[aarch64]: add intrinsics for vld1(q)_x4 and vst1(q)_x4
authorSylvia Taylor <sylvia.taylor@arm.com>
Thu, 22 Aug 2019 11:28:26 +0000 (11:28 +0000)
committerKyrylo Tkachov <ktkachov@gcc.gnu.org>
Thu, 22 Aug 2019 11:28:26 +0000 (11:28 +0000)
This patch adds the intrinsic functions for:
- vld1_<mode>_x4
- vst1_<mode>_x4
- vld1q_<mode>_x4
- vst1q_<mode>_x4

Bootstrapped and tested on aarch64-none-linux-gnu.

Committed on behalf of Sylvia Taylor.

2019-08-22  Sylvia Taylor  <sylvia.taylor@arm.com>

* config/aarch64/aarch64-simd-builtins.def:
(ld1x4): New.
(st1x4): Likewise.
* config/aarch64/aarch64-simd.md:
(aarch64_ld1x4<VALLDIF:mode>): New pattern.
(aarch64_st1x4<VALLDIF:mode>): Likewise.
(aarch64_ld1_x4_<mode>): Likewise.
(aarch64_st1_x4_<mode>): Likewise.
* config/aarch64/arm_neon.h:
(vld1_s8_x4): New function.
(vld1q_s8_x4): Likewise.
(vld1_s16_x4): Likewise.
(vld1q_s16_x4): Likewise.
(vld1_s32_x4): Likewise.
(vld1q_s32_x4): Likewise.
(vld1_u8_x4): Likewise.
(vld1q_u8_x4): Likewise.
(vld1_u16_x4): Likewise.
(vld1q_u16_x4): Likewise.
(vld1_u32_x4): Likewise.
(vld1q_u32_x4): Likewise.
(vld1_f16_x4): Likewise.
(vld1q_f16_x4): Likewise.
(vld1_f32_x4): Likewise.
(vld1q_f32_x4): Likewise.
(vld1_p8_x4): Likewise.
(vld1q_p8_x4): Likewise.
(vld1_p16_x4): Likewise.
(vld1q_p16_x4): Likewise.
(vld1_s64_x4): Likewise.
(vld1_u64_x4): Likewise.
(vld1_p64_x4): Likewise.
(vld1q_s64_x4): Likewise.
(vld1q_u64_x4): Likewise.
(vld1q_p64_x4): Likewise.
(vld1_f64_x4): Likewise.
(vld1q_f64_x4): Likewise.
(vst1_s8_x4): Likewise.
(vst1q_s8_x4): Likewise.
(vst1_s16_x4): Likewise.
(vst1q_s16_x4): Likewise.
(vst1_s32_x4): Likewise.
(vst1q_s32_x4): Likewise.
(vst1_u8_x4): Likewise.
(vst1q_u8_x4): Likewise.
(vst1_u16_x4): Likewise.
(vst1q_u16_x4): Likewise.
(vst1_u32_x4): Likewise.
(vst1q_u32_x4): Likewise.
(vst1_f16_x4): Likewise.
(vst1q_f16_x4): Likewise.
(vst1_f32_x4): Likewise.
(vst1q_f32_x4): Likewise.
(vst1_p8_x4): Likewise.
(vst1q_p8_x4): Likewise.
(vst1_p16_x4): Likewise.
(vst1q_p16_x4): Likewise.
(vst1_s64_x4): Likewise.
(vst1_u64_x4): Likewise.
(vst1_p64_x4): Likewise.
(vst1q_s64_x4): Likewise.
(vst1q_u64_x4): Likewise.
(vst1q_p64_x4): Likewise.
(vst1_f64_x4): Likewise.
(vst1q_f64_x4): Likewise.

* gcc.target/aarch64/advsimd-intrinsics/vld1x4.c: New test.
* gcc.target/aarch64/advsimd-intrinsics/vst1x4.c: New test.

From-SVN: r274820

gcc/ChangeLog
gcc/config/aarch64/aarch64-simd-builtins.def
gcc/config/aarch64/aarch64-simd.md
gcc/config/aarch64/arm_neon.h
gcc/testsuite/ChangeLog
gcc/testsuite/gcc.target/aarch64/advsimd-intrinsics/vld1x4.c [new file with mode: 0644]
gcc/testsuite/gcc.target/aarch64/advsimd-intrinsics/vst1x4.c [new file with mode: 0644]

index 603687e..271786f 100644 (file)
@@ -1,3 +1,71 @@
+2019-08-22  Sylvia Taylor  <sylvia.taylor@arm.com>
+
+       * config/aarch64/aarch64-simd-builtins.def:
+       (ld1x4): New.
+       (st1x4): Likewise.
+       * config/aarch64/aarch64-simd.md:
+       (aarch64_ld1x4<VALLDIF:mode>): New pattern.
+       (aarch64_st1x4<VALLDIF:mode>): Likewise.
+       (aarch64_ld1_x4_<mode>): Likewise.
+       (aarch64_st1_x4_<mode>): Likewise.
+       * config/aarch64/arm_neon.h:
+       (vld1_s8_x4): New function.
+       (vld1q_s8_x4): Likewise.
+       (vld1_s16_x4): Likewise.
+       (vld1q_s16_x4): Likewise.
+       (vld1_s32_x4): Likewise.
+       (vld1q_s32_x4): Likewise.
+       (vld1_u8_x4): Likewise.
+       (vld1q_u8_x4): Likewise.
+       (vld1_u16_x4): Likewise.
+       (vld1q_u16_x4): Likewise.
+       (vld1_u32_x4): Likewise.
+       (vld1q_u32_x4): Likewise.
+       (vld1_f16_x4): Likewise.
+       (vld1q_f16_x4): Likewise.
+       (vld1_f32_x4): Likewise.
+       (vld1q_f32_x4): Likewise.
+       (vld1_p8_x4): Likewise.
+       (vld1q_p8_x4): Likewise.
+       (vld1_p16_x4): Likewise.
+       (vld1q_p16_x4): Likewise.
+       (vld1_s64_x4): Likewise.
+       (vld1_u64_x4): Likewise.
+       (vld1_p64_x4): Likewise.
+       (vld1q_s64_x4): Likewise.
+       (vld1q_u64_x4): Likewise.
+       (vld1q_p64_x4): Likewise.
+       (vld1_f64_x4): Likewise.
+       (vld1q_f64_x4): Likewise.
+       (vst1_s8_x4): Likewise.
+       (vst1q_s8_x4): Likewise.
+       (vst1_s16_x4): Likewise.
+       (vst1q_s16_x4): Likewise.
+       (vst1_s32_x4): Likewise.
+       (vst1q_s32_x4): Likewise.
+       (vst1_u8_x4): Likewise.
+       (vst1q_u8_x4): Likewise.
+       (vst1_u16_x4): Likewise.
+       (vst1q_u16_x4): Likewise.
+       (vst1_u32_x4): Likewise.
+       (vst1q_u32_x4): Likewise.
+       (vst1_f16_x4): Likewise.
+       (vst1q_f16_x4): Likewise.
+       (vst1_f32_x4): Likewise.
+       (vst1q_f32_x4): Likewise.
+       (vst1_p8_x4): Likewise.
+       (vst1q_p8_x4): Likewise.
+       (vst1_p16_x4): Likewise.
+       (vst1q_p16_x4): Likewise.
+       (vst1_s64_x4): Likewise.
+       (vst1_u64_x4): Likewise.
+       (vst1_p64_x4): Likewise.
+       (vst1q_s64_x4): Likewise.
+       (vst1q_u64_x4): Likewise.
+       (vst1q_p64_x4): Likewise.
+       (vst1_f64_x4): Likewise.
+       (vst1q_f64_x4): Likewise.
+
 2019-08-22  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
 
        * config/aarch64/aarch64-sve.md (vcond_mask): Add "@".
index 01518fe..779111a 100644 (file)
   /* Implemented by aarch64_ld1x3<VALLDIF:mode>.  */
   BUILTIN_VALLDIF (LOADSTRUCT, ld1x3, 0)
 
+  /* Implemented by aarch64_ld1x4<VALLDIF:mode>.  */
+  BUILTIN_VALLDIF (LOADSTRUCT, ld1x4, 0)
+
   /* Implemented by aarch64_st1x2<VALLDIF:mode>.  */
   BUILTIN_VALLDIF (STORESTRUCT, st1x2, 0)
 
   /* Implemented by aarch64_st1x3<VALLDIF:mode>.  */
   BUILTIN_VALLDIF (STORESTRUCT, st1x3, 0)
 
+  /* Implemented by aarch64_st1x4<VALLDIF:mode>.  */
+  BUILTIN_VALLDIF (STORESTRUCT, st1x4, 0)
+
   /* Implemented by fma<mode>4.  */
   BUILTIN_VHSDF (TERNOP, fma, 4)
   VAR1 (TERNOP, fma, 4, hf)
index e33a009..6f7fb1c 100644 (file)
   [(set_attr "type" "neon_load1_3reg<q>")]
 )
 
+(define_expand "aarch64_ld1x4<VALLDIF:mode>"
+  [(match_operand:XI 0 "register_operand" "=w")
+   (match_operand:DI 1 "register_operand" "r")
+   (unspec:VALLDIF [(const_int 0)] UNSPEC_VSTRUCTDUMMY)]
+  "TARGET_SIMD"
+{
+  rtx mem = gen_rtx_MEM (XImode, operands[1]);
+  emit_insn (gen_aarch64_ld1_x4_<VALLDIF:mode> (operands[0], mem));
+  DONE;
+})
+
+(define_insn "aarch64_ld1_x4_<mode>"
+  [(set (match_operand:XI 0 "register_operand" "=w")
+       (unspec:XI
+         [(match_operand:XI 1 "aarch64_simd_struct_operand" "Utv")
+          (unspec:VALLDIF [(const_int 4)] UNSPEC_VSTRUCTDUMMY)]
+       UNSPEC_LD1))]
+  "TARGET_SIMD"
+  "ld1\\t{%S0.<Vtype> - %V0.<Vtype>}, %1"
+  [(set_attr "type" "neon_load1_4reg<q>")]
+)
+
 (define_expand "aarch64_st1x2<VALLDIF:mode>"
   [(match_operand:DI 0 "register_operand")
    (match_operand:OI 1 "register_operand")
   [(set_attr "type" "neon_store1_3reg<q>")]
 )
 
+(define_expand "aarch64_st1x4<VALLDIF:mode>"
+  [(match_operand:DI 0 "register_operand" "")
+   (match_operand:XI 1 "register_operand" "")
+   (unspec:VALLDIF [(const_int 0)] UNSPEC_VSTRUCTDUMMY)]
+  "TARGET_SIMD"
+{
+  rtx mem = gen_rtx_MEM (XImode, operands[0]);
+  emit_insn (gen_aarch64_st1_x4_<VALLDIF:mode> (mem, operands[1]));
+  DONE;
+})
+
+(define_insn "aarch64_st1_x4_<mode>"
+  [(set (match_operand:XI 0 "aarch64_simd_struct_operand" "=Utv")
+       (unspec:XI
+          [(match_operand:XI 1 "register_operand" "w")
+          (unspec:VALLDIF [(const_int 4)] UNSPEC_VSTRUCTDUMMY)]
+       UNSPEC_ST1))]
+  "TARGET_SIMD"
+  "st1\\t{%S1.<Vtype> - %V1.<Vtype>}, %0"
+  [(set_attr "type" "neon_store1_4reg<q>")]
+)
+
 (define_insn "*aarch64_mov<mode>"
   [(set (match_operand:VSTRUCT 0 "aarch64_simd_nonimmediate_operand" "=w,Utv,w")
        (match_operand:VSTRUCT 1 "aarch64_simd_general_operand" " w,w,Utv"))]
index 314ef30..9ecc00c 100644 (file)
@@ -17968,6 +17968,288 @@ vld1q_u64 (const uint64_t *a)
     __builtin_aarch64_ld1v2di ((const __builtin_aarch64_simd_di *) a);
 }
 
+/* vld1(q)_x4.  */
+
+__extension__ extern __inline int8x8x4_t
+__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
+vld1_s8_x4 (const int8_t *__a)
+{
+  union { int8x8x4_t __i; __builtin_aarch64_simd_xi __o; } __au;
+  __au.__o
+    = __builtin_aarch64_ld1x4v8qi ((const __builtin_aarch64_simd_qi *) __a);
+  return __au.__i;
+}
+
+__extension__ extern __inline int8x16x4_t
+__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
+vld1q_s8_x4 (const int8_t *__a)
+{
+  union { int8x16x4_t __i; __builtin_aarch64_simd_xi __o; } __au;
+  __au.__o
+    = __builtin_aarch64_ld1x4v16qi ((const __builtin_aarch64_simd_qi *) __a);
+  return __au.__i;
+}
+
+__extension__ extern __inline int16x4x4_t
+__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
+vld1_s16_x4 (const int16_t *__a)
+{
+  union { int16x4x4_t __i; __builtin_aarch64_simd_xi __o; } __au;
+  __au.__o
+    = __builtin_aarch64_ld1x4v4hi ((const __builtin_aarch64_simd_hi *) __a);
+  return __au.__i;
+}
+
+__extension__ extern __inline int16x8x4_t
+__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
+vld1q_s16_x4 (const int16_t *__a)
+{
+  union { int16x8x4_t __i; __builtin_aarch64_simd_xi __o; } __au;
+  __au.__o
+    = __builtin_aarch64_ld1x4v8hi ((const __builtin_aarch64_simd_hi *) __a);
+  return __au.__i;
+}
+
+__extension__ extern __inline int32x2x4_t
+__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
+vld1_s32_x4 (const int32_t *__a)
+{
+  union { int32x2x4_t __i; __builtin_aarch64_simd_xi __o; } __au;
+  __au.__o
+  = __builtin_aarch64_ld1x4v2si ((const __builtin_aarch64_simd_si *) __a);
+  return __au.__i;
+}
+
+__extension__ extern __inline int32x4x4_t
+__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
+vld1q_s32_x4 (const int32_t *__a)
+{
+  union { int32x4x4_t __i; __builtin_aarch64_simd_xi __o; } __au;
+  __au.__o
+  = __builtin_aarch64_ld1x4v4si ((const __builtin_aarch64_simd_si *) __a);
+  return __au.__i;
+}
+
+__extension__ extern __inline uint8x8x4_t
+__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
+vld1_u8_x4 (const uint8_t *__a)
+{
+  union { uint8x8x4_t __i; __builtin_aarch64_simd_xi __o; } __au;
+  __au.__o
+    = __builtin_aarch64_ld1x4v8qi ((const __builtin_aarch64_simd_qi *) __a);
+  return __au.__i;
+}
+
+__extension__ extern __inline uint8x16x4_t
+__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
+vld1q_u8_x4 (const uint8_t *__a)
+{
+  union { uint8x16x4_t __i; __builtin_aarch64_simd_xi __o; } __au;
+  __au.__o
+    = __builtin_aarch64_ld1x4v16qi ((const __builtin_aarch64_simd_qi *) __a);
+  return __au.__i;
+}
+
+__extension__ extern __inline uint16x4x4_t
+__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
+vld1_u16_x4 (const uint16_t *__a)
+{
+  union { uint16x4x4_t __i; __builtin_aarch64_simd_xi __o; } __au;
+  __au.__o
+    = __builtin_aarch64_ld1x4v4hi ((const __builtin_aarch64_simd_hi *) __a);
+  return __au.__i;
+}
+
+__extension__ extern __inline uint16x8x4_t
+__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
+vld1q_u16_x4 (const uint16_t *__a)
+{
+  union { uint16x8x4_t __i; __builtin_aarch64_simd_xi __o; } __au;
+  __au.__o
+    = __builtin_aarch64_ld1x4v8hi ((const __builtin_aarch64_simd_hi *) __a);
+  return __au.__i;
+}
+
+__extension__ extern __inline uint32x2x4_t
+__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
+vld1_u32_x4 (const uint32_t *__a)
+{
+  union { uint32x2x4_t __i; __builtin_aarch64_simd_xi __o; } __au;
+  __au.__o
+    = __builtin_aarch64_ld1x4v2si ((const __builtin_aarch64_simd_si *) __a);
+  return __au.__i;
+}
+
+__extension__ extern __inline uint32x4x4_t
+__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
+vld1q_u32_x4 (const uint32_t *__a)
+{
+  union { uint32x4x4_t __i; __builtin_aarch64_simd_xi __o; } __au;
+  __au.__o
+    = __builtin_aarch64_ld1x4v4si ((const __builtin_aarch64_simd_si *) __a);
+  return __au.__i;
+}
+
+__extension__ extern __inline float16x4x4_t
+__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
+vld1_f16_x4 (const float16_t *__a)
+{
+  union { float16x4x4_t __i; __builtin_aarch64_simd_xi __o; } __au;
+  __au.__o
+    = __builtin_aarch64_ld1x4v4hf ((const __builtin_aarch64_simd_hf *) __a);
+  return __au.__i;
+}
+
+__extension__ extern __inline float16x8x4_t
+__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
+vld1q_f16_x4 (const float16_t *__a)
+{
+  union { float16x8x4_t __i; __builtin_aarch64_simd_xi __o; } __au;
+  __au.__o
+    = __builtin_aarch64_ld1x4v8hf ((const __builtin_aarch64_simd_hf *) __a);
+  return __au.__i;
+}
+
+__extension__ extern __inline float32x2x4_t
+__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
+vld1_f32_x4 (const float32_t *__a)
+{
+  union { float32x2x4_t __i; __builtin_aarch64_simd_xi __o; } __au;
+  __au.__o
+    = __builtin_aarch64_ld1x4v2sf ((const __builtin_aarch64_simd_sf *) __a);
+  return __au.__i;
+}
+
+__extension__ extern __inline float32x4x4_t
+__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
+vld1q_f32_x4 (const float32_t *__a)
+{
+  union { float32x4x4_t __i; __builtin_aarch64_simd_xi __o; } __au;
+  __au.__o
+    = __builtin_aarch64_ld1x4v4sf ((const __builtin_aarch64_simd_sf *) __a);
+  return __au.__i;
+}
+
+__extension__ extern __inline poly8x8x4_t
+__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
+vld1_p8_x4 (const poly8_t *__a)
+{
+  union { poly8x8x4_t __i; __builtin_aarch64_simd_xi __o; } __au;
+  __au.__o
+    = __builtin_aarch64_ld1x4v8qi ((const __builtin_aarch64_simd_qi *) __a);
+  return __au.__i;
+}
+
+__extension__ extern __inline poly8x16x4_t
+__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
+vld1q_p8_x4 (const poly8_t *__a)
+{
+  union { poly8x16x4_t __i; __builtin_aarch64_simd_xi __o; } __au;
+  __au.__o
+    = __builtin_aarch64_ld1x4v16qi ((const __builtin_aarch64_simd_qi *) __a);
+  return __au.__i;
+}
+
+__extension__ extern __inline poly16x4x4_t
+__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
+vld1_p16_x4 (const poly16_t *__a)
+{
+  union { poly16x4x4_t __i; __builtin_aarch64_simd_xi __o; } __au;
+  __au.__o
+    = __builtin_aarch64_ld1x4v4hi ((const __builtin_aarch64_simd_hi *) __a);
+  return __au.__i;
+}
+
+__extension__ extern __inline poly16x8x4_t
+__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
+vld1q_p16_x4 (const poly16_t *__a)
+{
+  union { poly16x8x4_t __i; __builtin_aarch64_simd_xi __o; } __au;
+  __au.__o
+    = __builtin_aarch64_ld1x4v8hi ((const __builtin_aarch64_simd_hi *) __a);
+  return __au.__i;
+}
+
+__extension__ extern __inline int64x1x4_t
+__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
+vld1_s64_x4 (const int64_t *__a)
+{
+  union { int64x1x4_t __i; __builtin_aarch64_simd_xi __o; } __au;
+  __au.__o
+    = __builtin_aarch64_ld1x4di ((const __builtin_aarch64_simd_di *) __a);
+  return __au.__i;
+}
+
+__extension__ extern __inline uint64x1x4_t
+__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
+vld1_u64_x4 (const uint64_t *__a)
+{
+  union { uint64x1x4_t __i; __builtin_aarch64_simd_xi __o; } __au;
+  __au.__o
+    = __builtin_aarch64_ld1x4di ((const __builtin_aarch64_simd_di *) __a);
+  return __au.__i;
+}
+
+__extension__ extern __inline poly64x1x4_t
+__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
+vld1_p64_x4 (const poly64_t *__a)
+{
+  union { poly64x1x4_t __i; __builtin_aarch64_simd_xi __o; } __au;
+  __au.__o
+    = __builtin_aarch64_ld1x4di ((const __builtin_aarch64_simd_di *) __a);
+  return __au.__i;
+}
+
+__extension__ extern __inline int64x2x4_t
+__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
+vld1q_s64_x4 (const int64_t *__a)
+{
+  union { int64x2x4_t __i; __builtin_aarch64_simd_xi __o; } __au;
+  __au.__o
+    = __builtin_aarch64_ld1x4v2di ((const __builtin_aarch64_simd_di *) __a);
+  return __au.__i;
+}
+
+__extension__ extern __inline uint64x2x4_t
+__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
+vld1q_u64_x4 (const uint64_t *__a)
+{
+  union { uint64x2x4_t __i; __builtin_aarch64_simd_xi __o; } __au;
+  __au.__o
+    = __builtin_aarch64_ld1x4v2di ((const __builtin_aarch64_simd_di *) __a);
+  return __au.__i;
+}
+
+__extension__ extern __inline poly64x2x4_t
+__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
+vld1q_p64_x4 (const poly64_t *__a)
+{
+  union { poly64x2x4_t __i; __builtin_aarch64_simd_xi __o; } __au;
+  __au.__o
+    = __builtin_aarch64_ld1x4v2di ((const __builtin_aarch64_simd_di *) __a);
+  return __au.__i;
+}
+
+__extension__ extern __inline float64x1x4_t
+__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
+vld1_f64_x4 (const float64_t *__a)
+{
+  union { float64x1x4_t __i; __builtin_aarch64_simd_xi __o; } __au;
+  __au.__o
+    = __builtin_aarch64_ld1x4df ((const __builtin_aarch64_simd_df *) __a);
+  return __au.__i;
+}
+
+__extension__ extern __inline float64x2x4_t
+__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
+vld1q_f64_x4 (const float64_t *__a)
+{
+  union { float64x2x4_t __i; __builtin_aarch64_simd_xi __o; } __au;
+  __au.__o
+    = __builtin_aarch64_ld1x4v2df ((const __builtin_aarch64_simd_df *) __a);
+  return __au.__i;
+}
+
 /* vld1_dup  */
 
 __extension__ extern __inline float16x4_t
@@ -28596,6 +28878,232 @@ vst1q_p64_x3 (poly64_t * __a, poly64x2x3_t val)
   __builtin_aarch64_st1x3v2di ((__builtin_aarch64_simd_di *) __a, __o);
 }
 
+/* vst1(q)_x4.  */
+
+__extension__ extern __inline void
+__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
+vst1_s8_x4 (int8_t * __a, int8x8x4_t val)
+{
+  union { int8x8x4_t __i; __builtin_aarch64_simd_xi __o; } __u = { val };
+  __builtin_aarch64_st1x4v8qi ((__builtin_aarch64_simd_qi *) __a, __u.__o);
+}
+
+__extension__ extern __inline void
+__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
+vst1q_s8_x4 (int8_t * __a, int8x16x4_t val)
+{
+  union { int8x16x4_t __i; __builtin_aarch64_simd_xi __o; } __u = { val };
+  __builtin_aarch64_st1x4v16qi ((__builtin_aarch64_simd_qi *) __a, __u.__o);
+}
+
+__extension__ extern __inline void
+__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
+vst1_s16_x4 (int16_t * __a, int16x4x4_t val)
+{
+  union { int16x4x4_t __i; __builtin_aarch64_simd_xi __o; } __u = { val };
+  __builtin_aarch64_st1x4v4hi ((__builtin_aarch64_simd_hi *) __a, __u.__o);
+}
+
+__extension__ extern __inline void
+__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
+vst1q_s16_x4 (int16_t * __a, int16x8x4_t val)
+{
+  union { int16x8x4_t __i; __builtin_aarch64_simd_xi __o; } __u = { val };
+  __builtin_aarch64_st1x4v8hi ((__builtin_aarch64_simd_hi *) __a, __u.__o);
+}
+
+__extension__ extern __inline void
+__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
+vst1_s32_x4 (int32_t * __a, int32x2x4_t val)
+{
+  union { int32x2x4_t __i; __builtin_aarch64_simd_xi __o; } __u = { val };
+  __builtin_aarch64_st1x4v2si ((__builtin_aarch64_simd_si *) __a, __u.__o);
+}
+
+__extension__ extern __inline void
+__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
+vst1q_s32_x4 (int32_t * __a, int32x4x4_t val)
+{
+  union { int32x4x4_t __i; __builtin_aarch64_simd_xi __o; } __u = { val };
+  __builtin_aarch64_st1x4v4si ((__builtin_aarch64_simd_si *) __a, __u.__o);
+}
+
+__extension__ extern __inline void
+__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
+vst1_u8_x4 (uint8_t * __a, uint8x8x4_t val)
+{
+  union { uint8x8x4_t __i; __builtin_aarch64_simd_xi __o; } __u = { val };
+  __builtin_aarch64_st1x4v8qi ((__builtin_aarch64_simd_qi *) __a, __u.__o);
+}
+
+__extension__ extern __inline void
+__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
+vst1q_u8_x4 (uint8_t * __a, uint8x16x4_t val)
+{
+  union { uint8x16x4_t __i; __builtin_aarch64_simd_xi __o; } __u = { val };
+  __builtin_aarch64_st1x4v16qi ((__builtin_aarch64_simd_qi *) __a, __u.__o);
+}
+
+__extension__ extern __inline void
+__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
+vst1_u16_x4 (uint16_t * __a, uint16x4x4_t val)
+{
+  union { uint16x4x4_t __i; __builtin_aarch64_simd_xi __o; } __u = { val };
+  __builtin_aarch64_st1x4v4hi ((__builtin_aarch64_simd_hi *) __a, __u.__o);
+}
+
+__extension__ extern __inline void
+__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
+vst1q_u16_x4 (uint16_t * __a, uint16x8x4_t val)
+{
+  union { uint16x8x4_t __i; __builtin_aarch64_simd_xi __o; } __u = { val };
+  __builtin_aarch64_st1x4v8hi ((__builtin_aarch64_simd_hi *) __a, __u.__o);
+}
+
+__extension__ extern __inline void
+__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
+vst1_u32_x4 (uint32_t * __a, uint32x2x4_t val)
+{
+  union { uint32x2x4_t __i; __builtin_aarch64_simd_xi __o; } __u = { val };
+  __builtin_aarch64_st1x4v2si ((__builtin_aarch64_simd_si *) __a, __u.__o);
+}
+
+__extension__ extern __inline void
+__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
+vst1q_u32_x4 (uint32_t * __a, uint32x4x4_t val)
+{
+  union { uint32x4x4_t __i; __builtin_aarch64_simd_xi __o; } __u = { val };
+  __builtin_aarch64_st1x4v4si ((__builtin_aarch64_simd_si *) __a, __u.__o);
+}
+
+__extension__ extern __inline void
+__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
+vst1_f16_x4 (float16_t * __a, float16x4x4_t val)
+{
+  union { float16x4x4_t __i; __builtin_aarch64_simd_xi __o; } __u = { val };
+  __builtin_aarch64_st1x4v4hf ((__builtin_aarch64_simd_hf *) __a, __u.__o);
+}
+
+__extension__ extern __inline void
+__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
+vst1q_f16_x4 (float16_t * __a, float16x8x4_t val)
+{
+  union { float16x8x4_t __i; __builtin_aarch64_simd_xi __o; } __u = { val };
+  __builtin_aarch64_st1x4v8hf ((__builtin_aarch64_simd_hf *) __a, __u.__o);
+}
+
+__extension__ extern __inline void
+__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
+vst1_f32_x4 (float32_t * __a, float32x2x4_t val)
+{
+  union { float32x2x4_t __i; __builtin_aarch64_simd_xi __o; } __u = { val };
+  __builtin_aarch64_st1x4v2sf ((__builtin_aarch64_simd_sf *) __a, __u.__o);
+}
+
+__extension__ extern __inline void
+__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
+vst1q_f32_x4 (float32_t * __a, float32x4x4_t val)
+{
+  union { float32x4x4_t __i; __builtin_aarch64_simd_xi __o; } __u = { val };
+  __builtin_aarch64_st1x4v4sf ((__builtin_aarch64_simd_sf *) __a, __u.__o);
+}
+
+__extension__ extern __inline void
+__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
+vst1_p8_x4 (poly8_t * __a, poly8x8x4_t val)
+{
+  union { poly8x8x4_t __i; __builtin_aarch64_simd_xi __o; } __u = { val };
+  __builtin_aarch64_st1x4v8qi ((__builtin_aarch64_simd_qi *) __a, __u.__o);
+}
+
+__extension__ extern __inline void
+__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
+vst1q_p8_x4 (poly8_t * __a, poly8x16x4_t val)
+{
+  union { poly8x16x4_t __i; __builtin_aarch64_simd_xi __o; } __u = { val };
+  __builtin_aarch64_st1x4v16qi ((__builtin_aarch64_simd_qi *) __a, __u.__o);
+}
+
+__extension__ extern __inline void
+__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
+vst1_p16_x4 (poly16_t * __a, poly16x4x4_t val)
+{
+  union { poly16x4x4_t __i; __builtin_aarch64_simd_xi __o; } __u = { val };
+  __builtin_aarch64_st1x4v4hi ((__builtin_aarch64_simd_hi *) __a, __u.__o);
+}
+
+__extension__ extern __inline void
+__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
+vst1q_p16_x4 (poly16_t * __a, poly16x8x4_t val)
+{
+  union { poly16x8x4_t __i; __builtin_aarch64_simd_xi __o; } __u = { val };
+  __builtin_aarch64_st1x4v8hi ((__builtin_aarch64_simd_hi *) __a, __u.__o);
+}
+
+__extension__ extern __inline void
+__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
+vst1_s64_x4 (int64_t * __a, int64x1x4_t val)
+{
+  union { int64x1x4_t __i; __builtin_aarch64_simd_xi __o; } __u = { val };
+  __builtin_aarch64_st1x4di ((__builtin_aarch64_simd_di *) __a, __u.__o);
+}
+
+__extension__ extern __inline void
+__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
+vst1_u64_x4 (uint64_t * __a, uint64x1x4_t val)
+{
+  union { uint64x1x4_t __i; __builtin_aarch64_simd_xi __o; } __u = { val };
+  __builtin_aarch64_st1x4di ((__builtin_aarch64_simd_di *) __a, __u.__o);
+}
+
+__extension__ extern __inline void
+__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
+vst1_p64_x4 (poly64_t * __a, poly64x1x4_t val)
+{
+  union { poly64x1x4_t __i; __builtin_aarch64_simd_xi __o; } __u = { val };
+  __builtin_aarch64_st1x4di ((__builtin_aarch64_simd_di *) __a, __u.__o);
+}
+
+__extension__ extern __inline void
+__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
+vst1q_s64_x4 (int64_t * __a, int64x2x4_t val)
+{
+  union { int64x2x4_t __i; __builtin_aarch64_simd_xi __o; } __u = { val };
+  __builtin_aarch64_st1x4v2di ((__builtin_aarch64_simd_di *) __a, __u.__o);
+}
+
+__extension__ extern __inline void
+__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
+vst1q_u64_x4 (uint64_t * __a, uint64x2x4_t val)
+{
+  union { uint64x2x4_t __i; __builtin_aarch64_simd_xi __o; } __u = { val };
+  __builtin_aarch64_st1x4v2di ((__builtin_aarch64_simd_di *) __a, __u.__o);
+}
+
+__extension__ extern __inline void
+__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
+vst1q_p64_x4 (poly64_t * __a, poly64x2x4_t val)
+{
+  union { poly64x2x4_t __i; __builtin_aarch64_simd_xi __o; } __u = { val };
+  __builtin_aarch64_st1x4v2di ((__builtin_aarch64_simd_di *) __a, __u.__o);
+}
+
+__extension__ extern __inline void
+__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
+vst1_f64_x4 (float64_t * __a, float64x1x4_t val)
+{
+  union { float64x1x4_t __i; __builtin_aarch64_simd_xi __o; } __u = { val };
+  __builtin_aarch64_st1x4df ((__builtin_aarch64_simd_df *) __a, __u.__o);
+}
+
+__extension__ extern __inline void
+__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
+vst1q_f64_x4 (float64_t * __a, float64x2x4_t val)
+{
+  union { float64x2x4_t __i; __builtin_aarch64_simd_xi __o; } __u = { val };
+  __builtin_aarch64_st1x4v2df ((__builtin_aarch64_simd_df *) __a, __u.__o);
+}
+
 /* vstn */
 
 __extension__ extern __inline void
index 0d141dc..03fd832 100644 (file)
@@ -1,3 +1,8 @@
+2019-08-22  Sylvia Taylor  <sylvia.taylor@arm.com>
+
+       * gcc.target/aarch64/advsimd-intrinsics/vld1x4.c: New test.
+       * gcc.target/aarch64/advsimd-intrinsics/vst1x4.c: New test.
+
 2019-08-22  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
            Richard Sandiford  <richard.sandiford@arm.com>
 
diff --git a/gcc/testsuite/gcc.target/aarch64/advsimd-intrinsics/vld1x4.c b/gcc/testsuite/gcc.target/aarch64/advsimd-intrinsics/vld1x4.c
new file mode 100644 (file)
index 0000000..451a0af
--- /dev/null
@@ -0,0 +1,83 @@
+/* We haven't implemented these intrinsics for arm yet.  */
+/* { dg-xfail-if "" { arm*-*-* } } */
+/* { dg-do run } */
+/* { dg-options "-O3" } */
+
+#include <arm_neon.h>
+#include "arm-neon-ref.h"
+
+extern void abort (void);
+
+#define TESTMETH(BASE, ELTS, SUFFIX)                   \
+int __attribute__ ((noinline))                         \
+test_vld1##SUFFIX##_x4 ()                              \
+{                                                      \
+  BASE##_t data[ELTS * 4];                             \
+  BASE##_t temp[ELTS * 4];                             \
+  BASE##x##ELTS##x##4##_t vectors;                     \
+  int i,j;                                             \
+  for (i = 0; i < ELTS * 4; i++)                       \
+    data [i] = (BASE##_t) 4*i;                         \
+  asm volatile ("" : : : "memory");                    \
+  vectors = vld1##SUFFIX##_x4 (data);                  \
+  vst1##SUFFIX (temp, vectors.val[0]);                 \
+  vst1##SUFFIX (&temp[ELTS], vectors.val[1]);          \
+  vst1##SUFFIX (&temp[ELTS * 2], vectors.val[2]);      \
+  vst1##SUFFIX (&temp[ELTS * 3], vectors.val[3]);      \
+  asm volatile ("" : : : "memory");                    \
+  for (j = 0; j < ELTS * 4; j++)                       \
+    if (temp[j] != data[j])                            \
+      return 1;                                                \
+  return 0;                                            \
+}
+
+#define VARIANTS_1(VARIANT)    \
+VARIANT (uint8, 8, _u8)                \
+VARIANT (uint16, 4, _u16)      \
+VARIANT (uint32, 2, _u32)      \
+VARIANT (uint64, 1, _u64)      \
+VARIANT (int8, 8, _s8)         \
+VARIANT (int16, 4, _s16)       \
+VARIANT (int32, 2, _s32)       \
+VARIANT (int64, 1, _s64)       \
+VARIANT (poly8, 8, _p8)                \
+VARIANT (poly16, 4, _p16)      \
+VARIANT (poly64, 1, _p64)      \
+VARIANT (float16, 4, _f16)     \
+VARIANT (float32, 2, _f32)     \
+VARIANT (uint8, 16, q_u8)      \
+VARIANT (uint16, 8, q_u16)     \
+VARIANT (uint32, 4, q_u32)     \
+VARIANT (uint64, 2, q_u64)     \
+VARIANT (int8, 16, q_s8)       \
+VARIANT (int16, 8, q_s16)      \
+VARIANT (int32, 4, q_s32)      \
+VARIANT (int64, 2, q_s64)      \
+VARIANT (poly8, 16, q_p8)      \
+VARIANT (poly16, 8, q_p16)     \
+VARIANT (poly64, 2, q_p64)     \
+VARIANT (float16, 8, q_f16)    \
+VARIANT (float32, 4, q_f32)
+
+#ifdef __aarch64__
+#define VARIANTS(VARIANT) VARIANTS_1(VARIANT)  \
+VARIANT (float64, 1, _f64)                     \
+VARIANT (float64, 2, q_f64)
+#else
+#define VARIANTS(VARIANT) VARIANTS_1(VARIANT)
+#endif
+
+/* Tests of vld1_x4 and vld1q_x4.  */
+VARIANTS (TESTMETH)
+
+#define CHECKS(BASE, ELTS, SUFFIX)     \
+  if (test_vld1##SUFFIX##_x4 () != 0)  \
+    fprintf (stderr, "test_vld1##SUFFIX##_x4");
+
+int
+main (int argc, char **argv)
+{
+  VARIANTS (CHECKS)
+
+  return 0;
+}
diff --git a/gcc/testsuite/gcc.target/aarch64/advsimd-intrinsics/vst1x4.c b/gcc/testsuite/gcc.target/aarch64/advsimd-intrinsics/vst1x4.c
new file mode 100644 (file)
index 0000000..1f17b53
--- /dev/null
@@ -0,0 +1,83 @@
+/* We haven't implemented these intrinsics for arm yet.  */
+/* { dg-xfail-if "" { arm*-*-* } } */
+/* { dg-do run } */
+/* { dg-options "-O3" } */
+
+#include <arm_neon.h>
+#include "arm-neon-ref.h"
+
+extern void abort (void);
+
+#define TESTMETH(BASE, ELTS, SUFFIX)                   \
+int __attribute__ ((noinline))                         \
+test_vst1##SUFFIX##_x4 ()                              \
+{                                                      \
+  BASE##_t data[ELTS * 4];                             \
+  BASE##_t temp[ELTS * 4];                             \
+  BASE##x##ELTS##x##4##_t vectors;                     \
+  int i,j;                                             \
+  for (i = 0; i < ELTS * 4; i++)                       \
+    data [i] = (BASE##_t) 4*i;                         \
+  asm volatile ("" : : : "memory");                    \
+  vectors.val[0] = vld1##SUFFIX (data);                        \
+  vectors.val[1] = vld1##SUFFIX (&data[ELTS]);         \
+  vectors.val[2] = vld1##SUFFIX (&data[ELTS * 2]);     \
+  vectors.val[3] = vld1##SUFFIX (&data[ELTS * 3]);     \
+  vst1##SUFFIX##_x4 (temp, vectors);                   \
+  asm volatile ("" : : : "memory");                    \
+  for (j = 0; j < ELTS * 4; j++)                       \
+    if (temp[j] != data[j])                            \
+      return 1;                                                \
+  return 0;                                            \
+}
+
+#define VARIANTS_1(VARIANT)    \
+VARIANT (uint8, 8, _u8)                \
+VARIANT (uint16, 4, _u16)      \
+VARIANT (uint32, 2, _u32)      \
+VARIANT (uint64, 1, _u64)      \
+VARIANT (int8, 8, _s8)         \
+VARIANT (int16, 4, _s16)       \
+VARIANT (int32, 2, _s32)       \
+VARIANT (int64, 1, _s64)       \
+VARIANT (poly8, 8, _p8)                \
+VARIANT (poly16, 4, _p16)      \
+VARIANT (poly64, 1, _p64)      \
+VARIANT (float16, 4, _f16)     \
+VARIANT (float32, 2, _f32)     \
+VARIANT (uint8, 16, q_u8)      \
+VARIANT (uint16, 8, q_u16)     \
+VARIANT (uint32, 4, q_u32)     \
+VARIANT (uint64, 2, q_u64)     \
+VARIANT (int8, 16, q_s8)       \
+VARIANT (int16, 8, q_s16)      \
+VARIANT (int32, 4, q_s32)      \
+VARIANT (int64, 2, q_s64)      \
+VARIANT (poly8, 16, q_p8)      \
+VARIANT (poly16, 8, q_p16)     \
+VARIANT (poly64, 2, q_p64)     \
+VARIANT (float16, 8, q_f16)    \
+VARIANT (float32, 4, q_f32)
+
+#ifdef __aarch64__
+#define VARIANTS(VARIANT) VARIANTS_1(VARIANT)  \
+VARIANT (float64, 1, _f64)                     \
+VARIANT (float64, 2, q_f64)
+#else
+#define VARIANTS(VARIANT) VARIANTS_1(VARIANT)
+#endif
+
+/* Tests of vst1_x4 and vst1q_x4.  */
+VARIANTS (TESTMETH)
+
+#define CHECKS(BASE, ELTS, SUFFIX)     \
+  if (test_vst1##SUFFIX##_x4 () != 0)  \
+    fprintf (stderr, "test_vst1##SUFFIX##_x4");
+
+int
+main (int argc, char **argv)
+{
+  VARIANTS (CHECKS)
+
+  return 0;
+}