bug #1631: fix compilation with ARM NEON and clang, and cleanup the weird pshiftright_and_cast and pcast_and_shiftleft functions.

1 files changed, 16 insertions, 12 deletions

diff --git a/Eigen/src/Core/arch/NEON/PacketMath.h b/Eigen/src/Core/arch/NEON/PacketMath.h
index dc432f0d2..0a50153c7 100644
--- a/Eigen/src/Core/arch/NEON/PacketMath.h
+++ b/Eigen/src/Core/arch/NEON/PacketMath.h
@@ -140,8 +140,19 @@ EIGEN_STRONG_INLINE void        vst1q_f32(float* to, float32x4_t from) { ::vst1q
 EIGEN_STRONG_INLINE void        vst1_f32 (float* to, float32x2_t from) { ::vst1_f32 ((float32_t*)to,from); }
 #endif
 
-template<> struct unpacket_traits<Packet4f> { typedef float   type; enum {size=4, alignment=Aligned16}; typedef Packet4f half; };
-template<> struct unpacket_traits<Packet4i> { typedef int32_t type; enum {size=4, alignment=Aligned16}; typedef Packet4i half; };
+template<> struct unpacket_traits<Packet4f>
+{
+  typedef float     type;
+  typedef Packet4f  half;
+  typedef Packet4i  integer_packet;
+  enum {size=4, alignment=Aligned16};
+};
+template<> struct unpacket_traits<Packet4i>
+{
+  typedef int32_t   type;
+  typedef Packet4i  half;
+  enum {size=4, alignment=Aligned16};
+};
 
 template<> EIGEN_STRONG_INLINE Packet4f pset1<Packet4f>(const float&  from) { return vdupq_n_f32(from); }
 template<> EIGEN_STRONG_INLINE Packet4i pset1<Packet4i>(const int32_t&    from)   { return vdupq_n_s32(from); }
@@ -294,6 +305,9 @@ template<> EIGEN_STRONG_INLINE Packet4f pandnot<Packet4f>(const Packet4f& a, con
 }
 template<> EIGEN_STRONG_INLINE Packet4i pandnot<Packet4i>(const Packet4i& a, const Packet4i& b) { return vbicq_s32(a,b); }
 
+template<int N> EIGEN_STRONG_INLINE Packet4i pshiftright(Packet4i a) { return vshrq_n_s32(a,N); }
+template<int N> EIGEN_STRONG_INLINE Packet4i pshiftleft(Packet4i a) { return vshlq_n_s32(a,N); }
+
 template<> EIGEN_STRONG_INLINE Packet4f pload<Packet4f>(const float*    from) { EIGEN_DEBUG_ALIGNED_LOAD return vld1q_f32(from); }
 template<> EIGEN_STRONG_INLINE Packet4i pload<Packet4i>(const int32_t*  from) { EIGEN_DEBUG_ALIGNED_LOAD return vld1q_s32(from); }
 
@@ -384,20 +398,10 @@ template<> EIGEN_STRONG_INLINE Packet4i preverse(const Packet4i& a) {
 template<> EIGEN_STRONG_INLINE Packet4f pabs(const Packet4f& a) { return vabsq_f32(a); }
 template<> EIGEN_STRONG_INLINE Packet4i pabs(const Packet4i& a) { return vabsq_s32(a); }
 
-template<> EIGEN_STRONG_INLINE Packet4f pshiftright_and_cast(Packet4f a, int n) {
-  return vcvtq_f32_s32(vshrq_n_s32(vreinterpretq_s32_f32(a),n));
-}
-
 template<> EIGEN_STRONG_INLINE Packet4f pfrexp<Packet4f>(const Packet4f& a, Packet4f& exponent) {
   return pfrexp_float(a,exponent);
 }
 
-template<> EIGEN_STRONG_INLINE Packet4f pcast_and_shiftleft<Packet4f>(Packet4f v, int n)
-{
-  Packet4i vi = vcvtq_s32_f32(v);
-  return vreinterpretq_f32_s32(vshlq_n_s32(vi, n));
-}
-
 template<> EIGEN_STRONG_INLINE Packet4f pldexp<Packet4f>(const Packet4f& a, const Packet4f& exponent) {
   return pldexp_float(a,exponent);
 }