From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: by sourceware.org (Postfix, from userid 7835) id 3DC2C3892456; Wed, 27 Jan 2021 15:00:58 +0000 (GMT) DKIM-Filter: OpenDKIM Filter v2.11.0 sourceware.org 3DC2C3892456 MIME-Version: 1.0 Content-Transfer-Encoding: 7bit Content-Type: text/plain; charset="utf-8" From: Jonathan Wright To: gcc-cvs@gcc.gnu.org Subject: [gcc r11-6932] aarch64: Use RTL builtins for integer mls intrinsics X-Act-Checkin: gcc X-Git-Author: Jonathan Wright X-Git-Refname: refs/heads/master X-Git-Oldrev: a199da782fc165fd45f42a15cc9020994efd455d X-Git-Newrev: d2201ac0df461cb97a0187afc230029a6790b70e Message-Id: <20210127150058.3DC2C3892456@sourceware.org> Date: Wed, 27 Jan 2021 15:00:58 +0000 (GMT) X-BeenThere: gcc-cvs@gcc.gnu.org X-Mailman-Version: 2.1.29 Precedence: list List-Id: Gcc-cvs mailing list List-Unsubscribe: , List-Archive: List-Help: List-Subscribe: , X-List-Received-Date: Wed, 27 Jan 2021 15:00:58 -0000 https://gcc.gnu.org/g:d2201ac0df461cb97a0187afc230029a6790b70e commit r11-6932-gd2201ac0df461cb97a0187afc230029a6790b70e Author: Jonathan Wright Date: Mon Jan 11 17:52:45 2021 +0000 aarch64: Use RTL builtins for integer mls intrinsics Rewrite integer mls Neon intrinsics to use RTL builtins rather than inline assembly code, allowing for better scheduling and optimization. gcc/Changelog: 2021-01-11 Jonathan Wright * config/aarch64/aarch64-simd-builtins.def: Add mls builtin generator macro. * config/aarch64/arm_neon.h (vmls_s8): Use RTL builtin rather than asm. (vmls_s16): Likewise. (vmls_s32): Likewise. (vmls_u8): Likewise. (vmls_u16): Likewise. (vmls_u32): Likewise. (vmlsq_s8): Likewise. (vmlsq_s16): Likewise. (vmlsq_s32): Likewise. (vmlsq_u8): Likewise. (vmlsq_u16): Likewise. (vmlsq_u32): Likewise. Diff: --- gcc/config/aarch64/aarch64-simd-builtins.def | 3 + gcc/config/aarch64/arm_neon.h | 96 +++++++--------------------- 2 files changed, 27 insertions(+), 72 deletions(-) diff --git a/gcc/config/aarch64/aarch64-simd-builtins.def b/gcc/config/aarch64/aarch64-simd-builtins.def index ef83d5eee55..93a087987bb 100644 --- a/gcc/config/aarch64/aarch64-simd-builtins.def +++ b/gcc/config/aarch64/aarch64-simd-builtins.def @@ -183,6 +183,9 @@ /* Implemented by aarch64_mla_n. */ BUILTIN_VDQHS (TERNOP, mla_n, 0, NONE) + /* Implemented by aarch64_mls. */ + BUILTIN_VDQ_BHSI (TERNOP, mls, 0, NONE) + /* Implemented by aarch64_mlsl. */ BUILTIN_VD_BHSI (TERNOP, smlsl, 0, NONE) BUILTIN_VD_BHSI (TERNOPU, umlsl, 0, NONE) diff --git a/gcc/config/aarch64/arm_neon.h b/gcc/config/aarch64/arm_neon.h index ef865625a3d..45b3c125bab 100644 --- a/gcc/config/aarch64/arm_neon.h +++ b/gcc/config/aarch64/arm_neon.h @@ -7888,72 +7888,48 @@ __extension__ extern __inline int8x8_t __attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) vmls_s8 (int8x8_t __a, int8x8_t __b, int8x8_t __c) { - int8x8_t __result; - __asm__ ("mls %0.8b,%2.8b,%3.8b" - : "=w"(__result) - : "0"(__a), "w"(__b), "w"(__c) - : /* No clobbers */); - return __result; + return __builtin_aarch64_mlsv8qi (__a, __b, __c); } __extension__ extern __inline int16x4_t __attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) vmls_s16 (int16x4_t __a, int16x4_t __b, int16x4_t __c) { - int16x4_t __result; - __asm__ ("mls %0.4h,%2.4h,%3.4h" - : "=w"(__result) - : "0"(__a), "w"(__b), "w"(__c) - : /* No clobbers */); - return __result; + return __builtin_aarch64_mlsv4hi (__a, __b, __c); } __extension__ extern __inline int32x2_t __attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) vmls_s32 (int32x2_t __a, int32x2_t __b, int32x2_t __c) { - int32x2_t __result; - __asm__ ("mls %0.2s,%2.2s,%3.2s" - : "=w"(__result) - : "0"(__a), "w"(__b), "w"(__c) - : /* No clobbers */); - return __result; + return __builtin_aarch64_mlsv2si (__a, __b, __c); } __extension__ extern __inline uint8x8_t __attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) vmls_u8 (uint8x8_t __a, uint8x8_t __b, uint8x8_t __c) { - uint8x8_t __result; - __asm__ ("mls %0.8b,%2.8b,%3.8b" - : "=w"(__result) - : "0"(__a), "w"(__b), "w"(__c) - : /* No clobbers */); - return __result; + return (uint8x8_t) __builtin_aarch64_mlsv8qi ((int8x8_t) __a, + (int8x8_t) __b, + (int8x8_t) __c); } __extension__ extern __inline uint16x4_t __attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) vmls_u16 (uint16x4_t __a, uint16x4_t __b, uint16x4_t __c) { - uint16x4_t __result; - __asm__ ("mls %0.4h,%2.4h,%3.4h" - : "=w"(__result) - : "0"(__a), "w"(__b), "w"(__c) - : /* No clobbers */); - return __result; + return (uint16x4_t) __builtin_aarch64_mlsv4hi ((int16x4_t) __a, + (int16x4_t) __b, + (int16x4_t) __c); } __extension__ extern __inline uint32x2_t __attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) vmls_u32 (uint32x2_t __a, uint32x2_t __b, uint32x2_t __c) { - uint32x2_t __result; - __asm__ ("mls %0.2s,%2.2s,%3.2s" - : "=w"(__result) - : "0"(__a), "w"(__b), "w"(__c) - : /* No clobbers */); - return __result; + return (uint32x2_t) __builtin_aarch64_mlsv2si ((int32x2_t) __a, + (int32x2_t) __b, + (int32x2_t) __c); } #define vmlsl_high_lane_s16(a, b, c, d) \ @@ -8425,72 +8401,48 @@ __extension__ extern __inline int8x16_t __attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) vmlsq_s8 (int8x16_t __a, int8x16_t __b, int8x16_t __c) { - int8x16_t __result; - __asm__ ("mls %0.16b,%2.16b,%3.16b" - : "=w"(__result) - : "0"(__a), "w"(__b), "w"(__c) - : /* No clobbers */); - return __result; + return __builtin_aarch64_mlsv16qi (__a, __b, __c); } __extension__ extern __inline int16x8_t __attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) vmlsq_s16 (int16x8_t __a, int16x8_t __b, int16x8_t __c) { - int16x8_t __result; - __asm__ ("mls %0.8h,%2.8h,%3.8h" - : "=w"(__result) - : "0"(__a), "w"(__b), "w"(__c) - : /* No clobbers */); - return __result; + return __builtin_aarch64_mlsv8hi (__a, __b, __c); } __extension__ extern __inline int32x4_t __attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) vmlsq_s32 (int32x4_t __a, int32x4_t __b, int32x4_t __c) { - int32x4_t __result; - __asm__ ("mls %0.4s,%2.4s,%3.4s" - : "=w"(__result) - : "0"(__a), "w"(__b), "w"(__c) - : /* No clobbers */); - return __result; + return __builtin_aarch64_mlsv4si (__a, __b, __c); } __extension__ extern __inline uint8x16_t __attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) vmlsq_u8 (uint8x16_t __a, uint8x16_t __b, uint8x16_t __c) { - uint8x16_t __result; - __asm__ ("mls %0.16b,%2.16b,%3.16b" - : "=w"(__result) - : "0"(__a), "w"(__b), "w"(__c) - : /* No clobbers */); - return __result; + return (uint8x16_t) __builtin_aarch64_mlsv16qi ((int8x16_t) __a, + (int8x16_t) __b, + (int8x16_t) __c); } __extension__ extern __inline uint16x8_t __attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) vmlsq_u16 (uint16x8_t __a, uint16x8_t __b, uint16x8_t __c) { - uint16x8_t __result; - __asm__ ("mls %0.8h,%2.8h,%3.8h" - : "=w"(__result) - : "0"(__a), "w"(__b), "w"(__c) - : /* No clobbers */); - return __result; + return (uint16x8_t) __builtin_aarch64_mlsv8hi ((int16x8_t) __a, + (int16x8_t) __b, + (int16x8_t) __c); } __extension__ extern __inline uint32x4_t __attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) vmlsq_u32 (uint32x4_t __a, uint32x4_t __b, uint32x4_t __c) { - uint32x4_t __result; - __asm__ ("mls %0.4s,%2.4s,%3.4s" - : "=w"(__result) - : "0"(__a), "w"(__b), "w"(__c) - : /* No clobbers */); - return __result; + return (uint32x4_t) __builtin_aarch64_mlsv4si ((int32x4_t) __a, + (int32x4_t) __b, + (int32x4_t) __c); } __extension__ extern __inline int16x8_t