fix Linux build

Demos3/AllBullet2Demos/CMakeLists.txt

@@ -34,6 +34,10 @@ SET(App_AllBullet2Demos_SRCS
 	${BULLET_PHYSICS_SOURCE_DIR}/build3/bullet.rc
 )
 
+LINK_LIBRARIES(
+        Bullet3Common BulletSoftBody BulletDynamics BulletCollision LinearMath OpenGLWindow gwen ${OPENGL_gl_LIBRARY} ${OPENGL_glu_LIBRARY}
+)
+
 IF (WIN32)
 	SET(App_AllBullet2Demos_SRCS ${App_AllBullet2Demos_SRCS}  ${App_AllBullet2Demos_Common_SRCS})
 	INCLUDE_DIRECTORIES(
@@ -52,9 +56,6 @@ ENDIF(WIN32)
  
 	
 
-LINK_LIBRARIES(
-	Bullet3Common BulletSoftBody BulletDynamics BulletCollision LinearMath OpenGLWindow gwen ${OPENGL_gl_LIBRARY} ${OPENGL_glu_LIBRARY}
-)
 
 ADD_EXECUTABLE(App_AllBullet2Demos
 		${App_AllBullet2Demos_SRCS}

Demos3/SimpleOpenGL3/CMakeLists.txt

@@ -12,7 +12,7 @@ SET(AppSimpleOpenGL3_SRCS
 )
 
 LINK_LIBRARIES(
-	gwen OpenGLWindow Bullet3Common ${OPENGL_gl_LIBRARY} ${OPENGL_glu_LIBRARY}
+        gwen OpenGLWindow Bullet3Common ${OPENGL_gl_LIBRARY} ${OPENGL_glu_LIBRARY}
 )
 
 IF (WIN32)

src/BulletDynamics/ConstraintSolver/btSequentialImpulseConstraintSolver.cpp

@@ -22,6 +22,8 @@ subject to the following restrictions:
 #include "BulletCollision/NarrowPhaseCollision/btPersistentManifold.h"
 
 #include "LinearMath/btIDebugDraw.h"
+#include "LinearMath/btCpuFeatureUtility.h"
+
 //#include "btJacobianEntry.h"
 #include "LinearMath/btMinMax.h"
 #include "BulletDynamics/ConstraintSolver/btTypedConstraint.h"
@@ -38,11 +40,70 @@ int		gNumSplitImpulseRecoveries = 0;
 #include "BulletDynamics/Dynamics/btRigidBody.h"
 
 
+///This is the scalar reference implementation of solving a single constraint row, the innerloop of the Projected Gauss Seidel/Sequential Impulse constraint solver
+///Below are optional SSE2 and SSE4/FMA3 versions. We assume most hardware has SSE2. For SSE4/FMA3 we perform a CPU feature check.
+static btSimdScalar gResolveSingleConstraintRowGeneric_scalar_reference(btSolverBody& body1, btSolverBody& body2, const btSolverConstraint& c)
+{
+	btScalar deltaImpulse = c.m_rhs - btScalar(c.m_appliedImpulse)*c.m_cfm;
+	const btScalar deltaVel1Dotn = c.m_contactNormal1.dot(body1.internalGetDeltaLinearVelocity()) + c.m_relpos1CrossNormal.dot(body1.internalGetDeltaAngularVelocity());
+	const btScalar deltaVel2Dotn = c.m_contactNormal2.dot(body2.internalGetDeltaLinearVelocity()) + c.m_relpos2CrossNormal.dot(body2.internalGetDeltaAngularVelocity());
+
+	//	const btScalar delta_rel_vel	=	deltaVel1Dotn-deltaVel2Dotn;
+	deltaImpulse -= deltaVel1Dotn*c.m_jacDiagABInv;
+	deltaImpulse -= deltaVel2Dotn*c.m_jacDiagABInv;
+
+	const btScalar sum = btScalar(c.m_appliedImpulse) + deltaImpulse;
+	if (sum < c.m_lowerLimit)
+	{
+		deltaImpulse = c.m_lowerLimit - c.m_appliedImpulse;
+		c.m_appliedImpulse = c.m_lowerLimit;
+	}
+	else if (sum > c.m_upperLimit)
+	{
+		deltaImpulse = c.m_upperLimit - c.m_appliedImpulse;
+		c.m_appliedImpulse = c.m_upperLimit;
+	}
+	else
+	{
+		c.m_appliedImpulse = sum;
+	}
+
+	body1.internalApplyImpulse(c.m_contactNormal1*body1.internalGetInvMass(), c.m_angularComponentA, deltaImpulse);
+	body2.internalApplyImpulse(c.m_contactNormal2*body2.internalGetInvMass(), c.m_angularComponentB, deltaImpulse);
+
+	return deltaImpulse;
+}
+
+
+static btSimdScalar gResolveSingleConstraintRowLowerLimit_scalar_reference(btSolverBody& body1, btSolverBody& body2, const btSolverConstraint& c)
+{
+	btScalar deltaImpulse = c.m_rhs - btScalar(c.m_appliedImpulse)*c.m_cfm;
+	const btScalar deltaVel1Dotn = c.m_contactNormal1.dot(body1.internalGetDeltaLinearVelocity()) + c.m_relpos1CrossNormal.dot(body1.internalGetDeltaAngularVelocity());
+	const btScalar deltaVel2Dotn = c.m_contactNormal2.dot(body2.internalGetDeltaLinearVelocity()) + c.m_relpos2CrossNormal.dot(body2.internalGetDeltaAngularVelocity());
+
+	deltaImpulse -= deltaVel1Dotn*c.m_jacDiagABInv;
+	deltaImpulse -= deltaVel2Dotn*c.m_jacDiagABInv;
+	const btScalar sum = btScalar(c.m_appliedImpulse) + deltaImpulse;
+	if (sum < c.m_lowerLimit)
+	{
+		deltaImpulse = c.m_lowerLimit - c.m_appliedImpulse;
+		c.m_appliedImpulse = c.m_lowerLimit;
+	}
+	else
+	{
+		c.m_appliedImpulse = sum;
+	}
+	body1.internalApplyImpulse(c.m_contactNormal1*body1.internalGetInvMass(), c.m_angularComponentA, deltaImpulse);
+	body2.internalApplyImpulse(c.m_contactNormal2*body2.internalGetInvMass(), c.m_angularComponentB, deltaImpulse);
+
+	return deltaImpulse;
+}
+
+
 
 #ifdef USE_SIMD
 #include <emmintrin.h>
 
-//#include <intrin.h>//is it safe to include this header?
 
 #define btVecSplat(x, e) _mm_shuffle_ps(x, x, _MM_SHUFFLE(e,e,e,e))
 static inline __m128 btSimdDot3( __m128 vec0, __m128 vec1 )
@@ -51,12 +112,12 @@ static inline __m128 btSimdDot3( __m128 vec0, __m128 vec1 )
 	return _mm_add_ps( btVecSplat( result, 0 ), _mm_add_ps( btVecSplat( result, 1 ), btVecSplat( result, 2 ) ) );
 }
 
-#if defined (BT_USE_SSE4)
+#if defined (BT_ALLOW_SSE4)
 #include <intrin.h>
 
 #define USE_FMA					1
 #define USE_FMA3_INSTEAD_FMA4	1
-#define USE_SSE4_DOT			0
+#define USE_SSE4_DOT			1
 
 #define SSE4_DP(a, b)			_mm_dp_ps(a, b, 0x7f)
 #define SSE4_DP_FP(a, b)		_mm_cvtss_f32(_mm_dp_ps(a, b, 0x7f))
@@ -147,6 +208,7 @@ static btSimdScalar gResolveSingleConstraintRowGeneric_sse4_1_fma3(btSolverBody&
 }
 
 
+
 static btSimdScalar gResolveSingleConstraintRowLowerLimit_sse2(btSolverBody& body1, btSolverBody& body2, const btSolverConstraint& c)
 {
 	__m128 cpAppliedImp = _mm_set1_ps(c.m_appliedImpulse);
@@ -217,34 +279,7 @@ btSimdScalar btSequentialImpulseConstraintSolver::resolveSingleConstraintRowGene
 // Project Gauss Seidel or the equivalent Sequential Impulse
 btSimdScalar btSequentialImpulseConstraintSolver::resolveSingleConstraintRowGeneric(btSolverBody& body1,btSolverBody& body2,const btSolverConstraint& c)
 {
-	btScalar deltaImpulse = c.m_rhs-btScalar(c.m_appliedImpulse)*c.m_cfm;
+	return gResolveSingleConstraintRowGeneric_scalar_reference(body1, body2, c);
-	const btScalar deltaVel1Dotn	=	c.m_contactNormal1.dot(body1.internalGetDeltaLinearVelocity()) 	+ c.m_relpos1CrossNormal.dot(body1.internalGetDeltaAngularVelocity());
-	const btScalar deltaVel2Dotn	=	c.m_contactNormal2.dot(body2.internalGetDeltaLinearVelocity())  + c.m_relpos2CrossNormal.dot(body2.internalGetDeltaAngularVelocity());
-
-//	const btScalar delta_rel_vel	=	deltaVel1Dotn-deltaVel2Dotn;
-	deltaImpulse	-=	deltaVel1Dotn*c.m_jacDiagABInv;
-	deltaImpulse	-=	deltaVel2Dotn*c.m_jacDiagABInv;
-
-	const btScalar sum = btScalar(c.m_appliedImpulse) + deltaImpulse;
-	if (sum < c.m_lowerLimit)
-	{
-		deltaImpulse = c.m_lowerLimit-c.m_appliedImpulse;
-		c.m_appliedImpulse = c.m_lowerLimit;
-	}
-	else if (sum > c.m_upperLimit) 
-	{
-		deltaImpulse = c.m_upperLimit-c.m_appliedImpulse;
-		c.m_appliedImpulse = c.m_upperLimit;
-	}
-	else
-	{
-		c.m_appliedImpulse = sum;
-	}
-
-	body1.internalApplyImpulse(c.m_contactNormal1*body1.internalGetInvMass(),c.m_angularComponentA,deltaImpulse);
-	body2.internalApplyImpulse(c.m_contactNormal2*body2.internalGetInvMass(),c.m_angularComponentB,deltaImpulse);
-
-	return deltaImpulse;
 }
 
 btSimdScalar btSequentialImpulseConstraintSolver::resolveSingleConstraintRowLowerLimitSIMD(btSolverBody& body1,btSolverBody& body2,const btSolverConstraint& c)
@@ -259,26 +294,7 @@ btSimdScalar btSequentialImpulseConstraintSolver::resolveSingleConstraintRowLowe
 
 btSimdScalar btSequentialImpulseConstraintSolver::resolveSingleConstraintRowLowerLimit(btSolverBody& body1,btSolverBody& body2,const btSolverConstraint& c)
 {
-	btScalar deltaImpulse = c.m_rhs-btScalar(c.m_appliedImpulse)*c.m_cfm;
+	return gResolveSingleConstraintRowLowerLimit_scalar_reference(body1,body2,c);
-	const btScalar deltaVel1Dotn	=	c.m_contactNormal1.dot(body1.internalGetDeltaLinearVelocity()) + c.m_relpos1CrossNormal.dot(body1.internalGetDeltaAngularVelocity());
-	const btScalar deltaVel2Dotn	=	c.m_contactNormal2.dot(body2.internalGetDeltaLinearVelocity()) + c.m_relpos2CrossNormal.dot(body2.internalGetDeltaAngularVelocity());
-
-	deltaImpulse	-=	deltaVel1Dotn*c.m_jacDiagABInv;
-	deltaImpulse	-=	deltaVel2Dotn*c.m_jacDiagABInv;
-	const btScalar sum = btScalar(c.m_appliedImpulse) + deltaImpulse;
-	if (sum < c.m_lowerLimit)
-	{
-		deltaImpulse = c.m_lowerLimit-c.m_appliedImpulse;
-		c.m_appliedImpulse = c.m_lowerLimit;
-	}
-	else
-	{
-		c.m_appliedImpulse = sum;
-	}
-	body1.internalApplyImpulse(c.m_contactNormal1*body1.internalGetInvMass(),c.m_angularComponentA,deltaImpulse);
-	body2.internalApplyImpulse(c.m_contactNormal2*body2.internalGetInvMass(),c.m_angularComponentB,deltaImpulse);
-
-	return deltaImpulse;
 }
 
 
@@ -349,34 +365,61 @@ void	btSequentialImpulseConstraintSolver::resolveSplitPenetrationImpulseCacheFri
 
  btSequentialImpulseConstraintSolver::btSequentialImpulseConstraintSolver()
 	 :m_btSeed2(0),
-	 m_resolveSingleConstraintRowGeneric(gResolveSingleConstraintRowGeneric_sse2),
+	 m_resolveSingleConstraintRowGeneric(gResolveSingleConstraintRowGeneric_scalar_reference),
-	 m_resolveSingleConstraintRowLowerLimit(gResolveSingleConstraintRowLowerLimit_sse2)
+	 m_resolveSingleConstraintRowLowerLimit(gResolveSingleConstraintRowLowerLimit_scalar_reference)
  {
 
+#ifdef USE_SIMD
+	 m_resolveSingleConstraintRowGeneric = gResolveSingleConstraintRowGeneric_sse2;
+	 m_resolveSingleConstraintRowLowerLimit=gResolveSingleConstraintRowLowerLimit_sse2;
+#endif //USE_SIMD
+
+#ifdef BT_ALLOW_SSE4
+	 int cpuFeatures = btCpuFeatureUtility::getCpuFeatures();
+	 if ((cpuFeatures & btCpuFeatureUtility::CPU_FEATURE_FMA3) && (cpuFeatures & btCpuFeatureUtility::CPU_FEATURE_SSE4_1))
+	 {
+		m_resolveSingleConstraintRowGeneric = gResolveSingleConstraintRowGeneric_sse4_1_fma3;
+		m_resolveSingleConstraintRowLowerLimit = gResolveSingleConstraintRowLowerLimit_sse4_1_fma3;
+	 }
+#endif//BT_ALLOW_SSE4
+
  }
 
  btSequentialImpulseConstraintSolver::~btSequentialImpulseConstraintSolver()
  {
  }
 
+ btSingleConstraintRowSolver	btSequentialImpulseConstraintSolver::getScalarConstraintRowSolverGeneric()
+ {
+	 return gResolveSingleConstraintRowGeneric_scalar_reference;
+ }
 
+ btSingleConstraintRowSolver	btSequentialImpulseConstraintSolver::getScalarConstraintRowSolverLowerLimit()
+ {
+	 return gResolveSingleConstraintRowLowerLimit_scalar_reference;
+ }
+
+
+#ifdef USE_SIMD
  btSingleConstraintRowSolver	btSequentialImpulseConstraintSolver::getSSE2ConstraintRowSolverGeneric()
  {
 	 return gResolveSingleConstraintRowGeneric_sse2;
  }
- btSingleConstraintRowSolver	btSequentialImpulseConstraintSolver::getSSE4_1ConstraintRowSolverGeneric()
- {
-	 return gResolveSingleConstraintRowGeneric_sse4_1_fma3;
- }
  btSingleConstraintRowSolver	btSequentialImpulseConstraintSolver::getSSE2ConstraintRowSolverLowerLimit()
  {
 	 return gResolveSingleConstraintRowLowerLimit_sse2;
  }
+#ifdef BT_ALLOW_SSE4
+ btSingleConstraintRowSolver	btSequentialImpulseConstraintSolver::getSSE4_1ConstraintRowSolverGeneric()
+ {
+	 return gResolveSingleConstraintRowGeneric_sse4_1_fma3;
+ }
  btSingleConstraintRowSolver	btSequentialImpulseConstraintSolver::getSSE4_1ConstraintRowSolverLowerLimit()
  {
 	 return gResolveSingleConstraintRowLowerLimit_sse4_1_fma3;
  }
-
+#endif //BT_ALLOW_SSE4
+#endif //USE_SIMD
 
 unsigned long btSequentialImpulseConstraintSolver::btRand2()
 {

src/BulletDynamics/ConstraintSolver/btSequentialImpulseConstraintSolver.h

@@ -158,16 +158,15 @@ public:
 		m_resolveSingleConstraintRowLowerLimit = rowSolver;
 	}
 
-
+	///Various implementations of solving a single constraint row using a generic equality constraint, using scalar reference, SSE2 or SSE4
-	//	btSingleConstraintRowSolver	getScalarConstraintRowSolverLowerLimit();
+	btSingleConstraintRowSolver	getScalarConstraintRowSolverGeneric();
 	btSingleConstraintRowSolver	getSSE2ConstraintRowSolverGeneric();
 	btSingleConstraintRowSolver	getSSE4_1ConstraintRowSolverGeneric();
+
+	///Various implementations of solving a single constraint row using an inequality (lower limit) constraint, using scalar reference, SSE2 or SSE4
+	btSingleConstraintRowSolver	getScalarConstraintRowSolverLowerLimit();
 	btSingleConstraintRowSolver	getSSE2ConstraintRowSolverLowerLimit();
 	btSingleConstraintRowSolver	getSSE4_1ConstraintRowSolverLowerLimit();
-
-
-
-
 };