From 7fe08e7158a950c3f3866a81ddc284b6420d2097 Mon Sep 17 00:00:00 2001 From: Alan Lawrence Date: Tue, 8 Sep 2015 18:38:26 +0000 Subject: [PATCH] [ARM] float16x8_t intrinsics in arm_neon.h * config/arm/arm_neon.h (vgetq_lane_f16, vsetq_lane_f16, vld1q_lane_f16, vld1q_dup_f16, vreinterpretq_p8_f16, vreinterpretq_p16_f16, vreinterpretq_f16_p8, vreinterpretq_f16_p16, vreinterpretq_f16_f32, vreinterpretq_f16_p64, vreinterpretq_f16_p128, vreinterpretq_f16_s64, vreinterpretq_f16_u64, vreinterpretq_f16_s8, vreinterpretq_f16_s16, vreinterpretq_f16_s32, vreinterpretq_f16_u8, vreinterpretq_f16_u16, vreinterpretq_f16_u32, vreinterpretq_f32_f16, vreinterpretq_p64_f16, vreinterpretq_p128_f16, vreinterpretq_s64_f16, vreinterpretq_u64_f16, vreinterpretq_s8_f16, vreinterpretq_s16_f16, vreinterpretq_s32_f16, vreinterpretq_u8_f16, vreinterpretq_u16_f16, vreinterpretq_u32_f16): New. From-SVN: r227538 --- gcc/ChangeLog | 14 +++ gcc/config/arm/arm_neon.h | 257 ++++++++++++++++++++++++++++++++++++++ 2 files changed, 271 insertions(+) diff --git a/gcc/ChangeLog b/gcc/ChangeLog index e4aa8bc50a6..b80b3bb34d8 100644 --- a/gcc/ChangeLog +++ b/gcc/ChangeLog @@ -1,3 +1,17 @@ +2015-09-07 Alan Lawrence + + * config/arm/arm_neon.h (vgetq_lane_f16, vsetq_lane_f16, vld1q_lane_f16, + vld1q_dup_f16, vreinterpretq_p8_f16, vreinterpretq_p16_f16, + vreinterpretq_f16_p8, vreinterpretq_f16_p16, vreinterpretq_f16_f32, + vreinterpretq_f16_p64, vreinterpretq_f16_p128, vreinterpretq_f16_s64, + vreinterpretq_f16_u64, vreinterpretq_f16_s8, vreinterpretq_f16_s16, + vreinterpretq_f16_s32, vreinterpretq_f16_u8, vreinterpretq_f16_u16, + vreinterpretq_f16_u32, vreinterpretq_f32_f16, vreinterpretq_p64_f16, + vreinterpretq_p128_f16, vreinterpretq_s64_f16, vreinterpretq_u64_f16, + vreinterpretq_s8_f16, vreinterpretq_s16_f16, vreinterpretq_s32_f16, + vreinterpretq_u8_f16, vreinterpretq_u16_f16, vreinterpretq_u32_f16): + New. + 2015-09-07 Alan Lawrence * config/arm/arm.h (VALID_NEON_QREG_MODE): Add V8HFmode. diff --git a/gcc/config/arm/arm_neon.h b/gcc/config/arm/arm_neon.h index 10d2de945e1..b1c9cc76a4c 100644 --- a/gcc/config/arm/arm_neon.h +++ b/gcc/config/arm/arm_neon.h @@ -5288,6 +5288,17 @@ vgetq_lane_s32 (int32x4_t __a, const int __b) return (int32_t)__builtin_neon_vget_lanev4si (__a, __b); } +#if defined (__ARM_FP16_FORMAT_IEEE) || defined (__ARM_FP16_FORMAT_ALTERNATIVE) +#define vgetq_lane_f16(__v, __idx) \ + __extension__ \ + ({ \ + float16x8_t __vec = (__v); \ + __builtin_arm_lane_check (8, __idx); \ + float16_t __res = __vec[__idx]; \ + __res; \ + }) +#endif + __extension__ static __inline float32_t __attribute__ ((__always_inline__)) vgetq_lane_f32 (float32x4_t __a, const int __b) { @@ -5432,6 +5443,18 @@ vsetq_lane_s32 (int32_t __a, int32x4_t __b, const int __c) return (int32x4_t)__builtin_neon_vset_lanev4si ((__builtin_neon_si) __a, __b, __c); } +#if defined (__ARM_FP16_FORMAT_IEEE) || defined (__ARM_FP16_FORMAT_ALTERNATIVE) +#define vsetq_lane_f16(__e, __v, __idx) \ + __extension__ \ + ({ \ + float16_t __elem = (__e); \ + float16x8_t __vec = (__v); \ + __builtin_arm_lane_check (8, __idx); \ + __vec[__idx] = __elem; \ + __vec; \ + }) +#endif + __extension__ static __inline float32x4_t __attribute__ ((__always_inline__)) vsetq_lane_f32 (float32_t __a, float32x4_t __b, const int __c) { @@ -8923,6 +8946,14 @@ vld1q_lane_s32 (const int32_t * __a, int32x4_t __b, const int __c) return (int32x4_t)__builtin_neon_vld1_lanev4si ((const __builtin_neon_si *) __a, __b, __c); } +#if defined (__ARM_FP16_FORMAT_IEEE) || defined (__ARM_FP16_FORMAT_ALTERNATIVE) +__extension__ static __inline float16x8_t __attribute__ ((__always_inline__)) +vld1q_lane_f16 (const float16_t * __a, float16x8_t __b, const int __c) +{ + return vsetq_lane_f16 (*__a, __b, __c); +} +#endif + __extension__ static __inline float32x4_t __attribute__ ((__always_inline__)) vld1q_lane_f32 (const float32_t * __a, float32x4_t __b, const int __c) { @@ -9080,6 +9111,15 @@ vld1q_dup_s32 (const int32_t * __a) return (int32x4_t)__builtin_neon_vld1_dupv4si ((const __builtin_neon_si *) __a); } +#if defined (__ARM_FP16_FORMAT_IEEE) || defined (__ARM_FP16_FORMAT_ALTERNATIVE) +__extension__ static __inline float16x8_t __attribute__ ((__always_inline__)) +vld1q_dup_f16 (const float16_t * __a) +{ + float16_t __f = *__a; + return (float16x8_t) { __f, __f, __f, __f, __f, __f, __f, __f }; +} +#endif + __extension__ static __inline float32x4_t __attribute__ ((__always_inline__)) vld1q_dup_f32 (const float32_t * __a) { @@ -12922,6 +12962,14 @@ vreinterpretq_p8_p16 (poly16x8_t __a) return (poly8x16_t)__builtin_neon_vreinterpretv16qiv8hi ((int16x8_t) __a); } +#if defined (__ARM_FP16_FORMAT_IEEE) || defined (__ARM_FP16_FORMAT_ALTERNATIVE) +__extension__ static __inline poly8x16_t __attribute__ ((__always_inline__)) +vreinterpretq_p8_f16 (float16x8_t __a) +{ + return (poly8x16_t) __a; +} +#endif + __extension__ static __inline poly8x16_t __attribute__ ((__always_inline__)) vreinterpretq_p8_f32 (float32x4_t __a) { @@ -12998,6 +13046,14 @@ vreinterpretq_p16_p8 (poly8x16_t __a) return (poly16x8_t)__builtin_neon_vreinterpretv8hiv16qi ((int8x16_t) __a); } +#if defined (__ARM_FP16_FORMAT_IEEE) || defined (__ARM_FP16_FORMAT_ALTERNATIVE) +__extension__ static __inline poly16x8_t __attribute__ ((__always_inline__)) +vreinterpretq_p16_f16 (float16x8_t __a) +{ + return (poly16x8_t) __a; +} +#endif + __extension__ static __inline poly16x8_t __attribute__ ((__always_inline__)) vreinterpretq_p16_f32 (float32x4_t __a) { @@ -13068,6 +13124,114 @@ vreinterpretq_p16_u32 (uint32x4_t __a) return (poly16x8_t)__builtin_neon_vreinterpretv8hiv4si ((int32x4_t) __a); } +#if defined (__ARM_FP16_FORMAT_IEEE) || defined (__ARM_FP16_FORMAT_ALTERNATIVE) +__extension__ static __inline float16x8_t __attribute__ ((__always_inline__)) +vreinterpretq_f16_p8 (poly8x16_t __a) +{ + return (float16x8_t) __a; +} +#endif + +#if defined (__ARM_FP16_FORMAT_IEEE) || defined (__ARM_FP16_FORMAT_ALTERNATIVE) +__extension__ static __inline float16x8_t __attribute__ ((__always_inline__)) +vreinterpretq_f16_p16 (poly16x8_t __a) +{ + return (float16x8_t) __a; +} +#endif + +#if defined (__ARM_FP16_FORMAT_IEEE) || defined (__ARM_FP16_FORMAT_ALTERNATIVE) +__extension__ static __inline float16x8_t __attribute__ ((__always_inline__)) +vreinterpretq_f16_f32 (float32x4_t __a) +{ + return (float16x8_t) __a; +} +#endif + +#if defined (__ARM_FP16_FORMAT_IEEE) || defined (__ARM_FP16_FORMAT_ALTERNATIVE) +#ifdef __ARM_FEATURE_CRYPTO +__extension__ static __inline float16x8_t __attribute__ ((__always_inline__)) +vreinterpretq_f16_p64 (poly64x2_t __a) +{ + return (float16x8_t) __a; +} +#endif +#endif + +#if defined (__ARM_FP16_FORMAT_IEEE) || defined (__ARM_FP16_FORMAT_ALTERNATIVE) +#ifdef __ARM_FEATURE_CRYPTO +__extension__ static __inline float16x8_t __attribute__ ((__always_inline__)) +vreinterpretq_f16_p128 (poly128_t __a) +{ + return (float16x8_t) __a; +} +#endif +#endif + +#if defined (__ARM_FP16_FORMAT_IEEE) || defined (__ARM_FP16_FORMAT_ALTERNATIVE) +__extension__ static __inline float16x8_t __attribute__ ((__always_inline__)) +vreinterpretq_f16_s64 (int64x2_t __a) +{ + return (float16x8_t) __a; +} +#endif + +#if defined (__ARM_FP16_FORMAT_IEEE) || defined (__ARM_FP16_FORMAT_ALTERNATIVE) +__extension__ static __inline float16x8_t __attribute__ ((__always_inline__)) +vreinterpretq_f16_u64 (uint64x2_t __a) +{ + return (float16x8_t) __a; +} +#endif + +#if defined (__ARM_FP16_FORMAT_IEEE) || defined (__ARM_FP16_FORMAT_ALTERNATIVE) +__extension__ static __inline float16x8_t __attribute__ ((__always_inline__)) +vreinterpretq_f16_s8 (int8x16_t __a) +{ + return (float16x8_t) __a; +} +#endif + +#if defined (__ARM_FP16_FORMAT_IEEE) || defined (__ARM_FP16_FORMAT_ALTERNATIVE) +__extension__ static __inline float16x8_t __attribute__ ((__always_inline__)) +vreinterpretq_f16_s16 (int16x8_t __a) +{ + return (float16x8_t) __a; +} +#endif + +#if defined (__ARM_FP16_FORMAT_IEEE) || defined (__ARM_FP16_FORMAT_ALTERNATIVE) +__extension__ static __inline float16x8_t __attribute__ ((__always_inline__)) +vreinterpretq_f16_s32 (int32x4_t __a) +{ + return (float16x8_t) __a; +} +#endif + +#if defined (__ARM_FP16_FORMAT_IEEE) || defined (__ARM_FP16_FORMAT_ALTERNATIVE) +__extension__ static __inline float16x8_t __attribute__ ((__always_inline__)) +vreinterpretq_f16_u8 (uint8x16_t __a) +{ + return (float16x8_t) __a; +} +#endif + +#if defined (__ARM_FP16_FORMAT_IEEE) || defined (__ARM_FP16_FORMAT_ALTERNATIVE) +__extension__ static __inline float16x8_t __attribute__ ((__always_inline__)) +vreinterpretq_f16_u16 (uint16x8_t __a) +{ + return (float16x8_t) __a; +} +#endif + +#if defined (__ARM_FP16_FORMAT_IEEE) || defined (__ARM_FP16_FORMAT_ALTERNATIVE) +__extension__ static __inline float16x8_t __attribute__ ((__always_inline__)) +vreinterpretq_f16_u32 (uint32x4_t __a) +{ + return (float16x8_t) __a; +} +#endif + __extension__ static __inline float32x4_t __attribute__ ((__always_inline__)) vreinterpretq_f32_p8 (poly8x16_t __a) { @@ -13080,6 +13244,14 @@ vreinterpretq_f32_p16 (poly16x8_t __a) return (float32x4_t)__builtin_neon_vreinterpretv4sfv8hi ((int16x8_t) __a); } +#if defined (__ARM_FP16_FORMAT_IEEE) || defined (__ARM_FP16_FORMAT_ALTERNATIVE) +__extension__ static __inline float32x4_t __attribute__ ((__always_inline__)) +vreinterpretq_f32_f16 (float16x8_t __a) +{ + return (float32x4_t) __a; +} +#endif + #ifdef __ARM_FEATURE_CRYPTO __extension__ static __inline float32x4_t __attribute__ ((__always_inline__)) vreinterpretq_f32_p64 (poly64x2_t __a) @@ -13160,6 +13332,17 @@ vreinterpretq_p64_p16 (poly16x8_t __a) } #endif + +#ifdef __ARM_FEATURE_CRYPTO +#if defined (__ARM_FP16_FORMAT_IEEE) || defined (__ARM_FP16_FORMAT_ALTERNATIVE) +__extension__ static __inline poly64x2_t __attribute__ ((__always_inline__)) +vreinterpretq_p64_f16 (float16x8_t __a) +{ + return (poly64x2_t) __a; +} +#endif +#endif + #ifdef __ARM_FEATURE_CRYPTO __extension__ static __inline poly64x2_t __attribute__ ((__always_inline__)) vreinterpretq_p64_f32 (float32x4_t __a) @@ -13254,8 +13437,18 @@ vreinterpretq_p128_p16 (poly16x8_t __a) { return (poly128_t)__builtin_neon_vreinterprettiv8hi ((int16x8_t) __a); } +#endif +#ifdef __ARM_FEATURE_CRYPTO +#if defined (__ARM_FP16_FORMAT_IEEE) || defined (__ARM_FP16_FORMAT_ALTERNATIVE) +__extension__ static __inline poly128_t __attribute__ ((__always_inline__)) +vreinterpretq_p128_f16 (float16x8_t __a) +{ + return (poly128_t) __a; +} +#endif #endif + #ifdef __ARM_FEATURE_CRYPTO __extension__ static __inline poly128_t __attribute__ ((__always_inline__)) vreinterpretq_p128_f32 (float32x4_t __a) @@ -13348,6 +13541,14 @@ vreinterpretq_s64_p16 (poly16x8_t __a) return (int64x2_t)__builtin_neon_vreinterpretv2div8hi ((int16x8_t) __a); } +#if defined (__ARM_FP16_FORMAT_IEEE) || defined (__ARM_FP16_FORMAT_ALTERNATIVE) +__extension__ static __inline int64x2_t __attribute__ ((__always_inline__)) +vreinterpretq_s64_f16 (float16x8_t __a) +{ + return (int64x2_t) __a; +} +#endif + __extension__ static __inline int64x2_t __attribute__ ((__always_inline__)) vreinterpretq_s64_f32 (float32x4_t __a) { @@ -13424,6 +13625,14 @@ vreinterpretq_u64_p16 (poly16x8_t __a) return (uint64x2_t)__builtin_neon_vreinterpretv2div8hi ((int16x8_t) __a); } +#if defined (__ARM_FP16_FORMAT_IEEE) || defined (__ARM_FP16_FORMAT_ALTERNATIVE) +__extension__ static __inline uint64x2_t __attribute__ ((__always_inline__)) +vreinterpretq_u64_f16 (float16x8_t __a) +{ + return (uint64x2_t) __a; +} +#endif + __extension__ static __inline uint64x2_t __attribute__ ((__always_inline__)) vreinterpretq_u64_f32 (float32x4_t __a) { @@ -13500,6 +13709,14 @@ vreinterpretq_s8_p16 (poly16x8_t __a) return (int8x16_t)__builtin_neon_vreinterpretv16qiv8hi ((int16x8_t) __a); } +#if defined (__ARM_FP16_FORMAT_IEEE) || defined (__ARM_FP16_FORMAT_ALTERNATIVE) +__extension__ static __inline int8x16_t __attribute__ ((__always_inline__)) +vreinterpretq_s8_f16 (float16x8_t __a) +{ + return (int8x16_t) __a; +} +#endif + __extension__ static __inline int8x16_t __attribute__ ((__always_inline__)) vreinterpretq_s8_f32 (float32x4_t __a) { @@ -13576,6 +13793,14 @@ vreinterpretq_s16_p16 (poly16x8_t __a) return (int16x8_t)__builtin_neon_vreinterpretv8hiv8hi ((int16x8_t) __a); } +#if defined (__ARM_FP16_FORMAT_IEEE) || defined (__ARM_FP16_FORMAT_ALTERNATIVE) +__extension__ static __inline int16x8_t __attribute__ ((__always_inline__)) +vreinterpretq_s16_f16 (float16x8_t __a) +{ + return (int16x8_t) __a; +} +#endif + __extension__ static __inline int16x8_t __attribute__ ((__always_inline__)) vreinterpretq_s16_f32 (float32x4_t __a) { @@ -13652,6 +13877,14 @@ vreinterpretq_s32_p16 (poly16x8_t __a) return (int32x4_t)__builtin_neon_vreinterpretv4siv8hi ((int16x8_t) __a); } +#if defined (__ARM_FP16_FORMAT_IEEE) || defined (__ARM_FP16_FORMAT_ALTERNATIVE) +__extension__ static __inline int32x4_t __attribute__ ((__always_inline__)) +vreinterpretq_s32_f16 (float16x8_t __a) +{ + return (int32x4_t)__builtin_neon_vreinterpretv4siv8hi ((int16x8_t) __a); +} +#endif + __extension__ static __inline int32x4_t __attribute__ ((__always_inline__)) vreinterpretq_s32_f32 (float32x4_t __a) { @@ -13728,6 +13961,14 @@ vreinterpretq_u8_p16 (poly16x8_t __a) return (uint8x16_t)__builtin_neon_vreinterpretv16qiv8hi ((int16x8_t) __a); } +#if defined (__ARM_FP16_FORMAT_IEEE) || defined (__ARM_FP16_FORMAT_ALTERNATIVE) +__extension__ static __inline uint8x16_t __attribute__ ((__always_inline__)) +vreinterpretq_u8_f16 (float16x8_t __a) +{ + return (uint8x16_t) __a; +} +#endif + __extension__ static __inline uint8x16_t __attribute__ ((__always_inline__)) vreinterpretq_u8_f32 (float32x4_t __a) { @@ -13804,6 +14045,14 @@ vreinterpretq_u16_p16 (poly16x8_t __a) return (uint16x8_t)__builtin_neon_vreinterpretv8hiv8hi ((int16x8_t) __a); } +#if defined (__ARM_FP16_FORMAT_IEEE) || defined (__ARM_FP16_FORMAT_ALTERNATIVE) +__extension__ static __inline uint16x8_t __attribute__ ((__always_inline__)) +vreinterpretq_u16_f16 (float16x8_t __a) +{ + return (uint16x8_t) __a; +} +#endif + __extension__ static __inline uint16x8_t __attribute__ ((__always_inline__)) vreinterpretq_u16_f32 (float32x4_t __a) { @@ -13880,6 +14129,14 @@ vreinterpretq_u32_p16 (poly16x8_t __a) return (uint32x4_t)__builtin_neon_vreinterpretv4siv8hi ((int16x8_t) __a); } +#if defined (__ARM_FP16_FORMAT_IEEE) || defined (__ARM_FP16_FORMAT_ALTERNATIVE) +__extension__ static __inline uint32x4_t __attribute__ ((__always_inline__)) +vreinterpretq_u32_f16 (float16x8_t __a) +{ + return (uint32x4_t) __a; +} +#endif + __extension__ static __inline uint32x4_t __attribute__ ((__always_inline__)) vreinterpretq_u32_f32 (float32x4_t __a) { -- 2.30.2