From 955c099eb58347c8d333cbd3ae274511f8050f54 Mon Sep 17 00:00:00 2001
From: Gael Guennebaud <g.gael@free.fr>
Date: Wed, 23 Feb 2011 18:20:55 +0300
Subject: [PATCH] implement ploaddup for altivec and add respective unit test

---
 Eigen/src/Core/arch/AltiVec/Complex.h    |  5 +++++
 Eigen/src/Core/arch/AltiVec/PacketMath.h | 14 ++++++++++++++
 test/packetmath.cpp                      | 12 ++++++++++--
 3 files changed, 29 insertions(+), 2 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); }
 
diff --git a/test/packetmath.cpp b/test/packetmath.cpp
index c89fd7314..d8a785519 100644
--- a/test/packetmath.cpp
+++ b/test/packetmath.cpp
@@ -126,8 +126,8 @@ template<typename Scalar> void packetmath()
   RealScalar refvalue = 0;
   for (int i=0; i<size; ++i)
   {
-    data1[i] = internal::random<Scalar>()/PacketSize;
-    data2[i] = internal::random<Scalar>()/PacketSize;
+    data1[i] = internal::random<Scalar>()/RealScalar(PacketSize);
+    data2[i] = internal::random<Scalar>()/RealScalar(PacketSize);
     refvalue = std::max(refvalue,internal::abs(data1[i]));
   }
 
@@ -179,6 +179,14 @@ template<typename Scalar> void packetmath()
   VERIFY(areApprox(ref, data2, PacketSize) && "internal::pset1");
 
   VERIFY(internal::isApprox(data1[0], internal::pfirst(internal::pload<Packet>(data1))) && "internal::pfirst");
+  
+  if(PacketSize>1)
+  {
+    for(int i=0;i<PacketSize/2;++i)
+      ref[2*i+0] = ref[2*i+1] = data1[i];
+    internal::pstore(data2,internal::ploaddup<Packet>(data1));
+    VERIFY(areApprox(ref, data2, PacketSize) && "ploaddup");
+  }
 
   ref[0] = 0;
   for (int i=0; i<PacketSize; ++i)