[PATCH AArch64 2/2] Replace temporary inline assembler for vget_high

	* config/aarch64/arm_neon.h (__GET_HIGH): New macro.
	(vget_high_f32, vget_high_f64, vget_high_p8, vget_high_p16,
	vget_high_s8, vget_high_s16, vget_high_s32, vget_high_s64,
	vget_high_u8, vget_high_u16, vget_high_u32, vget_high_u64):
	Remove temporary __asm__ and reimplement.

From-SVN: r214952

2 files changed, 87 insertions, 132 deletions

diff --git a/gcc/ChangeLog b/gcc/ChangeLog
index 853832b..cb7d78b 100644
--- a/gcc/ChangeLog
+++ b/gcc/ChangeLog
@@ -1,5 +1,13 @@
 2014-09-05  Alan Lawrence  <alan.lawrence@arm.com>
 
+	* config/aarch64/arm_neon.h (__GET_HIGH): New macro.
+	(vget_high_f32, vget_high_f64, vget_high_p8, vget_high_p16,
+	vget_high_s8, vget_high_s16, vget_high_s32, vget_high_s64,
+	vget_high_u8, vget_high_u16, vget_high_u32, vget_high_u64):
+	Remove temporary __asm__ and reimplement.
+
+2014-09-05  Alan Lawrence  <alan.lawrence@arm.com>
+
 	* config/aarch64/aarch64-builtins.c (aarch64_fold_builtin): Remove code
 	handling cmge, cmgt, cmeq, cmtst.
 
diff --git a/gcc/config/aarch64/arm_neon.h b/gcc/config/aarch64/arm_neon.h
index b33dc5c..4598deb 100644
--- a/gcc/config/aarch64/arm_neon.h
+++ b/gcc/config/aarch64/arm_neon.h
@@ -4328,6 +4328,85 @@ vget_low_u64 (uint64x2_t __a)
 
 #undef __GET_LOW
 
+#define __GET_HIGH(__TYPE)					\
+  uint64x2_t tmp = vreinterpretq_u64_##__TYPE (__a);		\
+  uint64x1_t hi = vcreate_u64 (vgetq_lane_u64 (tmp, 1));	\
+  return vreinterpret_##__TYPE##_u64 (hi);
+
+__extension__ static __inline float32x2_t __attribute__ ((__always_inline__))
+vget_high_f32 (float32x4_t __a)
+{
+  __GET_HIGH (f32);
+}
+
+__extension__ static __inline float64x1_t __attribute__ ((__always_inline__))
+vget_high_f64 (float64x2_t __a)
+{
+  __GET_HIGH (f64);
+}
+
+__extension__ static __inline poly8x8_t __attribute__ ((__always_inline__))
+vget_high_p8 (poly8x16_t __a)
+{
+  __GET_HIGH (p8);
+}
+
+__extension__ static __inline poly16x4_t __attribute__ ((__always_inline__))
+vget_high_p16 (poly16x8_t __a)
+{
+  __GET_HIGH (p16);
+}
+
+__extension__ static __inline int8x8_t __attribute__ ((__always_inline__))
+vget_high_s8 (int8x16_t __a)
+{
+  __GET_HIGH (s8);
+}
+
+__extension__ static __inline int16x4_t __attribute__ ((__always_inline__))
+vget_high_s16 (int16x8_t __a)
+{
+  __GET_HIGH (s16);
+}
+
+__extension__ static __inline int32x2_t __attribute__ ((__always_inline__))
+vget_high_s32 (int32x4_t __a)
+{
+  __GET_HIGH (s32);
+}
+
+__extension__ static __inline int64x1_t __attribute__ ((__always_inline__))
+vget_high_s64 (int64x2_t __a)
+{
+  __GET_HIGH (s64);
+}
+
+__extension__ static __inline uint8x8_t __attribute__ ((__always_inline__))
+vget_high_u8 (uint8x16_t __a)
+{
+  __GET_HIGH (u8);
+}
+
+__extension__ static __inline uint16x4_t __attribute__ ((__always_inline__))
+vget_high_u16 (uint16x8_t __a)
+{
+  __GET_HIGH (u16);
+}
+
+__extension__ static __inline uint32x2_t __attribute__ ((__always_inline__))
+vget_high_u32 (uint32x4_t __a)
+{
+  __GET_HIGH (u32);
+}
+
+#undef __GET_HIGH
+
+__extension__ static __inline uint64x1_t __attribute__ ((__always_inline__))
+vget_high_u64 (uint64x2_t __a)
+{
+  return vcreate_u64 (vgetq_lane_u64 (__a, 1));
+}
+
 __extension__ static __inline int8x16_t __attribute__ ((__always_inline__))
 vcombine_s8 (int8x8_t __a, int8x8_t __b)
 {
@@ -5770,138 +5849,6 @@ vfmsq_f64 (float64x2_t a, float64x2_t b, float64x2_t c)
   return result;
 }
 
-__extension__ static __inline float32x2_t __attribute__ ((__always_inline__))
-vget_high_f32 (float32x4_t a)
-{
-  float32x2_t result;
-  __asm__ ("ins %0.d[0], %1.d[1]"
-           : "=w"(result)
-           : "w"(a)
-           : /* No clobbers */);
-  return result;
-}
-
-__extension__ static __inline float64x1_t __attribute__ ((__always_inline__))
-vget_high_f64 (float64x2_t a)
-{
-  float64x1_t result;
-  __asm__ ("ins %0.d[0], %1.d[1]"
-           : "=w"(result)
-           : "w"(a)
-           : /* No clobbers */);
-  return result;
-}
-
-__extension__ static __inline poly8x8_t __attribute__ ((__always_inline__))
-vget_high_p8 (poly8x16_t a)
-{
-  poly8x8_t result;
-  __asm__ ("ins %0.d[0], %1.d[1]"
-           : "=w"(result)
-           : "w"(a)
-           : /* No clobbers */);
-  return result;
-}
-
-__extension__ static __inline poly16x4_t __attribute__ ((__always_inline__))
-vget_high_p16 (poly16x8_t a)
-{
-  poly16x4_t result;
-  __asm__ ("ins %0.d[0], %1.d[1]"
-           : "=w"(result)
-           : "w"(a)
-           : /* No clobbers */);
-  return result;
-}
-
-__extension__ static __inline int8x8_t __attribute__ ((__always_inline__))
-vget_high_s8 (int8x16_t a)
-{
-  int8x8_t result;
-  __asm__ ("ins %0.d[0], %1.d[1]"
-           : "=w"(result)
-           : "w"(a)
-           : /* No clobbers */);
-  return result;
-}
-
-__extension__ static __inline int16x4_t __attribute__ ((__always_inline__))
-vget_high_s16 (int16x8_t a)
-{
-  int16x4_t result;
-  __asm__ ("ins %0.d[0], %1.d[1]"
-           : "=w"(result)
-           : "w"(a)
-           : /* No clobbers */);
-  return result;
-}
-
-__extension__ static __inline int32x2_t __attribute__ ((__always_inline__))
-vget_high_s32 (int32x4_t a)
-{
-  int32x2_t result;
-  __asm__ ("ins %0.d[0], %1.d[1]"
-           : "=w"(result)
-           : "w"(a)
-           : /* No clobbers */);
-  return result;
-}
-
-__extension__ static __inline int64x1_t __attribute__ ((__always_inline__))
-vget_high_s64 (int64x2_t a)
-{
-  int64x1_t result;
-  __asm__ ("ins %0.d[0], %1.d[1]"
-           : "=w"(result)
-           : "w"(a)
-           : /* No clobbers */);
-  return result;
-}
-
-__extension__ static __inline uint8x8_t __attribute__ ((__always_inline__))
-vget_high_u8 (uint8x16_t a)
-{
-  uint8x8_t result;
-  __asm__ ("ins %0.d[0], %1.d[1]"
-           : "=w"(result)
-           : "w"(a)
-           : /* No clobbers */);
-  return result;
-}
-
-__extension__ static __inline uint16x4_t __attribute__ ((__always_inline__))
-vget_high_u16 (uint16x8_t a)
-{
-  uint16x4_t result;
-  __asm__ ("ins %0.d[0], %1.d[1]"
-           : "=w"(result)
-           : "w"(a)
-           : /* No clobbers */);
-  return result;
-}
-
-__extension__ static __inline uint32x2_t __attribute__ ((__always_inline__))
-vget_high_u32 (uint32x4_t a)
-{
-  uint32x2_t result;
-  __asm__ ("ins %0.d[0], %1.d[1]"
-           : "=w"(result)
-           : "w"(a)
-           : /* No clobbers */);
-  return result;
-}
-
-__extension__ static __inline uint64x1_t __attribute__ ((__always_inline__))
-vget_high_u64 (uint64x2_t a)
-{
-  uint64x1_t result;
-  __asm__ ("ins %0.d[0], %1.d[1]"
-           : "=w"(result)
-           : "w"(a)
-           : /* No clobbers */);
-  return result;
-}
-
 __extension__ static __inline int8x8_t __attribute__ ((__always_inline__))
 vhsub_s8 (int8x8_t a, int8x8_t b)
 {