aarch64: Use type-qualified builtins for UADD[LW][2] Neon intrinsics

Declare unsigned type-qualified builtins and use them to implement
widening-add Neon intrinsics. This removes the need for many casts in
arm_neon.h.

gcc/ChangeLog:

2021-11-09  Jonathan Wright  <jonathan.wright@arm.com>

	* config/aarch64/aarch64-simd-builtins.def: Use BINOPU type
	qualifiers in generator macros for uadd[lw][2] builtins.
	* config/aarch64/arm_neon.h (vaddl_s8): Remove unnecessary
	cast.
	(vaddl_s16): Likewise.
	(vaddl_s32): Likewise.
	(vaddl_u8): Use type-qualified builtin and remove casts.
	(vaddl_u16): Likewise.
	(vaddl_u32): Likewise.
	(vaddl_high_s8): Remove unnecessary cast.
	(vaddl_high_s16): Likewise.
	(vaddl_high_s32): Likewise.
	(vaddl_high_u8): Use type-qualified builtin and remove casts.
	(vaddl_high_u16): Likewise.
	(vaddl_high_u32): Likewise.
	(vaddw_s8): Remove unnecessary cast.
	(vaddw_s16): Likewise.
	(vaddw_s32): Likewise.
	(vaddw_u8): Use type-qualified builtin and remove casts.
	(vaddw_u16): Likewise.
	(vaddw_u32): Likewise.
	(vaddw_high_s8): Remove unnecessary cast.
	(vaddw_high_s16): Likewise.
	(vaddw_high_s32): Likewise.
	(vaddw_high_u8): Use type-qualified builtin and remove casts.
	(vaddw_high_u16): Likewise.
	(vaddw_high_u32): Likewise.

diff --git a/gcc/config/aarch64/aarch64-simd-builtins.def b/gcc/config/aarch64/aarch64-simd-builtins.def
index 42dfc4cf35464d000f66c9b5afe8a32ece881c1d..879662eb0d9f5cfc80919b7dc8b0adc43aee85ac 100644 (file)
--- a/gcc/config/aarch64/aarch64-simd-builtins.def
+++ b/gcc/config/aarch64/aarch64-simd-builtins.def
@@ -158,21 +158,21 @@
   BUILTIN_VALLP (STORESTRUCT_LANE_P, st4_lane, 0, ALL)
 
   BUILTIN_VQW (BINOP, saddl2, 0, NONE)
-  BUILTIN_VQW (BINOP, uaddl2, 0, NONE)
+  BUILTIN_VQW (BINOPU, uaddl2, 0, NONE)
   BUILTIN_VQW (BINOP, ssubl2, 0, NONE)
   BUILTIN_VQW (BINOP, usubl2, 0, NONE)
   BUILTIN_VQW (BINOP, saddw2, 0, NONE)
-  BUILTIN_VQW (BINOP, uaddw2, 0, NONE)
+  BUILTIN_VQW (BINOPU, uaddw2, 0, NONE)
   BUILTIN_VQW (BINOP, ssubw2, 0, NONE)
   BUILTIN_VQW (BINOP, usubw2, 0, NONE)
   /* Implemented by aarch64_<ANY_EXTEND:su><ADDSUB:optab>l<mode>.  */
   BUILTIN_VD_BHSI (BINOP, saddl, 0, NONE)
-  BUILTIN_VD_BHSI (BINOP, uaddl, 0, NONE)
+  BUILTIN_VD_BHSI (BINOPU, uaddl, 0, NONE)
   BUILTIN_VD_BHSI (BINOP, ssubl, 0, NONE)
   BUILTIN_VD_BHSI (BINOP, usubl, 0, NONE)
   /* Implemented by aarch64_<ANY_EXTEND:su><ADDSUB:optab>w<mode>.  */
   BUILTIN_VD_BHSI (BINOP, saddw, 0, NONE)
-  BUILTIN_VD_BHSI (BINOP, uaddw, 0, NONE)
+  BUILTIN_VD_BHSI (BINOPU, uaddw, 0, NONE)
   BUILTIN_VD_BHSI (BINOP, ssubw, 0, NONE)
   BUILTIN_VD_BHSI (BINOP, usubw, 0, NONE)
   /* Implemented by aarch64_<sur>h<addsub><mode>.  */

diff --git a/gcc/config/aarch64/arm_neon.h b/gcc/config/aarch64/arm_neon.h
index 944b9662ed7fdae5750668c6279aeb6d2185fb6b..43a89b3a2b37aa8aff7ff5969b81edd05808055a 100644 (file)
--- a/gcc/config/aarch64/arm_neon.h
+++ b/gcc/config/aarch64/arm_neon.h
@@ -377,180 +377,168 @@ __extension__ extern __inline int16x8_t
 __attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
 vaddl_s8 (int8x8_t __a, int8x8_t __b)
 {
-  return (int16x8_t) __builtin_aarch64_saddlv8qi (__a, __b);
+  return __builtin_aarch64_saddlv8qi (__a, __b);
 }
 
 __extension__ extern __inline int32x4_t
 __attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
 vaddl_s16 (int16x4_t __a, int16x4_t __b)
 {
-  return (int32x4_t) __builtin_aarch64_saddlv4hi (__a, __b);
+  return __builtin_aarch64_saddlv4hi (__a, __b);
 }
 
 __extension__ extern __inline int64x2_t
 __attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
 vaddl_s32 (int32x2_t __a, int32x2_t __b)
 {
-  return (int64x2_t) __builtin_aarch64_saddlv2si (__a, __b);
+  return __builtin_aarch64_saddlv2si (__a, __b);
 }
 
 __extension__ extern __inline uint16x8_t
 __attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
 vaddl_u8 (uint8x8_t __a, uint8x8_t __b)
 {
-  return (uint16x8_t) __builtin_aarch64_uaddlv8qi ((int8x8_t) __a,
-                                                  (int8x8_t) __b);
+  return __builtin_aarch64_uaddlv8qi_uuu (__a, __b);
 }
 
 __extension__ extern __inline uint32x4_t
 __attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
 vaddl_u16 (uint16x4_t __a, uint16x4_t __b)
 {
-  return (uint32x4_t) __builtin_aarch64_uaddlv4hi ((int16x4_t) __a,
-                                                  (int16x4_t) __b);
+  return __builtin_aarch64_uaddlv4hi_uuu (__a, __b);
 }
 
 __extension__ extern __inline uint64x2_t
 __attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
 vaddl_u32 (uint32x2_t __a, uint32x2_t __b)
 {
-  return (uint64x2_t) __builtin_aarch64_uaddlv2si ((int32x2_t) __a,
-                                                  (int32x2_t) __b);
+  return __builtin_aarch64_uaddlv2si_uuu (__a, __b);
 }
 
 __extension__ extern __inline int16x8_t
 __attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
 vaddl_high_s8 (int8x16_t __a, int8x16_t __b)
 {
-  return (int16x8_t) __builtin_aarch64_saddl2v16qi (__a, __b);
+  return __builtin_aarch64_saddl2v16qi (__a, __b);
 }
 
 __extension__ extern __inline int32x4_t
 __attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
 vaddl_high_s16 (int16x8_t __a, int16x8_t __b)
 {
-  return (int32x4_t) __builtin_aarch64_saddl2v8hi (__a, __b);
+  return __builtin_aarch64_saddl2v8hi (__a, __b);
 }
 
 __extension__ extern __inline int64x2_t
 __attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
 vaddl_high_s32 (int32x4_t __a, int32x4_t __b)
 {
-  return (int64x2_t) __builtin_aarch64_saddl2v4si (__a, __b);
+  return __builtin_aarch64_saddl2v4si (__a, __b);
 }
 
 __extension__ extern __inline uint16x8_t
 __attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
 vaddl_high_u8 (uint8x16_t __a, uint8x16_t __b)
 {
-  return (uint16x8_t) __builtin_aarch64_uaddl2v16qi ((int8x16_t) __a,
-                                                    (int8x16_t) __b);
+  return __builtin_aarch64_uaddl2v16qi_uuu (__a, __b);
 }
 
 __extension__ extern __inline uint32x4_t
 __attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
 vaddl_high_u16 (uint16x8_t __a, uint16x8_t __b)
 {
-  return (uint32x4_t) __builtin_aarch64_uaddl2v8hi ((int16x8_t) __a,
-                                                   (int16x8_t) __b);
+  return __builtin_aarch64_uaddl2v8hi_uuu (__a, __b);
 }
 
 __extension__ extern __inline uint64x2_t
 __attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
 vaddl_high_u32 (uint32x4_t __a, uint32x4_t __b)
 {
-  return (uint64x2_t) __builtin_aarch64_uaddl2v4si ((int32x4_t) __a,
-                                                   (int32x4_t) __b);
+  return __builtin_aarch64_uaddl2v4si_uuu (__a, __b);
 }
 
 __extension__ extern __inline int16x8_t
 __attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
 vaddw_s8 (int16x8_t __a, int8x8_t __b)
 {
-  return (int16x8_t) __builtin_aarch64_saddwv8qi (__a, __b);
+  return __builtin_aarch64_saddwv8qi (__a, __b);
 }
 
 __extension__ extern __inline int32x4_t
 __attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
 vaddw_s16 (int32x4_t __a, int16x4_t __b)
 {
-  return (int32x4_t) __builtin_aarch64_saddwv4hi (__a, __b);
+  return __builtin_aarch64_saddwv4hi (__a, __b);
 }
 
 __extension__ extern __inline int64x2_t
 __attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
 vaddw_s32 (int64x2_t __a, int32x2_t __b)
 {
-  return (int64x2_t) __builtin_aarch64_saddwv2si (__a, __b);
+  return __builtin_aarch64_saddwv2si (__a, __b);
 }
 
 __extension__ extern __inline uint16x8_t
 __attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
 vaddw_u8 (uint16x8_t __a, uint8x8_t __b)
 {
-  return (uint16x8_t) __builtin_aarch64_uaddwv8qi ((int16x8_t) __a,
-                                                  (int8x8_t) __b);
+  return __builtin_aarch64_uaddwv8qi_uuu (__a, __b);
 }
 
 __extension__ extern __inline uint32x4_t
 __attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
 vaddw_u16 (uint32x4_t __a, uint16x4_t __b)
 {
-  return (uint32x4_t) __builtin_aarch64_uaddwv4hi ((int32x4_t) __a,
-                                                  (int16x4_t) __b);
+  return __builtin_aarch64_uaddwv4hi_uuu (__a, __b);
 }
 
 __extension__ extern __inline uint64x2_t
 __attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
 vaddw_u32 (uint64x2_t __a, uint32x2_t __b)
 {
-  return (uint64x2_t) __builtin_aarch64_uaddwv2si ((int64x2_t) __a,
-                                                  (int32x2_t) __b);
+  return __builtin_aarch64_uaddwv2si_uuu (__a, __b);
 }
 
 __extension__ extern __inline int16x8_t
 __attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
 vaddw_high_s8 (int16x8_t __a, int8x16_t __b)
 {
-  return (int16x8_t) __builtin_aarch64_saddw2v16qi (__a, __b);
+  return __builtin_aarch64_saddw2v16qi (__a, __b);
 }
 
 __extension__ extern __inline int32x4_t
 __attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
 vaddw_high_s16 (int32x4_t __a, int16x8_t __b)
 {
-  return (int32x4_t) __builtin_aarch64_saddw2v8hi (__a, __b);
+  return __builtin_aarch64_saddw2v8hi (__a, __b);
 }
 
 __extension__ extern __inline int64x2_t
 __attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
 vaddw_high_s32 (int64x2_t __a, int32x4_t __b)
 {
-  return (int64x2_t) __builtin_aarch64_saddw2v4si (__a, __b);
+  return __builtin_aarch64_saddw2v4si (__a, __b);
 }
 
 __extension__ extern __inline uint16x8_t
 __attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
 vaddw_high_u8 (uint16x8_t __a, uint8x16_t __b)
 {
-  return (uint16x8_t) __builtin_aarch64_uaddw2v16qi ((int16x8_t) __a,
-                                                    (int8x16_t) __b);
+  return __builtin_aarch64_uaddw2v16qi_uuu (__a, __b);
 }
 
 __extension__ extern __inline uint32x4_t
 __attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
 vaddw_high_u16 (uint32x4_t __a, uint16x8_t __b)
 {
-  return (uint32x4_t) __builtin_aarch64_uaddw2v8hi ((int32x4_t) __a,
-                                                   (int16x8_t) __b);
+  return __builtin_aarch64_uaddw2v8hi_uuu (__a, __b);
 }
 
 __extension__ extern __inline uint64x2_t
 __attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
 vaddw_high_u32 (uint64x2_t __a, uint32x4_t __b)
 {
-  return (uint64x2_t) __builtin_aarch64_uaddw2v4si ((int64x2_t) __a,
-                                                   (int32x4_t) __b);
+  return __builtin_aarch64_uaddw2v4si_uuu (__a, __b);
 }
 
 __extension__ extern __inline int8x8_t