* [PATCH] aarch64: Use type-qualified builtins for UADD[LW][2] Neon intrinsics
@ 2021-11-11 10:17 Jonathan Wright
2021-11-11 13:15 ` Kyrylo Tkachov
0 siblings, 1 reply; 2+ messages in thread
From: Jonathan Wright @ 2021-11-11 10:17 UTC (permalink / raw)
To: gcc-patches; +Cc: Richard Sandiford, Kyrylo Tkachov
[-- Attachment #1: Type: text/plain, Size: 1377 bytes --]
Hi,
This patch declares unsigned type-qualified builtins and uses them to
implement widening-add Neon intrinsics. This removes the need for
many casts in arm_neon.h.
Bootstrapped and regression tested on aarch64-none-linux-gnu - no
issues.
Ok for master?
Thanks,
Jonathan
---
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.
[-- Attachment #2: rb15033.patch --]
[-- Type: application/octet-stream, Size: 8835 bytes --]
diff --git a/gcc/config/aarch64/aarch64-simd-builtins.def b/gcc/config/aarch64/aarch64-simd-builtins.def
index 46ec2f9bfc509e5e460334d4c5324ddf18703639..ccd194978f948201698aec16d74baa82c187cad4 100644
--- 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 37f02e2a24fbc85f23ea73e2fd0e06deac7db87e..a3d742f25a896f8e736a5fb01535d372cd4b20db 100644
--- 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
^ permalink raw reply [flat|nested] 2+ messages in thread
* RE: [PATCH] aarch64: Use type-qualified builtins for UADD[LW][2] Neon intrinsics
2021-11-11 10:17 [PATCH] aarch64: Use type-qualified builtins for UADD[LW][2] Neon intrinsics Jonathan Wright
@ 2021-11-11 13:15 ` Kyrylo Tkachov
0 siblings, 0 replies; 2+ messages in thread
From: Kyrylo Tkachov @ 2021-11-11 13:15 UTC (permalink / raw)
To: Jonathan Wright, gcc-patches; +Cc: Richard Sandiford
Hi Jonathan,
> -----Original Message-----
> From: Jonathan Wright <Jonathan.Wright@arm.com>
> Sent: Thursday, November 11, 2021 10:18 AM
> To: gcc-patches@gcc.gnu.org
> Cc: Richard Sandiford <Richard.Sandiford@arm.com>; Kyrylo Tkachov
> <Kyrylo.Tkachov@arm.com>
> Subject: [PATCH] aarch64: Use type-qualified builtins for UADD[LW][2] Neon
> intrinsics
>
> Hi,
>
> This patch declares unsigned type-qualified builtins and uses them to
> implement widening-add Neon intrinsics. This removes the need for
> many casts in arm_neon.h.
>
> Bootstrapped and regression tested on aarch64-none-linux-gnu - no
> issues.
>
> Ok for master?
>
> Thanks,
> Jonathan
>
> ---
>
> 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.
Ok.
Thanks,
Kyrill
^ permalink raw reply [flat|nested] 2+ messages in thread
end of thread, other threads:[~2021-11-11 13:15 UTC | newest]
Thread overview: 2+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2021-11-11 10:17 [PATCH] aarch64: Use type-qualified builtins for UADD[LW][2] Neon intrinsics Jonathan Wright
2021-11-11 13:15 ` Kyrylo Tkachov
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for read-only IMAP folder(s) and NNTP newsgroup(s).