Fixed split impulse SIMD issue in the constraint solver (doesn't affect default behaviour, as split impulse is switched off by default)

Thanks to Laurent Coulon for the report/fix. http://bulletphysics.org/Bullet/phpBB3/viewtopic.php?f=9&t=7515&p=25766#p25766
This commit is contained in:
erwin.coumans
2011-10-12 20:22:55 +00:00
parent 961fb12985
commit 8b849e0d07

View File

@@ -225,7 +225,7 @@ void btSequentialImpulseConstraintSolver::resolveSplitPenetrationImpulseCacheFri
resultUpperLess = _mm_cmplt_ps(sum,upperLimit1);
__m128 lowMinApplied = _mm_sub_ps(lowerLimit1,cpAppliedImp);
deltaImpulse = _mm_or_ps( _mm_and_ps(resultLowerLess, lowMinApplied), _mm_andnot_ps(resultLowerLess, deltaImpulse) );
c.m_appliedImpulse = _mm_or_ps( _mm_and_ps(resultLowerLess, lowerLimit1), _mm_andnot_ps(resultLowerLess, sum) );
c.m_appliedPushImpulse = _mm_or_ps( _mm_and_ps(resultLowerLess, lowerLimit1), _mm_andnot_ps(resultLowerLess, sum) );
__m128 linearComponentA = _mm_mul_ps(c.m_contactNormal.mVec128,body1.internalGetInvMass().mVec128);
__m128 linearComponentB = _mm_mul_ps((c.m_contactNormal).mVec128,body2.internalGetInvMass().mVec128);
__m128 impulseMagnitude = deltaImpulse;