implement ploaddup for altivec and add respective unit test

2 files changed, 19 insertions, 0 deletions

diff --git a/Eigen/src/Core/arch/AltiVec/Complex.h b/Eigen/src/Core/arch/AltiVec/Complex.h
index 53fb1ba31..fd118ccff 100644
--- a/Eigen/src/Core/arch/AltiVec/Complex.h
+++ b/Eigen/src/Core/arch/AltiVec/Complex.h
@@ -112,6 +112,11 @@ template<> EIGEN_STRONG_INLINE Packet2cf pandnot<Packet2cf>(const Packet2cf& a,
 template<> EIGEN_STRONG_INLINE Packet2cf pload <Packet2cf>(const std::complex<float>* from) { EIGEN_DEBUG_ALIGNED_LOAD return Packet2cf(pload<Packet4f>((const float*)from)); }
 template<> EIGEN_STRONG_INLINE Packet2cf ploadu<Packet2cf>(const std::complex<float>* from) { EIGEN_DEBUG_UNALIGNED_LOAD return Packet2cf(ploadu<Packet4f>((const float*)from)); }
 
+template<> EIGEN_STRONG_INLINE Packet2cf ploaddup<Packet2cf>(const std::complex<float>*     from)
+{
+  return pset1<Packet2cf>(*from);
+}
+
 template<> EIGEN_STRONG_INLINE void pstore <std::complex<float> >(std::complex<float> *   to, const Packet2cf& from) { EIGEN_DEBUG_ALIGNED_STORE pstore((float*)to, from.v); }
 template<> EIGEN_STRONG_INLINE void pstoreu<std::complex<float> >(std::complex<float> *   to, const Packet2cf& from) { EIGEN_DEBUG_UNALIGNED_STORE pstoreu((float*)to, from.v); }
 
diff --git a/Eigen/src/Core/arch/AltiVec/PacketMath.h b/Eigen/src/Core/arch/AltiVec/PacketMath.h
index 96cb54283..efb4f62d1 100644
--- a/Eigen/src/Core/arch/AltiVec/PacketMath.h
+++ b/Eigen/src/Core/arch/AltiVec/PacketMath.h
@@ -73,6 +73,7 @@ static Packet4f p4f_COUNTDOWN = { 3.0, 2.0, 1.0, 0.0 };
 static Packet4i p4i_COUNTDOWN = { 3, 2, 1, 0 };
 static Packet16uc p16uc_REVERSE = {12,13,14,15, 8,9,10,11, 4,5,6,7, 0,1,2,3};
 static Packet16uc p16uc_FORWARD = vec_lvsl(0, (float*)0);
+static Packet16uc p16uc_DUPLICATE = {0,1,2,3, 0,1,2,3, 4,5,6,7, 4,5,6,7};
 
 static _EIGEN_DECLARE_CONST_FAST_Packet4f(ZERO, 0);
 static _EIGEN_DECLARE_CONST_FAST_Packet4i(ZERO, 0);
@@ -292,6 +293,19 @@ template<> EIGEN_STRONG_INLINE Packet4i ploadu<Packet4i>(const int* from)
   return (Packet4i) vec_perm(MSQ, LSQ, mask);    // align the data
 }
 
+template<> EIGEN_STRONG_INLINE Packet4f ploaddup<Packet4f>(const float*   from)
+{
+  Packet16uc tmp;
+  tmp = vec_ld(0, (unsigned char *)from);          // most significant quadword
+  return (Packet4f) vec_perm(tmp, tmp, p16uc_DUPLICATE);
+}
+template<> EIGEN_STRONG_INLINE Packet4i ploaddup<Packet4i>(const int*     from)
+{
+  Packet16uc tmp;
+  tmp = vec_ld(0, (unsigned char *)from);          // most significant quadword
+  return (Packet4i) vec_perm(tmp, tmp, p16uc_DUPLICATE);
+}
+
 template<> EIGEN_STRONG_INLINE void pstore<float>(float*   to, const Packet4f& from) { EIGEN_DEBUG_ALIGNED_STORE vec_st(from, 0, to); }
 template<> EIGEN_STRONG_INLINE void pstore<int>(int*       to, const Packet4i& from) { EIGEN_DEBUG_ALIGNED_STORE vec_st(from, 0, to); }