From cbbf88c4d7bf5c7c74658ae3294e3880106b83b0 Mon Sep 17 00:00:00 2001
From: Gael Guennebaud <g.gael@free.fr>
Date: Fri, 17 Feb 2017 14:39:02 +0100
Subject: Use int32_t instead of int in NEON code. Some platforms with 16 bytes
 int supports ARM NEON.

---
 Eigen/src/Core/arch/NEON/PacketMath.h | 52 +++++++++++++++++------------------
 1 file changed, 26 insertions(+), 26 deletions(-)

(limited to 'Eigen/src/Core/arch/NEON/PacketMath.h')

diff --git a/Eigen/src/Core/arch/NEON/PacketMath.h b/Eigen/src/Core/arch/NEON/PacketMath.h
index d392bf3ff..84a56bdcc 100644
--- a/Eigen/src/Core/arch/NEON/PacketMath.h
+++ b/Eigen/src/Core/arch/NEON/PacketMath.h
@@ -46,7 +46,7 @@ typedef uint32x4_t  Packet4ui;
   const Packet4f p4f_##NAME = pset1<Packet4f>(X)
 
 #define _EIGEN_DECLARE_CONST_Packet4f_FROM_INT(NAME,X) \
-  const Packet4f p4f_##NAME = vreinterpretq_f32_u32(pset1<int>(X))
+  const Packet4f p4f_##NAME = vreinterpretq_f32_u32(pset1<int32_t>(X))
 
 #define _EIGEN_DECLARE_CONST_Packet4i(NAME,X) \
   const Packet4i p4i_##NAME = pset1<Packet4i>(X)
@@ -83,7 +83,7 @@ template<> struct packet_traits<float>  : default_packet_traits
     HasSqrt = 0
   };
 };
-template<> struct packet_traits<int>    : default_packet_traits
+template<> struct packet_traits<int32_t>    : default_packet_traits
 {
   typedef Packet4i type;
   typedef Packet4i half; // Packet2i intrinsics not implemented yet
@@ -105,11 +105,11 @@ 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 int    type; enum {size=4, alignment=Aligned16}; typedef Packet4i half; };
+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<> EIGEN_STRONG_INLINE Packet4f pset1<Packet4f>(const float&  from) { return vdupq_n_f32(from); }
-template<> EIGEN_STRONG_INLINE Packet4i pset1<Packet4i>(const int&    from)   { return vdupq_n_s32(from); }
+template<> EIGEN_STRONG_INLINE Packet4i pset1<Packet4i>(const int32_t&    from)   { return vdupq_n_s32(from); }
 
 template<> EIGEN_STRONG_INLINE Packet4f plset<Packet4f>(const float& a)
 {
@@ -117,7 +117,7 @@ template<> EIGEN_STRONG_INLINE Packet4f plset<Packet4f>(const float& a)
   Packet4f countdown = vld1q_f32(f);
   return vaddq_f32(pset1<Packet4f>(a), countdown);
 }
-template<> EIGEN_STRONG_INLINE Packet4i plset<Packet4i>(const int& a)
+template<> EIGEN_STRONG_INLINE Packet4i plset<Packet4i>(const int32_t& a)
 {
   const int32_t i[] = {0, 1, 2, 3};
   Packet4i countdown = vld1q_s32(i);
@@ -240,20 +240,20 @@ 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<> 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 int*   from) { EIGEN_DEBUG_ALIGNED_LOAD return vld1q_s32(from); }
+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); }
 
-template<> EIGEN_STRONG_INLINE Packet4f ploadu<Packet4f>(const float* from) { EIGEN_DEBUG_UNALIGNED_LOAD return vld1q_f32(from); }
-template<> EIGEN_STRONG_INLINE Packet4i ploadu<Packet4i>(const int* from)   { EIGEN_DEBUG_UNALIGNED_LOAD return vld1q_s32(from); }
+template<> EIGEN_STRONG_INLINE Packet4f ploadu<Packet4f>(const float*   from) { EIGEN_DEBUG_UNALIGNED_LOAD return vld1q_f32(from); }
+template<> EIGEN_STRONG_INLINE Packet4i ploadu<Packet4i>(const int32_t* from) { EIGEN_DEBUG_UNALIGNED_LOAD return vld1q_s32(from); }
 
-template<> EIGEN_STRONG_INLINE Packet4f ploaddup<Packet4f>(const float*   from)
+template<> EIGEN_STRONG_INLINE Packet4f ploaddup<Packet4f>(const float* from)
 {
   float32x2_t lo, hi;
   lo = vld1_dup_f32(from);
   hi = vld1_dup_f32(from+1);
   return vcombine_f32(lo, hi);
 }
-template<> EIGEN_STRONG_INLINE Packet4i ploaddup<Packet4i>(const int*     from)
+template<> EIGEN_STRONG_INLINE Packet4i ploaddup<Packet4i>(const int32_t* from)
 {
   int32x2_t lo, hi;
   lo = vld1_dup_s32(from);
@@ -261,11 +261,11 @@ template<> EIGEN_STRONG_INLINE Packet4i ploaddup<Packet4i>(const int*     from)
   return vcombine_s32(lo, hi);
 }
 
-template<> EIGEN_STRONG_INLINE void pstore<float>(float*   to, const Packet4f& from) { EIGEN_DEBUG_ALIGNED_STORE vst1q_f32(to, from); }
-template<> EIGEN_STRONG_INLINE void pstore<int>(int*       to, const Packet4i& from) { EIGEN_DEBUG_ALIGNED_STORE vst1q_s32(to, from); }
+template<> EIGEN_STRONG_INLINE void pstore<float>  (float*    to, const Packet4f& from) { EIGEN_DEBUG_ALIGNED_STORE vst1q_f32(to, from); }
+template<> EIGEN_STRONG_INLINE void pstore<int32_t>(int32_t*  to, const Packet4i& from) { EIGEN_DEBUG_ALIGNED_STORE vst1q_s32(to, from); }
 
-template<> EIGEN_STRONG_INLINE void pstoreu<float>(float*  to, const Packet4f& from) { EIGEN_DEBUG_UNALIGNED_STORE vst1q_f32(to, from); }
-template<> EIGEN_STRONG_INLINE void pstoreu<int>(int*      to, const Packet4i& from) { EIGEN_DEBUG_UNALIGNED_STORE vst1q_s32(to, from); }
+template<> EIGEN_STRONG_INLINE void pstoreu<float>  (float*   to, const Packet4f& from) { EIGEN_DEBUG_UNALIGNED_STORE vst1q_f32(to, from); }
+template<> EIGEN_STRONG_INLINE void pstoreu<int32_t>(int32_t* to, const Packet4i& from) { EIGEN_DEBUG_UNALIGNED_STORE vst1q_s32(to, from); }
 
 template<> EIGEN_DEVICE_FUNC inline Packet4f pgather<float, Packet4f>(const float* from, Index stride)
 {
@@ -276,7 +276,7 @@ template<> EIGEN_DEVICE_FUNC inline Packet4f pgather<float, Packet4f>(const floa
   res = vsetq_lane_f32(from[3*stride], res, 3);
   return res;
 }
-template<> EIGEN_DEVICE_FUNC inline Packet4i pgather<int, Packet4i>(const int* from, Index stride)
+template<> EIGEN_DEVICE_FUNC inline Packet4i pgather<int32_t, Packet4i>(const int32_t* from, Index stride)
 {
   Packet4i res = pset1<Packet4i>(0);
   res = vsetq_lane_s32(from[0*stride], res, 0);
@@ -293,7 +293,7 @@ template<> EIGEN_DEVICE_FUNC inline void pscatter<float, Packet4f>(float* to, co
   to[stride*2] = vgetq_lane_f32(from, 2);
   to[stride*3] = vgetq_lane_f32(from, 3);
 }
-template<> EIGEN_DEVICE_FUNC inline void pscatter<int, Packet4i>(int* to, const Packet4i& from, Index stride)
+template<> EIGEN_DEVICE_FUNC inline void pscatter<int32_t, Packet4i>(int32_t* to, const Packet4i& from, Index stride)
 {
   to[stride*0] = vgetq_lane_s32(from, 0);
   to[stride*1] = vgetq_lane_s32(from, 1);
@@ -301,12 +301,12 @@ template<> EIGEN_DEVICE_FUNC inline void pscatter<int, Packet4i>(int* to, const
   to[stride*3] = vgetq_lane_s32(from, 3);
 }
 
-template<> EIGEN_STRONG_INLINE void prefetch<float>(const float* addr) { EIGEN_ARM_PREFETCH(addr); }
-template<> EIGEN_STRONG_INLINE void prefetch<int>(const int*     addr) { EIGEN_ARM_PREFETCH(addr); }
+template<> EIGEN_STRONG_INLINE void prefetch<float>  (const float*    addr) { EIGEN_ARM_PREFETCH(addr); }
+template<> EIGEN_STRONG_INLINE void prefetch<int32_t>(const int32_t*  addr) { EIGEN_ARM_PREFETCH(addr); }
 
 // FIXME only store the 2 first elements ?
-template<> EIGEN_STRONG_INLINE float  pfirst<Packet4f>(const Packet4f& a) { float EIGEN_ALIGN16 x[4]; vst1q_f32(x, a); return x[0]; }
-template<> EIGEN_STRONG_INLINE int    pfirst<Packet4i>(const Packet4i& a) { int   EIGEN_ALIGN16 x[4]; vst1q_s32(x, a); return x[0]; }
+template<> EIGEN_STRONG_INLINE float   pfirst<Packet4f>(const Packet4f& a) { float   EIGEN_ALIGN16 x[4]; vst1q_f32(x, a); return x[0]; }
+template<> EIGEN_STRONG_INLINE int32_t pfirst<Packet4i>(const Packet4i& a) { int32_t EIGEN_ALIGN16 x[4]; vst1q_s32(x, a); return x[0]; }
 
 template<> EIGEN_STRONG_INLINE Packet4f preverse(const Packet4f& a) {
   float32x2_t a_lo, a_hi;
@@ -361,7 +361,7 @@ template<> EIGEN_STRONG_INLINE Packet4f preduxp<Packet4f>(const Packet4f* vecs)
   return sum;
 }
 
-template<> EIGEN_STRONG_INLINE int predux<Packet4i>(const Packet4i& a)
+template<> EIGEN_STRONG_INLINE int32_t predux<Packet4i>(const Packet4i& a)
 {
   int32x2_t a_lo, a_hi, sum;
 
@@ -408,7 +408,7 @@ template<> EIGEN_STRONG_INLINE float predux_mul<Packet4f>(const Packet4f& a)
 
   return vget_lane_f32(prod, 0);
 }
-template<> EIGEN_STRONG_INLINE int predux_mul<Packet4i>(const Packet4i& a)
+template<> EIGEN_STRONG_INLINE int32_t predux_mul<Packet4i>(const Packet4i& a)
 {
   int32x2_t a_lo, a_hi, prod;
 
@@ -436,7 +436,7 @@ template<> EIGEN_STRONG_INLINE float predux_min<Packet4f>(const Packet4f& a)
   return vget_lane_f32(min, 0);
 }
 
-template<> EIGEN_STRONG_INLINE int predux_min<Packet4i>(const Packet4i& a)
+template<> EIGEN_STRONG_INLINE int32_t predux_min<Packet4i>(const Packet4i& a)
 {
   int32x2_t a_lo, a_hi, min;
 
@@ -461,7 +461,7 @@ template<> EIGEN_STRONG_INLINE float predux_max<Packet4f>(const Packet4f& a)
   return vget_lane_f32(max, 0);
 }
 
-template<> EIGEN_STRONG_INLINE int predux_max<Packet4i>(const Packet4i& a)
+template<> EIGEN_STRONG_INLINE int32_t predux_max<Packet4i>(const Packet4i& a)
 {
   int32x2_t a_lo, a_hi, max;
 
-- 
cgit v1.2.3