<bool mins> inline void init_foos(LLMatrix4a& foos) const
+	{
+		static bool done(false);
+		if (done) return;
+		const LLVector4a delta(0.0001f);
+		foos.setIdentity();
+		foos.getRow<0>().sub(delta);
+		foos.getRow<1>().sub(delta);
+		foos.getRow<2>().sub(delta);
+		foos.getRow<3>().sub(delta);
+		done = true;
+	}
+
+public:
+	inline bool isIdentity() const
+	{
+		static LLMatrix4a mins;
+		static LLMatrix4a maxs;
+
+		init_foos<false>(mins);
+		init_foos<true>(maxs);
+
+		LLVector4a mask1 = _mm_and_ps(_mm_cmpgt_ps(mMatrix[0],mins.getRow<0>()), _mm_cmplt_ps(mMatrix[0],maxs.getRow<0>()));
+		LLVector4a mask2 = _mm_and_ps(_mm_cmpgt_ps(mMatrix[1],mins.getRow<1>()), _mm_cmplt_ps(mMatrix[1],maxs.getRow<1>()));
+		LLVector4a mask3 = _mm_and_ps(_mm_cmpgt_ps(mMatrix[2],mins.getRow<2>()), _mm_cmplt_ps(mMatrix[2],maxs.getRow<2>()));
+		LLVector4a mask4 = _mm_and_ps(_mm_cmpgt_ps(mMatrix[3],mins.getRow<3>()), _mm_cmplt_ps(mMatrix[3],maxs.getRow<3>()));
+
+		mask1 = _mm_and_ps(mask1,mask2);
+		mask2 = _mm_and_ps(mask3,mask4);
+
+		return _mm_movemask_epi8(_mm_castps_si128(_mm_and_ps(mask1, mask2))) == 0xFFFF;
+	}
+} LL_ALIGN_POSTFIX(16);
 
 #endif

indra/llmath/llmodularmath.cpp

@@ -2,36 +2,28 @@
  * @file llmodularmath.cpp
  * @brief LLModularMath class implementation
  *
- * $LicenseInfo:firstyear=2001&license=viewergpl$
+ * $LicenseInfo:firstyear=2001&license=viewerlgpl$
- * 
- * Copyright (c) 2001-2010, Linden Research, Inc.
- * 
  * Second Life Viewer Source Code
- * The source code in this file ("Source Code") is provided by Linden Lab
+ * Copyright (C) 2010, Linden Research, Inc.
- * to you under the terms of the GNU General Public License, version 2.0
+ *
- * ("GPL"), unless you have obtained a separate licensing agreement
+ * This library is free software; you can redistribute it and/or
- * ("Other License"), formally executed by you and Linden Lab.  Terms of
+ * modify it under the terms of the GNU Lesser General Public
- * the GPL can be found in doc/GPL-license.txt in this distribution, or
+ * License as published by the Free Software Foundation;
- * online at http://secondlife.com/developers/opensource/gplv2
+ * version 2.1 of the License only.
  * 
- * There are special exceptions to the terms and conditions of the GPL as
+ * This library is distributed in the hope that it will be useful,
- * it is applied to this Source Code. View the full text of the exception
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
- * in the file doc/FLOSS-exception.txt in this software distribution, or
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
- * online at
+ * Lesser General Public License for more details.
- * http://secondlife.com/developers/opensource/flossexception
  * 
- * By copying, modifying or distributing this software, you acknowledge
+ * You should have received a copy of the GNU Lesser General Public
- * that you have read and understood your obligations described above,
+ * License along with this library; if not, write to the Free Software
- * and agree to abide by those obligations.
+ * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA  02110-1301  USA
  * 
- * ALL LINDEN LAB SOURCE CODE IS PROVIDED "AS IS." LINDEN LAB MAKES NO
+ * Linden Research, Inc., 945 Battery Street, San Francisco, CA  94111  USA
- * WARRANTIES, EXPRESS, IMPLIED OR OTHERWISE, REGARDING ITS ACCURACY,
- * COMPLETENESS OR PERFORMANCE.
  * $/LicenseInfo$
- * 
  */
 
-
 #include "linden_common.h"
 
 // implementation is all in the header, this include dep ensures the unit test is rerun if the implementation changes.

indra/llmath/llmodularmath.h

@@ -2,31 +2,25 @@
  * @file llmodularmath.h
  * @brief Useful modular math functions.
  *
- * $LicenseInfo:firstyear=2008&license=viewergpl$
+ * $LicenseInfo:firstyear=2008&license=viewerlgpl$
- * 
- * Copyright (c) 2008-2009, Linden Research, Inc.
- * 
  * Second Life Viewer Source Code
- * The source code in this file ("Source Code") is provided by Linden Lab
+ * Copyright (C) 2010, Linden Research, Inc.
- * to you under the terms of the GNU General Public License, version 2.0
- * ("GPL"), unless you have obtained a separate licensing agreement
- * ("Other License"), formally executed by you and Linden Lab.  Terms of
- * the GPL can be found in doc/GPL-license.txt in this distribution, or
- * online at http://secondlifegrid.net/programs/open_source/licensing/gplv2
  * 
- * There are special exceptions to the terms and conditions of the GPL as
+ * This library is free software; you can redistribute it and/or
- * it is applied to this Source Code. View the full text of the exception
+ * modify it under the terms of the GNU Lesser General Public
- * in the file doc/FLOSS-exception.txt in this software distribution, or
+ * License as published by the Free Software Foundation;
- * online at
+ * version 2.1 of the License only.
- * http://secondlifegrid.net/programs/open_source/licensing/flossexception
  * 
- * By copying, modifying or distributing this software, you acknowledge
+ * This library is distributed in the hope that it will be useful,
- * that you have read and understood your obligations described above,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
- * and agree to abide by those obligations.
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * Lesser General Public License for more details.
  * 
- * ALL LINDEN LAB SOURCE CODE IS PROVIDED "AS IS." LINDEN LAB MAKES NO
+ * You should have received a copy of the GNU Lesser General Public
- * WARRANTIES, EXPRESS, IMPLIED OR OTHERWISE, REGARDING ITS ACCURACY,
+ * License along with this library; if not, write to the Free Software
- * COMPLETENESS OR PERFORMANCE.
+ * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA  02110-1301  USA
+ *
+ * Linden Research, Inc., 945 Battery Street, San Francisco, CA  94111  USA
  * $/LicenseInfo$
  */
 

indra/llmath/lloctree.h

@@ -932,10 +932,10 @@ protected:
 		MIN = 3
 	} eDName;
 
-	LLVector4a mCenter;
+	LL_ALIGN_16(LLVector4a mCenter);
-	LLVector4a mSize;
+	LL_ALIGN_16(LLVector4a mSize);
-	LLVector4a mMax;
+	LL_ALIGN_16(LLVector4a mMax);
-	LLVector4a mMin;
+	LL_ALIGN_16(LLVector4a mMin);
 	
 	oct_node* mParent;
 	U8 mOctant;
@@ -964,6 +964,26 @@ public:
 	:	BaseType(center, size, parent)
 	{
 	}
+
+#ifdef LL_OCTREE_POOLS
+	void* operator new(size_t size)
+	{
+		return LLOctreeNode<T>::getPool(size).malloc();
+	}
+	void operator delete(void* ptr)
+	{
+		LLOctreeNode<T>::getPool(sizeof(LLOctreeNode<T>)).free(ptr);
+	}
+#else
+	void* operator new(size_t size)
+	{
+		return ll_aligned_malloc_16(size);
+	}
+	void operator delete(void* ptr)
+	{
+		ll_aligned_free_16(ptr);
+	}
+#endif
 	
 	bool balance()
 	{	

indra/llmath/llperlin.cpp

@@ -1,31 +1,25 @@
 /** 
  * @file llperlin.cpp
  *
- * $LicenseInfo:firstyear=2001&license=viewergpl$
+ * $LicenseInfo:firstyear=2001&license=viewerlgpl$
- * 
- * Copyright (c) 2001-2009, Linden Research, Inc.
- * 
  * Second Life Viewer Source Code
- * The source code in this file ("Source Code") is provided by Linden Lab
+ * Copyright (C) 2010, Linden Research, Inc.
- * to you under the terms of the GNU General Public License, version 2.0
- * ("GPL"), unless you have obtained a separate licensing agreement
- * ("Other License"), formally executed by you and Linden Lab.  Terms of
- * the GPL can be found in doc/GPL-license.txt in this distribution, or
- * online at http://secondlifegrid.net/programs/open_source/licensing/gplv2
  * 
- * There are special exceptions to the terms and conditions of the GPL as
+ * This library is free software; you can redistribute it and/or
- * it is applied to this Source Code. View the full text of the exception
+ * modify it under the terms of the GNU Lesser General Public
- * in the file doc/FLOSS-exception.txt in this software distribution, or
+ * License as published by the Free Software Foundation;
- * online at
+ * version 2.1 of the License only.
- * http://secondlifegrid.net/programs/open_source/licensing/flossexception
  * 
- * By copying, modifying or distributing this software, you acknowledge
+ * This library is distributed in the hope that it will be useful,
- * that you have read and understood your obligations described above,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
- * and agree to abide by those obligations.
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * Lesser General Public License for more details.
  * 
- * ALL LINDEN LAB SOURCE CODE IS PROVIDED "AS IS." LINDEN LAB MAKES NO
+ * You should have received a copy of the GNU Lesser General Public
- * WARRANTIES, EXPRESS, IMPLIED OR OTHERWISE, REGARDING ITS ACCURACY,
+ * License along with this library; if not, write to the Free Software
- * COMPLETENESS OR PERFORMANCE.
+ * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA  02110-1301  USA
+ *
+ * Linden Research, Inc., 945 Battery Street, San Francisco, CA  94111  USA
  * $/LicenseInfo$
  */
 

indra/llmath/llperlin.h

@@ -1,31 +1,25 @@
 /** 
  * @file llperlin.h
  *
- * $LicenseInfo:firstyear=2001&license=viewergpl$
+ * $LicenseInfo:firstyear=2001&license=viewerlgpl$
- * 
- * Copyright (c) 2001-2009, Linden Research, Inc.
- * 
  * Second Life Viewer Source Code
- * The source code in this file ("Source Code") is provided by Linden Lab
+ * Copyright (C) 2010, Linden Research, Inc.
- * to you under the terms of the GNU General Public License, version 2.0
- * ("GPL"), unless you have obtained a separate licensing agreement
- * ("Other License"), formally executed by you and Linden Lab.  Terms of
- * the GPL can be found in doc/GPL-license.txt in this distribution, or
- * online at http://secondlifegrid.net/programs/open_source/licensing/gplv2
  * 
- * There are special exceptions to the terms and conditions of the GPL as
+ * This library is free software; you can redistribute it and/or
- * it is applied to this Source Code. View the full text of the exception
+ * modify it under the terms of the GNU Lesser General Public
- * in the file doc/FLOSS-exception.txt in this software distribution, or
+ * License as published by the Free Software Foundation;
- * online at
+ * version 2.1 of the License only.
- * http://secondlifegrid.net/programs/open_source/licensing/flossexception
  * 
- * By copying, modifying or distributing this software, you acknowledge
+ * This library is distributed in the hope that it will be useful,
- * that you have read and understood your obligations described above,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
- * and agree to abide by those obligations.
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * Lesser General Public License for more details.
  * 
- * ALL LINDEN LAB SOURCE CODE IS PROVIDED "AS IS." LINDEN LAB MAKES NO
+ * You should have received a copy of the GNU Lesser General Public
- * WARRANTIES, EXPRESS, IMPLIED OR OTHERWISE, REGARDING ITS ACCURACY,
+ * License along with this library; if not, write to the Free Software
- * COMPLETENESS OR PERFORMANCE.
+ * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA  02110-1301  USA
+ *
+ * Linden Research, Inc., 945 Battery Street, San Francisco, CA  94111  USA
  * $/LicenseInfo$
  */
 

indra/llmath/llplane.h

@@ -101,7 +101,7 @@ public:
 	}
 
 private:
-	LLVector4a mV;
+	LL_ALIGN_16(LLVector4a mV);
 } LL_ALIGN_POSTFIX(16);
 
 

indra/llmath/llquaternion.cpp

@@ -58,34 +58,40 @@ LLQuaternion::LLQuaternion(const LLMatrix3 &mat)
 
 LLQuaternion::LLQuaternion(F32 angle, const LLVector4 &vec)
 {
-	LLVector3 v(vec.mV[VX], vec.mV[VY], vec.mV[VZ]);
+	F32 mag = sqrtf(vec.mV[VX] * vec.mV[VX] + vec.mV[VY] * vec.mV[VY] + vec.mV[VZ] * vec.mV[VZ]);
-	v.normalize();
+	if (mag > FP_MAG_THRESHOLD)
-
+	{
-	F32 c, s;
+		angle *= 0.5;
-	c = cosf(angle*0.5f);
+		F32 c = cosf(angle);
-	s = sinf(angle*0.5f);
+		F32 s = sinf(angle) / mag;
-
+		mQ[VX] = vec.mV[VX] * s;
-	mQ[VX] = v.mV[VX] * s;
+		mQ[VY] = vec.mV[VY] * s;
-	mQ[VY] = v.mV[VY] * s;
+		mQ[VZ] = vec.mV[VZ] * s;
-	mQ[VZ] = v.mV[VZ] * s;
+		mQ[VW] = c;
-	mQ[VW] = c;
+	}
-	normalize();
+	else
+	{
+		loadIdentity();
+	}
 }
 
 LLQuaternion::LLQuaternion(F32 angle, const LLVector3 &vec)
 {
-	LLVector3 v(vec);
+	F32 mag = sqrtf(vec.mV[VX] * vec.mV[VX] + vec.mV[VY] * vec.mV[VY] + vec.mV[VZ] * vec.mV[VZ]);
-	v.normalize();
+	if (mag > FP_MAG_THRESHOLD)
-
+	{
-	F32 c, s;
+		angle *= 0.5;
-	c = cosf(angle*0.5f);
+		F32 c = cosf(angle);
-	s = sinf(angle*0.5f);
+		F32 s = sinf(angle) / mag;
-
+		mQ[VX] = vec.mV[VX] * s;
-	mQ[VX] = v.mV[VX] * s;
+		mQ[VY] = vec.mV[VY] * s;
-	mQ[VY] = v.mV[VY] * s;
+		mQ[VZ] = vec.mV[VZ] * s;
-	mQ[VZ] = v.mV[VZ] * s;
+		mQ[VW] = c;
-	mQ[VW] = c;
+	}
-	normalize();
+	else
+	{
+		loadIdentity();
+	}
 }
 
 LLQuaternion::LLQuaternion(const LLVector3 &x_axis,
@@ -136,57 +142,61 @@ void	LLQuaternion::quantize8(F32 lower, F32 upper)
 
 const LLQuaternion&	LLQuaternion::setAngleAxis(F32 angle, F32 x, F32 y, F32 z)
 {
-	LLVector3 vec(x, y, z);
+	F32 mag = sqrtf(x * x + y * y + z * z);
-	vec.normalize();
+	if (mag > FP_MAG_THRESHOLD)
-
+	{
-	angle *= 0.5f;
+		angle *= 0.5;
-	F32 c, s;
+		F32 c = cosf(angle);
-	c = cosf(angle);
+		F32 s = sinf(angle) / mag;
-	s = sinf(angle);
+		mQ[VX] = x * s;
-
+		mQ[VY] = y * s;
-	mQ[VX] = vec.mV[VX]*s;
+		mQ[VZ] = z * s;
-	mQ[VY] = vec.mV[VY]*s;
+		mQ[VW] = c;
-	mQ[VZ] = vec.mV[VZ]*s;
+	}
-	mQ[VW] = c;
+	else
-
+	{
-	normalize();
+		loadIdentity();
+	}
 	return (*this);
 }
 
 const LLQuaternion&	LLQuaternion::setAngleAxis(F32 angle, const LLVector3 &vec)
 {
-	LLVector3 v(vec);
+	F32 mag = sqrtf(vec.mV[VX] * vec.mV[VX] + vec.mV[VY] * vec.mV[VY] + vec.mV[VZ] * vec.mV[VZ]);
-	v.normalize();
+	if (mag > FP_MAG_THRESHOLD)
-
+	{
-	angle *= 0.5f;
+		angle *= 0.5;
-	F32 c, s;
+		F32 c = cosf(angle);
-	c = cosf(angle);
+		F32 s = sinf(angle) / mag;
-	s = sinf(angle);
+		mQ[VX] = vec.mV[VX] * s;
-
+		mQ[VY] = vec.mV[VY] * s;
-	mQ[VX] = v.mV[VX]*s;
+		mQ[VZ] = vec.mV[VZ] * s;
-	mQ[VY] = v.mV[VY]*s;
+		mQ[VW] = c;
-	mQ[VZ] = v.mV[VZ]*s;
+	}
-	mQ[VW] = c;
+	else
-
+	{
-	normalize();
+		loadIdentity();
+	}
 	return (*this);
 }
 
 const LLQuaternion&	LLQuaternion::setAngleAxis(F32 angle, const LLVector4 &vec)
 {
-	LLVector3 v(vec.mV[VX], vec.mV[VY], vec.mV[VZ]);
+	F32 mag = sqrtf(vec.mV[VX] * vec.mV[VX] + vec.mV[VY] * vec.mV[VY] + vec.mV[VZ] * vec.mV[VZ]);
-	v.normalize();
+	if (mag > FP_MAG_THRESHOLD)
-
+	{
-	F32 c, s;
+		angle *= 0.5;
-	c = cosf(angle*0.5f);
+		F32 c = cosf(angle);
-	s = sinf(angle*0.5f);
+		F32 s = sinf(angle) / mag;
-
+		mQ[VX] = vec.mV[VX] * s;
-	mQ[VX] = v.mV[VX]*s;
+		mQ[VY] = vec.mV[VY] * s;
-	mQ[VY] = v.mV[VY]*s;
+		mQ[VZ] = vec.mV[VZ] * s;
-	mQ[VZ] = v.mV[VZ]*s;
+		mQ[VW] = c;
-	mQ[VW] = c;
+	}
-
+	else
-	normalize();
+	{
+		loadIdentity();
+	}
 	return (*this);
 }
 
@@ -219,68 +229,80 @@ const LLQuaternion&	LLQuaternion::set(const LLMatrix4 &mat)
 // deprecated
 const LLQuaternion&	LLQuaternion::setQuat(F32 angle, F32 x, F32 y, F32 z)
 {
-	LLVector3 vec(x, y, z);
+	F32 mag = sqrtf(x * x + y * y + z * z);
-	vec.normalize();
+	if (mag > FP_MAG_THRESHOLD)
-
+	{
-	angle *= 0.5f;
+		angle *= 0.5;
-	F32 c, s;
+		F32 c = cosf(angle);
-	c = cosf(angle);
+		F32 s = sinf(angle) / mag;
-	s = sinf(angle);
+		mQ[VX] = x * s;
-
+		mQ[VY] = y * s;
-	mQ[VX] = vec.mV[VX]*s;
+		mQ[VZ] = z * s;
-	mQ[VY] = vec.mV[VY]*s;
+		mQ[VW] = c;
-	mQ[VZ] = vec.mV[VZ]*s;
+	}
-	mQ[VW] = c;
+	else
-
+	{
-	normalize();
+		loadIdentity();
+	}
 	return (*this);
 }
 
 // deprecated
 const LLQuaternion&	LLQuaternion::setQuat(F32 angle, const LLVector3 &vec)
 {
-	LLVector3 v(vec);
+	F32 mag = sqrtf(vec.mV[VX] * vec.mV[VX] + vec.mV[VY] * vec.mV[VY] + vec.mV[VZ] * vec.mV[VZ]);
-	v.normalize();
+	if (mag > FP_MAG_THRESHOLD)
-
+	{
-	angle *= 0.5f;
+		angle *= 0.5;
-	F32 c, s;
+		F32 c = cosf(angle);
-	c = cosf(angle);
+		F32 s = sinf(angle) / mag;
-	s = sinf(angle);
+		mQ[VX] = vec.mV[VX] * s;
-
+		mQ[VY] = vec.mV[VY] * s;
-	mQ[VX] = v.mV[VX]*s;
+		mQ[VZ] = vec.mV[VZ] * s;
-	mQ[VY] = v.mV[VY]*s;
+		mQ[VW] = c;
-	mQ[VZ] = v.mV[VZ]*s;
+	}
-	mQ[VW] = c;
+	else
-
+	{
-	normalize();
+		loadIdentity();
+	}
 	return (*this);
 }
 
 const LLQuaternion&	LLQuaternion::setQuat(F32 angle, const LLVector4 &vec)
 {
-	LLVector3 v(vec.mV[VX], vec.mV[VY], vec.mV[VZ]);
+	F32 mag = sqrtf(vec.mV[VX] * vec.mV[VX] + vec.mV[VY] * vec.mV[VY] + vec.mV[VZ] * vec.mV[VZ]);
-	v.normalize();
+	if (mag > FP_MAG_THRESHOLD)
-
+	{
-	F32 c, s;
+		angle *= 0.5;
-	c = cosf(angle*0.5f);
+		F32 c = cosf(angle);
-	s = sinf(angle*0.5f);
+		F32 s = sinf(angle) / mag;
-
+		mQ[VX] = vec.mV[VX] * s;
-	mQ[VX] = v.mV[VX]*s;
+		mQ[VY] = vec.mV[VY] * s;
-	mQ[VY] = v.mV[VY]*s;
+		mQ[VZ] = vec.mV[VZ] * s;
-	mQ[VZ] = v.mV[VZ]*s;
+		mQ[VW] = c;
-	mQ[VW] = c;
+	}
-
+	else
-	normalize();
+	{
+		loadIdentity();
+	}
 	return (*this);
 }
 
 const LLQuaternion&	LLQuaternion::setQuat(F32 roll, F32 pitch, F32 yaw)
 {
-	LLMatrix3 rot_mat(roll, pitch, yaw);
+	roll  *= 0.5f;
-	rot_mat.orthogonalize();
+	pitch *= 0.5f;
-	*this = rot_mat.quaternion();
+	yaw   *= 0.5f;
-		
+	F32 sinX = sinf(roll);
-	normalize();
+	F32 cosX = cosf(roll);
+	F32 sinY = sinf(pitch);
+	F32 cosY = cosf(pitch);
+	F32 sinZ = sinf(yaw);
+	F32 cosZ = cosf(yaw);
+	mQ[VW] = cosX * cosY * cosZ - sinX * sinY * sinZ;
+	mQ[VX] = sinX * cosY * cosZ + cosX * sinY * sinZ;
+	mQ[VY] = cosX * sinY * cosZ - sinX * cosY * sinZ;
+	mQ[VZ] = cosX * cosY * sinZ + sinX * sinY * cosZ;
 	return (*this);
 }
 
@@ -425,68 +447,44 @@ LLMatrix4	LLQuaternion::getMatrix4(void) const
 // calculate the shortest rotation from a to b
 void LLQuaternion::shortestArc(const LLVector3 &a, const LLVector3 &b)
 {
-	// Make a local copy of both vectors.
+	F32 ab = a * b; // dotproduct
-	LLVector3 vec_a = a;
+	LLVector3 c = a % b; // crossproduct
-	LLVector3 vec_b = b;
+	F32 cc = c * c; // squared length of the crossproduct
-
+	if (ab * ab + cc) // test if the arguments have sufficient magnitude
-	// Make sure neither vector is zero length.  Also normalize
-	// the vectors while we are at it.
-	F32 vec_a_mag = vec_a.normalize();
-	F32 vec_b_mag = vec_b.normalize();
-	if (vec_a_mag < F_APPROXIMATELY_ZERO ||
-		vec_b_mag < F_APPROXIMATELY_ZERO)
 	{
-		// Can't calculate a rotation from this.
+		if (cc > 0.0f) // test if the arguments are (anti)parallel
-		// Just return ZERO_ROTATION instead.
-		loadIdentity();
-		return;
-	}
-
-	// Create an axis to rotate around, and the cos of the angle to rotate.
-	LLVector3 axis = vec_a % vec_b;
-	F32 cos_theta  = vec_a * vec_b;
-
-	// Check the angle between the vectors to see if they are parallel or anti-parallel.
-	if (cos_theta > 1.0 - F_APPROXIMATELY_ZERO)
-	{
-		// a and b are parallel.  No rotation is necessary.
-		loadIdentity();
-	}
-	else if (cos_theta < -1.0 + F_APPROXIMATELY_ZERO)
-	{
-		// a and b are anti-parallel.
-		// Rotate 180 degrees around some orthogonal axis.
-		// Find the projection of the x-axis onto a, and try
-		// using the vector between the projection and the x-axis
-		// as the orthogonal axis.
-		LLVector3 proj = vec_a.mV[VX] / (vec_a * vec_a) * vec_a;
-		LLVector3 ortho_axis(1.f, 0.f, 0.f);
-		ortho_axis -= proj;
-		
-		// Turn this into an orthonormal axis.
-		F32 ortho_length = ortho_axis.normalize();
-		// If the axis' length is 0, then our guess at an orthogonal axis
-		// was wrong (a is parallel to the x-axis).
-		if (ortho_length < F_APPROXIMATELY_ZERO)
 		{
-			// Use the z-axis instead.
+			F32 s = sqrtf(ab * ab + cc) + ab; // note: don't try to optimize this line
-			ortho_axis.setVec(0.f, 0.f, 1.f);
+			F32 m = 1.0f / sqrtf(cc + s * s); // the inverted magnitude of the quaternion
+			mQ[VX] = c.mV[VX] * m;
+			mQ[VY] = c.mV[VY] * m;
+			mQ[VZ] = c.mV[VZ] * m;
+			mQ[VW] = s * m;
+			return;
+		}
+		if (ab < 0.0f) // test if the angle is bigger than PI/2 (anti parallel)
+		{
+			c = a - b; // the arguments are anti-parallel, we have to choose an axis
+			F32 m = sqrtf(c.mV[VX] * c.mV[VX] +  c.mV[VY] * c.mV[VY]); // the length projected on the XY-plane
+			if (m > FP_MAG_THRESHOLD)
+			{
+				mQ[VX] = -c.mV[VY] / m; // return the quaternion with the axis in the XY-plane
+				mQ[VY] =  c.mV[VX] / m;
+				mQ[VZ] = 0.0f;
+				mQ[VW] = 0.0f;
+				return;
+			}
+			else // the vectors are parallel to the Z-axis
+			{
+				mQ[VX] = 1.0f; // rotate around the X-axis
+				mQ[VY] = 0.0f;
+				mQ[VZ] = 0.0f;
+				mQ[VW] = 0.0f;
+				return;
+			}
 		}
-
-		// Construct a quaternion from this orthonormal axis.
-		mQ[VX] = ortho_axis.mV[VX];
-		mQ[VY] = ortho_axis.mV[VY];
-		mQ[VZ] = ortho_axis.mV[VZ];
-		mQ[VW] = 0.f;
-	}
-	else
-	{
-		// a and b are NOT parallel or anti-parallel.
-		// Return the rotation between these vectors.
-		F32 theta = (F32)acos(cos_theta);
-
-		setAngleAxis(theta, axis);
 	}
+	loadIdentity();
 }
 
 // constrains rotation to a cone angle specified in radians
@@ -838,79 +836,82 @@ LLQuaternion::Order StringToOrder( const char *str )
 
 void LLQuaternion::getAngleAxis(F32* angle, LLVector3 &vec) const
 {
-	F32 cos_a = mQ[VW];
+	F32 v = sqrtf(mQ[VX] * mQ[VX] + mQ[VY] * mQ[VY] + mQ[VZ] * mQ[VZ]); // length of the vector-component
-	if (cos_a > 1.0f) cos_a = 1.0f;
+	if (v > FP_MAG_THRESHOLD)
-	if (cos_a < -1.0f) cos_a = -1.0f;
-
-    F32 sin_a = (F32) sqrt( 1.0f - cos_a * cos_a );
-
-    if ( fabs( sin_a ) < 0.0005f )
-		sin_a = 1.0f;
-	else
-		sin_a = 1.f/sin_a;
-
-    F32 temp_angle = 2.0f * (F32) acos( cos_a );
-	if (temp_angle > F_PI)
 	{
-		// The (angle,axis) pair should never have angles outside [PI, -PI]
+		F32 oomag = 1.0f / v;
-		// since we want the _shortest_ (angle,axis) solution.
+		F32 w = mQ[VW];
-		// Since acos is defined for [0, PI], and we multiply by 2.0, we
+		if (mQ[VW] < 0.0f)
-		// can push the angle outside the acceptible range.
+		{
-		// When this happens we set the angle to the other portion of a 
+			w = -w; // make VW positive
-		// full 2PI rotation, and negate the axis, which reverses the 
+			oomag = -oomag; // invert the axis
-		// direction of the rotation (by the right-hand rule).
+		}
-		*angle = 2.f * F_PI - temp_angle;
+		vec.mV[VX] = mQ[VX] * oomag; // normalize the axis
-    	vec.mV[VX] = - mQ[VX] * sin_a;
+		vec.mV[VY] = mQ[VY] * oomag;
-    	vec.mV[VY] = - mQ[VY] * sin_a;
+		vec.mV[VZ] = mQ[VZ] * oomag;
-    	vec.mV[VZ] = - mQ[VZ] * sin_a;
+		*angle = 2.0f * atan2f(v, w); // get the angle
 	}
 	else
 	{
-		*angle = temp_angle;
+		*angle = 0.0f; // no rotation
-    	vec.mV[VX] = mQ[VX] * sin_a;
+		vec.mV[VX] = 0.0f; // around some dummy axis
-    	vec.mV[VY] = mQ[VY] * sin_a;
+		vec.mV[VY] = 0.0f;
-    	vec.mV[VZ] = mQ[VZ] * sin_a;
+		vec.mV[VZ] = 1.0f;
 	}
 }
 
-
 // quaternion does not need to be normalized
 void LLQuaternion::getEulerAngles(F32 *roll, F32 *pitch, F32 *yaw) const
 {
-	LLMatrix3 rot_mat(*this);
+	F32 sx = 2 * (mQ[VX] * mQ[VW] - mQ[VY] * mQ[VZ]); // sine of the roll
-	rot_mat.orthogonalize();
+	F32 sy = 2 * (mQ[VY] * mQ[VW] + mQ[VX] * mQ[VZ]); // sine of the pitch
-	rot_mat.getEulerAngles(roll, pitch, yaw);
+	F32 ys = mQ[VW] * mQ[VW] - mQ[VY] * mQ[VY]; // intermediate cosine 1
-
+	F32 xz = mQ[VX] * mQ[VX] - mQ[VZ] * mQ[VZ]; // intermediate cosine 2
-//	// NOTE: LLQuaternion's are actually inverted with respect to
+	F32 cx = ys - xz; // cosine of the roll
-//	// the matrices, so this code also assumes inverted quaternions
+	F32 cy = sqrtf(sx * sx + cx * cx); // cosine of the pitch
-//	// (-x, -y, -z, w). The result is that roll,pitch,yaw are applied
+	if (cy > GIMBAL_THRESHOLD) // no gimbal lock
-//	// in reverse order (yaw,pitch,roll).
+	{
-//	F32 x = -mQ[VX], y = -mQ[VY], z = -mQ[VZ], w = mQ[VW];
+		*roll  = atan2f(sx, cx);
-//	F64 m20 = 2.0*(x*z-y*w);
+		*pitch = atan2f(sy, cy);
-//	if (1.0f - fabsf(m20) < F_APPROXIMATELY_ZERO)
+		*yaw   = atan2f(2 * (mQ[VZ] * mQ[VW] - mQ[VX] * mQ[VY]), ys + xz);
-//	{
+	}
-//		*roll = 0.0f;
+	else // gimbal lock
-//		*pitch = (F32)asin(m20);
+	{
-//		*yaw = (F32)atan2(2.0*(x*y-z*w), 1.0 - 2.0*(x*x+z*z));
+		if (sy > 0)
-//	}
+		{
-//	else
+			*pitch = F_PI_BY_TWO;
-//	{
+			*yaw = 2 * atan2f(mQ[VZ] + mQ[VX], mQ[VW] + mQ[VY]);
-//		*roll  = (F32)atan2(-2.0*(y*z+x*w), 1.0-2.0*(x*x+y*y));
+		}
-//		*pitch = (F32)asin(m20);
+		else
-//		*yaw   = (F32)atan2(-2.0*(x*y+z*w), 1.0-2.0*(y*y+z*z));
+		{
-//	}
+			*pitch = -F_PI_BY_TWO;
+			*yaw = 2 * atan2f(mQ[VZ] - mQ[VX], mQ[VW] - mQ[VY]);
+		}
+		*roll = 0;
+	}
 }
 
 // Saves space by using the fact that our quaternions are normalized
 LLVector3 LLQuaternion::packToVector3() const
 {
+	F32 x = mQ[VX];
+	F32 y = mQ[VY];
+	F32 z = mQ[VZ];
+	F32 w = mQ[VW];
+	F32 mag = sqrtf(x * x + y * y + z * z + w * w);
+	if (mag > FP_MAG_THRESHOLD)
+	{
+		x /= mag;
+		y /= mag;
+		z /= mag; // no need to normalize w, it's not used
+	}
 	if( mQ[VW] >= 0 )
 	{
-		return LLVector3( mQ[VX], mQ[VY], mQ[VZ] );
+		return LLVector3( x, y , z );
 	}
 	else
 	{
-		return LLVector3( -mQ[VX], -mQ[VY], -mQ[VZ] );
+		return LLVector3( -x, -y, -z );
 	}
 }
 

indra/llmath/llquaternion.h

@@ -304,43 +304,29 @@ inline const LLQuaternion&	LLQuaternion::setQuat(const F32 *q)
 	return (*this);
 }
 
-// There may be a cheaper way that avoids the sqrt.
-// Does sin_a = VX*VX + VY*VY + VZ*VZ?
-// Copied from Matrix and Quaternion FAQ 1.12
 inline void LLQuaternion::getAngleAxis(F32* angle, F32* x, F32* y, F32* z) const
 {
-	F32 cos_a = mQ[VW];
+	F32 v = sqrtf(mQ[VX] * mQ[VX] + mQ[VY] * mQ[VY] + mQ[VZ] * mQ[VZ]); // length of the vector-component
-	if (cos_a > 1.0f) cos_a = 1.0f;
+	if (v > FP_MAG_THRESHOLD)
-	if (cos_a < -1.0f) cos_a = -1.0f;
-
-    F32 sin_a = (F32) sqrt( 1.0f - cos_a * cos_a );
-
-    if ( fabs( sin_a ) < 0.0005f )
-		sin_a = 1.0f;
-	else
-		sin_a = 1.f/sin_a;
-
-    F32 temp_angle = 2.0f * (F32) acos( cos_a );
-	if (temp_angle > F_PI)
 	{
-		// The (angle,axis) pair should never have angles outside [PI, -PI]
+		F32 oomag = 1.0f / v;
-		// since we want the _shortest_ (angle,axis) solution.
+		F32 w = mQ[VW];
-		// Since acos is defined for [0, PI], and we multiply by 2.0, we
+		if (w < 0.0f)
-		// can push the angle outside the acceptible range.
+		{
-		// When this happens we set the angle to the other portion of a 
+			w = -w; // make VW positive
-		// full 2PI rotation, and negate the axis, which reverses the 
+			oomag = -oomag; // invert the axis
-		// direction of the rotation (by the right-hand rule).
+		}
-		*angle = 2.f * F_PI - temp_angle;
+		*x = mQ[VX] * oomag; // normalize the axis
-    	*x = - mQ[VX] * sin_a;
+		*y = mQ[VY] * oomag;
-    	*y = - mQ[VY] * sin_a;
+		*z = mQ[VZ] * oomag;
-    	*z = - mQ[VZ] * sin_a;
+		*angle = 2.0f * atan2f(v, w); // get the angle
 	}
 	else
 	{
-		*angle = temp_angle;
+		*angle = 0.0f; // no rotation
-    	*x = mQ[VX] * sin_a;
+		*x = 0.0f; // around some dummy axis
-    	*y = mQ[VY] * sin_a;
+		*y = 0.0f;
-    	*z = mQ[VZ] * sin_a;
+		*z = 1.0f;
 	}
 }
 

indra/llmath/llquaternion2.h

@@ -40,6 +40,7 @@
 /////////////////////////////
 #include "llquaternion.h"
 
+LL_ALIGN_PREFIX(16)
 class LLQuaternion2
 {
 public:
@@ -84,6 +85,8 @@ public:
 	// Quantize this quaternion to 16 bit precision
 	inline void quantize16();
 
+	inline void mul(const LLQuaternion2& b);
+
 	/////////////////////////
 	// Quaternion inspection
 	/////////////////////////
@@ -98,8 +101,8 @@ public:
 
 protected:
 
-	LLVector4a mQ;
+	LL_ALIGN_16(LLVector4a mQ);
 
-};
+} LL_ALIGN_POSTFIX(16);
 
 #endif

indra/llmath/llquaternion2.inl

@@ -50,6 +50,39 @@ inline LLVector4a& LLQuaternion2::getVector4aRw()
 	return mQ;
 }
 
+inline void LLQuaternion2::mul(const LLQuaternion2& b)
+{
+	static LL_ALIGN_16(const unsigned int signMask[4]) = { 0x0, 0x0, 0x0, 0x80000000 };
+
+	LLVector4a sum1, sum2, prod1, prod2, prod3, prod4;
+	const LLVector4a& va = mQ;
+	const LLVector4a& vb = b.getVector4a();
+
+	//			[VX] [VY] [VZ] [VW]
+	//prod1:	+wx  +wy  +wz  +ww  Bwwww*Axyzw 
+	//prod2:	+xw  +yw  +zw  -xx  Bxyzx*Awwwx		[VW] sign flip
+	//prod3:	+yz  +zx  +xy  -yy  Byzxy*Azxyy		[VW] sign flip
+	//prod4:	-zy  -xz  -yx  -zz  Bzxyz*Ayzzz
+
+	const LLVector4a Bwwww = _mm_shuffle_ps(vb,vb,_MM_SHUFFLE(3,3,3,3));
+	const LLVector4a Bxyzx = _mm_shuffle_ps(vb,vb,_MM_SHUFFLE(0,2,1,0));
+	const LLVector4a Awwwx = _mm_shuffle_ps(va,va,_MM_SHUFFLE(0,3,3,3));
+	const LLVector4a Byzxy = _mm_shuffle_ps(vb,vb,_MM_SHUFFLE(1,0,2,1));
+	const LLVector4a Azxyy = _mm_shuffle_ps(va,va,_MM_SHUFFLE(1,1,0,2));
+	const LLVector4a Bzxyz = _mm_shuffle_ps(vb,vb,_MM_SHUFFLE(2,1,0,2));
+	const LLVector4a Ayzxz = _mm_shuffle_ps(va,va,_MM_SHUFFLE(2,0,2,1));
+
+	prod1.setMul(Bwwww,va);
+	prod2.setMul(Bxyzx,Awwwx);
+	prod3.setMul(Byzxy,Azxyy);
+	prod4.setMul(Bzxyz,Ayzxz);
+
+	sum1.setAdd(prod2,prod3);
+	sum1 = _mm_xor_ps(sum1, _mm_load_ps((const float*)signMask));	
+	sum2.setSub(prod1,prod4);
+	mQ.setAdd(sum1,sum2);
+}
+
 /////////////////////////
 // Quaternion modification
 /////////////////////////

indra/llmath/llvector4a.h

@@ -128,7 +128,7 @@ public:
 	inline void loadua(const F32* src);
 	
 	// Load only three floats beginning at address 'src'. Slowest method.
-	inline void load3(const F32* src);
+	inline void load3(const F32* src, const F32 w=0.f);
 	
 	// Store to a 16-byte aligned memory address
 	inline void store4a(F32* dst) const;
@@ -170,6 +170,9 @@ public:
 	
 	// Set all 4 elements to element i of v, with i NOT known at compile time
 	inline void splat(const LLVector4a& v, U32 i);
+
+	// Sets element N to that of src's element N. Much cleaner than.. {LLVector4Logical mask; mask.clear(); mask.setElement<N>(); target.setSelectWithMask(mask,src,target);}
+	template <int N> inline void copyComponent(const LLVector4a& src);
 	
 	// Select bits from sourceIfTrue and sourceIfFalse according to bits in mask
 	inline void setSelectWithMask( const LLVector4Logical& mask, const LLVector4a& sourceIfTrue, const LLVector4a& sourceIfFalse );
@@ -282,6 +285,8 @@ public:
 	void quantize8( const LLVector4a& low, const LLVector4a& high );
 	void quantize16( const LLVector4a& low, const LLVector4a& high );
 
+	void negate();
+
 	////////////////////////////////////
 	// LOGICAL
 	////////////////////////////////////	

indra/llmath/llvector4a.inl

@@ -41,11 +41,11 @@ inline void LLVector4a::loadua(const F32* src)
 }
 
 // Load only three floats beginning at address 'src'. Slowest method.
-inline void LLVector4a::load3(const F32* src)
+inline void LLVector4a::load3(const F32* src, const F32 w)
 {
 	// mQ = { 0.f, src[2], src[1], src[0] } = { W, Z, Y, X }
 	// NB: This differs from the convention of { Z, Y, X, W }
-	mQ = _mm_set_ps(0.f, src[2], src[1], src[0]);
+	mQ = _mm_set_ps(w, src[2], src[1], src[0]);
 }	
 
 // Store to a 16-byte aligned memory address
@@ -154,6 +154,13 @@ inline void LLVector4a::splat(const LLVector4a& v, U32 i)
 	}
 }
 
+// Sets element N to that of src's element N
+template <int N> inline void LLVector4a::copyComponent(const LLVector4a& src)
+{
+	static const LLVector4Logical mask = _mm_load_ps((F32*)&S_V4LOGICAL_MASK_TABLE[N*4]);
+	setSelectWithMask(mask,src,mQ);
+}
+
 // Select bits from sourceIfTrue and sourceIfFalse according to bits in mask
 inline void LLVector4a::setSelectWithMask( const LLVector4Logical& mask, const LLVector4a& sourceIfTrue, const LLVector4a& sourceIfFalse )
 {
@@ -529,6 +536,11 @@ inline void LLVector4a::clamp( const LLVector4a& low, const LLVector4a& high )
 	setSelectWithMask( lowMask, low, *this );
 }
 
+inline void LLVector4a::negate()
+{
+	static LL_ALIGN_16(const U32 signMask[4]) = {0x80000000, 0x80000000, 0x80000000, 0x80000000 };
+	mQ = _mm_xor_ps(*reinterpret_cast<const LLQuad*>(signMask), mQ);
+}
 
 ////////////////////////////////////
 // LOGICAL

indra/llmath/llvector4logical.h

@@ -79,7 +79,7 @@ public:
 	{
 		static const LL_ALIGN_16(U32 allOnes[4]) = { 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF };
 		ll_assert_aligned(allOnes,16);
-		mQ = _mm_andnot_ps( mQ, *(LLQuad*)(allOnes) );
+		mQ = _mm_andnot_ps( mQ, _mm_load_ps((F32*)(allOnes)));
 		return *this;
 	}
 	
@@ -115,7 +115,7 @@ public:
 
 	template<int N> void setElement()
 	{
-		mQ = _mm_or_ps( mQ, *reinterpret_cast<const LLQuad*>(S_V4LOGICAL_MASK_TABLE + 4*N) );
+		mQ = _mm_or_ps( mQ, _mm_load_ps( (F32*)&S_V4LOGICAL_MASK_TABLE[4*N] ) );
 	}
 	
 private:

indra/llmath/llvolume.cpp

@@ -50,7 +50,6 @@
 #include "llstl.h"
 #include "llsdserialize.h"
 #include "llvector4a.h"
-#include "llmatrix4a.h"
 #include "lltimer.h"
 
 #define DEBUG_SILHOUETTE_BINORMALS 0
@@ -1649,7 +1648,7 @@ BOOL LLPath::generate(const LLPathParams& params, F32 detail, S32 split,
 			F32 t = (F32)i * mStep;
 			mPath[i].mPos.set(0,
 								lerp(0,   -sin(F_PI*params.getTwist()*t)*0.5f,t),
-								lerp(-0.5, cos(F_PI*params.getTwist()*t)*0.5f,t));
+								lerp(-0.5f, cos(F_PI*params.getTwist()*t)*0.5f,t));
 			mPath[i].mScale.set(lerp(1,params.getScale().mV[0],t),
 								lerp(1,params.getScale().mV[1],t), 0,1);
 			mPath[i].mTexT  = t;
@@ -2184,7 +2183,7 @@ BOOL LLVolume::generate()
 				0, 0, scale[2], 0,
 					0, 0, 0, 1 };
 			
-			LLMatrix4 rot((F32*) mPathp->mPath[s].mRot.mMatrix);
+			LLMatrix4 rot(mPathp->mPath[s].mRot.getF32ptr());
 			LLMatrix4 scale_mat(sc);
 			
 			scale_mat *= rot;
@@ -3670,16 +3669,14 @@ S32 LLVolume::getNumTriangles(S32* vcount) const
 void LLVolume::generateSilhouetteVertices(std::vector<LLVector3> &vertices,
 										  std::vector<LLVector3> &normals,
 										  const LLVector3& obj_cam_vec_in,
-										  const LLMatrix4& mat_in,
+										  const LLMatrix4a& mat_in,
-										  const LLMatrix3& norm_mat_in,
+										  const LLMatrix4a& norm_mat_in,
 										  S32 face_mask)
 {
-	LLMatrix4a mat;
+	const LLMatrix4a& mat = mat_in;
-	mat.loadu(mat_in);
+
+	const LLMatrix4a& norm_mat = norm_mat_in;
 
-	LLMatrix4a norm_mat;
-	norm_mat.loadu(norm_mat_in);
-		
 	LLVector4a obj_cam_vec;
 	obj_cam_vec.load3(obj_cam_vec_in.mV);
 

indra/llmath/llvolume.h

@@ -27,6 +27,9 @@
 #ifndef LL_LLVOLUME_H
 #define LL_LLVOLUME_H
 
+#ifdef IN_PCH
+#error "llvolume.h should not be in pch include chain."
+#endif
 #include <iostream>
 
 class LLProfileParams;
@@ -747,10 +750,10 @@ public:
 	class PathPt
 	{
 	public:
-		LLMatrix4a   mRot;
+		LL_ALIGN_16(LLMatrix4a   mRot);
-		LLVector4a	 mPos;
+		LL_ALIGN_16(LLVector4a	 mPos);
 		
-		LLVector4a   mScale;
+		LL_ALIGN_16(LLVector4a   mScale);
 		F32			 mTexT;
 		F32 pad[3]; //for alignment
 		PathPt() 
@@ -1017,8 +1020,8 @@ public:
 	void generateSilhouetteVertices(std::vector<LLVector3> &vertices, 
 									std::vector<LLVector3> &normals, 
 									const LLVector3& view_vec,
-									const LLMatrix4& mat,
+									const LLMatrix4a& mat,
-									const LLMatrix3& norm_mat,
+									const LLMatrix4a& norm_mat,
 									S32 face_index);
 
 	//get the face index of the face that intersects with the given line segment at the point 

indra/llmath/llvolumemgr.h

@@ -2,31 +2,25 @@
  * @file llvolumemgr.h
  * @brief LLVolumeMgr class.
  *
- * $LicenseInfo:firstyear=2002&license=viewergpl$
+ * $LicenseInfo:firstyear=2002&license=viewerlgpl$
- * 
- * Copyright (c) 2002-2009, Linden Research, Inc.
- * 
  * Second Life Viewer Source Code
- * The source code in this file ("Source Code") is provided by Linden Lab
+ * Copyright (C) 2010, Linden Research, Inc.
- * to you under the terms of the GNU General Public License, version 2.0
- * ("GPL"), unless you have obtained a separate licensing agreement
- * ("Other License"), formally executed by you and Linden Lab.  Terms of
- * the GPL can be found in doc/GPL-license.txt in this distribution, or
- * online at http://secondlifegrid.net/programs/open_source/licensing/gplv2
  * 
- * There are special exceptions to the terms and conditions of the GPL as
+ * This library is free software; you can redistribute it and/or
- * it is applied to this Source Code. View the full text of the exception
+ * modify it under the terms of the GNU Lesser General Public
- * in the file doc/FLOSS-exception.txt in this software distribution, or
+ * License as published by the Free Software Foundation;
- * online at
+ * version 2.1 of the License only.
- * http://secondlifegrid.net/programs/open_source/licensing/flossexception
  * 
- * By copying, modifying or distributing this software, you acknowledge
+ * This library is distributed in the hope that it will be useful,
- * that you have read and understood your obligations described above,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
- * and agree to abide by those obligations.
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * Lesser General Public License for more details.
  * 
- * ALL LINDEN LAB SOURCE CODE IS PROVIDED "AS IS." LINDEN LAB MAKES NO
+ * You should have received a copy of the GNU Lesser General Public
- * WARRANTIES, EXPRESS, IMPLIED OR OTHERWISE, REGARDING ITS ACCURACY,
+ * License along with this library; if not, write to the Free Software
- * COMPLETENESS OR PERFORMANCE.
+ * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA  02110-1301  USA
+ *
+ * Linden Research, Inc., 945 Battery Street, San Francisco, CA  94111  USA
  * $/LicenseInfo$
  */
 

indra/llmath/llvolumeoctree.h

@@ -127,13 +127,14 @@ public:
 	LL_ALIGN_16(LLVector4a mExtents[2]); // extents (min, max) of this node and all its children
 };
 
+LL_ALIGN_PREFIX(16)
 class LLOctreeTriangleRayIntersect : public LLOctreeTraveler<LLVolumeTriangle>
 {
 public:
 	const LLVolumeFace* mFace;
-	LLVector4a mStart;
+	LL_ALIGN_16(LLVector4a mStart);
-	LLVector4a mDir;
+	LL_ALIGN_16(LLVector4a mDir);
-	LLVector4a mEnd;
+	LL_ALIGN_16(LLVector4a mEnd);
 	LLVector4a* mIntersection;
 	LLVector2* mTexCoord;
 	LLVector4a* mNormal;
@@ -148,7 +149,7 @@ public:
 	void traverse(const LLOctreeNode<LLVolumeTriangle>* node);
 
 	virtual void visit(const LLOctreeNode<LLVolumeTriangle>* node);
-};
+} LL_ALIGN_POSTFIX(16);
 
 class LLVolumeOctreeValidate : public LLOctreeTraveler<LLVolumeTriangle>
 {

indra/llmath/m3math.cpp

@@ -2,31 +2,25 @@
  * @file m3math.cpp
  * @brief LLMatrix3 class implementation.
  *
- * $LicenseInfo:firstyear=2000&license=viewergpl$
+ * $LicenseInfo:firstyear=2000&license=viewerlgpl$
- * 
- * Copyright (c) 2000-2009, Linden Research, Inc.
- * 
  * Second Life Viewer Source Code
- * The source code in this file ("Source Code") is provided by Linden Lab
+ * Copyright (C) 2010, Linden Research, Inc.
- * to you under the terms of the GNU General Public License, version 2.0
- * ("GPL"), unless you have obtained a separate licensing agreement
- * ("Other License"), formally executed by you and Linden Lab.  Terms of
- * the GPL can be found in doc/GPL-license.txt in this distribution, or
- * online at http://secondlifegrid.net/programs/open_source/licensing/gplv2
  * 
- * There are special exceptions to the terms and conditions of the GPL as
+ * This library is free software; you can redistribute it and/or
- * it is applied to this Source Code. View the full text of the exception
+ * modify it under the terms of the GNU Lesser General Public
- * in the file doc/FLOSS-exception.txt in this software distribution, or
+ * License as published by the Free Software Foundation;
- * online at
+ * version 2.1 of the License only.
- * http://secondlifegrid.net/programs/open_source/licensing/flossexception
  * 
- * By copying, modifying or distributing this software, you acknowledge
+ * This library is distributed in the hope that it will be useful,
- * that you have read and understood your obligations described above,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
- * and agree to abide by those obligations.
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * Lesser General Public License for more details.
  * 
- * ALL LINDEN LAB SOURCE CODE IS PROVIDED "AS IS." LINDEN LAB MAKES NO
+ * You should have received a copy of the GNU Lesser General Public
- * WARRANTIES, EXPRESS, IMPLIED OR OTHERWISE, REGARDING ITS ACCURACY,
+ * License along with this library; if not, write to the Free Software
- * COMPLETENESS OR PERFORMANCE.
+ * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA  02110-1301  USA
+ *
+ * Linden Research, Inc., 945 Battery Street, San Francisco, CA  94111  USA
  * $/LicenseInfo$
  */
 

indra/llmath/m3math.h

@@ -2,31 +2,25 @@
  * @file m3math.h
  * @brief LLMatrix3 class header file.
  *
- * $LicenseInfo:firstyear=2000&license=viewergpl$
+ * $LicenseInfo:firstyear=2000&license=viewerlgpl$
- * 
- * Copyright (c) 2000-2009, Linden Research, Inc.
- * 
  * Second Life Viewer Source Code
- * The source code in this file ("Source Code") is provided by Linden Lab
+ * Copyright (C) 2010, Linden Research, Inc.
- * to you under the terms of the GNU General Public License, version 2.0
- * ("GPL"), unless you have obtained a separate licensing agreement
- * ("Other License"), formally executed by you and Linden Lab.  Terms of
- * the GPL can be found in doc/GPL-license.txt in this distribution, or
- * online at http://secondlifegrid.net/programs/open_source/licensing/gplv2
  * 
- * There are special exceptions to the terms and conditions of the GPL as
+ * This library is free software; you can redistribute it and/or
- * it is applied to this Source Code. View the full text of the exception
+ * modify it under the terms of the GNU Lesser General Public
- * in the file doc/FLOSS-exception.txt in this software distribution, or
+ * License as published by the Free Software Foundation;
- * online at
+ * version 2.1 of the License only.
- * http://secondlifegrid.net/programs/open_source/licensing/flossexception
  * 
- * By copying, modifying or distributing this software, you acknowledge
+ * This library is distributed in the hope that it will be useful,
- * that you have read and understood your obligations described above,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
- * and agree to abide by those obligations.
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * Lesser General Public License for more details.
  * 
- * ALL LINDEN LAB SOURCE CODE IS PROVIDED "AS IS." LINDEN LAB MAKES NO
+ * You should have received a copy of the GNU Lesser General Public
- * WARRANTIES, EXPRESS, IMPLIED OR OTHERWISE, REGARDING ITS ACCURACY,
+ * License along with this library; if not, write to the Free Software
- * COMPLETENESS OR PERFORMANCE.
+ * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA  02110-1301  USA
+ *
+ * Linden Research, Inc., 945 Battery Street, San Francisco, CA  94111  USA
  * $/LicenseInfo$
  */
 

indra/llmath/m4math.cpp

@@ -2,31 +2,25 @@
  * @file m4math.cpp
  * @brief LLMatrix4 class implementation.
  *
- * $LicenseInfo:firstyear=2000&license=viewergpl$
+ * $LicenseInfo:firstyear=2000&license=viewerlgpl$
- * 
- * Copyright (c) 2000-2009, Linden Research, Inc.
- * 
  * Second Life Viewer Source Code
- * The source code in this file ("Source Code") is provided by Linden Lab
+ * Copyright (C) 2010, Linden Research, Inc.
- * to you under the terms of the GNU General Public License, version 2.0
+ *
- * ("GPL"), unless you have obtained a separate licensing agreement
+ * This library is free software; you can redistribute it and/or
- * ("Other License"), formally executed by you and Linden Lab.  Terms of
+ * modify it under the terms of the GNU Lesser General Public
- * the GPL can be found in doc/GPL-license.txt in this distribution, or
+ * License as published by the Free Software Foundation;
- * online at http://secondlifegrid.net/programs/open_source/licensing/gplv2
+ * version 2.1 of the License only.
  * 
- * There are special exceptions to the terms and conditions of the GPL as
+ * This library is distributed in the hope that it will be useful,
- * it is applied to this Source Code. View the full text of the exception
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
- * in the file doc/FLOSS-exception.txt in this software distribution, or
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
- * online at
+ * Lesser General Public License for more details.
- * http://secondlifegrid.net/programs/open_source/licensing/flossexception
  * 
- * By copying, modifying or distributing this software, you acknowledge
+ * You should have received a copy of the GNU Lesser General Public
- * that you have read and understood your obligations described above,
+ * License along with this library; if not, write to the Free Software
- * and agree to abide by those obligations.
+ * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA  02110-1301  USA
  * 
- * ALL LINDEN LAB SOURCE CODE IS PROVIDED "AS IS." LINDEN LAB MAKES NO
+ * Linden Research, Inc., 945 Battery Street, San Francisco, CA  94111  USA
- * WARRANTIES, EXPRESS, IMPLIED OR OTHERWISE, REGARDING ITS ACCURACY,
- * COMPLETENESS OR PERFORMANCE.
  * $/LicenseInfo$
  */
 
@@ -684,37 +678,6 @@ const LLMatrix4&  	LLMatrix4::initMatrix(const LLMatrix3 &mat, const LLVector4 &
 
 // LLMatrix4 Operators
 
-
-/* Not implemented to help enforce code consistency with the syntax of
-   row-major notation.  This is a Good Thing.
-LLVector4 operator*(const LLMatrix4 &a, const LLVector4 &b)
-{
-	// Operate "to the right" on column-vector b
-	LLVector4	vec;
-	vec.mV[VX] = a.mMatrix[VX][VX] * b.mV[VX] + 
-				 a.mMatrix[VY][VX] * b.mV[VY] + 
- 				 a.mMatrix[VZ][VX] * b.mV[VZ] +
-				 a.mMatrix[VW][VX] * b.mV[VW];
-
-	vec.mV[VY] = a.mMatrix[VX][VY] * b.mV[VX] + 
-				 a.mMatrix[VY][VY] * b.mV[VY] + 
-				 a.mMatrix[VZ][VY] * b.mV[VZ] +
-				 a.mMatrix[VW][VY] * b.mV[VW];
-
-	vec.mV[VZ] = a.mMatrix[VX][VZ] * b.mV[VX] + 
-			  	 a.mMatrix[VY][VZ] * b.mV[VY] + 
-				 a.mMatrix[VZ][VZ] * b.mV[VZ] +
-				 a.mMatrix[VW][VZ] * b.mV[VW];
-
-	vec.mV[VW] = a.mMatrix[VX][VW] * b.mV[VX] + 
-				 a.mMatrix[VY][VW] * b.mV[VY] + 
-				 a.mMatrix[VZ][VW] * b.mV[VZ] +
-				 a.mMatrix[VW][VW] * b.mV[VW];
-	return vec;
-}
-*/
-
-
 LLVector4 operator*(const LLVector4 &a, const LLMatrix4 &b)
 {
 	// Operate "to the left" on row-vector a

indra/llmath/m4math.h

@@ -2,31 +2,25 @@
  * @file m4math.h
  * @brief LLMatrix4 class header file.
  *
- * $LicenseInfo:firstyear=2000&license=viewergpl$
+ * $LicenseInfo:firstyear=2000&license=viewerlgpl$
- * 
- * Copyright (c) 2000-2009, Linden Research, Inc.
- * 
  * Second Life Viewer Source Code
- * The source code in this file ("Source Code") is provided by Linden Lab
+ * Copyright (C) 2010, Linden Research, Inc.
- * to you under the terms of the GNU General Public License, version 2.0
- * ("GPL"), unless you have obtained a separate licensing agreement
- * ("Other License"), formally executed by you and Linden Lab.  Terms of
- * the GPL can be found in doc/GPL-license.txt in this distribution, or
- * online at http://secondlifegrid.net/programs/open_source/licensing/gplv2
  * 
- * There are special exceptions to the terms and conditions of the GPL as
+ * This library is free software; you can redistribute it and/or
- * it is applied to this Source Code. View the full text of the exception
+ * modify it under the terms of the GNU Lesser General Public
- * in the file doc/FLOSS-exception.txt in this software distribution, or
+ * License as published by the Free Software Foundation;
- * online at
+ * version 2.1 of the License only.
- * http://secondlifegrid.net/programs/open_source/licensing/flossexception
  * 
- * By copying, modifying or distributing this software, you acknowledge
+ * This library is distributed in the hope that it will be useful,
- * that you have read and understood your obligations described above,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
- * and agree to abide by those obligations.
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * Lesser General Public License for more details.
  * 
- * ALL LINDEN LAB SOURCE CODE IS PROVIDED "AS IS." LINDEN LAB MAKES NO
+ * You should have received a copy of the GNU Lesser General Public
- * WARRANTIES, EXPRESS, IMPLIED OR OTHERWISE, REGARDING ITS ACCURACY,
+ * License along with this library; if not, write to the Free Software
- * COMPLETENESS OR PERFORMANCE.
+ * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA  02110-1301  USA
+ *
+ * Linden Research, Inc., 945 Battery Street, San Francisco, CA  94111  USA
  * $/LicenseInfo$
  */
 
@@ -229,9 +223,6 @@ public:
 	// Operators
 	//
 
-// Not implemented to enforce code that agrees with symbolic syntax
-//		friend LLVector4 operator*(const LLMatrix4 &a, const LLVector4 &b);		// Apply rotation a to vector b
-
 //	friend inline LLMatrix4 operator*(const LLMatrix4 &a, const LLMatrix4 &b);		// Return a * b
 	friend LLVector4 operator*(const LLVector4 &a, const LLMatrix4 &b);		// Return transform of vector a by matrix b
 	friend const LLVector3 operator*(const LLVector3 &a, const LLMatrix4 &b);		// Return full transform of a by matrix b

indra/llmath/raytrace.cpp

@@ -2,31 +2,25 @@
  * @file raytrace.cpp
  * @brief Functions called by box object scripts.
  *
- * $LicenseInfo:firstyear=2001&license=viewergpl$
+ * $LicenseInfo:firstyear=2001&license=viewerlgpl$
- * 
- * Copyright (c) 2001-2009, Linden Research, Inc.
- * 
  * Second Life Viewer Source Code
- * The source code in this file ("Source Code") is provided by Linden Lab
+ * Copyright (C) 2010, Linden Research, Inc.
- * to you under the terms of the GNU General Public License, version 2.0
- * ("GPL"), unless you have obtained a separate licensing agreement
- * ("Other License"), formally executed by you and Linden Lab.  Terms of
- * the GPL can be found in doc/GPL-license.txt in this distribution, or
- * online at http://secondlifegrid.net/programs/open_source/licensing/gplv2
  * 
- * There are special exceptions to the terms and conditions of the GPL as
+ * This library is free software; you can redistribute it and/or
- * it is applied to this Source Code. View the full text of the exception
+ * modify it under the terms of the GNU Lesser General Public
- * in the file doc/FLOSS-exception.txt in this software distribution, or
+ * License as published by the Free Software Foundation;
- * online at
+ * version 2.1 of the License only.
- * http://secondlifegrid.net/programs/open_source/licensing/flossexception
  * 
- * By copying, modifying or distributing this software, you acknowledge
+ * This library is distributed in the hope that it will be useful,
- * that you have read and understood your obligations described above,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
- * and agree to abide by those obligations.
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * Lesser General Public License for more details.
  * 
- * ALL LINDEN LAB SOURCE CODE IS PROVIDED "AS IS." LINDEN LAB MAKES NO
+ * You should have received a copy of the GNU Lesser General Public
- * WARRANTIES, EXPRESS, IMPLIED OR OTHERWISE, REGARDING ITS ACCURACY,
+ * License along with this library; if not, write to the Free Software
- * COMPLETENESS OR PERFORMANCE.
+ * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA  02110-1301  USA
+ *
+ * Linden Research, Inc., 945 Battery Street, San Francisco, CA  94111  USA
  * $/LicenseInfo$
  */
 

indra/llmath/raytrace.h

@@ -2,31 +2,25 @@
  * @file raytrace.h
  * @brief Ray intersection tests for primitives.
  *
- * $LicenseInfo:firstyear=2001&license=viewergpl$
+ * $LicenseInfo:firstyear=2001&license=viewerlgpl$
- * 
- * Copyright (c) 2001-2009, Linden Research, Inc.
- * 
  * Second Life Viewer Source Code
- * The source code in this file ("Source Code") is provided by Linden Lab
+ * Copyright (C) 2010, Linden Research, Inc.
- * to you under the terms of the GNU General Public License, version 2.0
- * ("GPL"), unless you have obtained a separate licensing agreement
- * ("Other License"), formally executed by you and Linden Lab.  Terms of
- * the GPL can be found in doc/GPL-license.txt in this distribution, or
- * online at http://secondlifegrid.net/programs/open_source/licensing/gplv2
  * 
- * There are special exceptions to the terms and conditions of the GPL as
+ * This library is free software; you can redistribute it and/or
- * it is applied to this Source Code. View the full text of the exception
+ * modify it under the terms of the GNU Lesser General Public
- * in the file doc/FLOSS-exception.txt in this software distribution, or
+ * License as published by the Free Software Foundation;
- * online at
+ * version 2.1 of the License only.
- * http://secondlifegrid.net/programs/open_source/licensing/flossexception
  * 
- * By copying, modifying or distributing this software, you acknowledge
+ * This library is distributed in the hope that it will be useful,
- * that you have read and understood your obligations described above,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
- * and agree to abide by those obligations.
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * Lesser General Public License for more details.
  * 
- * ALL LINDEN LAB SOURCE CODE IS PROVIDED "AS IS." LINDEN LAB MAKES NO
+ * You should have received a copy of the GNU Lesser General Public
- * WARRANTIES, EXPRESS, IMPLIED OR OTHERWISE, REGARDING ITS ACCURACY,
+ * License along with this library; if not, write to the Free Software
- * COMPLETENESS OR PERFORMANCE.
+ * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA  02110-1301  USA
+ *
+ * Linden Research, Inc., 945 Battery Street, San Francisco, CA  94111  USA
  * $/LicenseInfo$
  */
 

indra/llmath/v2math.cpp

@@ -2,31 +2,25 @@
  * @file v2math.cpp
  * @brief LLVector2 class implementation.
  *
- * $LicenseInfo:firstyear=2000&license=viewergpl$
+ * $LicenseInfo:firstyear=2000&license=viewerlgpl$
- * 
- * Copyright (c) 2000-2009, Linden Research, Inc.
- * 
  * Second Life Viewer Source Code
- * The source code in this file ("Source Code") is provided by Linden Lab
+ * Copyright (C) 2010, Linden Research, Inc.
- * to you under the terms of the GNU General Public License, version 2.0
- * ("GPL"), unless you have obtained a separate licensing agreement
- * ("Other License"), formally executed by you and Linden Lab.  Terms of
- * the GPL can be found in doc/GPL-license.txt in this distribution, or
- * online at http://secondlifegrid.net/programs/open_source/licensing/gplv2
  * 
- * There are special exceptions to the terms and conditions of the GPL as
+ * This library is free software; you can redistribute it and/or
- * it is applied to this Source Code. View the full text of the exception
+ * modify it under the terms of the GNU Lesser General Public
- * in the file doc/FLOSS-exception.txt in this software distribution, or
+ * License as published by the Free Software Foundation;
- * online at
+ * version 2.1 of the License only.
- * http://secondlifegrid.net/programs/open_source/licensing/flossexception
  * 
- * By copying, modifying or distributing this software, you acknowledge
+ * This library is distributed in the hope that it will be useful,
- * that you have read and understood your obligations described above,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
- * and agree to abide by those obligations.
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * Lesser General Public License for more details.
  * 
- * ALL LINDEN LAB SOURCE CODE IS PROVIDED "AS IS." LINDEN LAB MAKES NO
+ * You should have received a copy of the GNU Lesser General Public
- * WARRANTIES, EXPRESS, IMPLIED OR OTHERWISE, REGARDING ITS ACCURACY,
+ * License along with this library; if not, write to the Free Software
- * COMPLETENESS OR PERFORMANCE.
+ * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA  02110-1301  USA
+ *
+ * Linden Research, Inc., 945 Battery Street, San Francisco, CA  94111  USA
  * $/LicenseInfo$
  */
 

indra/llmath/v2math.h

@@ -2,31 +2,25 @@
  * @file v2math.h
  * @brief LLVector2 class header file.
  *
- * $LicenseInfo:firstyear=2000&license=viewergpl$
+ * $LicenseInfo:firstyear=2000&license=viewerlgpl$
- * 
- * Copyright (c) 2000-2009, Linden Research, Inc.
- * 
  * Second Life Viewer Source Code
- * The source code in this file ("Source Code") is provided by Linden Lab
+ * Copyright (C) 2010, Linden Research, Inc.
- * to you under the terms of the GNU General Public License, version 2.0
- * ("GPL"), unless you have obtained a separate licensing agreement
- * ("Other License"), formally executed by you and Linden Lab.  Terms of
- * the GPL can be found in doc/GPL-license.txt in this distribution, or
- * online at http://secondlifegrid.net/programs/open_source/licensing/gplv2
  * 
- * There are special exceptions to the terms and conditions of the GPL as
+ * This library is free software; you can redistribute it and/or
- * it is applied to this Source Code. View the full text of the exception
+ * modify it under the terms of the GNU Lesser General Public
- * in the file doc/FLOSS-exception.txt in this software distribution, or
+ * License as published by the Free Software Foundation;
- * online at
+ * version 2.1 of the License only.
- * http://secondlifegrid.net/programs/open_source/licensing/flossexception
  * 
- * By copying, modifying or distributing this software, you acknowledge
+ * This library is distributed in the hope that it will be useful,
- * that you have read and understood your obligations described above,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
- * and agree to abide by those obligations.
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * Lesser General Public License for more details.
  * 
- * ALL LINDEN LAB SOURCE CODE IS PROVIDED "AS IS." LINDEN LAB MAKES NO
+ * You should have received a copy of the GNU Lesser General Public
- * WARRANTIES, EXPRESS, IMPLIED OR OTHERWISE, REGARDING ITS ACCURACY,
+ * License along with this library; if not, write to the Free Software
- * COMPLETENESS OR PERFORMANCE.
+ * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA  02110-1301  USA
+ *
+ * Linden Research, Inc., 945 Battery Street, San Francisco, CA  94111  USA
  * $/LicenseInfo$
  */
 

indra/llmath/v3color.cpp

@@ -2,31 +2,25 @@
  * @file v3color.cpp
  * @brief LLColor3 class implementation.
  *
- * $LicenseInfo:firstyear=2000&license=viewergpl$
+ * $LicenseInfo:firstyear=2000&license=viewerlgpl$
- * 
- * Copyright (c) 2000-2009, Linden Research, Inc.
- * 
  * Second Life Viewer Source Code
- * The source code in this file ("Source Code") is provided by Linden Lab
+ * Copyright (C) 2010, Linden Research, Inc.
- * to you under the terms of the GNU General Public License, version 2.0
- * ("GPL"), unless you have obtained a separate licensing agreement
- * ("Other License"), formally executed by you and Linden Lab.  Terms of
- * the GPL can be found in doc/GPL-license.txt in this distribution, or
- * online at http://secondlifegrid.net/programs/open_source/licensing/gplv2
  * 
- * There are special exceptions to the terms and conditions of the GPL as
+ * This library is free software; you can redistribute it and/or
- * it is applied to this Source Code. View the full text of the exception
+ * modify it under the terms of the GNU Lesser General Public
- * in the file doc/FLOSS-exception.txt in this software distribution, or
+ * License as published by the Free Software Foundation;
- * online at
+ * version 2.1 of the License only.
- * http://secondlifegrid.net/programs/open_source/licensing/flossexception
  * 
- * By copying, modifying or distributing this software, you acknowledge
+ * This library is distributed in the hope that it will be useful,
- * that you have read and understood your obligations described above,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
- * and agree to abide by those obligations.
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * Lesser General Public License for more details.
  * 
- * ALL LINDEN LAB SOURCE CODE IS PROVIDED "AS IS." LINDEN LAB MAKES NO
+ * You should have received a copy of the GNU Lesser General Public
- * WARRANTIES, EXPRESS, IMPLIED OR OTHERWISE, REGARDING ITS ACCURACY,
+ * License along with this library; if not, write to the Free Software
- * COMPLETENESS OR PERFORMANCE.
+ * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA  02110-1301  USA
+ *
+ * Linden Research, Inc., 945 Battery Street, San Francisco, CA  94111  USA
  * $/LicenseInfo$
  */
 

indra/llmath/v3color.h

@@ -2,31 +2,25 @@
  * @file v3color.h
  * @brief LLColor3 class header file.
  *
- * $LicenseInfo:firstyear=2001&license=viewergpl$
+ * $LicenseInfo:firstyear=2001&license=viewerlgpl$
- * 
- * Copyright (c) 2001-2009, Linden Research, Inc.
- * 
  * Second Life Viewer Source Code
- * The source code in this file ("Source Code") is provided by Linden Lab
+ * Copyright (C) 2010, Linden Research, Inc.
- * to you under the terms of the GNU General Public License, version 2.0
- * ("GPL"), unless you have obtained a separate licensing agreement
- * ("Other License"), formally executed by you and Linden Lab.  Terms of
- * the GPL can be found in doc/GPL-license.txt in this distribution, or
- * online at http://secondlifegrid.net/programs/open_source/licensing/gplv2
  * 
- * There are special exceptions to the terms and conditions of the GPL as
+ * This library is free software; you can redistribute it and/or
- * it is applied to this Source Code. View the full text of the exception
+ * modify it under the terms of the GNU Lesser General Public
- * in the file doc/FLOSS-exception.txt in this software distribution, or
+ * License as published by the Free Software Foundation;
- * online at
+ * version 2.1 of the License only.
- * http://secondlifegrid.net/programs/open_source/licensing/flossexception
  * 
- * By copying, modifying or distributing this software, you acknowledge
+ * This library is distributed in the hope that it will be useful,
- * that you have read and understood your obligations described above,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
- * and agree to abide by those obligations.
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * Lesser General Public License for more details.
  * 
- * ALL LINDEN LAB SOURCE CODE IS PROVIDED "AS IS." LINDEN LAB MAKES NO
+ * You should have received a copy of the GNU Lesser General Public
- * WARRANTIES, EXPRESS, IMPLIED OR OTHERWISE, REGARDING ITS ACCURACY,
+ * License along with this library; if not, write to the Free Software
- * COMPLETENESS OR PERFORMANCE.
+ * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA  02110-1301  USA
+ *
+ * Linden Research, Inc., 945 Battery Street, San Francisco, CA  94111  USA
  * $/LicenseInfo$
  */
 
@@ -39,6 +33,7 @@ class LLVector4;
 #include "llerror.h"
 #include "llmath.h"
 #include "llsd.h"
+#include <string.h>
 
 //  LLColor3 = |r g b|
 

indra/llmath/v3dmath.cpp

@@ -2,31 +2,25 @@
  * @file v3dmath.cpp
  * @brief LLVector3d class implementation.
  *
- * $LicenseInfo:firstyear=2000&license=viewergpl$
+ * $LicenseInfo:firstyear=2000&license=viewerlgpl$
- * 
- * Copyright (c) 2000-2009, Linden Research, Inc.
- * 
  * Second Life Viewer Source Code
- * The source code in this file ("Source Code") is provided by Linden Lab
+ * Copyright (C) 2010, Linden Research, Inc.
- * to you under the terms of the GNU General Public License, version 2.0
- * ("GPL"), unless you have obtained a separate licensing agreement
- * ("Other License"), formally executed by you and Linden Lab.  Terms of
- * the GPL can be found in doc/GPL-license.txt in this distribution, or
- * online at http://secondlifegrid.net/programs/open_source/licensing/gplv2
  * 
- * There are special exceptions to the terms and conditions of the GPL as
+ * This library is free software; you can redistribute it and/or
- * it is applied to this Source Code. View the full text of the exception
+ * modify it under the terms of the GNU Lesser General Public
- * in the file doc/FLOSS-exception.txt in this software distribution, or
+ * License as published by the Free Software Foundation;
- * online at
+ * version 2.1 of the License only.
- * http://secondlifegrid.net/programs/open_source/licensing/flossexception
  * 
- * By copying, modifying or distributing this software, you acknowledge
+ * This library is distributed in the hope that it will be useful,
- * that you have read and understood your obligations described above,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
- * and agree to abide by those obligations.
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * Lesser General Public License for more details.
+ *
+ * You should have received a copy of the GNU Lesser General Public
+ * License along with this library; if not, write to the Free Software
+ * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA  02110-1301  USA
  * 
- * ALL LINDEN LAB SOURCE CODE IS PROVIDED "AS IS." LINDEN LAB MAKES NO
+ * Linden Research, Inc., 945 Battery Street, San Francisco, CA  94111  USA
- * WARRANTIES, EXPRESS, IMPLIED OR OTHERWISE, REGARDING ITS ACCURACY,
- * COMPLETENESS OR PERFORMANCE.
  * $/LicenseInfo$
  */
 

indra/llmath/v3dmath.h

@@ -72,17 +72,22 @@ class LLVector3d
 		BOOL		clamp(const F64 min, const F64 max);		// Clamps all values to (min,max), returns TRUE if data changed
 		BOOL		abs();						// sets all values to absolute value of original value (first octant), returns TRUE if changed
 
-		inline const LLVector3d&	clearVec();		// Clears LLVector3d to (0, 0, 0, 1)
+		inline const LLVector3d&	clear();		// Clears LLVector3d to (0, 0, 0, 1)
+		inline const LLVector3d&	clearVec();		// deprecated
 		inline const LLVector3d&	setZero();		// Zero LLVector3d to (0, 0, 0, 0)
 		inline const LLVector3d&	zeroVec();		// deprecated
-		inline const LLVector3d&	setVec(const F64 x, const F64 y, const F64 z);	// Sets LLVector3d to (x, y, z, 1)
+		inline const LLVector3d&	set(const F64 x, const F64 y, const F64 z);	// Sets LLVector3d to (x, y, z, 1)
-		inline const LLVector3d&	setVec(const LLVector3d &vec);	// Sets LLVector3d to vec
+		inline const LLVector3d&	set(const LLVector3d &vec);	// Sets LLVector3d to vec
-		inline const LLVector3d&	setVec(const F64 *vec);			// Sets LLVector3d to vec
+		inline const LLVector3d&	set(const F64 *vec);		// Sets LLVector3d to vec
-		inline const LLVector3d&	setVec(const LLVector3 &vec);
+		inline const LLVector3d&	set(const LLVector3 &vec);
+		inline const LLVector3d&	setVec(const F64 x, const F64 y, const F64 z);	// deprecated
+		inline const LLVector3d&	setVec(const LLVector3d &vec);	// deprecated
+		inline const LLVector3d&	setVec(const F64 *vec);			// deprecated
+		inline const LLVector3d&	setVec(const LLVector3 &vec);	// deprecated
 
-		F64		magVec() const;				// Returns magnitude of LLVector3d
+		F64		magVec() const;				// deprecated
-		F64		magVecSquared() const;		// Returns magnitude squared of LLVector3d
+		F64		magVecSquared() const;		// deprecated
-		inline F64		normVec();					// Normalizes and returns the magnitude of LLVector3d
+		inline F64		normVec();					// deprecated
 
 		F64 length() const;			// Returns magnitude of LLVector3d
 		F64 lengthSquared() const;	// Returns magnitude squared of LLVector3d
@@ -127,7 +132,15 @@ class LLVector3d
 
 typedef LLVector3d LLGlobalVec;
 
-const LLVector3d &LLVector3d::setVec(const LLVector3 &vec)
+inline const LLVector3d &LLVector3d::set(const LLVector3 &vec)
+{
+	mdV[0] = vec.mV[0];
+	mdV[1] = vec.mV[1];
+	mdV[2] = vec.mV[2];
+	return *this;
+}
+
+inline const LLVector3d &LLVector3d::setVec(const LLVector3 &vec)
 {
 	mdV[0] = vec.mV[0];
 	mdV[1] = vec.mV[1];
@@ -184,6 +197,14 @@ inline BOOL LLVector3d::isFinite() const
 
 // Clear and Assignment Functions
 
+inline const LLVector3d&	LLVector3d::clear(void)
+{
+	mdV[0] = 0.f;
+	mdV[1] = 0.f;
+	mdV[2]= 0.f;
+	return (*this);
+}
+
 inline const LLVector3d&	LLVector3d::clearVec(void)
 {
 	mdV[0] = 0.f;
@@ -208,6 +229,30 @@ inline const LLVector3d&	LLVector3d::zeroVec(void)
 	return (*this);
 }
 
+inline const LLVector3d&	LLVector3d::set(const F64 x, const F64 y, const F64 z)
+{
+	mdV[VX] = x;
+	mdV[VY] = y;
+	mdV[VZ] = z;
+	return (*this);
+}
+
+inline const LLVector3d&	LLVector3d::set(const LLVector3d &vec)
+{
+	mdV[0] = vec.mdV[0];
+	mdV[1] = vec.mdV[1];
+	mdV[2] = vec.mdV[2];
+	return (*this);
+}
+
+inline const LLVector3d&	LLVector3d::set(const F64 *vec)
+{
+	mdV[0] = vec[0];
+	mdV[1] = vec[1];
+	mdV[2] = vec[2];
+	return (*this);
+}
+
 inline const LLVector3d&	LLVector3d::setVec(const F64 x, const F64 y, const F64 z)
 {
 	mdV[VX] = x;
@@ -472,4 +517,15 @@ inline LLVector3d projected_vec(const LLVector3d &a, const LLVector3d &b)
 	return project_axis * (a * project_axis);
 }
 
+inline LLVector3d inverse_projected_vec(const LLVector3d& a, const LLVector3d& b)
+{
+	LLVector3d normalized_a = a;
+	normalized_a.normalize();
+	LLVector3d normalized_b = b;
+	F64 b_length = normalized_b.normalize();
+
+	F64 dot_product = normalized_a * normalized_b;
+	return normalized_a * (b_length / dot_product);
+}
+
 #endif // LL_V3DMATH_H

indra/llmath/v3math.cpp

@@ -2,31 +2,25 @@
  * @file v3math.cpp
  * @brief LLVector3 class implementation.
  *
- * $LicenseInfo:firstyear=2000&license=viewergpl$
+ * $LicenseInfo:firstyear=2000&license=viewerlgpl$
- * 
- * Copyright (c) 2000-2009, Linden Research, Inc.
- * 
  * Second Life Viewer Source Code
- * The source code in this file ("Source Code") is provided by Linden Lab
+ * Copyright (C) 2010, Linden Research, Inc.
- * to you under the terms of the GNU General Public License, version 2.0
+ *
- * ("GPL"), unless you have obtained a separate licensing agreement
+ * This library is free software; you can redistribute it and/or
- * ("Other License"), formally executed by you and Linden Lab.  Terms of
+ * modify it under the terms of the GNU Lesser General Public
- * the GPL can be found in doc/GPL-license.txt in this distribution, or
+ * License as published by the Free Software Foundation;
- * online at http://secondlifegrid.net/programs/open_source/licensing/gplv2
+ * version 2.1 of the License only.
  * 
- * There are special exceptions to the terms and conditions of the GPL as
+ * This library is distributed in the hope that it will be useful,
- * it is applied to this Source Code. View the full text of the exception
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
- * in the file doc/FLOSS-exception.txt in this software distribution, or
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
- * online at
+ * Lesser General Public License for more details.
- * http://secondlifegrid.net/programs/open_source/licensing/flossexception
  * 
- * By copying, modifying or distributing this software, you acknowledge
+ * You should have received a copy of the GNU Lesser General Public
- * that you have read and understood your obligations described above,
+ * License along with this library; if not, write to the Free Software
- * and agree to abide by those obligations.
+ * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA  02110-1301  USA
  * 
- * ALL LINDEN LAB SOURCE CODE IS PROVIDED "AS IS." LINDEN LAB MAKES NO
+ * Linden Research, Inc., 945 Battery Street, San Francisco, CA  94111  USA
- * WARRANTIES, EXPRESS, IMPLIED OR OTHERWISE, REGARDING ITS ACCURACY,
- * COMPLETENESS OR PERFORMANCE.
  * $/LicenseInfo$
  */
 

indra/llmath/v3math.h

@@ -159,6 +159,7 @@ F32	dist_vec(const LLVector3 &a, const LLVector3 &b);		// Returns distance betwe
 F32	dist_vec_squared(const LLVector3 &a, const LLVector3 &b);// Returns distance squared between a and b
 F32	dist_vec_squared2D(const LLVector3 &a, const LLVector3 &b);// Returns distance squared between a and b ignoring Z component
 LLVector3 projected_vec(const LLVector3 &a, const LLVector3 &b); // Returns vector a projected on vector b
+LLVector3 inverse_projected_vec(const LLVector3 &a, const LLVector3 &b); // Returns vector a scaled such that projected_vec(inverse_projected_vec(a, b), b) == b;
 LLVector3 parallel_component(const LLVector3 &a, const LLVector3 &b); // Returns vector a projected on vector b (same as projected_vec)
 LLVector3 orthogonal_component(const LLVector3 &a, const LLVector3 &b); // Returns component of vector a not parallel to vector b (same as projected_vec)
 LLVector3 lerp(const LLVector3 &a, const LLVector3 &b, F32 u); // Returns a vector that is a linear interpolation between a and b
@@ -490,9 +491,27 @@ inline F32	dist_vec_squared2D(const LLVector3 &a, const LLVector3 &b)
 
 inline LLVector3 projected_vec(const LLVector3 &a, const LLVector3 &b)
 {
-	LLVector3 project_axis = b;
+	F32 bb = b * b;
-	project_axis.normalize();
+	if (bb > FP_MAG_THRESHOLD * FP_MAG_THRESHOLD)
-	return project_axis * (a * project_axis);
+	{
+		return ((a * b) / bb) * b;
+	}
+	else
+	{
+		return b.zero;
+	}
+}
+
+inline LLVector3 inverse_projected_vec(const LLVector3& a, const LLVector3& b)
+{
+	LLVector3 normalized_a = a;
+	normalized_a.normalize();
+	LLVector3 normalized_b = b;
+	F32 b_length = normalized_b.normalize();
+
+	F32 dot_product = normalized_a * normalized_b;
+	//NB: if a _|_ b, then returns an infinite vector
+	return normalized_a * (b_length / dot_product);
 }
 
 inline LLVector3 parallel_component(const LLVector3 &a, const LLVector3 &b)
@@ -556,15 +575,13 @@ inline void update_min_max(LLVector3& min, LLVector3& max, const F32* pos)
 
 inline F32 angle_between(const LLVector3& a, const LLVector3& b)
 {
-	LLVector3 an = a;
+	F32 ab = a * b; // dotproduct
-	LLVector3 bn = b;
+	if (ab == -0.0f)
-	an.normalize();
+	{
-	bn.normalize();
+		ab = 0.0f; // get rid of negative zero
-	F32 cosine = an * bn;
+	}
-	F32 angle = (cosine >= 1.0f) ? 0.0f :
+	LLVector3 c = a % b; // crossproduct
-				(cosine <= -1.0f) ? F_PI :
+	return atan2f(sqrtf(c * c), ab); // return the angle
-				(F32)acos(cosine);
-	return angle;
 }
 
 inline BOOL are_parallel(const LLVector3 &a, const LLVector3 &b, F32 epsilon)

indra/llmath/v4coloru.cpp

@@ -2,31 +2,25 @@
  * @file v4coloru.cpp
  * @brief LLColor4U class implementation.
  *
- * $LicenseInfo:firstyear=2001&license=viewergpl$
+ * $LicenseInfo:firstyear=2001&license=viewerlgpl$
- * 
- * Copyright (c) 2001-2009, Linden Research, Inc.
- * 
  * Second Life Viewer Source Code
- * The source code in this file ("Source Code") is provided by Linden Lab
+ * Copyright (C) 2010, Linden Research, Inc.
- * to you under the terms of the GNU General Public License, version 2.0
+ *
- * ("GPL"), unless you have obtained a separate licensing agreement
+ * This library is free software; you can redistribute it and/or
- * ("Other License"), formally executed by you and Linden Lab.  Terms of
+ * modify it under the terms of the GNU Lesser General Public
- * the GPL can be found in doc/GPL-license.txt in this distribution, or
+ * License as published by the Free Software Foundation;
- * online at http://secondlifegrid.net/programs/open_source/licensing/gplv2
+ * version 2.1 of the License only.
  * 
- * There are special exceptions to the terms and conditions of the GPL as
+ * This library is distributed in the hope that it will be useful,
- * it is applied to this Source Code. View the full text of the exception
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
- * in the file doc/FLOSS-exception.txt in this software distribution, or
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
- * online at
+ * Lesser General Public License for more details.
- * http://secondlifegrid.net/programs/open_source/licensing/flossexception
  * 
- * By copying, modifying or distributing this software, you acknowledge
+ * You should have received a copy of the GNU Lesser General Public
- * that you have read and understood your obligations described above,
+ * License along with this library; if not, write to the Free Software
- * and agree to abide by those obligations.
+ * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA  02110-1301  USA
  * 
- * ALL LINDEN LAB SOURCE CODE IS PROVIDED "AS IS." LINDEN LAB MAKES NO
+ * Linden Research, Inc., 945 Battery Street, San Francisco, CA  94111  USA
- * WARRANTIES, EXPRESS, IMPLIED OR OTHERWISE, REGARDING ITS ACCURACY,
- * COMPLETENESS OR PERFORMANCE.
  * $/LicenseInfo$
  */
 

indra/llmath/v4math.cpp

@@ -2,31 +2,25 @@
  * @file v4math.cpp
  * @brief LLVector4 class implementation.
  *
- * $LicenseInfo:firstyear=2000&license=viewergpl$
+ * $LicenseInfo:firstyear=2000&license=viewerlgpl$
- * 
- * Copyright (c) 2000-2009, Linden Research, Inc.
- * 
  * Second Life Viewer Source Code
- * The source code in this file ("Source Code") is provided by Linden Lab
+ * Copyright (C) 2010, Linden Research, Inc.
- * to you under the terms of the GNU General Public License, version 2.0
- * ("GPL"), unless you have obtained a separate licensing agreement
- * ("Other License"), formally executed by you and Linden Lab.  Terms of
- * the GPL can be found in doc/GPL-license.txt in this distribution, or
- * online at http://secondlifegrid.net/programs/open_source/licensing/gplv2
  * 
- * There are special exceptions to the terms and conditions of the GPL as
+ * This library is free software; you can redistribute it and/or
- * it is applied to this Source Code. View the full text of the exception
+ * modify it under the terms of the GNU Lesser General Public
- * in the file doc/FLOSS-exception.txt in this software distribution, or
+ * License as published by the Free Software Foundation;
- * online at
+ * version 2.1 of the License only.
- * http://secondlifegrid.net/programs/open_source/licensing/flossexception
  * 
- * By copying, modifying or distributing this software, you acknowledge
+ * This library is distributed in the hope that it will be useful,
- * that you have read and understood your obligations described above,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
- * and agree to abide by those obligations.
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * Lesser General Public License for more details.
  * 
- * ALL LINDEN LAB SOURCE CODE IS PROVIDED "AS IS." LINDEN LAB MAKES NO
+ * You should have received a copy of the GNU Lesser General Public
- * WARRANTIES, EXPRESS, IMPLIED OR OTHERWISE, REGARDING ITS ACCURACY,
+ * License along with this library; if not, write to the Free Software
- * COMPLETENESS OR PERFORMANCE.
+ * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA  02110-1301  USA
+ *
+ * Linden Research, Inc., 945 Battery Street, San Francisco, CA  94111  USA
  * $/LicenseInfo$
  */
 

indra/llmath/v4math.h

@@ -2,31 +2,25 @@
  * @file v4math.h
  * @brief LLVector4 class header file.
  *
- * $LicenseInfo:firstyear=2000&license=viewergpl$
+ * $LicenseInfo:firstyear=2000&license=viewerlgpl$
- * 
- * Copyright (c) 2000-2009, Linden Research, Inc.
- * 
  * Second Life Viewer Source Code
- * The source code in this file ("Source Code") is provided by Linden Lab
+ * Copyright (C) 2010, Linden Research, Inc.
- * to you under the terms of the GNU General Public License, version 2.0
- * ("GPL"), unless you have obtained a separate licensing agreement
- * ("Other License"), formally executed by you and Linden Lab.  Terms of
- * the GPL can be found in doc/GPL-license.txt in this distribution, or
- * online at http://secondlifegrid.net/programs/open_source/licensing/gplv2
  * 
- * There are special exceptions to the terms and conditions of the GPL as
+ * This library is free software; you can redistribute it and/or
- * it is applied to this Source Code. View the full text of the exception
+ * modify it under the terms of the GNU Lesser General Public
- * in the file doc/FLOSS-exception.txt in this software distribution, or
+ * License as published by the Free Software Foundation;
- * online at
+ * version 2.1 of the License only.
- * http://secondlifegrid.net/programs/open_source/licensing/flossexception
  * 
- * By copying, modifying or distributing this software, you acknowledge
+ * This library is distributed in the hope that it will be useful,
- * that you have read and understood your obligations described above,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
- * and agree to abide by those obligations.
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * Lesser General Public License for more details.
  * 
- * ALL LINDEN LAB SOURCE CODE IS PROVIDED "AS IS." LINDEN LAB MAKES NO
+ * You should have received a copy of the GNU Lesser General Public
- * WARRANTIES, EXPRESS, IMPLIED OR OTHERWISE, REGARDING ITS ACCURACY,
+ * License along with this library; if not, write to the Free Software
- * COMPLETENESS OR PERFORMANCE.
+ * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA  02110-1301  USA
+ *
+ * Linden Research, Inc., 945 Battery Street, San Francisco, CA  94111  USA
  * $/LicenseInfo$
  */
 

indra/llmath/xform.cpp

@@ -1,31 +1,25 @@
 /** 
  * @file xform.cpp
  *
- * $LicenseInfo:firstyear=2001&license=viewergpl$
+ * $LicenseInfo:firstyear=2001&license=viewerlgpl$
- * 
- * Copyright (c) 2001-2009, Linden Research, Inc.
- * 
  * Second Life Viewer Source Code
- * The source code in this file ("Source Code") is provided by Linden Lab
+ * Copyright (C) 2010, Linden Research, Inc.
- * to you under the terms of the GNU General Public License, version 2.0
+ *
- * ("GPL"), unless you have obtained a separate licensing agreement
+ * This library is free software; you can redistribute it and/or
- * ("Other License"), formally executed by you and Linden Lab.  Terms of
+ * modify it under the terms of the GNU Lesser General Public
- * the GPL can be found in doc/GPL-license.txt in this distribution, or
+ * License as published by the Free Software Foundation;
- * online at http://secondlifegrid.net/programs/open_source/licensing/gplv2
+ * version 2.1 of the License only.
  * 
- * There are special exceptions to the terms and conditions of the GPL as
+ * This library is distributed in the hope that it will be useful,
- * it is applied to this Source Code. View the full text of the exception
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
- * in the file doc/FLOSS-exception.txt in this software distribution, or
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
- * online at
+ * Lesser General Public License for more details.
- * http://secondlifegrid.net/programs/open_source/licensing/flossexception
  * 
- * By copying, modifying or distributing this software, you acknowledge
+ * You should have received a copy of the GNU Lesser General Public
- * that you have read and understood your obligations described above,
+ * License along with this library; if not, write to the Free Software
- * and agree to abide by those obligations.
+ * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA  02110-1301  USA
  * 
- * ALL LINDEN LAB SOURCE CODE IS PROVIDED "AS IS." LINDEN LAB MAKES NO
+ * Linden Research, Inc., 945 Battery Street, San Francisco, CA  94111  USA
- * WARRANTIES, EXPRESS, IMPLIED OR OTHERWISE, REGARDING ITS ACCURACY,
- * COMPLETENESS OR PERFORMANCE.
  * $/LicenseInfo$
  */
 
@@ -42,7 +36,7 @@ LLXform::~LLXform()
 {
 }
 
-// Link optimization - don't inline these llwarns
+// Link optimization - don't inline these LL_WARNS()
 void LLXform::warn(const char* const msg)
 {
 	llwarns << msg << llendl;
@@ -96,30 +90,29 @@ void LLXformMatrix::updateMatrix(BOOL update_bounds)
 {
 	update();
 
-	mWorldMatrix.initAll(mScale, mWorldRotation, mWorldPosition);
+	LLMatrix4 world_matrix;
+	world_matrix.initAll(mScale, mWorldRotation, mWorldPosition);
+	mWorldMatrix.loadu(world_matrix);
 
 	if (update_bounds && (mChanged & MOVED))
 	{
-		mMin.mV[0] = mMax.mV[0] = mWorldMatrix.mMatrix[3][0];
+		mMax = mMin = mWorldMatrix.getRow<3>();
-		mMin.mV[1] = mMax.mV[1] = mWorldMatrix.mMatrix[3][1];
-		mMin.mV[2] = mMax.mV[2] = mWorldMatrix.mMatrix[3][2];
 
-		F32 f0 = (fabs(mWorldMatrix.mMatrix[0][0])+fabs(mWorldMatrix.mMatrix[1][0])+fabs(mWorldMatrix.mMatrix[2][0])) * 0.5f;
+		LLVector4a total_sum,sum1,sum2;
-		F32 f1 = (fabs(mWorldMatrix.mMatrix[0][1])+fabs(mWorldMatrix.mMatrix[1][1])+fabs(mWorldMatrix.mMatrix[2][1])) * 0.5f;
+		total_sum.setAbs(mWorldMatrix.getRow<0>());
-		F32 f2 = (fabs(mWorldMatrix.mMatrix[0][2])+fabs(mWorldMatrix.mMatrix[1][2])+fabs(mWorldMatrix.mMatrix[2][2])) * 0.5f;
+		sum1.setAbs(mWorldMatrix.getRow<1>());
+		sum2.setAbs(mWorldMatrix.getRow<2>());
+		sum1.add(sum2);
+		total_sum.add(sum1);
+		total_sum.mul(.5f);
 
-		mMin.mV[0] -= f0; 
+		mMax.add(total_sum);
-		mMin.mV[1] -= f1; 
+		mMin.sub(total_sum);
-		mMin.mV[2] -= f2; 
-
-		mMax.mV[0] += f0; 
-		mMax.mV[1] += f1; 
-		mMax.mV[2] += f2; 
 	}
 }
 
 void LLXformMatrix::getMinMax(LLVector3& min, LLVector3& max) const
 {
-	min = mMin;
+	min.set(mMin.getF32ptr());
-	max = mMax;
+	max.set(mMax.getF32ptr());
 }

indra/llmath/xform.h

@@ -28,6 +28,7 @@
 
 #include "v3math.h"
 #include "m4math.h"
+#include "llmatrix4a.h"
 #include "llquaternion.h"
 
 const F32 MAX_OBJECT_Z 		= 4096.f; // should match REGION_HEIGHT_METERS, Pre-havok4: 768.f
@@ -103,9 +104,9 @@ public:
 	inline void setRotation(const F32 x, const F32 y, const F32 z, const F32 s);
 
 	// Above functions must be inline for speed, but also
-	// need to emit warnings.  llwarns causes inline LLError::CallSite
+	// need to emit warnings.  LL_WARNS() causes inline LLError::CallSite
 	// static objects that make more work for the linker.
-	// Avoid inline llwarns by calling this function.
+	// Avoid inline LL_WARNS() by calling this function.
 	void warn(const char* const msg);
 	
 	void 		setChanged(const U32 bits)					{ mChanged |= bits; }
@@ -130,20 +131,21 @@ public:
 	const LLVector3&	getWorldPosition() const	{ return mWorldPosition; }
 };
 
+LL_ALIGN_PREFIX(16)
 class LLXformMatrix : public LLXform
 {
 public:
 	LLXformMatrix() : LLXform() {};
 	virtual ~LLXformMatrix();
 
-	const LLMatrix4&    getWorldMatrix() const      { return mWorldMatrix; }
+	const LLMatrix4a&    getWorldMatrix() const      { return mWorldMatrix; }
-	void setWorldMatrix (const LLMatrix4& mat)   { mWorldMatrix = mat; }
+	void setWorldMatrix (const LLMatrix4a& mat)   { mWorldMatrix = mat; }
 
 	void init()
 	{
 		mWorldMatrix.setIdentity();
-		mMin.clearVec();
+		mMin.clear();
-		mMax.clearVec();
+		mMax.clear();
 
 		LLXform::init();
 	}
@@ -153,11 +155,11 @@ public:
 	void getMinMax(LLVector3& min,LLVector3& max) const;
 
 protected:
-	LLMatrix4	mWorldMatrix;
+	LL_ALIGN_16(LLMatrix4a	mWorldMatrix);
-	LLVector3	mMin;
+	LL_ALIGN_16(LLVector4a	mMin);
-	LLVector3	mMax;
+	LL_ALIGN_16(LLVector4a	mMax);
 
-};
+} LL_ALIGN_POSTFIX(16);
 
 BOOL LLXform::setParent(LLXform* parent)
 {

indra/llmessage/aicurlperservice.cpp

@@ -610,7 +610,7 @@ void AIPerService::purge(void)
 	  per_service_w->mCapabilityType[i].mQueuedRequests.clear();
 	  if (is_approved((AICapabilityType)i))
 	  {
-		llassert(total_queued_w->approved >= s);
+		llassert(total_queued_w->approved >= (S32)s);
 		total_queued_w->approved -= s;
 	  }
 	}

indra/llmessage/aihttptimeoutpolicy.cpp

@@ -936,6 +936,7 @@ P(fetchScriptLimitsRegionSummaryResponder);
 P(fnPtrResponder);
 P(floaterPermsResponder);
 P2(gamingDataReceived,							transfer_22s_connect_10s);
+P(groupBanDataResponder);
 P2(groupMemberDataResponder,					transfer_300s);
 P2(groupProposalBallotResponder,				transfer_300s);
 P(homeLocationResponder);

indra/llmessage/llavatarnamecache.cpp

@@ -794,6 +794,7 @@ void LLAvatarNameCache::setUseDisplayNames(bool use)
 	if (use != sUseDisplayNames)
 	{
 		sUseDisplayNames = use;
+		LL_DEBUGS("AvNameCache") << "Display names are now: " << (use ? "on" : "off") << LL_ENDL;
 		// flush our cache
 		sCache.clear();
 

indra/llmessage/llregionflags.h

@@ -78,6 +78,8 @@ const U64 REGION_FLAGS_DENY_ANONYMOUS			= (1 << 23);
 
 const U64 REGION_FLAGS_ALLOW_PARCEL_CHANGES		= (1 << 26);
 
+const U64 REGION_FLAGS_BLOCK_FLYOVER = (1 << 27);
+
 const U64 REGION_FLAGS_ALLOW_VOICE = (1 << 28);
 
 const U64 REGION_FLAGS_BLOCK_PARCEL_SEARCH = (1 << 29);

indra/llmessage/lltemplatemessagereader.cpp

@@ -91,15 +91,17 @@ void LLTemplateMessageReader::getData(const char *blockname, const char *varname
 	}
 
 	LLMsgBlkData *msg_block_data = iter->second;
-	LLMsgVarData& vardata = msg_block_data->mMemberVarData[vnamep];
+	LLMsgBlkData::msg_var_data_map_t &var_data_map = msg_block_data->mMemberVarData;
 
-	if (!vardata.getName())
+	if (var_data_map.find(vnamep) == var_data_map.end())
 	{
 		llerrs << "Variable "<< vnamep << " not in message "
 			<< mCurrentRMessageData->mName<< " block " << bnamep << llendl;
 		return;
 	}
 
+	LLMsgVarData& vardata = msg_block_data->mMemberVarData[vnamep];
+
 	if (size && size != vardata.getSize())
 	{
 		llerrs << "Msg " << mCurrentRMessageData->mName 
@@ -284,7 +286,7 @@ void LLTemplateMessageReader::getU8(const char *block, const char *var,
 void LLTemplateMessageReader::getBOOL(const char *block, const char *var, 
 										  BOOL &b, S32 blocknum )
 {
-	U8 value;
+	U8 value(0);
 	getData(block, var, &value, sizeof(U8), blocknum);
 	b = (BOOL) value;
 }

indra/llmessage/llurlrequest.cpp

@@ -78,7 +78,7 @@ LLURLRequest::LLURLRequest(LLURLRequest::ERequestAction action, std::string cons
 	LLHTTPClient::ResponderPtr responder, AIHTTPHeaders& headers, AIPerService::Approvement* approved,
 	bool keepalive, bool is_auth, bool compression) :
     mAction(action), mURL(url), mKeepAlive(keepalive), mIsAuth(is_auth), mNoCompression(!compression),
-	mBody(body), mResponder(responder), mHeaders(headers), mResponderNameCache(responder ? responder->getName() : "<uninitialized>")
+	mBody(body), mResponder(responder), mHeaders(headers), mResponderNameCache(std::string("LLURLRequest:") + std::string(responder ? responder->getName() : "<uninitialized>"))
 {
 	if (approved)
 	{
@@ -276,32 +276,4 @@ bool LLURLRequest::configure(AICurlEasyRequest_wat const& curlEasyRequest_w)
 	return rv;
 }
 
-// Called from AIStateMachine::mainloop, but put here because we don't want to include llurlrequest.h there of course.
-void print_statemachine_diagnostics(U64 total_clocks, U64 max_delta, AIEngine::queued_type::const_reference slowest_element)
-{
-  AIStateMachine const& slowest_state_machine = slowest_element.statemachine();
-  LLURLRequest const* request = dynamic_cast<LLURLRequest const*>(&slowest_state_machine);
-  F64 const tfactor = 1000 / calc_clock_frequency();
-  std::ostringstream msg;
-  if (total_clocks > max_delta)
-  {
-	  msg << "AIStateMachine::mainloop did run for " << (total_clocks * tfactor) << " ms. The slowest ";
-  }
-  else
-  {
-	  msg << "AIStateMachine::mainloop: A ";
-  }
-  msg << "state machine ";
-  if (request)
-  {
-	  msg << "(" << request->getResponderName() << ") ";
-  }
-  msg << "ran for " << (max_delta * tfactor) << " ms";
-  if (slowest_state_machine.getRuntime() > max_delta)
-  {
-	  msg << " (" << (slowest_state_machine.getRuntime() * tfactor) << " ms in total now)";
-  }
-  msg << ".";
-  llwarns << msg.str() << llendl;
-}
 

indra/llmessage/llurlrequest.h

@@ -72,7 +72,7 @@ class LLURLRequest : public AICurlEasyRequestStateMachine {
 	/**
 	 * @brief Cached value of responder->getName() as passed to the constructor.
 	 */
-	char const* getResponderName(void) const { return mResponderNameCache; }
+	/*virtual*/ const char* getName() const { return mResponderNameCache.c_str(); }
 
   protected:
 	// Call abort(), not delete.
@@ -113,7 +113,7 @@ class LLURLRequest : public AICurlEasyRequestStateMachine {
 	U32 mBodySize;
 	LLHTTPClient::ResponderPtr mResponder;
 	AIHTTPHeaders mHeaders;
-	char const* mResponderNameCache;
+	std::string mResponderNameCache;
 
   protected:
 	// Handle initializing the object.

indra/llprimitive/llmodel.cpp

@@ -31,6 +31,7 @@
 #include "llconvexdecomposition.h"
 #include "llsdserialize.h"
 #include "llvector4a.h"
+#include "llmatrix4a.h"
 #if LL_MSVC
 #pragma warning (push)
 #pragma warning (disable : 4068)
@@ -172,6 +173,11 @@ LLModel::EModelStatus load_face_from_dom_triangles(std::vector<LLVolumeFace>& fa
 		return LLModel::BAD_ELEMENT;
 	}
 
+	if (!pos_source)
+	{
+		llwarns << "Unable to process mesh without position data; invalid model;  invalid model." << llendl;
+		return LLModel::BAD_ELEMENT;
+	}
 	
 	domPRef p = tri->getP();
 	domListOfUInts& idx = p->getValue();
@@ -181,19 +187,22 @@ LLModel::EModelStatus load_face_from_dom_triangles(std::vector<LLVolumeFace>& fa
 	domListOfFloats& tc = tc_source ? tc_source->getFloat_array()->getValue() : dummy ;
 	domListOfFloats& n = norm_source ? norm_source->getFloat_array()->getValue() : dummy ;
 
-	if (pos_source)
-	{
-		face.mExtents[0].set(v[0], v[1], v[2]);
-		face.mExtents[1].set(v[0], v[1], v[2]);
-	}
-	
 	LLVolumeFace::VertexMapData::PointMap point_map;
-	
+
 	U32 index_count  = idx.getCount();
 	U32 vertex_count = pos_source  ? v.getCount()  : 0;
 	U32 tc_count     = tc_source   ? tc.getCount() : 0;
 	U32 norm_count   = norm_source ? n.getCount()  : 0;
 
+	if (vertex_count == 0)
+	{
+		llwarns << "Unable to process mesh with empty position array; invalid model." << llendl;
+		return LLModel::BAD_ELEMENT;
+	}
+
+	face.mExtents[0].set(v[0], v[1], v[2]);
+	face.mExtents[1].set(v[0], v[1], v[2]);
+
 	for (U32 i = 0; i < index_count; i += idx_stride)
 	{
 		LLVolumeFace::VertexData cv;

indra/llrender/CMakeLists.txt

@@ -74,6 +74,11 @@ set(llrender_HEADER_FILES
 set_source_files_properties(${llrender_HEADER_FILES}
                             PROPERTIES HEADER_FILE_ONLY TRUE)
 
+# Workaround hack for clang bugs
+if (DARWIN)
+    set_property(SOURCE llgl.cpp PROPERTY COMPILE_FLAGS -O1)
+endif (DARWIN)
+
 list(APPEND llrender_SOURCE_FILES ${llrender_HEADER_FILES})
 
 add_library (llrender ${llrender_SOURCE_FILES})

indra/llrender/llcubemap.cpp

@@ -34,6 +34,7 @@
 #include "v3dmath.h"
 #include "m3math.h"
 #include "m4math.h"
+#include "llmatrix4a.h"
 
 #include "llrender.h"
 #include "llglslshader.h"
@@ -265,18 +266,19 @@ void LLCubeMap::setMatrix(S32 stage)
 		gGL.getTexUnit(stage)->activate();
 	}
 
-	LLVector3 x(gGLModelView+0);
+	LLVector3 x(gGLModelView.getRow<0>().getF32ptr());
-	LLVector3 y(gGLModelView+4);
+	LLVector3 y(gGLModelView.getRow<1>().getF32ptr());
-	LLVector3 z(gGLModelView+8);
+	LLVector3 z(gGLModelView.getRow<2>().getF32ptr());
 
 	LLMatrix3 mat3;
 	mat3.setRows(x,y,z);
-	LLMatrix4 trans(mat3);
+	LLMatrix4a trans;
+	trans.loadu(mat3);
 	trans.transpose();
 
 	gGL.matrixMode(LLRender::MM_TEXTURE);
 	gGL.pushMatrix();
-	gGL.loadMatrix((F32 *)trans.mMatrix);
+	gGL.loadMatrix(trans);
 	gGL.matrixMode(LLRender::MM_MODELVIEW);
 	
 	/*if (stage > 0)

indra/llrender/llgl.cpp

@@ -444,6 +444,7 @@ LLGLManager::LLGLManager() :
 	mHasDebugOutput(FALSE),
 
 	mHasAdaptiveVsync(FALSE),
+	mHasTextureSwizzle(FALSE),
 
 	mIsATI(FALSE),
 	mIsNVIDIA(FALSE),
@@ -1382,6 +1383,35 @@ void flush_glerror()
 	glGetError();
 }
 
+const std::string getGLErrorString(GLenum error)
+{
+	switch(error)
+	{
+	case GL_NO_ERROR:
+		return "No Error";
+	case GL_INVALID_ENUM:
+		return "Invalid Enum";
+	case GL_INVALID_VALUE:
+		return "Invalid Value";
+	case GL_INVALID_OPERATION:
+		return "Invalid Operation";
+	case GL_INVALID_FRAMEBUFFER_OPERATION:
+		return "Invalid Framebuffer Operation";
+	case GL_OUT_OF_MEMORY:
+		return "Out of Memory";
+	case GL_STACK_UNDERFLOW:
+		return "Stack Underflow";
+	case GL_STACK_OVERFLOW:
+		return "Stack Overflow";
+#ifdef GL_TABLE_TOO_LARGE
+	case GL_TABLE_TOO_LARGE:
+		return "Table too large";
+#endif
+	default:
+		return "UNKNOWN ERROR";
+	}
+}
+
 //this function outputs gl error to the log file, does not crash the code.
 void log_glerror()
 {
@@ -1394,17 +1424,8 @@ void log_glerror()
 	error = glGetError();
 	while (LL_UNLIKELY(error))
 	{
-		GLubyte const * gl_error_msg = gluErrorString(error);
+		std::string gl_error_msg = getGLErrorString(error);
-		if (NULL != gl_error_msg)
+		llwarns << "GL Error: 0x" << std::hex << error << std::dec << " GL Error String: " << gl_error_msg << llendl;			
-		{
-			llwarns << "GL Error: " << error << " GL Error String: " << gl_error_msg << llendl ;			
-		}
-		else
-		{
-			// gluErrorString returns NULL for some extensions' error codes.
-			// you'll probably have to grep for the number in glext.h.
-			llwarns << "GL Error: UNKNOWN 0x" << std::hex << error << std::dec << llendl;
-		}
 		error = glGetError();
 	}
 }
@@ -1418,27 +1439,13 @@ void do_assert_glerror()
 	while (LL_UNLIKELY(error))
 	{
 		quit = TRUE;
-		GLubyte const * gl_error_msg = gluErrorString(error);
+		
-		if (NULL != gl_error_msg)
+		std::string gl_error_msg = getGLErrorString(error);
+		LL_WARNS("RenderState") << "GL Error: 0x" << std::hex << error << std::dec << LL_ENDL;		
+		LL_WARNS("RenderState") << "GL Error String: " << gl_error_msg << LL_ENDL;
+		if (gDebugSession)
 		{
-			LL_WARNS("RenderState") << "GL Error:" << error<< LL_ENDL;
+			gFailLog << "GL Error: 0x" << std::hex << error << std::dec << " GL Error String: " << gl_error_msg << std::endl;
-			LL_WARNS("RenderState") << "GL Error String:" << gl_error_msg << LL_ENDL;
-
-			if (gDebugSession)
-			{
-				gFailLog << "GL Error:" << gl_error_msg << std::endl;
-			}
-		}
-		else
-		{
-			// gluErrorString returns NULL for some extensions' error codes.
-			// you'll probably have to grep for the number in glext.h.
-			LL_WARNS("RenderState") << "GL Error: UNKNOWN 0x" << std::hex << error << std::dec << LL_ENDL;
-
-			if (gDebugSession)
-			{
-				gFailLog << "GL Error: UNKNOWN 0x" << std::hex << error << std::dec << std::endl;
-			}
 		}
 		error = glGetError();
 	}
@@ -1662,10 +1669,6 @@ void LLGLState::checkTextureChannels(const std::string& msg)
 
 	GLint stackDepth = 0;
 
-	glh::matrix4f mat;
-	glh::matrix4f identity;
-	identity.identity();
-
 	for (GLint i = 1; i < gGLManager.mNumTextureUnits; i++)
 	{
 		gGL.getTexUnit(i)->activate();
@@ -1685,10 +1688,11 @@ void LLGLState::checkTextureChannels(const std::string& msg)
 			}
 		}
 
-		glGetFloatv(GL_TEXTURE_MATRIX, (GLfloat*) mat.m);
+		LLMatrix4a mat;
+		glGetFloatv(GL_TEXTURE_MATRIX, (GLfloat*) mat.mMatrix);
 		stop_glerror();
 
-		if (mat != identity)
+		if (!mat.isIdentity())
 		{
 			error = TRUE;
 			LL_WARNS("RenderState") << "Texture matrix in channel " << i << " corrupt." << LL_ENDL;
@@ -2179,7 +2183,7 @@ void parse_glsl_version(S32& major, S32& minor)
 	LLStringUtil::convertToS32(minor_str, minor);
 }
 
-LLGLUserClipPlane::LLGLUserClipPlane(const LLPlane& p, const glh::matrix4f& modelview, const glh::matrix4f& projection, bool apply)
+LLGLUserClipPlane::LLGLUserClipPlane(const LLPlane& p, const LLMatrix4a& modelview, const LLMatrix4a& projection, bool apply)
 {
 	mApply = apply;
 
@@ -2194,27 +2198,42 @@ LLGLUserClipPlane::LLGLUserClipPlane(const LLPlane& p, const glh::matrix4f& mode
 
 void LLGLUserClipPlane::setPlane(F32 a, F32 b, F32 c, F32 d)
 {
-	glh::matrix4f& P = mProjection;
+    LLMatrix4a& P = mProjection;
-	glh::matrix4f& M = mModelview;
+	LLMatrix4a& M = mModelview;
-    
-	glh::matrix4f invtrans_MVP = (P * M).inverse().transpose();
-    glh::vec4f oplane(a,b,c,d);
-    glh::vec4f cplane;
-    invtrans_MVP.mult_matrix_vec(oplane, cplane);
 
-    cplane /= fabs(cplane[2]); // normalize such that depth is not scaled
+	LLMatrix4a invtrans_MVP;
-    cplane[3] -= 1;
+	invtrans_MVP.setMul(P,M);
+	invtrans_MVP.invert();
+	invtrans_MVP.transpose();
 
-    if(cplane[2] < 0)
+	LLVector4a oplane(a,b,c,d);
-        cplane *= -1;
+	LLVector4a cplane;
+	LLVector4a cplane_splat;
+	LLVector4a cplane_neg;
+
+	invtrans_MVP.rotate4(oplane,cplane);
+	
+	cplane_splat.splat<2>(cplane);
+	cplane_splat.setAbs(cplane_splat);
+	cplane.div(cplane_splat);
+	cplane.sub(LLVector4a(0.f,0.f,0.f,1.f));
+
+	cplane_splat.splat<2>(cplane);
+	cplane_neg = cplane;
+	cplane_neg.negate();
+
+	cplane.setSelectWithMask( cplane_splat.lessThan( _mm_setzero_ps() ), cplane_neg, cplane );
+
+	LLMatrix4a suffix;
+	suffix.setIdentity();
+	suffix.setColumn<2>(cplane);
+	LLMatrix4a newP;
+	newP.setMul(suffix,P);
 
-    glh::matrix4f suffix;
-    suffix.set_row(2, cplane);
-    glh::matrix4f newP = suffix * P;
     gGL.matrixMode(LLRender::MM_PROJECTION);
 	gGL.pushMatrix();
-    gGL.loadMatrix(newP.m);
+    gGL.loadMatrix(newP);
-	gGLObliqueProjectionInverse = LLMatrix4(newP.inverse().transpose().m);
+	//gGLObliqueProjectionInverse = LLMatrix4(newP.inverse().transpose().m);
     gGL.matrixMode(LLRender::MM_MODELVIEW);
 }
 
@@ -2403,19 +2422,18 @@ void LLGLDepthTest::checkState()
 	}
 }
 
-LLGLSquashToFarClip::LLGLSquashToFarClip(glh::matrix4f P, U32 layer)
+LLGLSquashToFarClip::LLGLSquashToFarClip(const LLMatrix4a& P_in, U32 layer)
 {
-
+	LLMatrix4a P = P_in;
 	F32 depth = 0.99999f - 0.0001f * layer;
 
-	for (U32 i = 0; i < 4; i++)
+	LLVector4a col = P.getColumn<3>();
-	{
+	col.mul(depth);
-		P.element(2, i) = P.element(3, i) * depth;
+	P.setColumn<2>(col);
-	}
 
 	gGL.matrixMode(LLRender::MM_PROJECTION);
 	gGL.pushMatrix();
-	gGL.loadMatrix(P.m);
+	gGL.loadMatrix(P);
 	gGL.matrixMode(LLRender::MM_MODELVIEW);
 }
 

indra/llrender/llgl.h

@@ -38,12 +38,12 @@
 #include "llstring.h"
 #include "stdtypes.h"
 #include "v4math.h"
+#include "llmatrix4a.h"
 #include "llplane.h"
 #include "llgltypes.h"
 #include "llinstancetracker.h"
 
 #include "llglheaders.h"
-#include "glh/glh_linear.h"
 
 extern BOOL gDebugGL;
 extern BOOL gDebugSession;
@@ -321,21 +321,23 @@ public:
   Does not stack.
   Caches inverse of projection matrix used in gGLObliqueProjectionInverse
 */
+LL_ALIGN_PREFIX(16)
 class LLGLUserClipPlane 
 {
 public:
 	
-	LLGLUserClipPlane(const LLPlane& plane, const glh::matrix4f& modelview, const glh::matrix4f& projection, bool apply = true);
+	LLGLUserClipPlane(const LLPlane& plane, const LLMatrix4a& modelview, const LLMatrix4a& projection, bool apply = true);
 	~LLGLUserClipPlane();
 
 	void setPlane(F32 a, F32 b, F32 c, F32 d);
 
 private:
-	bool mApply;
 
-	glh::matrix4f mProjection;
+	LL_ALIGN_16(LLMatrix4a mProjection);
-	glh::matrix4f mModelview;
+	LL_ALIGN_16(LLMatrix4a mModelview);
-};
+
+	bool mApply;
+} LL_ALIGN_POSTFIX(16);
 
 /*
   Modify and load projection matrix to push depth values to far clip plane.
@@ -348,7 +350,7 @@ private:
 class LLGLSquashToFarClip
 {
 public:
-	LLGLSquashToFarClip(glh::matrix4f projection, U32 layer = 0);
+	LLGLSquashToFarClip(const LLMatrix4a& projection, U32 layer = 0);
 	~LLGLSquashToFarClip();
 };
 
@@ -455,8 +457,6 @@ public:
 	void wait();
 };
 
-extern LLMatrix4 gGLObliqueProjectionInverse;
-
 #include "llglstates.h"
 
 void init_glstates();

indra/llrender/llglheaders.h

@@ -41,7 +41,6 @@
 #   include "GL/glx.h"
 #   define  GL_GLEXT_PROTOTYPES 1
 #   include "GL/glext.h"
-#   include "GL/glu.h"
 #   include "GL/glx.h"
 #   define  GLX_GLXEXT_PROTOTYPES 1
 #   include "GL/glxext.h"
@@ -266,7 +265,6 @@ extern PFNGLGENERATEMIPMAPEXTPROC glGenerateMipmapEXT;
 #define GL_GLEXT_PROTOTYPES 1
 #include "GL/gl.h"
 #include "GL/glext.h"
-#include "GL/glu.h"
 
 // The __APPLE__ kludge is to make glh_extensions.h not symbol-clash horribly
 # define __APPLE__
@@ -282,7 +280,6 @@ extern PFNGLGENERATEMIPMAPEXTPROC glGenerateMipmapEXT;
 // quotes so we get libraries/.../GL/ version
 #include "GL/gl.h"
 #include "GL/glext.h"
-#include "GL/glu.h"
 
 
 #if LL_LINUX && !LL_MESA_HEADLESS
@@ -551,7 +548,6 @@ extern PFNGLBINDBUFFERRANGEPROC glBindBufferRange;
 
 //----------------------------------------------------------------------------
 #include <GL/gl.h>
-#include <GL/glu.h>
 
 // quotes so we get libraries/.../GL/ version
 #include "GL/glext.h"
@@ -789,7 +785,6 @@ extern PFNGLGETDEBUGMESSAGELOGARBPROC glGetDebugMessageLogARB;
 // LL_DARWIN
 
 #include <OpenGL/gl.h>
-#include <OpenGL/glu.h>
 
 #define GL_EXT_separate_specular_color 1
 #include <OpenGL/glext.h>

indra/llrender/llimagegl.cpp

@@ -468,6 +468,8 @@ LLImageGL::~LLImageGL()
 	sCount--;
 }
 
+const S8 INVALID_OFFSET = -99 ;
+
 void LLImageGL::init(BOOL usemipmaps)
 {
 	// keep these members in the same order as declared in llimagehl.h
@@ -484,14 +486,12 @@ void LLImageGL::init(BOOL usemipmaps)
 	mHasExplicitFormat = FALSE;
 	mAutoGenMips = FALSE;
 
-	mCanMask = TRUE;
 	mIsMask = FALSE;
 	mMaskRMSE = 1.f ;
 
-
+	mNeedsAlphaAndPickMask = FALSE ;
-	mNeedsAlphaAndPickMask = TRUE ;
 	mAlphaStride = 0 ;
-	mAlphaOffset = 0 ;
+	mAlphaOffset = INVALID_OFFSET ;
 
 	mGLTextureCreated = FALSE ;
 	mTexName = 0;
@@ -1709,7 +1709,6 @@ void LLImageGL::setTarget(const LLGLenum target, const LLTexUnit::eTextureType b
 }
 
 //Used by media in V2
-const S8 INVALID_OFFSET = -99 ;
 void LLImageGL::setNeedsAlphaAndPickMask(BOOL need_mask) 
 {
 	if(mNeedsAlphaAndPickMask != need_mask)
@@ -1723,7 +1722,6 @@ void LLImageGL::setNeedsAlphaAndPickMask(BOOL need_mask)
 		else //do not need alpha mask
 		{
 			mAlphaOffset = INVALID_OFFSET ;
-			mCanMask = FALSE;
 		}
 	}
 }
@@ -1746,8 +1744,7 @@ void LLImageGL::calcAlphaChannelOffsetAndStride()
 		mAlphaStride = 2;
 		break;
 	case GL_RGB:
-		mNeedsAlphaAndPickMask = FALSE ;
+		setNeedsAlphaAndPickMask(FALSE);
-		mCanMask = FALSE;
 		return ; //no alpha channel.
 	case GL_RGBA:
 		mAlphaStride = 4;
@@ -1793,15 +1790,14 @@ void LLImageGL::calcAlphaChannelOffsetAndStride()
 	{
 		llwarns << "Cannot analyze alpha for image with format type " << std::hex << mFormatType << std::dec << llendl;
 
-		mNeedsAlphaAndPickMask = FALSE ;
+		setNeedsAlphaAndPickMask(FALSE);
-		mCanMask = FALSE;
 	}
 }
 
 //std::map<LLGLuint, std::list<std::pair<std::string,std::string> > > sTextureMaskMap;
 void LLImageGL::analyzeAlpha(const void* data_in, U32 w, U32 h)
 {
-	if(!mNeedsAlphaAndPickMask || !mCanMask)
+	if(!mNeedsAlphaAndPickMask)
 	{
 		return ;
 	}

indra/llrender/llimagegl.h

@@ -135,7 +135,7 @@ public:
 	BOOL getHasGLTexture() const { return mTexName != 0; }
 	LLGLuint getTexName() const { return mTexName; }
 
-	BOOL getIsAlphaMask(const F32 max_rmse) const { return mCanMask && (max_rmse < 0.f ? (bool)mIsMask : (mMaskRMSE <= max_rmse)); }
+	BOOL getIsAlphaMask(const F32 max_rmse) const { return mNeedsAlphaAndPickMask && (max_rmse < 0.f ? (bool)mIsMask : (mMaskRMSE <= max_rmse)); }
 
 	BOOL getIsResident(BOOL test_now = FALSE); // not const
 
@@ -185,7 +185,6 @@ private:
 	S8 mHasExplicitFormat; // If false (default), GL format is f(mComponents)
 	S8 mAutoGenMips;
 
-	BOOL mCanMask;
 	BOOL mIsMask;
 	F32  mMaskRMSE;
 	BOOL mNeedsAlphaAndPickMask;

indra/llrender/llpostprocess.cpp

@@ -43,6 +43,7 @@
 #include "llsdutil_math.h"
 #include "llvertexbuffer.h"
 #include "llfasttimer.h"
+#include "llmatrix4a.h"
 
 extern LLGLSLShader			gPostColorFilterProgram;
 extern LLGLSLShader			gPostNightVisionProgram;
@@ -305,21 +306,21 @@ public:
 	{
 		addSetting(mStrength);
 	}
-	/*virtual*/ bool isEnabled()		const	{ return LLPostProcessShader::isEnabled() && llabs(gGLModelView[0] - gGLPreviousModelView[0]) > .0000001; }
+	/*virtual*/ bool isEnabled()		const	{ return LLPostProcessShader::isEnabled() && llabs(gGLModelView.getF32ptr()[0] - gGLPreviousModelView.getF32ptr()[0]) > .0000001; }
 	/*virtual*/ S32 getColorChannel()	const	{ return 0; }
 	/*virtual*/ S32 getDepthChannel()	const	{ return 1; }
 	/*virtual*/ QuadType preDraw()
 	{
-		glh::matrix4f inv_proj(gGLModelView);
+		LLMatrix4a inv_proj;
-		inv_proj.mult_left(gGLProjection);
+		inv_proj.setMul(gGLProjection,gGLModelView);
-		inv_proj = inv_proj.inverse();
+		inv_proj.invert();
-		glh::matrix4f prev_proj(gGLPreviousModelView);
+		LLMatrix4a prev_proj;
-		prev_proj.mult_left(gGLProjection);
+		prev_proj.setMul(gGLProjection,gGLPreviousModelView);
 
 		LLVector2 screen_rect = LLPostProcess::getInstance()->getDimensions();
 
-		getShader().uniformMatrix4fv(sPrevProj, 1, GL_FALSE, prev_proj.m);
+		getShader().uniformMatrix4fv(sPrevProj, 1, GL_FALSE, prev_proj.getF32ptr());
-		getShader().uniformMatrix4fv(sInvProj, 1, GL_FALSE, inv_proj.m);
+		getShader().uniformMatrix4fv(sInvProj, 1, GL_FALSE, inv_proj.getF32ptr());
 		getShader().uniform2fv(sScreenRes, 1, screen_rect.mV);
 		getShader().uniform1i(sBlurStrength, mStrength);
 		

indra/llrender/llrender.cpp

@@ -35,17 +35,18 @@
 #include "llrendertarget.h"
 #include "lltexture.h"
 #include "llshadermgr.h"
+#include "llmatrix4a.h"
 
 LLRender gGL;
 
 // Handy copies of last good GL matrices
 //Would be best to migrate these to LLMatrix4a and LLVector4a, but that's too divergent right now.
-LL_ALIGN_16(F32	gGLModelView[16]);
+LLMatrix4a gGLModelView;
-LL_ALIGN_16(F32	gGLLastModelView[16]);
+LLMatrix4a gGLLastModelView;
-LL_ALIGN_16(F32 gGLPreviousModelView[16]);
+LLMatrix4a gGLPreviousModelView;
-LL_ALIGN_16(F32 gGLLastProjection[16]);
+LLMatrix4a gGLLastProjection;
-LL_ALIGN_16(F32 gGLProjection[16]);
+LLMatrix4a gGLProjection;
-LL_ALIGN_16(S32	gGLViewport[4]);
+S32 gGLViewport[4];
 
 U32 LLRender::sUICalls = 0;
 U32 LLRender::sUIVerts = 0;
@@ -928,12 +929,12 @@ void LLLightState::setPosition(const LLVector4& position)
 	}
 	else
 	{ //transform position by current modelview matrix
-		glh::vec4f pos(position.mV);
+		LLVector4a pos;
+		pos.loadua(position.mV);
 
-		const glh::matrix4f& mat = gGL.getModelviewMatrix();
+		gGL.getModelviewMatrix().rotate4(pos,pos);
-		mat.mult_matrix_vec(pos);
 
-		mPosition.set(pos.v);
+		mPosition.set(pos.getF32ptr());
 	}
 
 }
@@ -1014,12 +1015,12 @@ void LLLightState::setSpotDirection(const LLVector3& direction)
 	}
 	else
 	{ //transform direction by current modelview matrix
-		glh::vec3f dir(direction.mV);
+		LLVector4a dir;
+		dir.load3(direction.mV);
 
-		const glh::matrix4f& mat = gGL.getModelviewMatrix();
+		gGL.getModelviewMatrix().rotate(dir,dir);
-		mat.mult_matrix_dir(dir);
 
-		mSpotDirection.set(dir.v);
+		mSpotDirection.set(dir.getF32ptr());
 	}
 }
 
@@ -1066,6 +1067,18 @@ LLRender::LLRender()
 	}
 
 	mLightHash = 0;
+	
+	//Init base matrix for each mode
+	for(S32 i = 0; i < NUM_MATRIX_MODES; ++i)
+	{
+		mMatrix[i][0].setIdentity();
+	}
+
+	gGLModelView.setIdentity();
+	gGLLastModelView.setIdentity();
+	gGLPreviousModelView.setIdentity();
+	gGLLastProjection.setIdentity();
+	gGLProjection.setIdentity();
 }
 
 LLRender::~LLRender()
@@ -1188,12 +1201,11 @@ void LLRender::syncMatrices()
 	};
 
 	LLGLSLShader* shader = LLGLSLShader::sCurBoundShaderPtr;
-
+	static LLMatrix4a cached_mvp;
-	static glh::matrix4f cached_mvp;
 	static U32 cached_mvp_mdv_hash = 0xFFFFFFFF;
 	static U32 cached_mvp_proj_hash = 0xFFFFFFFF;
 	
-	static glh::matrix4f cached_normal;
+	static LLMatrix4a cached_normal;
 	static U32 cached_normal_hash = 0xFFFFFFFF;
 
 	if (shader)
@@ -1205,9 +1217,9 @@ void LLRender::syncMatrices()
 		U32 i = MM_MODELVIEW;
 		if (mMatHash[i] != shader->mMatHash[i])
 		{ //update modelview, normal, and MVP
-			glh::matrix4f& mat = mMatrix[i][mMatIdx[i]];
+			const LLMatrix4a& mat = mMatrix[i][mMatIdx[i]];
-
+			
-			shader->uniformMatrix4fv(name[i], 1, GL_FALSE, mat.m);
+			shader->uniformMatrix4fv(name[i], 1, GL_FALSE, mat.getF32ptr());
 			shader->mMatHash[i] = mMatHash[i];
 
 			//update normal matrix
@@ -1216,20 +1228,20 @@ void LLRender::syncMatrices()
 			{
 				if (cached_normal_hash != mMatHash[i])
 				{
-					cached_normal = mat.inverse().transpose();
+					cached_normal = mat;
+					cached_normal.invert();
+					cached_normal.transpose();
 					cached_normal_hash = mMatHash[i];
 				}
+				
+				const LLMatrix4a& norm = cached_normal;
 
-				glh::matrix4f& norm = cached_normal;
+				LLVector3 norms[3];
+				norms[0].set(norm.getRow<0>().getF32ptr());
+				norms[1].set(norm.getRow<1>().getF32ptr());
+				norms[2].set(norm.getRow<2>().getF32ptr());
 
-				F32 norm_mat[] = 
+				shader->uniformMatrix3fv(LLShaderMgr::NORMAL_MATRIX, 1, GL_FALSE, norms[0].mV);
-				{
-					norm.m[0], norm.m[1], norm.m[2],
-					norm.m[4], norm.m[5], norm.m[6],
-					norm.m[8], norm.m[9], norm.m[10] 
-				};
-
-				shader->uniformMatrix3fv(LLShaderMgr::NORMAL_MATRIX, 1, GL_FALSE, norm_mat);
 			}
 
 			//update MVP matrix
@@ -1241,13 +1253,12 @@ void LLRender::syncMatrices()
 
 				if (cached_mvp_mdv_hash != mMatHash[i] || cached_mvp_proj_hash != mMatHash[MM_PROJECTION])
 				{
-					cached_mvp = mat;
+					cached_mvp.setMul(mMatrix[proj][mMatIdx[proj]], mat);
-					cached_mvp.mult_left(mMatrix[proj][mMatIdx[proj]]);
 					cached_mvp_mdv_hash = mMatHash[i];
 					cached_mvp_proj_hash = mMatHash[MM_PROJECTION];
 				}
 
-				shader->uniformMatrix4fv(LLShaderMgr::MODELVIEW_PROJECTION_MATRIX, 1, GL_FALSE, cached_mvp.m);
+				shader->uniformMatrix4fv(LLShaderMgr::MODELVIEW_PROJECTION_MATRIX, 1, GL_FALSE, cached_mvp.getF32ptr());
 			}
 		}
 
@@ -1255,9 +1266,9 @@ void LLRender::syncMatrices()
 		i = MM_PROJECTION;
 		if (mMatHash[i] != shader->mMatHash[i])
 		{ //update projection matrix, normal, and MVP
-			glh::matrix4f& mat = mMatrix[i][mMatIdx[i]];
+			const LLMatrix4a& mat = mMatrix[i][mMatIdx[i]];
-
+			
-			shader->uniformMatrix4fv(name[i], 1, GL_FALSE, mat.m);
+			shader->uniformMatrix4fv(name[i], 1, GL_FALSE, mat.getF32ptr());
 			shader->mMatHash[i] = mMatHash[i];
 
 			if (!mvp_done)
@@ -1269,13 +1280,12 @@ void LLRender::syncMatrices()
 					if (cached_mvp_mdv_hash != mMatHash[i] || cached_mvp_proj_hash != mMatHash[MM_PROJECTION])
 					{
 						U32 mdv = MM_MODELVIEW;
-						cached_mvp = mat;
+						cached_mvp.setMul(mat,mMatrix[mdv][mMatIdx[mdv]]);
-						cached_mvp.mult_right(mMatrix[mdv][mMatIdx[mdv]]);
 						cached_mvp_mdv_hash = mMatHash[MM_MODELVIEW];
 						cached_mvp_proj_hash = mMatHash[MM_PROJECTION];
 					}
-									
+
-					shader->uniformMatrix4fv(LLShaderMgr::MODELVIEW_PROJECTION_MATRIX, 1, GL_FALSE, cached_mvp.m);
+					shader->uniformMatrix4fv(LLShaderMgr::MODELVIEW_PROJECTION_MATRIX, 1, GL_FALSE, cached_mvp.getF32ptr());
 				}
 			}
 		}
@@ -1284,7 +1294,7 @@ void LLRender::syncMatrices()
 		{
 			if (mMatHash[i] != shader->mMatHash[i])
 			{
-				shader->uniformMatrix4fv(name[i], 1, GL_FALSE, mMatrix[i][mMatIdx[i]].m);
+				shader->uniformMatrix4fv(name[i], 1, GL_FALSE, mMatrix[i][mMatIdx[i]].getF32ptr());
 				shader->mMatHash[i] = mMatHash[i];
 			}
 		}
@@ -1312,7 +1322,7 @@ void LLRender::syncMatrices()
 			if (mMatHash[i] != mCurMatHash[i])
 			{
 				glMatrixMode(mode[i]);
-				glLoadMatrixf(mMatrix[i][mMatIdx[i]].m);
+				glLoadMatrixf(mMatrix[i][mMatIdx[i]].getF32ptr());
 				mCurMatHash[i] = mMatHash[i];
 			}
 		}
@@ -1323,7 +1333,7 @@ void LLRender::syncMatrices()
 			{
 				gGL.getTexUnit(i-2)->activate();
 				glMatrixMode(mode[i]);
-				glLoadMatrixf(mMatrix[i][mMatIdx[i]].m);
+				glLoadMatrixf(mMatrix[i][mMatIdx[i]].getF32ptr());
 				mCurMatHash[i] = mMatHash[i];
 			}
 		}
@@ -1332,32 +1342,143 @@ void LLRender::syncMatrices()
 	stop_glerror();
 }
 
+LLMatrix4a LLRender::genRot(const GLfloat& a, const LLVector4a& axis) const
+{
+	F32 r = a * DEG_TO_RAD;
+
+	F32 c = cosf(r);
+	F32 s = sinf(r);
+
+	F32 ic = 1.f-c;
+
+	const LLVector4a add1(c,axis[VZ]*s,-axis[VY]*s);	//1,z,-y
+	const LLVector4a add2(-axis[VZ]*s,c,axis[VX]*s);	//-z,1,x
+	const LLVector4a add3(axis[VY]*s,-axis[VX]*s,c);	//y,-x,1
+
+	LLVector4a axis_x;
+	axis_x.splat<0>(axis);
+	LLVector4a axis_y;
+	axis_y.splat<1>(axis);
+	LLVector4a axis_z;
+	axis_z.splat<2>(axis);
+
+	LLVector4a c_axis;
+	c_axis.setMul(axis,ic);
+
+	LLMatrix4a rot_mat;
+	rot_mat.getRow<0>().setMul(c_axis,axis_x);
+	rot_mat.getRow<0>().add(add1);
+	rot_mat.getRow<1>().setMul(c_axis,axis_y);
+	rot_mat.getRow<1>().add(add2);
+	rot_mat.getRow<2>().setMul(c_axis,axis_z);
+	rot_mat.getRow<2>().add(add3);
+	rot_mat.setRow<3>(LLVector4a(0,0,0,1));
+
+	return rot_mat;
+}
+LLMatrix4a LLRender::genOrtho(const GLfloat& left, const GLfloat& right, const GLfloat& bottom, const GLfloat&  top, const GLfloat& zNear, const GLfloat& zFar) const
+{
+	LLMatrix4a ortho_mat;
+	ortho_mat.setRow<0>(LLVector4a(2.f/(right-left),0,0));
+	ortho_mat.setRow<1>(LLVector4a(0,2.f/(top-bottom),0));
+	ortho_mat.setRow<2>(LLVector4a(0,0,-2.f/(zFar-zNear)));
+	ortho_mat.setRow<3>(LLVector4a(-(right+left)/(right-left),-(top+bottom)/(top-bottom),-(zFar+zNear)/(zFar-zNear),1));
+
+	return ortho_mat;
+}
+
+LLMatrix4a LLRender::genPersp(const GLfloat& fovy, const GLfloat& aspect, const GLfloat& zNear, const GLfloat& zFar) const
+{
+	GLfloat f = 1.f/tanf(DEG_TO_RAD*fovy/2.f);
+
+	LLMatrix4a persp_mat;
+	persp_mat.setRow<0>(LLVector4a(f/aspect,0,0));
+	persp_mat.setRow<1>(LLVector4a(0,f,0));
+	persp_mat.setRow<2>(LLVector4a(0,0,(zFar+zNear)/(zNear-zFar),-1.f));
+	persp_mat.setRow<3>(LLVector4a(0,0,(2.f*zFar*zNear)/(zNear-zFar),0));
+
+	return persp_mat;
+}
+
+LLMatrix4a LLRender::genLook(const LLVector3& pos_in, const LLVector3& dir_in, const LLVector3& up_in) const
+{
+	const LLVector4a pos(pos_in.mV[VX],pos_in.mV[VY],pos_in.mV[VZ],1.f);
+	LLVector4a dir(dir_in.mV[VX],dir_in.mV[VY],dir_in.mV[VZ]);
+	const LLVector4a up(up_in.mV[VX],up_in.mV[VY],up_in.mV[VZ]);
+
+	LLVector4a left_norm;
+	left_norm.setCross3(dir,up);
+	left_norm.normalize3fast();
+	LLVector4a up_norm;
+	up_norm.setCross3(left_norm,dir);
+	up_norm.normalize3fast();
+	LLVector4a& dir_norm = dir;
+	dir.normalize3fast();
+	
+	LLVector4a left_dot;
+	left_dot.setAllDot3(left_norm,pos);
+	left_dot.negate();
+	LLVector4a up_dot;
+	up_dot.setAllDot3(up_norm,pos);
+	up_dot.negate();
+	LLVector4a dir_dot;
+	dir_dot.setAllDot3(dir_norm,pos);
+
+	dir_norm.negate();
+
+	LLMatrix4a lookat_mat;
+	lookat_mat.setRow<0>(left_norm);
+	lookat_mat.setRow<1>(up_norm);
+	lookat_mat.setRow<2>(dir_norm);
+	lookat_mat.setRow<3>(LLVector4a(0,0,0,1));
+
+	lookat_mat.getRow<0>().copyComponent<3>(left_dot);
+	lookat_mat.getRow<1>().copyComponent<3>(up_dot);
+	lookat_mat.getRow<2>().copyComponent<3>(dir_dot);
+
+	lookat_mat.transpose();
+
+	return lookat_mat;
+}
+
+const LLMatrix4a& LLRender::genNDCtoWC() const
+{
+	static LLMatrix4a mat(
+		LLVector4a(.5f,0,0,0),
+		LLVector4a(0,.5f,0,0),
+		LLVector4a(0,0,.5f,0),
+		LLVector4a(.5f,.5f,.5f,1.f));
+	return mat;
+}
+
 void LLRender::translatef(const GLfloat& x, const GLfloat& y, const GLfloat& z)
 {
+	if(	llabs(x) < F_APPROXIMATELY_ZERO &&
+		llabs(y) < F_APPROXIMATELY_ZERO &&
+		llabs(z) < F_APPROXIMATELY_ZERO)
+	{
+		return;
+	}
+
 	flush();
 
-	{
+	mMatrix[mMatrixMode][mMatIdx[mMatrixMode]].applyTranslation_affine(x,y,z);
-		glh::matrix4f trans_mat(1,0,0,x,
+	mMatHash[mMatrixMode]++;
-								0,1,0,y,
+
-								0,0,1,z,
-								0,0,0,1);
-	
-		mMatrix[mMatrixMode][mMatIdx[mMatrixMode]].mult_right(trans_mat);
-		mMatHash[mMatrixMode]++;
-	}
 }
 
 void LLRender::scalef(const GLfloat& x, const GLfloat& y, const GLfloat& z)
 {
+	if(	(llabs(x-1.f)) < F_APPROXIMATELY_ZERO &&
+		(llabs(y-1.f)) < F_APPROXIMATELY_ZERO &&
+		(llabs(z-1.f)) < F_APPROXIMATELY_ZERO)
+	{
+		return;
+	}
 	flush();
 	
 	{
-		glh::matrix4f scale_mat(x,0,0,0,
+		mMatrix[mMatrixMode][mMatIdx[mMatrixMode]].applyScale_affine(x,y,z);
-								0,y,0,0,
-								0,0,z,0,
-								0,0,0,1);
-	
-		mMatrix[mMatrixMode][mMatIdx[mMatrixMode]].mult_right(scale_mat);
 		mMatHash[mMatrixMode]++;
 	}
 }
@@ -1366,38 +1487,156 @@ void LLRender::ortho(F32 left, F32 right, F32 bottom, F32 top, F32 zNear, F32 zF
 {
 	flush();
 
-	{
+	LLMatrix4a ortho_mat;
+	ortho_mat.setRow<0>(LLVector4a(2.f/(right-left),0,0));
+	ortho_mat.setRow<1>(LLVector4a(0,2.f/(top-bottom),0));
+	ortho_mat.setRow<2>(LLVector4a(0,0,-2.f/(zFar-zNear)));
+	ortho_mat.setRow<3>(LLVector4a(-(right+left)/(right-left),-(top+bottom)/(top-bottom),-(zFar+zNear)/(zFar-zNear),1));	
 
-		glh::matrix4f ortho_mat(2.f/(right-left),0,0,	-(right+left)/(right-left),
+	mMatrix[mMatrixMode][mMatIdx[mMatrixMode]].mul_affine(ortho_mat);
-								0,2.f/(top-bottom),0,	-(top+bottom)/(top-bottom),
+	mMatHash[mMatrixMode]++;
-								0,0,-2.f/(zFar-zNear),	-(zFar+zNear)/(zFar-zNear),
+}
-								0,0,0,1);
+
-	
+void LLRender::rotatef(const LLMatrix4a& rot)
-		mMatrix[mMatrixMode][mMatIdx[mMatrixMode]].mult_right(ortho_mat);
+{
-		mMatHash[mMatrixMode]++;
+	flush();
-	}
+
+	mMatrix[mMatrixMode][mMatIdx[mMatrixMode]].mul_affine(rot);
+	mMatHash[mMatrixMode]++;
 }
 
 void LLRender::rotatef(const GLfloat& a, const GLfloat& x, const GLfloat& y, const GLfloat& z)
 {
+	if(	llabs(a) < F_APPROXIMATELY_ZERO ||
+		llabs(a-360.f) < F_APPROXIMATELY_ZERO)
+	{
+		return;
+	}
+	
 	flush();
 
-	{
+	rotatef(genRot(a,x,y,z));
-		F32 r = a * DEG_TO_RAD;
+}
 
-		F32 c = cosf(r);
+//LLRender::projectf & LLRender::unprojectf adapted from gluProject & gluUnproject in Mesa's GLU 9.0 library.
-		F32 s = sinf(r);
+//  License/Copyright Statement:
+/*
+ * SGI FREE SOFTWARE LICENSE B (Version 2.0, Sept. 18, 2008)
+ * Copyright (C) 1991-2000 Silicon Graphics, Inc. All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice including the dates of first publication and
+ * either this permission notice or a reference to
+ * http://oss.sgi.com/projects/FreeB/
+ * shall be included in all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
+ * SILICON GRAPHICS, INC. BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY,
+ * WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF
+ * OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+ * SOFTWARE.
+ *
+ * Except as contained in this notice, the name of Silicon Graphics, Inc.
+ * shall not be used in advertising or otherwise to promote the sale, use or
+ * other dealings in this Software without prior written authorization from
+ * Silicon Graphics, Inc.
+ */
 
-		F32 ic = 1.f-c;
+bool LLRender::projectf(const LLVector3& object, const LLMatrix4a& modelview, const LLMatrix4a& projection, const LLRect& viewport, LLVector3& windowCoordinate)
+{
+	//Begin SSE intrinsics
 
-		glh::matrix4f rot_mat(x*x*ic+c,		x*y*ic-z*s,		x*z*ic+y*s,		0,
+	// Declare locals
-							  x*y*ic+z*s,	y*y*ic+c,		y*z*ic-x*s,		0,
+	const LLVector4a obj_vector(object.mV[VX],object.mV[VY],object.mV[VZ]);
-							  x*z*ic-y*s,	y*z*ic+x*s,		z*z*ic+c,		0,
+	const LLVector4a one(1.f);
-							  0,0,0,1);
+	LLVector4a temp_vec;								//Scratch vector
+	LLVector4a w;										//Splatted W-component.
 	
-		mMatrix[mMatrixMode][mMatIdx[mMatrixMode]].mult_right(rot_mat);
+	modelview.affineTransform(obj_vector, temp_vec);	//temp_vec = modelview * obj_vector;
-		mMatHash[mMatrixMode]++;
+
-	}
+	//Passing temp_matrix as v and res is safe. res not altered until after all other calculations
+	projection.rotate4(temp_vec, temp_vec);				//temp_vec = projection * temp_vec
+
+	w.splat<3>(temp_vec);								//w = temp_vec.wwww
+
+	//If w == 0.f, use 1.f instead.
+	LLVector4a div;
+	div.setSelectWithMask( w.equal( _mm_setzero_ps() ), one, w );	//float div = (w[N] == 0.f ? 1.f : w[N]);
+	temp_vec.div(div);									//temp_vec /= div;
+
+	//Map x, y to range 0-1
+	temp_vec.mul(.5f);
+	temp_vec.add(.5f);
+
+	LLVector4Logical mask = temp_vec.equal(_mm_setzero_ps());
+	if(mask.areAllSet(LLVector4Logical::MASK_W))
+		return false;
+
+	//End SSE intrinsics
+
+	//Window coordinates
+	windowCoordinate[0]=temp_vec[VX]*viewport.getWidth()+viewport.mLeft;
+	windowCoordinate[1]=temp_vec[VY]*viewport.getHeight()+viewport.mBottom;
+	//This is only correct when glDepthRange(0.0, 1.0)
+	windowCoordinate[2]=temp_vec[VZ];	
+	
+	return true;
+}
+
+bool LLRender::unprojectf(const LLVector3& windowCoordinate, const LLMatrix4a& modelview, const LLMatrix4a& projection, const LLRect& viewport, LLVector3& object)
+{
+	//Begin SSE intrinsics
+
+	// Declare locals
+	static const LLVector4a one(1.f);
+	static const LLVector4a two(2.f);
+	LLVector4a norm_view(
+		((windowCoordinate.mV[VX] - (F32)viewport.mLeft) / (F32)viewport.getWidth()),
+		((windowCoordinate.mV[VY] - (F32)viewport.mBottom) / (F32)viewport.getHeight()),
+		windowCoordinate.mV[VZ],
+		1.f);
+	
+	LLMatrix4a inv_mat;								//Inverse transformation matrix
+	LLVector4a temp_vec;							//Scratch vector
+	LLVector4a w;									//Splatted W-component.
+
+	inv_mat.setMul(projection,modelview);			//inv_mat = projection*modelview
+
+	float det = inv_mat.invert();
+
+	//Normalize. -1.0 : +1.0
+	norm_view.mul(two);								// norm_view *= vec4(.2f)
+	norm_view.sub(one);								// norm_view -= vec4(1.f)
+
+	inv_mat.rotate4(norm_view,temp_vec);			//inv_mat * norm_view
+
+	w.splat<3>(temp_vec);							//w = temp_vec.wwww
+
+	//If w == 0.f, use 1.f instead. Defer return if temp_vec.w == 0.f until after all SSE intrinsics.
+	LLVector4a div;
+	div.setSelectWithMask( w.equal( _mm_setzero_ps() ), one, w );	//float div = (w[N] == 0.f ? 1.f : w[N]);
+	temp_vec.div(div);								//temp_vec /= div;
+
+	LLVector4Logical mask = temp_vec.equal(_mm_setzero_ps());
+	if(mask.areAllSet(LLVector4Logical::MASK_W))
+		return false;
+
+	//End SSE intrinsics
+
+	if(det == 0.f)	
+       return false;
+
+	object.set(temp_vec.getF32ptr());
+
+	return true;
 }
 
 void LLRender::pushMatrix()
@@ -1433,24 +1672,21 @@ void LLRender::popMatrix()
 	}
 }
 
-void LLRender::loadMatrix(const GLfloat* m)
+void LLRender::loadMatrix(const LLMatrix4a& mat)
 {
 	flush();
-	{
+	
-		mMatrix[mMatrixMode][mMatIdx[mMatrixMode]].set_value((GLfloat*) m);
+	mMatrix[mMatrixMode][mMatIdx[mMatrixMode]] = mat;
-		mMatHash[mMatrixMode]++;
+	mMatHash[mMatrixMode]++;
-	}
 }
 
-void LLRender::multMatrix(const GLfloat* m)
+void LLRender::multMatrix(const LLMatrix4a& mat)
 {
 	flush();
-	{
+
-		glh::matrix4f mat((GLfloat*) m);
+	mMatrix[mMatrixMode][mMatIdx[mMatrixMode]].mul_affine(mat); 
+	mMatHash[mMatrixMode]++;
 	
-		mMatrix[mMatrixMode][mMatIdx[mMatrixMode]].mult_right(mat);
-		mMatHash[mMatrixMode]++;
-	}
 }
 
 void LLRender::matrixMode(U32 mode)
@@ -1479,20 +1715,16 @@ void LLRender::loadIdentity()
 {
 	flush();
 
-	{
+	mMatrix[mMatrixMode][mMatIdx[mMatrixMode]].setIdentity();
-		llassert_always(mMatrixMode < NUM_MATRIX_MODES) ;
+	mMatHash[mMatrixMode]++;
-
-		mMatrix[mMatrixMode][mMatIdx[mMatrixMode]].make_identity();
-		mMatHash[mMatrixMode]++;
-	}
 }
 
-const glh::matrix4f& LLRender::getModelviewMatrix()
+const LLMatrix4a& LLRender::getModelviewMatrix()
 {
 	return mMatrix[MM_MODELVIEW][mMatIdx[MM_MODELVIEW]];
 }
 
-const glh::matrix4f& LLRender::getProjectionMatrix()
+const LLMatrix4a& LLRender::getProjectionMatrix()
 {
 	return mMatrix[MM_PROJECTION][mMatIdx[MM_PROJECTION]];
 }

indra/llrender/llrender.h

@@ -38,18 +38,19 @@
 #include "v3math.h"
 #include "v4coloru.h"
 #include "v4math.h"
+#include "llmatrix4a.h"
 #include "llalignedarray.h"
 #include "llstrider.h"
 #include "llpointer.h"
 #include "llglheaders.h"
-#include "llmatrix4a.h"
+#include "llrect.h"
-#include "glh/glh_linear.h"
 
 class LLVertexBuffer;
 class LLCubeMap;
 class LLImageGL;
 class LLRenderTarget;
 class LLTexture ;
+class LLMatrix4a;
 
 #define LL_MATRIX_STACK_DEPTH 32
 
@@ -257,6 +258,8 @@ protected:
 	F32 mSpotExponent;
 	F32 mSpotCutoff;
 };
+
+LL_ALIGN_PREFIX(16)
 class LLRender
 {
 	friend class LLTexUnit;
@@ -343,21 +346,32 @@ public:
 	// Needed when the render context has changed and invalidated the current state
 	void refreshState(void);
 
+	LLMatrix4a genRot(const GLfloat& a, const LLVector4a& axis) const;
+	LLMatrix4a genRot(const GLfloat& a, const GLfloat& x, const GLfloat& y, const GLfloat& z) const { return genRot(a,LLVector4a(x,y,z)); }
+	LLMatrix4a genOrtho(const GLfloat& left, const GLfloat& right, const GLfloat& bottom, const GLfloat&  top, const GLfloat& znear, const GLfloat& zfar) const;
+	LLMatrix4a genPersp(const GLfloat& fovy, const GLfloat& aspect, const GLfloat& znear, const GLfloat& zfar) const;
+	LLMatrix4a genLook(const LLVector3& pos_in, const LLVector3& dir_in, const LLVector3& up_in) const;
+	const LLMatrix4a& genNDCtoWC() const;
+
 	void translatef(const GLfloat& x, const GLfloat& y, const GLfloat& z);
 	void scalef(const GLfloat& x, const GLfloat& y, const GLfloat& z);
+	//rotatef requires generation of a transform matrix involving sine/cosine. If rotating by a constant value, use genRot, store the result in a static variable, and pass that var to rotatef.
+	void rotatef(const LLMatrix4a& rot);
 	void rotatef(const GLfloat& a, const GLfloat& x, const GLfloat& y, const GLfloat& z);
 	void ortho(F32 left, F32 right, F32 bottom, F32 top, F32 zNear, F32 zFar);
+	bool projectf(const LLVector3& object, const LLMatrix4a& modelview, const LLMatrix4a& projection, const LLRect& viewport, LLVector3& windowCoordinate);
+	bool unprojectf(const LLVector3& windowCoordinate, const LLMatrix4a& modelview, const LLMatrix4a& projection, const LLRect& viewport, LLVector3& object);
 
 	void pushMatrix();
 	void popMatrix();
-	void loadMatrix(const GLfloat* m);
+	void loadMatrix(const LLMatrix4a& mat);
 	void loadIdentity();
-	void multMatrix(const GLfloat* m);
+	void multMatrix(const LLMatrix4a& mat);
 	void matrixMode(U32 mode);	
 	U32 getMatrixMode();
 
-	const glh::matrix4f& getModelviewMatrix();
+	const LLMatrix4a& getModelviewMatrix();
-	const glh::matrix4f& getProjectionMatrix();
+	const LLMatrix4a& getProjectionMatrix();
 
 	void syncMatrices();
 	void syncLightState();
@@ -447,7 +461,7 @@ private:
 	U32 mMatrixMode;
 	U32 mMatIdx[NUM_MATRIX_MODES];
 	U32 mMatHash[NUM_MATRIX_MODES];
-	glh::matrix4f mMatrix[NUM_MATRIX_MODES][LL_MATRIX_STACK_DEPTH];
+	LL_ALIGN_16(LLMatrix4a mMatrix[NUM_MATRIX_MODES][LL_MATRIX_STACK_DEPTH]);
 	U32 mCurMatHash[NUM_MATRIX_MODES];
 	U32 mLightHash;
 	LLColor4 mAmbientLightColor;
@@ -478,13 +492,14 @@ private:
 
 	LLAlignedArray<LLVector4a, 64> mUIOffset;
 	LLAlignedArray<LLVector4a, 64> mUIScale;
-};
+} LL_ALIGN_POSTFIX(16);
 
-extern F32 gGLModelView[16];
+
-extern F32 gGLLastModelView[16];
+extern LLMatrix4a gGLModelView;
-extern F32 gGLLastProjection[16];
+extern LLMatrix4a gGLLastModelView;
-extern F32 gGLPreviousModelView[16];
+extern LLMatrix4a gGLLastProjection;
-extern F32 gGLProjection[16];
+extern LLMatrix4a gGLPreviousModelView;
+extern LLMatrix4a gGLProjection;
 extern S32 gGLViewport[4];
 
 extern LLRender gGL;

indra/llrender/llvertexbuffer.cpp

@@ -2044,6 +2044,10 @@ bool LLVertexBuffer::getNormalStrider(LLStrider<LLVector3>& strider, S32 index,
 {
 	return VertexBufferStrider<LLVector3,TYPE_NORMAL>::get(*this, strider, index, count, map_range);
 }
+bool LLVertexBuffer::getNormalStrider(LLStrider<LLVector4a>& strider, S32 index, S32 count, bool map_range)
+{
+	return VertexBufferStrider<LLVector4a,TYPE_NORMAL>::get(*this, strider, index, count, map_range);
+}
 bool LLVertexBuffer::getTangentStrider(LLStrider<LLVector3>& strider, S32 index, S32 count, bool map_range)
 {
 	return VertexBufferStrider<LLVector3,TYPE_TANGENT>::get(*this, strider, index, count, map_range);

indra/llrender/llvertexbuffer.h

@@ -250,6 +250,7 @@ public:
 	bool getTexCoord1Strider(LLStrider<LLVector2>& strider, S32 index=0, S32 count = -1, bool map_range = false);
 	bool getTexCoord2Strider(LLStrider<LLVector2>& strider, S32 index=0, S32 count = -1, bool map_range = false);
 	bool getNormalStrider(LLStrider<LLVector3>& strider, S32 index=0, S32 count = -1, bool map_range = false);
+	bool getNormalStrider(LLStrider<LLVector4a>& strider, S32 index=0, S32 count = -1, bool map_range = false);
 	bool getTangentStrider(LLStrider<LLVector3>& strider, S32 index=0, S32 count = -1, bool map_range = false);
 	bool getTangentStrider(LLStrider<LLVector4a>& strider, S32 index=0, S32 count = -1, bool map_range = false);
 	bool getColorStrider(LLStrider<LLColor4U>& strider, S32 index=0, S32 count = -1, bool map_range = false);

indra/llui/llcombobox.cpp

@@ -909,6 +909,19 @@ BOOL LLComboBox::handleUnicodeCharHere(llwchar uni_char)
 	return result;
 }
 
+BOOL LLComboBox::handleScrollWheel(S32 x, S32 y, S32 clicks)
+{
+	if (mList->getVisible()) return mList->handleScrollWheel(x, y, clicks);
+	if (mAllowTextEntry) // We might be editable
+		if (!mList->getFirstSelected()) // We aren't in the list, don't kill their text
+			return false;
+
+	setCurrentByIndex(llclamp(getCurrentIndex() + clicks, 0, getItemCount() - 1));
+	prearrangeList();
+	onCommit();
+	return true;
+}
+
 void LLComboBox::setAllowTextEntry(BOOL allow, S32 max_chars, BOOL set_tentative)
 {
 	mAllowTextEntry = allow;
@@ -924,6 +937,7 @@ void LLComboBox::setTextEntry(const LLStringExplicit& text)
 	if (mTextEntry)
 	{
 		mTextEntry->setText(text);
+		mTextEntry->setCursor(0); // Singu Note: Move the cursor over to the beginning
 		mHasAutocompletedText = FALSE;
 		updateSelection();
 	}

indra/llui/llcombobox.h

@@ -76,6 +76,7 @@ public:
 	virtual BOOL	handleToolTip(S32 x, S32 y, std::string& msg, LLRect* sticky_rect);
 	virtual BOOL	handleKeyHere(KEY key, MASK mask);
 	virtual BOOL	handleUnicodeCharHere(llwchar uni_char);
+	virtual BOOL	handleScrollWheel(S32 x, S32 y, S32 clicks);
 
 	// LLUICtrl interface
 	virtual void	clear();					// select nothing

indra/llui/llfloater.cpp

@@ -571,6 +571,9 @@ void LLFloater::open()	/* Flawfinder: ignore */
 		setVisibleAndFrontmost(mAutoFocus);
 	}
 
+	if (!getControlName().empty())
+		setControlValue(true);
+
 	onOpen();
 }
 
@@ -633,6 +636,9 @@ void LLFloater::close(bool app_quitting)
 			}
 		}
 
+		if (!app_quitting && !getControlName().empty())
+			setControlValue(false);
+
 		// Let floater do cleanup.
 		onClose(app_quitting);
 	}

indra/llui/llnotifications.cpp

@@ -1543,6 +1543,8 @@ class UpdateItemSM : public AIStateMachine
 
 	static void add(UpdateItem const& ui);
 
+	/*virtual*/ const char* getName() const { return "UpdateItemSM"; }
+
   private:
 	static UpdateItemSM* sSelf;
 	typedef std::deque<UpdateItem> updateQueue_type;

indra/llui/llradiogroup.cpp

@@ -46,16 +46,36 @@
 
 static LLRegisterWidget<LLRadioGroup> r("radio_group");
 
+/*
+ * A checkbox control with use_radio_style == true.
+ */
+class LLRadioCtrl : public LLCheckBoxCtrl
+{
+public:
+	LLRadioCtrl(const std::string& name, const LLRect& rect, const std::string& label, const std::string& value = "", const LLFontGL* font = NULL, commit_callback_t commit_callback = NULL);
+	/*virtual*/ ~LLRadioCtrl();
+
+	virtual LLXMLNodePtr getXML(bool save_children = true) const;
+	/*virtual*/ void setValue(const LLSD& value);
+
+	LLSD getPayload() { return mPayload; }
+protected:
+	friend class LLUICtrlFactory;
+
+	LLSD mPayload;	// stores data that this item represents in the radio group
+};
+
 LLRadioGroup::LLRadioGroup(const std::string& name, const LLRect& rect,
 			   S32 initial_index, commit_callback_t commit_callback,
-			   BOOL border) :
+			   bool border, bool allow_deselect) :
 	LLUICtrl(name, rect, TRUE, commit_callback, FOLLOWS_LEFT | FOLLOWS_TOP),
-	mSelectedIndex(initial_index)
+	mSelectedIndex(initial_index),
+	mAllowDeselect(allow_deselect)
 {
 	init(border);
 }
 
-void LLRadioGroup::init(BOOL border)
+void LLRadioGroup::init(bool border)
 {
 	if (border)
 	{
@@ -67,12 +87,19 @@ void LLRadioGroup::init(BOOL border)
 }
 
 
-
-
 LLRadioGroup::~LLRadioGroup()
 {
 }
 
+// virtual
+BOOL LLRadioGroup::postBuild()
+{
+	if (!mRadioButtons.empty())
+	{
+		mRadioButtons[0]->setTabStop(true);
+	}
+	return TRUE;
+}
 
 // virtual
 void LLRadioGroup::setEnabled(BOOL enabled)
@@ -133,16 +160,39 @@ void LLRadioGroup::setIndexEnabled(S32 index, BOOL enabled)
 
 BOOL LLRadioGroup::setSelectedIndex(S32 index, BOOL from_event)
 {
-	if (index < 0 || index >= (S32)mRadioButtons.size())
+	if ((S32)mRadioButtons.size() <= index )
 	{
 		return FALSE;
 	}
 
+	if (mSelectedIndex >= 0)
+	{
+		LLRadioCtrl* old_radio_item = mRadioButtons[mSelectedIndex];
+		old_radio_item->setTabStop(false);
+		old_radio_item->setValue( FALSE );
+	}
+	else
+	{
+		mRadioButtons[0]->setTabStop(false);
+	}
+
 	mSelectedIndex = index;
 
+	if (mSelectedIndex >= 0)
+	{
+		LLRadioCtrl* radio_item = mRadioButtons[mSelectedIndex];
+		radio_item->setTabStop(true);
+		radio_item->setValue( TRUE );
+
+		if (hasFocus())
+		{
+			radio_item->focusFirstItem(FALSE, FALSE);
+		}
+	}
+
 	if (!from_event)
 	{
-		setControlValue(getSelectedIndex());
+		setControlValue(getValue());
 	}
 
 	return TRUE;
@@ -207,41 +257,23 @@ BOOL LLRadioGroup::handleKeyHere(KEY key, MASK mask)
 	return handled;
 }
 
-void LLRadioGroup::draw()
+BOOL LLRadioGroup::handleMouseDown(S32 x, S32 y, MASK mask)
 {
-	S32 current_button = 0;
+	// grab focus preemptively, before child button takes mousecapture
-
+	//
-	BOOL take_focus = FALSE;
+	if (hasTabStop())
-	if (gFocusMgr.childHasKeyboardFocus(this))
 	{
-		take_focus = TRUE;
+		focusFirstItem(FALSE, FALSE);
 	}
 
-	for (button_list_t::iterator iter = mRadioButtons.begin();
+	return LLUICtrl::handleMouseDown(x, y, mask);
-		 iter != mRadioButtons.end(); ++iter)
-	{
-		LLRadioCtrl* radio = *iter;
-		BOOL selected = (current_button == mSelectedIndex);
-		radio->setValue( selected );
-		if (take_focus && selected && !gFocusMgr.childHasKeyboardFocus(radio))
-		{
-			// don't flash keyboard focus when navigating via keyboard
-			BOOL DONT_FLASH = FALSE;
-			radio->focusFirstItem(FALSE, DONT_FLASH);
-		}
-		current_button++;
-	}
-
-	LLView::draw();
 }
 
-
 // When adding a button, we need to ensure that the radio
 // group gets a message when the button is clicked.
-LLRadioCtrl* LLRadioGroup::addRadioButton(const std::string& name, const std::string& label, const LLRect& rect, const LLFontGL* font )
+LLRadioCtrl* LLRadioGroup::addRadioButton(const std::string& name, const std::string& label, const LLRect& rect, const LLFontGL* font, const std::string& payload)
 {
-	// Highlight will get fixed in draw method above
+	LLRadioCtrl* radio = new LLRadioCtrl(name, rect, label, payload, font, boost::bind(&LLRadioGroup::onClickButton, this, _1));
-	LLRadioCtrl* radio = new LLRadioCtrl(name, rect, label, font, boost::bind(&LLRadioGroup::onClickButton, this, _1));
 	addChild(radio);
 	mRadioButtons.push_back(radio);
 	return radio;
@@ -252,7 +284,7 @@ LLRadioCtrl* LLRadioGroup::addRadioButton(const std::string& name, const std::st
 
 void LLRadioGroup::onClickButton(LLUICtrl* ctrl)
 {
-	// llinfos << "LLRadioGroup::onClickButton" << llendl;
+	// LL_INFOS() << "LLRadioGroup::onClickButton" << LL_ENDL;
 	LLRadioCtrl* clicked_radio = dynamic_cast<LLRadioCtrl*>(ctrl);
 	if (!clicked_radio)
 	    return;
@@ -263,9 +295,15 @@ void LLRadioGroup::onClickButton(LLUICtrl* ctrl)
 		LLRadioCtrl* radio = *iter;
 		if (radio == clicked_radio)
 		{
-			// llinfos << "clicked button " << counter << llendl;
+			if (index == mSelectedIndex && mAllowDeselect)
-			setSelectedIndex(index);
+			{
-			setControlValue(index);
+				// don't select anything
+				setSelectedIndex(-1);
+			}
+			else
+			{
+				setSelectedIndex(index);
+			}
 			
 			// BUG: Calls click callback even if button didn't actually change
 			onCommit();
@@ -281,13 +319,12 @@ void LLRadioGroup::onClickButton(LLUICtrl* ctrl)
 
 void LLRadioGroup::setValue( const LLSD& value )
 {
-	std::string value_name = value.asString();
 	int idx = 0;
 	for (button_list_t::const_iterator iter = mRadioButtons.begin();
 		 iter != mRadioButtons.end(); ++iter)
 	{
 		LLRadioCtrl* radio = *iter;
-		if (radio->getName() == value_name)
+		if (radio->getPayload().asString() == value.asString())
 		{
 			setSelectedIndex(idx);
 			idx = -1;
@@ -304,8 +341,7 @@ void LLRadioGroup::setValue( const LLSD& value )
 		}
 		else
 		{
-			llwarns << "LLRadioGroup::setValue: radio_item with name=\"" << value_name << "\" not found, radio_group values are set by radio_item name not value.  Falling back on LLUICtrl::setValue." << llendl;
+			setSelectedIndex(-1, TRUE);
-			LLUICtrl::setValue(value);
 		}
 	}
 }
@@ -317,7 +353,7 @@ LLSD LLRadioGroup::getValue() const
 	for (button_list_t::const_iterator iter = mRadioButtons.begin();
 		 iter != mRadioButtons.end(); ++iter)
 	{
-		if (idx == index) return LLSD((*iter)->getName());
+		if (idx == index) return LLSD((*iter)->getPayload());
 		++idx;
 	}
 	return LLSD();
@@ -333,6 +369,7 @@ LLXMLNodePtr LLRadioGroup::getXML(bool save_children) const
 	// Attributes
 
 	node->createChild("draw_border", TRUE)->setBoolValue(mHasBorder);
+	node->createChild("allow_deselect", TRUE)->setBoolValue(mAllowDeselect);
 
 	// Contents
 
@@ -355,17 +392,21 @@ LLView* LLRadioGroup::fromXML(LLXMLNodePtr node, LLView *parent, LLUICtrlFactory
 	U32 initial_value = 0;
 	node->getAttributeU32("initial_value", initial_value);
 
-	BOOL draw_border = TRUE;
+	bool draw_border = true;
-	node->getAttributeBOOL("draw_border", draw_border);
+	node->getAttribute_bool("draw_border", draw_border);
+
+	bool allow_deselect = false;
+	node->getAttribute_bool("allow_deselect", allow_deselect);
 
 	LLRect rect;
 	createRect(node, rect, parent, LLRect());
 
-	LLRadioGroup* radio_group = new LLRadioGroup("radio_group", 
+	LLRadioGroup* radio_group = new LLRadioGroup("radio_group",
 		rect,
 		initial_value,
 		NULL,
-		draw_border);
+		draw_border,
+		allow_deselect);
 
 	const std::string& contents = node->getValue();
 
@@ -406,7 +447,13 @@ LLView* LLRadioGroup::fromXML(LLXMLNodePtr node, LLView *parent, LLUICtrlFactory
 				createRect(child, item_rect, radio_group, rect);
 
 				std::string item_label = child->getTextContents();
-				LLRadioCtrl* radio = radio_group->addRadioButton("radio", item_label, item_rect, font);
+				child->getAttributeString("label", item_label);
+				std::string item_name("radio");
+				child->getAttributeString("name", item_name);
+				std::string payload(item_name); // Support old-style name as payload
+				child->getAttributeString("value", payload);
+				child->getAttributeString("initial_value", payload); // Synonym
+				LLRadioCtrl* radio = radio_group->addRadioButton(item_name, item_label, item_rect, font, payload);
 
 				radio->initFromXML(child, radio_group);
 			}
@@ -432,11 +479,10 @@ LLUUID	LLRadioGroup::getCurrentID() const
 BOOL	LLRadioGroup::setSelectedByValue(const LLSD& value, BOOL selected)
 {
 	S32 idx = 0;
-	std::string value_string = value.asString();
 	for (button_list_t::const_iterator iter = mRadioButtons.begin();
 		 iter != mRadioButtons.end(); ++iter)
 	{
-		if((*iter)->getName() == value_string)
+		if((*iter)->getPayload().asString() == value.asString())
 		{
 			setSelectedIndex(idx);
 			return TRUE;
@@ -455,11 +501,10 @@ LLSD	LLRadioGroup::getSelectedValue()
 BOOL	LLRadioGroup::isSelected(const LLSD& value) const
 {
 	S32 idx = 0;
-	std::string value_string = value.asString();
 	for (button_list_t::const_iterator iter = mRadioButtons.begin();
 		 iter != mRadioButtons.end(); ++iter)
 	{
-		if((*iter)->getName() == value_string)
+		if((*iter)->getPayload().asString() == value.asString())
 		{
 			if (idx == mSelectedIndex) 
 			{
@@ -481,6 +526,17 @@ BOOL	LLRadioGroup::operateOnAll(EOperation op)
 	return FALSE;
 }
 
+LLRadioCtrl::LLRadioCtrl(const std::string& name, const LLRect& rect, const std::string& label, const std::string& value, const LLFontGL* font, commit_callback_t commit_callback)
+:	LLCheckBoxCtrl(name, rect, label, font, commit_callback, FALSE, RADIO_STYLE),
+	mPayload(value)
+{
+	setTabStop(FALSE);
+	// use name as default "Value" for backwards compatibility
+	if (value.empty())
+	{
+		mPayload = name;
+	}
+}
 
 LLRadioCtrl::~LLRadioCtrl()
 {
@@ -499,7 +555,7 @@ LLXMLNodePtr LLRadioCtrl::getXML(bool save_children) const
 
 	node->setName(LL_RADIO_ITEM_TAG);
 
-	node->setStringValue(getLabel());
+	node->createChild("value", TRUE)->setStringValue(mPayload);
 
 	return node;
 }

indra/llui/llradiogroup.h

@@ -37,26 +37,6 @@
 #include "llcheckboxctrl.h"
 #include "llctrlselectioninterface.h"
 
-
-/*
- * A checkbox control with use_radio_style == true.
- */
-class LLRadioCtrl : public LLCheckBoxCtrl 
-{
-public:
-	LLRadioCtrl(const std::string& name, const LLRect& rect, const std::string& label, const LLFontGL* font = NULL,
-		commit_callback_t commit_callback = NULL) :
-				LLCheckBoxCtrl(name, rect, label, font, commit_callback, FALSE, RADIO_STYLE)
-	{
-		setTabStop(FALSE);
-	}
-	/*virtual*/ ~LLRadioCtrl();
-
-	virtual LLXMLNodePtr getXML(bool save_children = true) const;
-	/*virtual*/ void setValue(const LLSD& value);
-};
-
-
 /*
  * An invisible view containing multiple mutually exclusive toggling 
  * buttons (usually radio buttons).  Automatically handles the mutex
@@ -66,25 +46,32 @@ class LLRadioGroup
 :	public LLUICtrl, public LLCtrlSelectionInterface
 {
 public:
-	// Radio group constructor. Doesn't rely on
+
-	// needing a control
+	// Radio group constructor. Doesn't rely on needing a control
 	LLRadioGroup(const std::string& name, const LLRect& rect,
 				 S32 initial_index,
 				 commit_callback_t commit_callback,
-				 BOOL border = TRUE);
+				 bool border = true, bool allow_deselect = false);
+
+protected:
+	friend class LLUICtrlFactory;
+
+public:
 
 	virtual ~LLRadioGroup();
 
+	virtual BOOL postBuild();
+
+	virtual BOOL handleMouseDown(S32 x, S32 y, MASK mask);
+
 	virtual BOOL handleKeyHere(KEY key, MASK mask);
 
 	virtual void setEnabled(BOOL enabled);
 	virtual LLXMLNodePtr getXML(bool save_children = true) const;
 	static LLView* fromXML(LLXMLNodePtr node, LLView *parent, LLUICtrlFactory *factory);
 	void setIndexEnabled(S32 index, BOOL enabled);
-	
 	// return the index value of the selected item
 	S32 getSelectedIndex() const { return mSelectedIndex; }
-	
 	// set the index value programatically
 	BOOL setSelectedIndex(S32 index, BOOL from_event = FALSE);
 
@@ -92,14 +79,10 @@ public:
 	virtual void	setValue(const LLSD& value );
 	virtual LLSD	getValue() const;
 
-	// Draw the group, but also fix the highlighting based on the control.
-	void draw();
-
 	// You must use this method to add buttons to a radio group.
 	// Don't use addChild -- it won't set the callback function
 	// correctly.
-	LLRadioCtrl* addRadioButton(const std::string& name, const std::string& label, const LLRect& rect, const LLFontGL* font);
+	class LLRadioCtrl* addRadioButton(const std::string& name, const std::string& label, const LLRect& rect, const LLFontGL* font, const std::string& payload = "");
-	LLRadioCtrl* getRadioButton(const S32& index) { return mRadioButtons[index]; }
 	// Update the control as needed.  Userdata must be a pointer to the button.
 	void onClickButton(LLUICtrl* clicked_radio);
 	
@@ -123,14 +106,15 @@ public:
 
 private:
 	// protected function shared by the two constructors.
-	void init(BOOL border);
+	void init(bool border);
 
 	S32 mSelectedIndex;
-	typedef std::vector<LLRadioCtrl*> button_list_t;
+
+	typedef std::vector<class LLRadioCtrl*> button_list_t;
 	button_list_t mRadioButtons;
 
-	BOOL mHasBorder;
+	bool mHasBorder;
+	bool				mAllowDeselect;	// user can click on an already selected option to deselect it
 };
 
-
 #endif

indra/llvfs/lldir_win32.cpp

@@ -62,13 +62,13 @@ LLDir_Win32::LLDir_Win32()
 		if((*pSHGetKnownFolderPath)(FOLDERID_RoamingAppData, 0, NULL, &pPath) == S_OK)
 			wcscpy_s(w_str,pPath);
 		else
-			SHGetSpecialFolderPath(NULL, w_str, CSIDL_APPDATA, TRUE);
+			SHGetFolderPath(NULL, CSIDL_APPDATA | CSIDL_FLAG_CREATE, NULL, SHGFP_TYPE_DEFAULT, w_str );
 		if(pPath)
 			CoTaskMemFree(pPath);
 	}
 	else	//XP doesn't support SHGetKnownFolderPath
 	{
-		SHGetSpecialFolderPath(NULL, w_str, CSIDL_APPDATA, TRUE);
+		SHGetFolderPath(NULL, CSIDL_APPDATA | CSIDL_FLAG_CREATE, NULL, SHGFP_TYPE_DEFAULT, w_str );
 	}
 
 	mOSUserDir = utf16str_to_utf8str(llutf16string(w_str));
@@ -91,13 +91,13 @@ LLDir_Win32::LLDir_Win32()
 		if((*pSHGetKnownFolderPath)(FOLDERID_LocalAppData, 0, NULL, &pPath) == S_OK)
 			wcscpy_s(w_str,pPath);
 		else
-			SHGetSpecialFolderPath(NULL, w_str, CSIDL_LOCAL_APPDATA, TRUE);
+			SHGetFolderPath(NULL, CSIDL_LOCAL_APPDATA | CSIDL_FLAG_CREATE, NULL, SHGFP_TYPE_DEFAULT, w_str );
 		if(pPath)
 			CoTaskMemFree(pPath);
 	}
 	else	//XP doesn't support SHGetKnownFolderPath
 	{
-		SHGetSpecialFolderPath(NULL, w_str, CSIDL_LOCAL_APPDATA, TRUE);
+		SHGetFolderPath(NULL, CSIDL_LOCAL_APPDATA | CSIDL_FLAG_CREATE, NULL, SHGFP_TYPE_DEFAULT, w_str );
 	}
 
 	if(shell)

indra/llwindow/lldxhardware.cpp

@@ -2,31 +2,25 @@
  * @file lldxhardware.cpp
  * @brief LLDXHardware implementation
  *
- * $LicenseInfo:firstyear=2001&license=viewergpl$
+ * $LicenseInfo:firstyear=2001&license=viewerlgpl$
- * 
- * Copyright (c) 2001-2009, Linden Research, Inc.
- * 
  * Second Life Viewer Source Code
- * The source code in this file ("Source Code") is provided by Linden Lab
+ * Copyright (C) 2010, Linden Research, Inc.
- * to you under the terms of the GNU General Public License, version 2.0
+ *
- * ("GPL"), unless you have obtained a separate licensing agreement
+ * This library is free software; you can redistribute it and/or
- * ("Other License"), formally executed by you and Linden Lab.  Terms of
+ * modify it under the terms of the GNU Lesser General Public
- * the GPL can be found in doc/GPL-license.txt in this distribution, or
+ * License as published by the Free Software Foundation;
- * online at http://secondlifegrid.net/programs/open_source/licensing/gplv2
+ * version 2.1 of the License only.
  * 
- * There are special exceptions to the terms and conditions of the GPL as
+ * This library is distributed in the hope that it will be useful,
- * it is applied to this Source Code. View the full text of the exception
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
- * in the file doc/FLOSS-exception.txt in this software distribution, or
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
- * online at
+ * Lesser General Public License for more details.
- * http://secondlifegrid.net/programs/open_source/licensing/flossexception
  * 
- * By copying, modifying or distributing this software, you acknowledge
+ * You should have received a copy of the GNU Lesser General Public
- * that you have read and understood your obligations described above,
+ * License along with this library; if not, write to the Free Software
- * and agree to abide by those obligations.
+ * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA  02110-1301  USA
  * 
- * ALL LINDEN LAB SOURCE CODE IS PROVIDED "AS IS." LINDEN LAB MAKES NO
+ * Linden Research, Inc., 945 Battery Street, San Francisco, CA  94111  USA
- * WARRANTIES, EXPRESS, IMPLIED OR OTHERWISE, REGARDING ITS ACCURACY,
- * COMPLETENESS OR PERFORMANCE.
  * $/LicenseInfo$
  */
 
@@ -40,9 +34,12 @@
 #include <dxdiag.h>
 #undef INITGUID
 
+#include <wbemidl.h>
+
 #include <boost/tokenizer.hpp>
 
 #include "lldxhardware.h"
+
 #include "llerror.h"
 
 #include "llstring.h"
@@ -59,11 +56,160 @@ LLDXHardware gDXHardware;
 #define SAFE_DELETE_ARRAY(p) { if(p) { delete[] (p);   (p)=NULL; } }
 #define SAFE_RELEASE(p)      { if(p) { (p)->Release(); (p)=NULL; } }
 
-std::string get_string(IDxDiagContainer *containerp, WCHAR *wszPropName)
+typedef BOOL ( WINAPI* PfnCoSetProxyBlanket )( IUnknown* pProxy, DWORD dwAuthnSvc, DWORD dwAuthzSvc,
+											   OLECHAR* pServerPrincName, DWORD dwAuthnLevel, DWORD dwImpLevel,
+											   RPC_AUTH_IDENTITY_HANDLE pAuthInfo, DWORD dwCapabilities );
+
+HRESULT GetVideoMemoryViaWMI( WCHAR* strInputDeviceID, DWORD* pdwAdapterRam )
 {
-    HRESULT hr;
+	HRESULT hr;
+	bool bGotMemory = false;
+	HRESULT hrCoInitialize = S_OK;
+	IWbemLocator* pIWbemLocator = nullptr;
+	IWbemServices* pIWbemServices = nullptr;
+	BSTR pNamespace = nullptr;
+
+	*pdwAdapterRam = 0;
+	hrCoInitialize = CoInitialize( 0 );
+
+	hr = CoCreateInstance( CLSID_WbemLocator,
+						   nullptr,
+						   CLSCTX_INPROC_SERVER,
+						   IID_IWbemLocator,
+						   ( LPVOID* )&pIWbemLocator );
+#ifdef PRINTF_DEBUGGING
+	if( FAILED( hr ) ) wprintf( L"WMI: CoCreateInstance failed: 0x%0.8x\n", hr );
+#endif
+
+	if( SUCCEEDED( hr ) && pIWbemLocator )
+	{
+		// Using the locator, connect to WMI in the given namespace.
+		pNamespace = SysAllocString( L"\\\\.\\root\\cimv2" );
+
+		hr = pIWbemLocator->ConnectServer( pNamespace, nullptr, nullptr, 0L,
+										   0L, nullptr, nullptr, &pIWbemServices );
+#ifdef PRINTF_DEBUGGING
+		if( FAILED( hr ) ) wprintf( L"WMI: pIWbemLocator->ConnectServer failed: 0x%0.8x\n", hr );
+#endif
+		if( SUCCEEDED( hr ) && pIWbemServices != 0 )
+		{
+			HINSTANCE hinstOle32 = nullptr;
+
+			hinstOle32 = LoadLibraryW( L"ole32.dll" );
+			if( hinstOle32 )
+			{
+				PfnCoSetProxyBlanket pfnCoSetProxyBlanket = nullptr;
+
+				pfnCoSetProxyBlanket = ( PfnCoSetProxyBlanket )GetProcAddress( hinstOle32, "CoSetProxyBlanket" );
+				if( pfnCoSetProxyBlanket != 0 )
+				{
+					// Switch security level to IMPERSONATE.
+					pfnCoSetProxyBlanket( pIWbemServices, RPC_C_AUTHN_WINNT, RPC_C_AUTHZ_NONE, nullptr,
+										  RPC_C_AUTHN_LEVEL_CALL, RPC_C_IMP_LEVEL_IMPERSONATE, nullptr, 0 );
+				}
+
+				FreeLibrary( hinstOle32 );
+			}
+
+			IEnumWbemClassObject* pEnumVideoControllers = nullptr;
+			BSTR pClassName = nullptr;
+
+			pClassName = SysAllocString( L"Win32_VideoController" );
+
+			hr = pIWbemServices->CreateInstanceEnum( pClassName, 0,
+													 nullptr, &pEnumVideoControllers );
+#ifdef PRINTF_DEBUGGING
+			if( FAILED( hr ) ) wprintf( L"WMI: pIWbemServices->CreateInstanceEnum failed: 0x%0.8x\n", hr );
+#endif
+
+			if( SUCCEEDED( hr ) && pEnumVideoControllers )
+			{
+				IWbemClassObject* pVideoControllers[10] = {0};
+				DWORD uReturned = 0;
+				BSTR pPropName = nullptr;
+
+				// Get the first one in the list
+				pEnumVideoControllers->Reset();
+				hr = pEnumVideoControllers->Next( 5000,             // timeout in 5 seconds
+												  10,                  // return the first 10
+												  pVideoControllers,
+												  &uReturned );
+#ifdef PRINTF_DEBUGGING
+				if( FAILED( hr ) ) wprintf( L"WMI: pEnumVideoControllers->Next failed: 0x%0.8x\n", hr );
+				if( uReturned == 0 ) wprintf( L"WMI: pEnumVideoControllers uReturned == 0\n" );
+#endif
+
+				VARIANT var;
+				if( SUCCEEDED( hr ) )
+				{
+					bool bFound = false;
+					for( UINT iController = 0; iController < uReturned; iController++ )
+					{
+						if ( !pVideoControllers[iController] )
+							continue;
+
+						pPropName = SysAllocString( L"PNPDeviceID" );
+						hr = pVideoControllers[iController]->Get( pPropName, 0L, &var, nullptr, nullptr );
+#ifdef PRINTF_DEBUGGING
+						if( FAILED( hr ) )
+							wprintf( L"WMI: pVideoControllers[iController]->Get PNPDeviceID failed: 0x%0.8x\n", hr );
+#endif
+						if( SUCCEEDED( hr ) )
+						{
+							if( wcsstr( var.bstrVal, strInputDeviceID ) != 0 )
+								bFound = true;
+						}
+						VariantClear( &var );
+						if( pPropName ) SysFreeString( pPropName );
+
+						if( bFound )
+						{
+							pPropName = SysAllocString( L"AdapterRAM" );
+							hr = pVideoControllers[iController]->Get( pPropName, 0L, &var, nullptr, nullptr );
+#ifdef PRINTF_DEBUGGING
+							if( FAILED( hr ) )
+								wprintf( L"WMI: pVideoControllers[iController]->Get AdapterRAM failed: 0x%0.8x\n",
+										 hr );
+#endif
+							if( SUCCEEDED( hr ) )
+							{
+								bGotMemory = true;
+								*pdwAdapterRam = var.ulVal;
+							}
+							VariantClear( &var );
+							if( pPropName ) SysFreeString( pPropName );
+							break;
+						}
+						SAFE_RELEASE( pVideoControllers[iController] );
+					}
+				}
+			}
+
+			if( pClassName )
+				SysFreeString( pClassName );
+			SAFE_RELEASE( pEnumVideoControllers );
+		}
+
+		if( pNamespace )
+			SysFreeString( pNamespace );
+		SAFE_RELEASE( pIWbemServices );
+	}
+
+	SAFE_RELEASE( pIWbemLocator );
+
+	if( SUCCEEDED( hrCoInitialize ) )
+		CoUninitialize();
+
+	if( bGotMemory )
+		return S_OK;
+	else
+		return E_FAIL;
+}
+
+void get_wstring(IDxDiagContainer* containerp, WCHAR* wszPropName, WCHAR* wszPropValue, int outputSize)
+{
+	HRESULT hr;
 	VARIANT var;
-	WCHAR wszPropValue[256];
 
 	VariantInit( &var );
 	hr = containerp->GetProp(wszPropName, &var );
@@ -82,13 +228,19 @@ std::string get_string(IDxDiagContainer *containerp, WCHAR *wszPropName)
 				wcscpy( wszPropValue, (var.boolVal) ? L"true" : L"false" );	/* Flawfinder: ignore */
 				break;
 			case VT_BSTR:
-				wcsncpy( wszPropValue, var.bstrVal, 255 );	/* Flawfinder: ignore */
+				wcsncpy( wszPropValue, var.bstrVal, outputSize-1 );	/* Flawfinder: ignore */
-				wszPropValue[255] = 0;
+				wszPropValue[outputSize-1] = 0;
 				break;
 		}
 	}
 	// Clear the variant (this is needed to free BSTR memory)
 	VariantClear( &var );
+}
+
+std::string get_string(IDxDiagContainer *containerp, WCHAR *wszPropName)
+{
+	WCHAR wszPropValue[256];
+	get_wstring(containerp, wszPropName, wszPropValue, 256);
 
 	return utf16str_to_utf8str(wszPropValue);
 }
@@ -126,7 +278,7 @@ BOOL LLVersion::set(const std::string &version_string)
 	}
 	if (count < 4)
 	{
-		//llwarns << "Potentially bogus version string!" << version_string << llendl;
+		//LL_WARNS() << "Potentially bogus version string!" << version_string << LL_ENDL;
 		for (i = 0; i < 4; i++)
 		{
 			mFields[i] = 0;
@@ -177,6 +329,7 @@ std::string LLDXDriverFile::dump()
 LLDXDevice::~LLDXDevice()
 {
 	for_each(mDriverFiles.begin(), mDriverFiles.end(), DeletePairedPointer());
+	mDriverFiles.clear();
 }
 
 std::string LLDXDevice::dump()
@@ -236,6 +389,7 @@ LLDXHardware::LLDXHardware()
 void LLDXHardware::cleanup()
 {
   // for_each(mDevices.begin(), mDevices.end(), DeletePairedPointer());
+  // mDevices.clear();
 }
 
 /*
@@ -365,8 +519,18 @@ BOOL LLDXHardware::getInfo(BOOL vram_only)
             goto LCleanup;
 		}
 		
-		// Get the English VRAM string
+		DWORD vram = 0;
+
+		WCHAR deviceID[512];
+
+		get_wstring(device_containerp, L"szDeviceID", deviceID, 512);
+
+		if (SUCCEEDED(GetVideoMemoryViaWMI(deviceID, &vram)))
 		{
+			mVRAM = vram/(1024*1024);
+		}
+		else
+		{ // Get the English VRAM string
 		  std::string ram_str = get_string(device_containerp, L"szDisplayMemoryEnglish");
 
 		  // We don't need the device any more

indra/llwindow/llwindowmesaheadless.h

@@ -30,7 +30,6 @@
 #if LL_MESA_HEADLESS
 
 #include "llwindow.h"
-#include "GL/glu.h"
 #include "GL/osmesa.h"
 
 class LLWindowMesaHeadless : public LLWindow

indra/llwindow/llwindowwin32.cpp

@@ -38,6 +38,7 @@
 
 // Linden library includes
 #include "llerror.h"
+#include "llfasttimer.h"
 #include "llgl.h"
 #include "llstring.h"
 #include "lldir.h"
@@ -58,8 +59,6 @@
 #include <dinput.h>
 #include <Dbt.h.>
 
-#include "llfasttimer.h"
-
 // culled from winuser.h
 #ifndef WM_MOUSEWHEEL /* Added to be compatible with later SDK's */
 const S32	WM_MOUSEWHEEL = 0x020A;
@@ -87,6 +86,18 @@ void show_window_creation_error(const std::string& title)
 	LL_WARNS("Window") << title << LL_ENDL;
 }
 
+HGLRC SafeCreateContext(HDC hdc)
+{
+	__try
+	{
+		return wglCreateContext(hdc);
+	}
+	__except(EXCEPTION_EXECUTE_HANDLER)
+	{
+		return NULL;
+	}
+}
+
 //static
 BOOL LLWindowWin32::sIsClassRegistered = FALSE;
 
@@ -161,9 +172,8 @@ LLWinImm LLWinImm::sTheInstance;
 LLWinImm::LLWinImm() : mHImmDll(NULL)
 {
 	// Check system metrics 
-	if ( !GetSystemMetrics( SM_DBCSENABLED ) )
+	if ( !GetSystemMetrics( SM_IMMENABLED ) )
 		return;
-	
 
 	mHImmDll = LoadLibraryA("Imm32");
 	if (mHImmDll != NULL)
@@ -205,7 +215,7 @@ LLWinImm::LLWinImm() : mHImmDll(NULL)
 			// the case, since it is very unusual; these APIs are available from 
 			// the beginning, and all versions of IMM32.DLL should have them all.  
 			// Unfortunately, this code may be executed before initialization of 
-			// the logging channel (llwarns), and we can't do it here...  Yes, this 
+			// the logging channel (LL_WARNS()), and we can't do it here...  Yes, this
 			// is one of disadvantages to use static constraction to DLL loading. 
 			FreeLibrary(mHImmDll);
 			mHImmDll = NULL;
@@ -554,6 +564,7 @@ LLWindowWin32::LLWindowWin32(LLWindowCallbacks* callbacks,
 		if (closest_refresh == 0)
 		{
 			LL_WARNS("Window") << "Couldn't find display mode " << width << " by " << height << " at " << BITS_PER_PIXEL << " bits per pixel" << LL_ENDL;
+			//success = FALSE;
 
 			if (!EnumDisplaySettings(NULL, ENUM_CURRENT_SETTINGS, &dev_mode))
 			{
@@ -660,7 +671,7 @@ LLWindowWin32::~LLWindowWin32()
 	delete [] mSupportedResolutions;
 	mSupportedResolutions = NULL;
 
-	delete mWindowClassName;
+	delete [] mWindowClassName;
 	mWindowClassName = NULL;
 }
 
@@ -749,7 +760,7 @@ void LLWindowWin32::close()
 	LL_DEBUGS("Window") << "Destroying Window" << LL_ENDL;
 	
 	// Don't process events in our mainWindowProc any longer.
-	SetWindowLongPtr(mWindowHandle, GWLP_USERDATA, NULL);
+	SetWindowLongPtr(mWindowHandle, GWLP_USERDATA, NULL); // <alchemy/>
 
 	// Make sure we don't leave a blank toolbar button.
 	ShowWindow(mWindowHandle, SW_HIDE);
@@ -1012,7 +1023,8 @@ BOOL LLWindowWin32::switchContext(BOOL fullscreen, const LLCoordScreen &size, co
 			dw_ex_style = WS_EX_APPWINDOW;
 			dw_style = WS_POPUP;
 
-			// Move window borders out not to cover window contents
+			// Move window borders out not to cover window contents.
+			// This converts client rect to window rect, i.e. expands it by the window border size.
 			AdjustWindowRectEx(&window_rect, dw_style, FALSE, dw_ex_style);
 		}
 		// If it failed, we don't want to run fullscreen
@@ -1059,7 +1071,7 @@ BOOL LLWindowWin32::switchContext(BOOL fullscreen, const LLCoordScreen &size, co
 		mhInstance,
 		NULL);
 
-	LL_INFOS("Window") << "window is created." << llendl ;
+	LL_INFOS("Window") << "window is created." << LL_ENDL ;
 
 	//-----------------------------------------------------------------------
 	// Create GL drawing context
@@ -1092,7 +1104,7 @@ BOOL LLWindowWin32::switchContext(BOOL fullscreen, const LLCoordScreen &size, co
 		return FALSE;
 	}
 
-	LL_INFOS("Window") << "Device context retrieved." << llendl ;
+	LL_INFOS("Window") << "Device context retrieved." << LL_ENDL ;
 
 	if (!(pixel_format = ChoosePixelFormat(mhDC, &pfd)))
 	{
@@ -1102,7 +1114,7 @@ BOOL LLWindowWin32::switchContext(BOOL fullscreen, const LLCoordScreen &size, co
 		return FALSE;
 	}
 
-	LL_INFOS("Window") << "Pixel format chosen." << llendl ;
+	LL_INFOS("Window") << "Pixel format chosen." << LL_ENDL ;
 
 	// Verify what pixel format we actually received.
 	if (!DescribePixelFormat(mhDC, pixel_format, sizeof(PIXELFORMATDESCRIPTOR),
@@ -1115,35 +1127,35 @@ BOOL LLWindowWin32::switchContext(BOOL fullscreen, const LLCoordScreen &size, co
 	}
 
 	// (EXP-1765) dump pixel data to see if there is a pattern that leads to unreproducible crash
-	LL_INFOS("Window") << "--- begin pixel format dump ---" << llendl ;
+	LL_INFOS("Window") << "--- begin pixel format dump ---" << LL_ENDL ;
-	LL_INFOS("Window") << "pixel_format is " << pixel_format << llendl ;
+	LL_INFOS("Window") << "pixel_format is " << pixel_format << LL_ENDL ;
-	LL_INFOS("Window") << "pfd.nSize:            " << pfd.nSize << llendl ;
+	LL_INFOS("Window") << "pfd.nSize:            " << pfd.nSize << LL_ENDL ;
-	LL_INFOS("Window") << "pfd.nVersion:         " << pfd.nVersion << llendl ;
+	LL_INFOS("Window") << "pfd.nVersion:         " << pfd.nVersion << LL_ENDL ;
-	LL_INFOS("Window") << "pfd.dwFlags:          0x" << std::hex << pfd.dwFlags << std::dec << llendl ;
+	LL_INFOS("Window") << "pfd.dwFlags:          0x" << std::hex << pfd.dwFlags << std::dec << LL_ENDL ;
-	LL_INFOS("Window") << "pfd.iPixelType:       " << (int)pfd.iPixelType << llendl ;
+	LL_INFOS("Window") << "pfd.iPixelType:       " << (int)pfd.iPixelType << LL_ENDL ;
-	LL_INFOS("Window") << "pfd.cColorBits:       " << (int)pfd.cColorBits << llendl ;
+	LL_INFOS("Window") << "pfd.cColorBits:       " << (int)pfd.cColorBits << LL_ENDL ;
-	LL_INFOS("Window") << "pfd.cRedBits:         " << (int)pfd.cRedBits << llendl ;
+	LL_INFOS("Window") << "pfd.cRedBits:         " << (int)pfd.cRedBits << LL_ENDL ;
-	LL_INFOS("Window") << "pfd.cRedShift:        " << (int)pfd.cRedShift << llendl ;
+	LL_INFOS("Window") << "pfd.cRedShift:        " << (int)pfd.cRedShift << LL_ENDL ;
-	LL_INFOS("Window") << "pfd.cGreenBits:       " << (int)pfd.cGreenBits << llendl ;
+	LL_INFOS("Window") << "pfd.cGreenBits:       " << (int)pfd.cGreenBits << LL_ENDL ;
-	LL_INFOS("Window") << "pfd.cGreenShift:      " << (int)pfd.cGreenShift << llendl ;
+	LL_INFOS("Window") << "pfd.cGreenShift:      " << (int)pfd.cGreenShift << LL_ENDL ;
-	LL_INFOS("Window") << "pfd.cBlueBits:        " << (int)pfd.cBlueBits << llendl ;
+	LL_INFOS("Window") << "pfd.cBlueBits:        " << (int)pfd.cBlueBits << LL_ENDL ;
-	LL_INFOS("Window") << "pfd.cBlueShift:       " << (int)pfd.cBlueShift << llendl ;
+	LL_INFOS("Window") << "pfd.cBlueShift:       " << (int)pfd.cBlueShift << LL_ENDL ;
-	LL_INFOS("Window") << "pfd.cAlphaBits:       " << (int)pfd.cAlphaBits << llendl ;
+	LL_INFOS("Window") << "pfd.cAlphaBits:       " << (int)pfd.cAlphaBits << LL_ENDL ;
-	LL_INFOS("Window") << "pfd.cAlphaShift:      " << (int)pfd.cAlphaShift << llendl ;
+	LL_INFOS("Window") << "pfd.cAlphaShift:      " << (int)pfd.cAlphaShift << LL_ENDL ;
-	LL_INFOS("Window") << "pfd.cAccumBits:       " << (int)pfd.cAccumBits << llendl ;
+	LL_INFOS("Window") << "pfd.cAccumBits:       " << (int)pfd.cAccumBits << LL_ENDL ;
-	LL_INFOS("Window") << "pfd.cAccumRedBits:    " << (int)pfd.cAccumRedBits << llendl ;
+	LL_INFOS("Window") << "pfd.cAccumRedBits:    " << (int)pfd.cAccumRedBits << LL_ENDL ;
-	LL_INFOS("Window") << "pfd.cAccumGreenBits:  " << (int)pfd.cAccumGreenBits << llendl ;
+	LL_INFOS("Window") << "pfd.cAccumGreenBits:  " << (int)pfd.cAccumGreenBits << LL_ENDL ;
-	LL_INFOS("Window") << "pfd.cAccumBlueBits:   " << (int)pfd.cAccumBlueBits << llendl ;
+	LL_INFOS("Window") << "pfd.cAccumBlueBits:   " << (int)pfd.cAccumBlueBits << LL_ENDL ;
-	LL_INFOS("Window") << "pfd.cAccumAlphaBits:  " << (int)pfd.cAccumAlphaBits << llendl ;
+	LL_INFOS("Window") << "pfd.cAccumAlphaBits:  " << (int)pfd.cAccumAlphaBits << LL_ENDL ;
-	LL_INFOS("Window") << "pfd.cDepthBits:       " << (int)pfd.cDepthBits << llendl ;
+	LL_INFOS("Window") << "pfd.cDepthBits:       " << (int)pfd.cDepthBits << LL_ENDL ;
-	LL_INFOS("Window") << "pfd.cStencilBits:     " << (int)pfd.cStencilBits << llendl ;
+	LL_INFOS("Window") << "pfd.cStencilBits:     " << (int)pfd.cStencilBits << LL_ENDL ;
-	LL_INFOS("Window") << "pfd.cAuxBuffers:      " << (int)pfd.cAuxBuffers << llendl ;
+	LL_INFOS("Window") << "pfd.cAuxBuffers:      " << (int)pfd.cAuxBuffers << LL_ENDL ;
-	LL_INFOS("Window") << "pfd.iLayerType:       " << (int)pfd.iLayerType << llendl ;
+	LL_INFOS("Window") << "pfd.iLayerType:       " << (int)pfd.iLayerType << LL_ENDL ;
-	LL_INFOS("Window") << "pfd.bReserved:        " << (int)pfd.bReserved << llendl ;
+	LL_INFOS("Window") << "pfd.bReserved:        " << (int)pfd.bReserved << LL_ENDL ;
-	LL_INFOS("Window") << "pfd.dwLayerMask:      " << pfd.dwLayerMask << llendl ;
+	LL_INFOS("Window") << "pfd.dwLayerMask:      " << pfd.dwLayerMask << LL_ENDL ;
-	LL_INFOS("Window") << "pfd.dwVisibleMask:    " << pfd.dwVisibleMask << llendl ;
+	LL_INFOS("Window") << "pfd.dwVisibleMask:    " << pfd.dwVisibleMask << LL_ENDL ;
-	LL_INFOS("Window") << "pfd.dwDamageMask:     " << pfd.dwDamageMask << llendl ;
+	LL_INFOS("Window") << "pfd.dwDamageMask:     " << pfd.dwDamageMask << LL_ENDL ;
-	LL_INFOS("Window") << "--- end pixel format dump ---" << llendl ;
+	LL_INFOS("Window") << "--- end pixel format dump ---" << LL_ENDL ;
 
 	if (pfd.cColorBits < 32)
 	{
@@ -1169,7 +1181,7 @@ BOOL LLWindowWin32::switchContext(BOOL fullscreen, const LLCoordScreen &size, co
 		return FALSE;
 	}
 
-	if (!(mhRC = wglCreateContext(mhDC)))
+	if (!(mhRC = SafeCreateContext(mhDC)))
 	{
 		close();
 		OSMessageBox(mCallbacks->translateString("MBGLContextErr"),
@@ -1185,7 +1197,7 @@ BOOL LLWindowWin32::switchContext(BOOL fullscreen, const LLCoordScreen &size, co
 		return FALSE;
 	}
 
-	LL_INFOS("Window") << "Drawing context is created." << llendl ;
+	LL_INFOS("Window") << "Drawing context is created." << LL_ENDL ;
 
 	gGLManager.initWGL();
 
@@ -1323,142 +1335,15 @@ BOOL LLWindowWin32::switchContext(BOOL fullscreen, const LLCoordScreen &size, co
 			LL_INFOS("Window") << "Choosing pixel formats: " << num_formats << " pixel formats returned" << LL_ENDL;
 		}
 
-		LL_INFOS("Window") << "pixel formats done." << llendl ;
+		LL_INFOS("Window") << "pixel formats done." << LL_ENDL ;
-
-		/*for(int i = 0; i <= num_formats-1; ++i)
-		{
-			GLint query[] = {	WGL_SAMPLE_BUFFERS_ARB,
-								WGL_SAMPLES_ARB,
-								WGL_NUMBER_PIXEL_FORMATS_ARB,
-								WGL_DRAW_TO_WINDOW_ARB,
-								WGL_DRAW_TO_BITMAP_ARB,
-								WGL_ACCELERATION_ARB,
-								WGL_NEED_PALETTE_ARB,
-								WGL_NEED_SYSTEM_PALETTE_ARB,
-								WGL_SWAP_LAYER_BUFFERS_ARB,
-								WGL_SWAP_METHOD_ARB,
-								WGL_NUMBER_OVERLAYS_ARB,
-								WGL_NUMBER_UNDERLAYS_ARB,
-								WGL_TRANSPARENT_ARB,
-								WGL_TRANSPARENT_RED_VALUE_ARB,
-								WGL_TRANSPARENT_GREEN_VALUE_ARB,
-								WGL_TRANSPARENT_BLUE_VALUE_ARB,
-								WGL_TRANSPARENT_ALPHA_VALUE_ARB,
-								WGL_TRANSPARENT_INDEX_VALUE_ARB,
-								WGL_SHARE_DEPTH_ARB,
-								WGL_SHARE_STENCIL_ARB,
-								WGL_SHARE_ACCUM_ARB,
-								WGL_SUPPORT_GDI_ARB,
-								WGL_SUPPORT_OPENGL_ARB,
-								WGL_DOUBLE_BUFFER_ARB,
-								WGL_STEREO_ARB,
-								WGL_PIXEL_TYPE_ARB,
-								WGL_COLOR_BITS_ARB,
-								WGL_RED_BITS_ARB,
-								WGL_RED_SHIFT_ARB,
-								WGL_GREEN_BITS_ARB,
-								WGL_GREEN_SHIFT_ARB,
-								WGL_BLUE_BITS_ARB,
-								WGL_BLUE_SHIFT_ARB,
-								WGL_ALPHA_BITS_ARB,
-								WGL_ALPHA_SHIFT_ARB,
-								WGL_ACCUM_BITS_ARB,
-								WGL_ACCUM_RED_BITS_ARB,
-								WGL_ACCUM_GREEN_BITS_ARB,
-								WGL_ACCUM_BLUE_BITS_ARB,
-								WGL_ACCUM_ALPHA_BITS_ARB,
-								WGL_DEPTH_BITS_ARB,
-								WGL_STENCIL_BITS_ARB,
-								WGL_AUX_BUFFERS_ARB};
-			std::string names[] = {	"WGL_SAMPLE_BUFFERS_ARB",
-								"WGL_SAMPLES_ARB",
-								"WGL_NUMBER_PIXEL_FORMATS_ARB",
-								"WGL_DRAW_TO_WINDOW_ARB",
-								"WGL_DRAW_TO_BITMAP_ARB",
-								"WGL_ACCELERATION_ARB",
-								"WGL_NEED_PALETTE_ARB",
-								"WGL_NEED_SYSTEM_PALETTE_ARB",
-								"WGL_SWAP_LAYER_BUFFERS_ARB",
-								"WGL_SWAP_METHOD_ARB",
-								"WGL_NUMBER_OVERLAYS_ARB",
-								"WGL_NUMBER_UNDERLAYS_ARB",
-								"WGL_TRANSPARENT_ARB",
-								"WGL_TRANSPARENT_RED_VALUE_ARB",
-								"WGL_TRANSPARENT_GREEN_VALUE_ARB",
-								"WGL_TRANSPARENT_BLUE_VALUE_ARB",
-								"WGL_TRANSPARENT_ALPHA_VALUE_ARB",
-								"WGL_TRANSPARENT_INDEX_VALUE_ARB",
-								"WGL_SHARE_DEPTH_ARB",
-								"WGL_SHARE_STENCIL_ARB",
-								"WGL_SHARE_ACCUM_ARB",
-								"WGL_SUPPORT_GDI_ARB",
-								"WGL_SUPPORT_OPENGL_ARB",
-								"WGL_DOUBLE_BUFFER_ARB",
-								"WGL_STEREO_ARB",
-								"WGL_PIXEL_TYPE_ARB",
-								"WGL_COLOR_BITS_ARB",
-								"WGL_RED_BITS_ARB",
-								"WGL_RED_SHIFT_ARB",
-								"WGL_GREEN_BITS_ARB",
-								"WGL_GREEN_SHIFT_ARB",
-								"WGL_BLUE_BITS_ARB",
-								"WGL_BLUE_SHIFT_ARB",
-								"WGL_ALPHA_BITS_ARB",
-								"WGL_ALPHA_SHIFT_ARB",
-								"WGL_ACCUM_BITS_ARB",
-								"WGL_ACCUM_RED_BITS_ARB",
-								"WGL_ACCUM_GREEN_BITS_ARB",
-								"WGL_ACCUM_BLUE_BITS_ARB",
-								"WGL_ACCUM_ALPHA_BITS_ARB",
-								"WGL_DEPTH_BITS_ARB",
-								"WGL_STENCIL_BITS_ARB",
-								"WGL_AUX_BUFFERS_ARB"};
-			S32 results[sizeof(query)/sizeof(query[0])]={0};
-
-			if(wglGetPixelFormatAttribivARB(mhDC, pixel_formats[i], 0, sizeof(query)/sizeof(query[0]), query, results))
-			{
-				llinfos << i << ":" << llendl;
-				for(int j = 0; j < sizeof(query)/sizeof(query[0]); ++j)
-				{
-					switch(results[j])
-					{
-					case WGL_NO_ACCELERATION_ARB:
-						llinfos << "   " << names[j] << " = " << "WGL_NO_ACCELERATION_ARB" << llendl;
-						break;
-					case WGL_GENERIC_ACCELERATION_ARB:
-						llinfos << "   " << names[j] << " = " << "WGL_GENERIC_ACCELERATION_ARB" << llendl;
-						break;
-					case WGL_FULL_ACCELERATION_ARB:
-						llinfos << "   " << names[j] << " = " << "WGL_FULL_ACCELERATION_ARB" << llendl;
-						break;
-					case WGL_SWAP_EXCHANGE_ARB:
-						llinfos << "   " << names[j] << " = " << "WGL_SWAP_EXCHANGE_ARB" << llendl;
-						break;
-					case WGL_SWAP_COPY_ARB:
-						llinfos << "   " << names[j] << " = " << "WGL_SWAP_COPY_ARB" << llendl;
-						break;
-					case WGL_SWAP_UNDEFINED_ARB:
-						llinfos << "   " << names[j] << " = " << "WGL_SWAP_UNDEFINED_ARB" << llendl;
-						break;
-					case WGL_TYPE_RGBA_ARB:
-						llinfos << "   " << names[j] << " = " << "WGL_TYPE_RGBA_ARB" << llendl;
-						break;
-					case WGL_TYPE_COLORINDEX_ARB:
-						llinfos << "   " << names[j] << " = " << "WGL_TYPE_COLORINDEX_ARB" << llendl;
-						break;
-					default:
-						llinfos << "   " << names[j] << " = " << results[j] << llendl;
-					}
-					
-				}
-			}
-		}*/
 
 		//Singu note: Reversed order of this loop. Generally, choosepixelformat returns an array with the closer matches towards the start.
 		S32 swap_method = 0;
 		S32 cur_format = 0;//num_formats-1;
 		GLint swap_query = WGL_SWAP_METHOD_ARB;
+
 		BOOL found_format = FALSE;
+
 		while (!found_format && wglGetPixelFormatAttribivARB(mhDC, pixel_formats[cur_format], 0, 1, &swap_query, &swap_method))
 		{
 			if (swap_method == WGL_SWAP_UNDEFINED_ARB /*|| cur_format <= 0*/)
@@ -1507,7 +1392,7 @@ BOOL LLWindowWin32::switchContext(BOOL fullscreen, const LLCoordScreen &size, co
 			mhInstance,
 			NULL);
 
-		LL_INFOS("Window") << "recreate window done." << llendl ;
+		LL_INFOS("Window") << "recreate window done." << LL_ENDL ;
 
 		if (!(mhDC = GetDC(mWindowHandle)))
 		{
@@ -1747,6 +1632,7 @@ BOOL LLWindowWin32::setCursorPosition(const LLCoordWindow position)
 		return FALSE;
 	}
 
+
 	// Inform the application of the new mouse position (needed for per-frame
 	// hover/picking to function).
 	mCallbacks->handleMouseMove(this, position.convert(), (MASK)0);
@@ -1996,7 +1882,11 @@ LRESULT CALLBACK LLWindowWin32::mainWindowProc(HWND h_wnd, UINT u_msg, WPARAM w_
 	// This helps prevent avatar walking after maximizing the window by double-clicking the title bar.
 	static bool sHandleLeftMouseUp = true;
 
-	LLWindowWin32 *window_imp = (LLWindowWin32 *)GetWindowLongPtr(h_wnd, GWLP_USERDATA);
+	// Ignore the double click received right after activating app.
+	// This is to avoid triggering double click teleport after returning focus (see MAINT-3786).
+	static bool sHandleDoubleClick = true;
+
+	LLWindowWin32 *window_imp = (LLWindowWin32 *)GetWindowLongPtr(h_wnd, GWLP_USERDATA); // <alchemy/>
 
 
 	if (NULL != window_imp)
@@ -2123,6 +2013,11 @@ LRESULT CALLBACK LLWindowWin32::mainWindowProc(HWND h_wnd, UINT u_msg, WPARAM w_
 					}
 				}
 
+				if (!activating)
+				{
+					sHandleDoubleClick = false;
+				}
+
 				window_imp->mCallbacks->handleActivateApp(window_imp, activating);
 
 				break;
@@ -2347,6 +2242,7 @@ LRESULT CALLBACK LLWindowWin32::mainWindowProc(HWND h_wnd, UINT u_msg, WPARAM w_
 				window_imp->mCallbacks->handlePingWatchdog(window_imp, "Main:WM_NCLBUTTONDOWN");
 				// A click in a non-client area, e.g. title bar or window border.
 				sHandleLeftMouseUp = false;
+				sHandleDoubleClick = true;
 			}
 			break;
 
@@ -2391,6 +2287,13 @@ LRESULT CALLBACK LLWindowWin32::mainWindowProc(HWND h_wnd, UINT u_msg, WPARAM w_
 		//case WM_RBUTTONDBLCLK:
 			{
 				window_imp->mCallbacks->handlePingWatchdog(window_imp, "Main:WM_LBUTTONDBLCLK");
+
+				if (!sHandleDoubleClick)
+				{
+					sHandleDoubleClick = true;
+					//break; // <singu/> Broke first double click every session.
+				}
+
 				// Because we move the cursor position in the app, we need to query
 				// to find out where the cursor at the time the event is handled.
 				// If we don't do this, many clicks could get buffered up, and if the
@@ -3379,7 +3282,6 @@ void LLWindowWin32::spawnWebBrowser(const std::string& escaped_url, bool async)
 	{
 		sei.fMask = SEE_MASK_ASYNCOK;
 	}
-	
 	else
 	{
 		sei.fMask = SEE_MASK_FLAG_DDEWAIT;

indra/lscript/lscript_compile/indra.l

@@ -594,7 +594,8 @@ void parse_string();
 "REGION_FLAG_SANDBOX"		{ count(); yylval.ival = REGION_FLAGS_SANDBOX; return(INTEGER_CONSTANT); }
 "REGION_FLAG_DISABLE_COLLISIONS"		{ count(); yylval.ival = REGION_FLAGS_SKIP_COLLISIONS; return(INTEGER_CONSTANT); }
 "REGION_FLAG_DISABLE_PHYSICS"		{ count(); yylval.ival = REGION_FLAGS_SKIP_PHYSICS; return(INTEGER_CONSTANT); }
-"REGION_FLAG_BLOCK_FLY"		{ count(); yylval.ival = REGION_FLAGS_BLOCK_FLY; return(INTEGER_CONSTANT); }
+"REGION_FLAG_BLOCK_FLY"			{ count(); yylval.ival = REGION_FLAGS_BLOCK_FLY; return(INTEGER_CONSTANT); }
+"REGION_FLAG_BLOCK_FLYOVER"		{ count(); yylval.ival = REGION_FLAGS_BLOCK_FLYOVER; return(INTEGER_CONSTANT); }
 "REGION_FLAG_ALLOW_DIRECT_TELEPORT"		{ count(); yylval.ival = REGION_FLAGS_ALLOW_DIRECT_TELEPORT; return(INTEGER_CONSTANT); }
 "REGION_FLAG_RESTRICT_PUSHOBJECT"		{ count(); yylval.ival = REGION_FLAGS_RESTRICT_PUSHOBJECT; return(INTEGER_CONSTANT); }