aboutsummaryrefslogtreecommitdiffhomepage
path: root/Eigen/src/Core/arch
diff options
context:
space:
mode:
authorGravatar Benoit Steiner <benoit.steiner.goog@gmail.com>2014-02-24 13:45:32 -0800
committerGravatar Benoit Steiner <benoit.steiner.goog@gmail.com>2014-02-24 13:45:32 -0800
commitdb7d49efbb60b30f06f5fa97c767589f01e3df88 (patch)
tree6f90955592875b3a7c73e0e4d49d80231ec97281 /Eigen/src/Core/arch
parent7ed9441ea472031bb9357b5bec80151cae7ed2cb (diff)
Added support for FMA instructions
Diffstat (limited to 'Eigen/src/Core/arch')
-rw-r--r--Eigen/src/Core/arch/AVX/PacketMath.h5
-rw-r--r--Eigen/src/Core/arch/SSE/PacketMath.h4
2 files changed, 9 insertions, 0 deletions
diff --git a/Eigen/src/Core/arch/AVX/PacketMath.h b/Eigen/src/Core/arch/AVX/PacketMath.h
index 244e63e74..d1a134087 100644
--- a/Eigen/src/Core/arch/AVX/PacketMath.h
+++ b/Eigen/src/Core/arch/AVX/PacketMath.h
@@ -120,6 +120,11 @@ template<> EIGEN_STRONG_INLINE Packet8i pdiv<Packet8i>(const Packet8i& /*a*/, co
return pset1<Packet8i>(0);
}
+#ifdef EIGEN_VECTORIZE_FMA
+template<> EIGEN_STRONG_INLINE Packet8f pmadd(const Packet8f& a, const Packet8f& b, const Packet8f& c) { return _mm256_fmadd_ps(a,b,c); }
+template<> EIGEN_STRONG_INLINE Packet4d pmadd(const Packet4d& a, const Packet4d& b, const Packet4d& c) { return _mm256_fmadd_pd(a,b,c); }
+#endif
+
template<> EIGEN_STRONG_INLINE Packet8f pmin<Packet8f>(const Packet8f& a, const Packet8f& b) { return _mm256_min_ps(a,b); }
template<> EIGEN_STRONG_INLINE Packet4d pmin<Packet4d>(const Packet4d& a, const Packet4d& b) { return _mm256_min_pd(a,b); }
diff --git a/Eigen/src/Core/arch/SSE/PacketMath.h b/Eigen/src/Core/arch/SSE/PacketMath.h
index e913af650..a35f9ce98 100644
--- a/Eigen/src/Core/arch/SSE/PacketMath.h
+++ b/Eigen/src/Core/arch/SSE/PacketMath.h
@@ -179,6 +179,10 @@ template<> EIGEN_STRONG_INLINE Packet4i pdiv<Packet4i>(const Packet4i& /*a*/, co
// for some weird raisons, it has to be overloaded for packet of integers
template<> EIGEN_STRONG_INLINE Packet4i pmadd(const Packet4i& a, const Packet4i& b, const Packet4i& c) { return padd(pmul(a,b), c); }
+#ifdef EIGEN_VECTORIZE_FMA
+template<> EIGEN_STRONG_INLINE Packet4f pmadd(const Packet4f& a, const Packet4f& b, const Packet4f& c) { return _mm_fmadd_ps(a,b,c); }
+template<> EIGEN_STRONG_INLINE Packet2d pmadd(const Packet2d& a, const Packet2d& b, const Packet2d& c) { return _mm_fmadd_pd(a,b,c); }
+#endif
template<> EIGEN_STRONG_INLINE Packet4f pmin<Packet4f>(const Packet4f& a, const Packet4f& b) { return _mm_min_ps(a,b); }
template<> EIGEN_STRONG_INLINE Packet2d pmin<Packet2d>(const Packet2d& a, const Packet2d& b) { return _mm_min_pd(a,b); }