From: Alex Velenko Date: Tue, 22 Apr 2014 16:06:05 +0000 (+0000) Subject: [AArch64] 64-bit float vreinterpret implemention X-Git-Tag: releases/gcc-5.1.0~8004 X-Git-Url: https://gcc.gnu.org/git/?a=commitdiff_plain;h=bcd48995efefb23dbb8ccd44850065febcf1a7e8;p=gcc.git [AArch64] 64-bit float vreinterpret implemention This patch introduces vreinterpret implementation for vectors with 64-bit float lanes and adds testcase for those intrinsics. From-SVN: r209642 --- diff --git a/gcc/ChangeLog b/gcc/ChangeLog index 149c0a05753c..1811fa48be34 100644 --- a/gcc/ChangeLog +++ b/gcc/ChangeLog @@ -1,3 +1,68 @@ +2014-04-22 Alex Velenko + + * config/aarch64/aarch64-builtins.c (TYPES_REINTERP): Removed. + (aarch64_types_signed_unsigned_qualifiers): Qualifier added. + (aarch64_types_signed_poly_qualifiers): Likewise. + (aarch64_types_unsigned_signed_qualifiers): Likewise. + (aarch64_types_poly_signed_qualifiers): Likewise. + (TYPES_REINTERP_SS): Type macro added. + (TYPES_REINTERP_SU): Likewise. + (TYPES_REINTERP_SP): Likewise. + (TYPES_REINTERP_US): Likewise. + (TYPES_REINTERP_PS): Likewise. + (aarch64_fold_builtin): New expression folding added. + * config/aarch64/aarch64-simd-builtins.def (REINTERP): + Declarations removed. + (REINTERP_SS): Declarations added. + (REINTERP_US): Likewise. + (REINTERP_PS): Likewise. + (REINTERP_SU): Likewise. + (REINTERP_SP): Likewise. + * config/aarch64/arm_neon.h (vreinterpret_p8_f64): Implemented. + (vreinterpretq_p8_f64): Likewise. + (vreinterpret_p16_f64): Likewise. + (vreinterpretq_p16_f64): Likewise. + (vreinterpret_f32_f64): Likewise. + (vreinterpretq_f32_f64): Likewise. + (vreinterpret_f64_f32): Likewise. + (vreinterpret_f64_p8): Likewise. + (vreinterpret_f64_p16): Likewise. + (vreinterpret_f64_s8): Likewise. + (vreinterpret_f64_s16): Likewise. + (vreinterpret_f64_s32): Likewise. + (vreinterpret_f64_s64): Likewise. + (vreinterpret_f64_u8): Likewise. + (vreinterpret_f64_u16): Likewise. + (vreinterpret_f64_u32): Likewise. + (vreinterpret_f64_u64): Likewise. + (vreinterpretq_f64_f32): Likewise. + (vreinterpretq_f64_p8): Likewise. + (vreinterpretq_f64_p16): Likewise. + (vreinterpretq_f64_s8): Likewise. + (vreinterpretq_f64_s16): Likewise. + (vreinterpretq_f64_s32): Likewise. + (vreinterpretq_f64_s64): Likewise. + (vreinterpretq_f64_u8): Likewise. + (vreinterpretq_f64_u16): Likewise. + (vreinterpretq_f64_u32): Likewise. + (vreinterpretq_f64_u64): Likewise. + (vreinterpret_s64_f64): Likewise. + (vreinterpretq_s64_f64): Likewise. + (vreinterpret_u64_f64): Likewise. + (vreinterpretq_u64_f64): Likewise. + (vreinterpret_s8_f64): Likewise. + (vreinterpretq_s8_f64): Likewise. + (vreinterpret_s16_f64): Likewise. + (vreinterpretq_s16_f64): Likewise. + (vreinterpret_s32_f64): Likewise. + (vreinterpretq_s32_f64): Likewise. + (vreinterpret_u8_f64): Likewise. + (vreinterpretq_u8_f64): Likewise. + (vreinterpret_u16_f64): Likewise. + (vreinterpretq_u16_f64): Likewise. + (vreinterpret_u32_f64): Likewise. + (vreinterpretq_u32_f64): Likewise. + 2014-04-22 Alex Velenko * config/aarch64/aarch64/aarch64-builtins.c (TYPES_REINTERP): Removed. diff --git a/gcc/config/aarch64/aarch64-builtins.c b/gcc/config/aarch64/aarch64-builtins.c index 5903d2202752..591260f18bcc 100644 --- a/gcc/config/aarch64/aarch64-builtins.c +++ b/gcc/config/aarch64/aarch64-builtins.c @@ -147,6 +147,23 @@ aarch64_types_unopu_qualifiers[SIMD_MAX_BUILTIN_ARGS] = { qualifier_unsigned, qualifier_unsigned }; #define TYPES_UNOPU (aarch64_types_unopu_qualifiers) #define TYPES_CREATE (aarch64_types_unop_qualifiers) +#define TYPES_REINTERP_SS (aarch64_types_unop_qualifiers) +static enum aarch64_type_qualifiers +aarch64_types_unop_su_qualifiers[SIMD_MAX_BUILTIN_ARGS] + = { qualifier_none, qualifier_unsigned }; +#define TYPES_REINTERP_SU (aarch64_types_unop_su_qualifiers) +static enum aarch64_type_qualifiers +aarch64_types_unop_sp_qualifiers[SIMD_MAX_BUILTIN_ARGS] + = { qualifier_none, qualifier_poly }; +#define TYPES_REINTERP_SP (aarch64_types_unop_sp_qualifiers) +static enum aarch64_type_qualifiers +aarch64_types_unop_us_qualifiers[SIMD_MAX_BUILTIN_ARGS] + = { qualifier_unsigned, qualifier_none }; +#define TYPES_REINTERP_US (aarch64_types_unop_us_qualifiers) +static enum aarch64_type_qualifiers +aarch64_types_unop_ps_qualifiers[SIMD_MAX_BUILTIN_ARGS] + = { qualifier_poly, qualifier_none }; +#define TYPES_REINTERP_PS (aarch64_types_unop_ps_qualifiers) static enum aarch64_type_qualifiers aarch64_types_binop_qualifiers[SIMD_MAX_BUILTIN_ARGS] = { qualifier_none, qualifier_none, qualifier_maybe_immediate }; @@ -1128,6 +1145,25 @@ aarch64_fold_builtin (tree fndecl, int n_args ATTRIBUTE_UNUSED, tree *args, return fold_build2 (NE_EXPR, type, and_node, vec_zero_node); break; } + VAR1 (REINTERP_SS, reinterpretdi, 0, df) + VAR1 (REINTERP_SS, reinterpretv8qi, 0, df) + VAR1 (REINTERP_SS, reinterpretv4hi, 0, df) + VAR1 (REINTERP_SS, reinterpretv2si, 0, df) + VAR1 (REINTERP_SS, reinterpretv2sf, 0, df) + BUILTIN_VD (REINTERP_SS, reinterpretdf, 0) + BUILTIN_VD (REINTERP_SU, reinterpretdf, 0) + VAR1 (REINTERP_US, reinterpretdi, 0, df) + VAR1 (REINTERP_US, reinterpretv8qi, 0, df) + VAR1 (REINTERP_US, reinterpretv4hi, 0, df) + VAR1 (REINTERP_US, reinterpretv2si, 0, df) + VAR1 (REINTERP_US, reinterpretv2sf, 0, df) + BUILTIN_VD (REINTERP_SP, reinterpretdf, 0) + VAR1 (REINTERP_PS, reinterpretdi, 0, df) + VAR1 (REINTERP_PS, reinterpretv8qi, 0, df) + VAR1 (REINTERP_PS, reinterpretv4hi, 0, df) + VAR1 (REINTERP_PS, reinterpretv2si, 0, df) + VAR1 (REINTERP_PS, reinterpretv2sf, 0, df) + return fold_build1 (VIEW_CONVERT_EXPR, type, args[0]); VAR1 (UNOP, floatv2si, 2, v2sf) VAR1 (UNOP, floatv4si, 2, v4sf) VAR1 (UNOP, floatv2di, 2, v2df) diff --git a/gcc/config/aarch64/aarch64-simd-builtins.def b/gcc/config/aarch64/aarch64-simd-builtins.def index 4db8129edb9a..bfd2b750ba05 100644 --- a/gcc/config/aarch64/aarch64-simd-builtins.def +++ b/gcc/config/aarch64/aarch64-simd-builtins.def @@ -51,6 +51,28 @@ VAR1 (GETLANE, get_lane, 0, di) BUILTIN_VALL (GETLANE, be_checked_get_lane, 0) + VAR1 (REINTERP_SS, reinterpretdi, 0, df) + VAR1 (REINTERP_SS, reinterpretv8qi, 0, df) + VAR1 (REINTERP_SS, reinterpretv4hi, 0, df) + VAR1 (REINTERP_SS, reinterpretv2si, 0, df) + VAR1 (REINTERP_SS, reinterpretv2sf, 0, df) + BUILTIN_VD (REINTERP_SS, reinterpretdf, 0) + + BUILTIN_VD (REINTERP_SU, reinterpretdf, 0) + + VAR1 (REINTERP_US, reinterpretdi, 0, df) + VAR1 (REINTERP_US, reinterpretv8qi, 0, df) + VAR1 (REINTERP_US, reinterpretv4hi, 0, df) + VAR1 (REINTERP_US, reinterpretv2si, 0, df) + VAR1 (REINTERP_US, reinterpretv2sf, 0, df) + + BUILTIN_VD (REINTERP_SP, reinterpretdf, 0) + + VAR1 (REINTERP_PS, reinterpretdi, 0, df) + VAR1 (REINTERP_PS, reinterpretv8qi, 0, df) + VAR1 (REINTERP_PS, reinterpretv4hi, 0, df) + VAR1 (REINTERP_PS, reinterpretv2si, 0, df) + VAR1 (REINTERP_PS, reinterpretv2sf, 0, df) BUILTIN_VDQ_I (BINOP, dup_lane, 0) /* Implemented by aarch64_qshl. */ diff --git a/gcc/config/aarch64/aarch64-simd.md b/gcc/config/aarch64/aarch64-simd.md index 753ca1c9e9e6..285d8399fea5 100644 --- a/gcc/config/aarch64/aarch64-simd.md +++ b/gcc/config/aarch64/aarch64-simd.md @@ -2259,6 +2259,15 @@ DONE; }) +(define_expand "aarch64_reinterpretdf" + [(match_operand:DF 0 "register_operand" "") + (match_operand:VD_RE 1 "register_operand" "")] + "TARGET_SIMD" +{ + aarch64_simd_reinterpret (operands[0], operands[1]); + DONE; +}) + (define_expand "aarch64_reinterpretv16qi" [(match_operand:V16QI 0 "register_operand" "") (match_operand:VQ 1 "register_operand" "")] diff --git a/gcc/config/aarch64/arm_neon.h b/gcc/config/aarch64/arm_neon.h index 7caf220edc90..9f1fa98e6fb5 100644 --- a/gcc/config/aarch64/arm_neon.h +++ b/gcc/config/aarch64/arm_neon.h @@ -2654,6 +2654,12 @@ vgetq_lane_u64 (uint64x2_t __a, const int __b) /* vreinterpret */ +__extension__ static __inline poly8x8_t __attribute__ ((__always_inline__)) +vreinterpret_p8_f64 (float64x1_t __a) +{ + return __builtin_aarch64_reinterpretv8qidf_ps (__a); +} + __extension__ static __inline poly8x8_t __attribute__ ((__always_inline__)) vreinterpret_p8_s8 (int8x8_t __a) { @@ -2714,6 +2720,12 @@ vreinterpret_p8_p16 (poly16x4_t __a) return (poly8x8_t) __a; } +__extension__ static __inline poly8x16_t __attribute__ ((__always_inline__)) +vreinterpretq_p8_f64 (float64x2_t __a) +{ + return (poly8x16_t) __a; +} + __extension__ static __inline poly8x16_t __attribute__ ((__always_inline__)) vreinterpretq_p8_s8 (int8x16_t __a) { @@ -2774,6 +2786,12 @@ vreinterpretq_p8_p16 (poly16x8_t __a) return (poly8x16_t) __a; } +__extension__ static __inline poly16x4_t __attribute__ ((__always_inline__)) +vreinterpret_p16_f64 (float64x1_t __a) +{ + return __builtin_aarch64_reinterpretv4hidf_ps (__a); +} + __extension__ static __inline poly16x4_t __attribute__ ((__always_inline__)) vreinterpret_p16_s8 (int8x8_t __a) { @@ -2834,6 +2852,12 @@ vreinterpret_p16_p8 (poly8x8_t __a) return (poly16x4_t) __a; } +__extension__ static __inline poly16x8_t __attribute__ ((__always_inline__)) +vreinterpretq_p16_f64 (float64x2_t __a) +{ + return (poly16x8_t) __a; +} + __extension__ static __inline poly16x8_t __attribute__ ((__always_inline__)) vreinterpretq_p16_s8 (int8x16_t __a) { @@ -2894,6 +2918,12 @@ vreinterpretq_p16_p8 (poly8x16_t __a) return (poly16x8_t) __a; } +__extension__ static __inline float32x2_t __attribute__ ((__always_inline__)) +vreinterpret_f32_f64 (float64x1_t __a) +{ + return __builtin_aarch64_reinterpretv2sfdf (__a); +} + __extension__ static __inline float32x2_t __attribute__ ((__always_inline__)) vreinterpret_f32_s8 (int8x8_t __a) { @@ -2954,6 +2984,12 @@ vreinterpret_f32_p16 (poly16x4_t __a) return (float32x2_t) __a; } +__extension__ static __inline float32x4_t __attribute__ ((__always_inline__)) +vreinterpretq_f32_f64 (float64x2_t __a) +{ + return (float32x4_t) __a; +} + __extension__ static __inline float32x4_t __attribute__ ((__always_inline__)) vreinterpretq_f32_s8 (int8x16_t __a) { @@ -3014,6 +3050,144 @@ vreinterpretq_f32_p16 (poly16x8_t __a) return (float32x4_t) __a; } +__extension__ static __inline float64x1_t __attribute__((__always_inline__)) +vreinterpret_f64_f32 (float32x2_t __a) +{ + return __builtin_aarch64_reinterpretdfv2sf (__a); +} + +__extension__ static __inline float64x1_t __attribute__((__always_inline__)) +vreinterpret_f64_p8 (poly8x8_t __a) +{ + return __builtin_aarch64_reinterpretdfv8qi_sp (__a); +} + +__extension__ static __inline float64x1_t __attribute__((__always_inline__)) +vreinterpret_f64_p16 (poly16x4_t __a) +{ + return __builtin_aarch64_reinterpretdfv4hi_sp (__a); +} + +__extension__ static __inline float64x1_t __attribute__((__always_inline__)) +vreinterpret_f64_s8 (int8x8_t __a) +{ + return __builtin_aarch64_reinterpretdfv8qi (__a); +} + +__extension__ static __inline float64x1_t __attribute__((__always_inline__)) +vreinterpret_f64_s16 (int16x4_t __a) +{ + return __builtin_aarch64_reinterpretdfv4hi (__a); +} + +__extension__ static __inline float64x1_t __attribute__((__always_inline__)) +vreinterpret_f64_s32 (int32x2_t __a) +{ + return __builtin_aarch64_reinterpretdfv2si (__a); +} + +__extension__ static __inline float64x1_t __attribute__((__always_inline__)) +vreinterpret_f64_s64 (int64x1_t __a) +{ + return __builtin_aarch64_createdf ((uint64_t) vget_lane_s64 (__a, 0)); +} + +__extension__ static __inline float64x1_t __attribute__((__always_inline__)) +vreinterpret_f64_u8 (uint8x8_t __a) +{ + return __builtin_aarch64_reinterpretdfv8qi_su (__a); +} + +__extension__ static __inline float64x1_t __attribute__((__always_inline__)) +vreinterpret_f64_u16 (uint16x4_t __a) +{ + return __builtin_aarch64_reinterpretdfv4hi_su (__a); +} + +__extension__ static __inline float64x1_t __attribute__((__always_inline__)) +vreinterpret_f64_u32 (uint32x2_t __a) +{ + return __builtin_aarch64_reinterpretdfv2si_su (__a); +} + +__extension__ static __inline float64x1_t __attribute__((__always_inline__)) +vreinterpret_f64_u64 (uint64x1_t __a) +{ + return __builtin_aarch64_createdf (vget_lane_u64 (__a, 0)); +} + +__extension__ static __inline float64x2_t __attribute__((__always_inline__)) +vreinterpretq_f64_f32 (float32x4_t __a) +{ + return (float64x2_t) __a; +} + +__extension__ static __inline float64x2_t __attribute__((__always_inline__)) +vreinterpretq_f64_p8 (poly8x16_t __a) +{ + return (float64x2_t) __a; +} + +__extension__ static __inline float64x2_t __attribute__((__always_inline__)) +vreinterpretq_f64_p16 (poly16x8_t __a) +{ + return (float64x2_t) __a; +} + +__extension__ static __inline float64x2_t __attribute__((__always_inline__)) +vreinterpretq_f64_s8 (int8x16_t __a) +{ + return (float64x2_t) __a; +} + +__extension__ static __inline float64x2_t __attribute__((__always_inline__)) +vreinterpretq_f64_s16 (int16x8_t __a) +{ + return (float64x2_t) __a; +} + +__extension__ static __inline float64x2_t __attribute__((__always_inline__)) +vreinterpretq_f64_s32 (int32x4_t __a) +{ + return (float64x2_t) __a; +} + +__extension__ static __inline float64x2_t __attribute__((__always_inline__)) +vreinterpretq_f64_s64 (int64x2_t __a) +{ + return (float64x2_t) __a; +} + +__extension__ static __inline float64x2_t __attribute__((__always_inline__)) +vreinterpretq_f64_u8 (uint8x16_t __a) +{ + return (float64x2_t) __a; +} + +__extension__ static __inline float64x2_t __attribute__((__always_inline__)) +vreinterpretq_f64_u16 (uint16x8_t __a) +{ + return (float64x2_t) __a; +} + +__extension__ static __inline float64x2_t __attribute__((__always_inline__)) +vreinterpretq_f64_u32 (uint32x4_t __a) +{ + return (float64x2_t) __a; +} + +__extension__ static __inline float64x2_t __attribute__((__always_inline__)) +vreinterpretq_f64_u64 (uint64x2_t __a) +{ + return (float64x2_t) __a; +} + +__extension__ static __inline int64x1_t __attribute__ ((__always_inline__)) +vreinterpret_s64_f64 (float64x1_t __a) +{ + return __builtin_aarch64_reinterpretdidf (__a); +} + __extension__ static __inline int64x1_t __attribute__ ((__always_inline__)) vreinterpret_s64_s8 (int8x8_t __a) { @@ -3074,6 +3248,12 @@ vreinterpret_s64_p16 (poly16x4_t __a) return (int64x1_t) __a; } +__extension__ static __inline int64x2_t __attribute__ ((__always_inline__)) +vreinterpretq_s64_f64 (float64x2_t __a) +{ + return (int64x2_t) __a; +} + __extension__ static __inline int64x2_t __attribute__ ((__always_inline__)) vreinterpretq_s64_s8 (int8x16_t __a) { @@ -3134,6 +3314,12 @@ vreinterpretq_s64_p16 (poly16x8_t __a) return (int64x2_t) __a; } +__extension__ static __inline uint64x1_t __attribute__ ((__always_inline__)) +vreinterpret_u64_f64 (float64x1_t __a) +{ + return __builtin_aarch64_reinterpretdidf_us (__a); +} + __extension__ static __inline uint64x1_t __attribute__ ((__always_inline__)) vreinterpret_u64_s8 (int8x8_t __a) { @@ -3194,6 +3380,12 @@ vreinterpret_u64_p16 (poly16x4_t __a) return (uint64x1_t) __a; } +__extension__ static __inline uint64x2_t __attribute__ ((__always_inline__)) +vreinterpretq_u64_f64 (float64x2_t __a) +{ + return (uint64x2_t) __a; +} + __extension__ static __inline uint64x2_t __attribute__ ((__always_inline__)) vreinterpretq_u64_s8 (int8x16_t __a) { @@ -3254,6 +3446,12 @@ vreinterpretq_u64_p16 (poly16x8_t __a) return (uint64x2_t) __a; } +__extension__ static __inline int8x8_t __attribute__ ((__always_inline__)) +vreinterpret_s8_f64 (float64x1_t __a) +{ + return __builtin_aarch64_reinterpretv8qidf (__a); +} + __extension__ static __inline int8x8_t __attribute__ ((__always_inline__)) vreinterpret_s8_s16 (int16x4_t __a) { @@ -3314,6 +3512,12 @@ vreinterpret_s8_p16 (poly16x4_t __a) return (int8x8_t) __a; } +__extension__ static __inline int8x16_t __attribute__ ((__always_inline__)) +vreinterpretq_s8_f64 (float64x2_t __a) +{ + return (int8x16_t) __a; +} + __extension__ static __inline int8x16_t __attribute__ ((__always_inline__)) vreinterpretq_s8_s16 (int16x8_t __a) { @@ -3374,6 +3578,12 @@ vreinterpretq_s8_p16 (poly16x8_t __a) return (int8x16_t) __a; } +__extension__ static __inline int16x4_t __attribute__ ((__always_inline__)) +vreinterpret_s16_f64 (float64x1_t __a) +{ + return __builtin_aarch64_reinterpretv4hidf (__a); +} + __extension__ static __inline int16x4_t __attribute__ ((__always_inline__)) vreinterpret_s16_s8 (int8x8_t __a) { @@ -3434,6 +3644,12 @@ vreinterpret_s16_p16 (poly16x4_t __a) return (int16x4_t) __a; } +__extension__ static __inline int16x8_t __attribute__ ((__always_inline__)) +vreinterpretq_s16_f64 (float64x2_t __a) +{ + return (int16x8_t) __a; +} + __extension__ static __inline int16x8_t __attribute__ ((__always_inline__)) vreinterpretq_s16_s8 (int8x16_t __a) { @@ -3494,6 +3710,12 @@ vreinterpretq_s16_p16 (poly16x8_t __a) return (int16x8_t) __a; } +__extension__ static __inline int32x2_t __attribute__ ((__always_inline__)) +vreinterpret_s32_f64 (float64x1_t __a) +{ + return __builtin_aarch64_reinterpretv2sidf (__a); +} + __extension__ static __inline int32x2_t __attribute__ ((__always_inline__)) vreinterpret_s32_s8 (int8x8_t __a) { @@ -3554,6 +3776,12 @@ vreinterpret_s32_p16 (poly16x4_t __a) return (int32x2_t) __a; } +__extension__ static __inline int32x4_t __attribute__ ((__always_inline__)) +vreinterpretq_s32_f64 (float64x2_t __a) +{ + return (int32x4_t) __a; +} + __extension__ static __inline int32x4_t __attribute__ ((__always_inline__)) vreinterpretq_s32_s8 (int8x16_t __a) { @@ -3614,6 +3842,12 @@ vreinterpretq_s32_p16 (poly16x8_t __a) return (int32x4_t) __a; } +__extension__ static __inline uint8x8_t __attribute__ ((__always_inline__)) +vreinterpret_u8_f64 (float64x1_t __a) +{ + return __builtin_aarch64_reinterpretv8qidf_us (__a); +} + __extension__ static __inline uint8x8_t __attribute__ ((__always_inline__)) vreinterpret_u8_s8 (int8x8_t __a) { @@ -3674,6 +3908,12 @@ vreinterpret_u8_p16 (poly16x4_t __a) return (uint8x8_t) __a; } +__extension__ static __inline uint8x16_t __attribute__ ((__always_inline__)) +vreinterpretq_u8_f64 (float64x2_t __a) +{ + return (uint8x16_t) __a; +} + __extension__ static __inline uint8x16_t __attribute__ ((__always_inline__)) vreinterpretq_u8_s8 (int8x16_t __a) { @@ -3734,6 +3974,12 @@ vreinterpretq_u8_p16 (poly16x8_t __a) return (uint8x16_t) __a; } +__extension__ static __inline uint16x4_t __attribute__ ((__always_inline__)) +vreinterpret_u16_f64 (float64x1_t __a) +{ + return __builtin_aarch64_reinterpretv4hidf_us (__a); +} + __extension__ static __inline uint16x4_t __attribute__ ((__always_inline__)) vreinterpret_u16_s8 (int8x8_t __a) { @@ -3794,6 +4040,12 @@ vreinterpret_u16_p16 (poly16x4_t __a) return (uint16x4_t) __a; } +__extension__ static __inline uint16x8_t __attribute__ ((__always_inline__)) +vreinterpretq_u16_f64 (float64x2_t __a) +{ + return (uint16x8_t) __a; +} + __extension__ static __inline uint16x8_t __attribute__ ((__always_inline__)) vreinterpretq_u16_s8 (int8x16_t __a) { @@ -3854,6 +4106,12 @@ vreinterpretq_u16_p16 (poly16x8_t __a) return (uint16x8_t) __a; } +__extension__ static __inline uint32x2_t __attribute__ ((__always_inline__)) +vreinterpret_u32_f64 (float64x1_t __a) +{ + return __builtin_aarch64_reinterpretv2sidf_us (__a); +} + __extension__ static __inline uint32x2_t __attribute__ ((__always_inline__)) vreinterpret_u32_s8 (int8x8_t __a) { @@ -3914,6 +4172,12 @@ vreinterpret_u32_p16 (poly16x4_t __a) return (uint32x2_t) __a; } +__extension__ static __inline uint32x4_t __attribute__ ((__always_inline__)) +vreinterpretq_u32_f64 (float64x2_t __a) +{ + return (uint32x4_t) __a; +} + __extension__ static __inline uint32x4_t __attribute__ ((__always_inline__)) vreinterpretq_u32_s8 (int8x16_t __a) { diff --git a/gcc/testsuite/ChangeLog b/gcc/testsuite/ChangeLog index 908646d6823a..2dced65048de 100644 --- a/gcc/testsuite/ChangeLog +++ b/gcc/testsuite/ChangeLog @@ -1,3 +1,7 @@ +2014-04-22 Alex Velenko + + * gcc.target/aarch64/vreinterpret_f64_1.c: New. + 2014-04-22 Alex Velenko * gcc.target/aarch64/vqneg_s64_1.c: New testcase. diff --git a/gcc/testsuite/gcc.target/aarch64/vreinterpret_f64_1.c b/gcc/testsuite/gcc.target/aarch64/vreinterpret_f64_1.c new file mode 100644 index 000000000000..ca98b9f63c1c --- /dev/null +++ b/gcc/testsuite/gcc.target/aarch64/vreinterpret_f64_1.c @@ -0,0 +1,596 @@ +/* Test vreinterpret_f64_* and vreinterpret_*_f64 intrinsics work correctly. */ +/* { dg-do run } */ +/* { dg-options "-O3" } */ + +#include + +extern void abort (void); + +#define ABS(a) __builtin_fabs (a) +#define ISNAN(a) __builtin_isnan (a) + +#define DOUBLE_EQUALS(a, b, epsilon) \ +( \ + ((a) == (b)) \ + || (ISNAN (a) && ISNAN (b)) \ + || (ABS (a - b) < epsilon) \ +) + +/* Pi accurate up to 16 digits. + Further digits are a closest binary approximation. */ +#define PI_F64 3.14159265358979311599796346854 +/* Hex representation in Double (IEEE754 Double precision 64-bit) is: + 0x400921FB54442D18. */ + +/* E accurate up to 16 digits. + Further digits are a closest binary approximation. */ +#define E_F64 2.71828182845904509079559829843 +/* Hex representation in Double (IEEE754 Double precision 64-bit) is: + 0x4005BF0A8B145769. */ + +float32x2_t __attribute__ ((noinline)) +wrap_vreinterpret_f32_f64 (float64x1_t __a) +{ + return vreinterpret_f32_f64 (__a); +} + +int __attribute__ ((noinline)) +test_vreinterpret_f32_f64 () +{ + float64x1_t a; + float32x2_t b; + float64_t c[1] = { PI_F64 }; + /* Values { 0x54442D18, 0x400921FB } reinterpreted as f32. */ + float32_t d[2] = { 3.3702805504E12, 2.1426990032196044921875E0 }; + float32_t e[2]; + int i; + + a = vld1_f64 (c); + b = wrap_vreinterpret_f32_f64 (a); + vst1_f32 (e, b); + for (i = 0; i < 2; i++) + if (!DOUBLE_EQUALS (d[i], e[i], __FLT_EPSILON__)) + return 1; + return 0; +}; + +int8x8_t __attribute__ ((noinline)) +wrap_vreinterpret_s8_f64 (float64x1_t __a) +{ + return vreinterpret_s8_f64 (__a); +} + +int __attribute__ ((noinline)) +test_vreinterpret_s8_f64 () +{ + float64x1_t a; + int8x8_t b; + float64_t c[1] = { PI_F64 }; + int8_t d[8] = { 0x18, 0x2D, 0x44, 0x54, 0xFB, 0x21, 0x09, 0x40 }; + int8_t e[8]; + int i; + + a = vld1_f64 (c); + b = wrap_vreinterpret_s8_f64 (a); + vst1_s8 (e, b); + for (i = 0; i < 8; i++) + if (d[i] != e[i]) + return 1; + return 0; +}; + +int16x4_t __attribute__ ((noinline)) +wrap_vreinterpret_s16_f64 (float64x1_t __a) +{ + return vreinterpret_s16_f64 (__a); +} + +int __attribute__ ((noinline)) +test_vreinterpret_s16_f64 () +{ + float64x1_t a; + int16x4_t b; + float64_t c[1] = { PI_F64 }; + int16_t d[4] = { 0x2D18, 0x5444, 0x21FB, 0x4009 }; + int16_t e[4]; + int i; + + a = vld1_f64 (c); + b = wrap_vreinterpret_s16_f64 (a); + vst1_s16 (e, b); + for (i = 0; i < 4; i++) + if (d[i] != e[i]) + return 1; + return 0; +}; + +int32x2_t __attribute__ ((noinline)) +wrap_vreinterpret_s32_f64 (float64x1_t __a) +{ + return vreinterpret_s32_f64 (__a); +} + +int __attribute__ ((noinline)) +test_vreinterpret_s32_f64 () +{ + float64x1_t a; + int32x2_t b; + float64_t c[1] = { PI_F64 }; + int32_t d[2] = { 0x54442D18, 0x400921FB }; + int32_t e[2]; + int i; + + a = vld1_f64 (c); + b = wrap_vreinterpret_s32_f64 (a); + vst1_s32 (e, b); + for (i = 0; i < 2; i++) + if (d[i] != e[i]) + return 1; + return 0; +}; + +int64x1_t __attribute__ ((noinline)) +wrap_vreinterpret_s64_f64 (float64x1_t __a) +{ + return vreinterpret_s64_f64 (__a); +} + +int __attribute__ ((noinline)) +test_vreinterpret_s64_f64 () +{ + float64x1_t a; + int64x1_t b; + float64_t c[1] = { PI_F64 }; + int64_t d[1] = { 0x400921FB54442D18 }; + int64_t e[1]; + int i; + + a = vld1_f64 (c); + b = wrap_vreinterpret_s64_f64 (a); + vst1_s64 (e, b); + if (d[0] != e[0]) + return 1; + return 0; +}; + +float32x4_t __attribute__ ((noinline)) +wrap_vreinterpretq_f32_f64 (float64x2_t __a) +{ + return vreinterpretq_f32_f64 (__a); +} + +int __attribute__ ((noinline)) +test_vreinterpretq_f32_f64 () +{ + float64x2_t a; + float32x4_t b; + float64_t c[2] = { PI_F64, E_F64 }; + + /* Values corresponding to f32 reinterpret of + { 0x54442D18, 0x400921FB, 0x8B145769, 0x4005BF0A }. */ + float32_t d[4] = { 3.3702805504E12, + 2.1426990032196044921875E0, + -2.8569523269651966444143014594E-32, + 2.089785099029541015625E0 }; + float32_t e[4]; + int i; + + a = vld1q_f64 (c); + b = wrap_vreinterpretq_f32_f64 (a); + vst1q_f32 (e, b); + for (i = 0; i < 4; i++) + { + if (!DOUBLE_EQUALS (d[i], e[i], __FLT_EPSILON__)) + return 1; + } + return 0; +}; + +int8x16_t __attribute__ ((noinline)) +wrap_vreinterpretq_s8_f64 (float64x2_t __a) +{ + return vreinterpretq_s8_f64 (__a); +} + +int __attribute__ ((noinline)) +test_vreinterpretq_s8_f64 () +{ + float64x2_t a; + int8x16_t b; + float64_t c[2] = { PI_F64, E_F64 }; + int8_t d[16] = { 0x18, 0x2D, 0x44, 0x54, 0xFB, 0x21, 0x09, 0x40, + 0x69, 0x57, 0x14, 0x8B, 0x0A, 0xBF, 0x05, 0x40 }; + int8_t e[16]; + int i; + + a = vld1q_f64 (c); + b = wrap_vreinterpretq_s8_f64 (a); + vst1q_s8 (e, b); + for (i = 0; i < 16; i++) + if (d[i] != e[i]) + return 1; + return 0; +}; + +int16x8_t __attribute__ ((noinline)) +wrap_vreinterpretq_s16_f64 (float64x2_t __a) +{ + return vreinterpretq_s16_f64 (__a); +} + +int __attribute__ ((noinline)) +test_vreinterpretq_s16_f64 () +{ + float64x2_t a; + int16x8_t b; + float64_t c[2] = { PI_F64, E_F64 }; + int16_t d[8] = { 0x2D18, 0x5444, 0x21FB, 0x4009, + 0x5769, 0x8B14, 0xBF0A, 0x4005 }; + int16_t e[8]; + int i; + + a = vld1q_f64 (c); + b = wrap_vreinterpretq_s16_f64 (a); + vst1q_s16 (e, b); + for (i = 0; i < 8; i++) + if (d[i] != e[i]) + return 1; + return 0; +}; + +int32x4_t __attribute__ ((noinline)) +wrap_vreinterpretq_s32_f64 (float64x2_t __a) +{ + return vreinterpretq_s32_f64 (__a); +} + +int __attribute__ ((noinline)) +test_vreinterpretq_s32_f64 () +{ + float64x2_t a; + int32x4_t b; + float64_t c[2] = { PI_F64, E_F64 }; + int32_t d[4] = { 0x54442D18, 0x400921FB, 0x8B145769, 0x4005BF0A }; + int32_t e[4]; + int i; + + a = vld1q_f64 (c); + b = wrap_vreinterpretq_s32_f64 (a); + vst1q_s32 (e, b); + for (i = 0; i < 4; i++) + if (d[i] != e[i]) + return 1; + return 0; +}; + +int64x2_t __attribute__ ((noinline)) +wrap_vreinterpretq_s64_f64 (float64x2_t __a) +{ + return vreinterpretq_s64_f64 (__a); +} + +int __attribute__ ((noinline)) +test_vreinterpretq_s64_f64 () +{ + float64x2_t a; + int64x2_t b; + float64_t c[2] = { PI_F64, E_F64 }; + int64_t d[2] = { 0x400921FB54442D18, 0x4005BF0A8B145769 }; + int64_t e[2]; + int i; + + a = vld1q_f64 (c); + b = wrap_vreinterpretq_s64_f64 (a); + vst1q_s64 (e, b); + for (i = 0; i < 2; i++) + if (d[i] != e[i]) + return 1; + return 0; +}; + +float64x1_t __attribute__ ((noinline)) +wrap_vreinterpret_f64_f32 (float32x2_t __a) +{ + return vreinterpret_f64_f32 (__a); +} + +int __attribute__ ((noinline)) +test_vreinterpret_f64_f32 () +{ + float32x2_t a; + float64x1_t b; + /* Values { 0x54442D18, 0x400921FB } reinterpreted as f32. */ + float32_t c[2] = { 3.3702805504E12, 2.1426990032196044921875E0 }; + float64_t d[1] = { PI_F64 }; + float64_t e[1]; + int i; + + a = vld1_f32 (c); + b = wrap_vreinterpret_f64_f32 (a); + vst1_f64 (e, b); + if (!DOUBLE_EQUALS (d[0], e[0], __DBL_EPSILON__)) + return 1; + return 0; +}; + +float64x1_t __attribute__ ((noinline)) +wrap_vreinterpret_f64_s8 (int8x8_t __a) +{ + return vreinterpret_f64_s8 (__a); +} + +int __attribute__ ((noinline)) +test_vreinterpret_f64_s8 () +{ + int8x8_t a; + float64x1_t b; + int8_t c[8] = { 0x18, 0x2D, 0x44, 0x54, 0xFB, 0x21, 0x09, 0x40 }; + float64_t d[1] = { PI_F64 }; + float64_t e[1]; + int i; + + a = vld1_s8 (c); + b = wrap_vreinterpret_f64_s8 (a); + vst1_f64 (e, b); + if (!DOUBLE_EQUALS (d[0], e[0], __DBL_EPSILON__)) + return 1; + return 0; +}; + +float64x1_t __attribute__ ((noinline)) +wrap_vreinterpret_f64_s16 (int16x4_t __a) +{ + return vreinterpret_f64_s16 (__a); +} + +int __attribute__ ((noinline)) +test_vreinterpret_f64_s16 () +{ + int16x4_t a; + float64x1_t b; + int16_t c[4] = { 0x2D18, 0x5444, 0x21FB, 0x4009 }; + float64_t d[1] = { PI_F64 }; + float64_t e[1]; + int i; + + a = vld1_s16 (c); + b = wrap_vreinterpret_f64_s16 (a); + vst1_f64 (e, b); + if (!DOUBLE_EQUALS (d[0], e[0], __DBL_EPSILON__)) + return 1; + return 0; +}; + +float64x1_t __attribute__ ((noinline)) +wrap_vreinterpret_f64_s32 (int32x2_t __a) +{ + return vreinterpret_f64_s32 (__a); +} + +int __attribute__ ((noinline)) +test_vreinterpret_f64_s32 () +{ + int32x2_t a; + float64x1_t b; + int32_t c[2] = { 0x54442D18, 0x400921FB }; + float64_t d[1] = { PI_F64 }; + float64_t e[1]; + int i; + + a = vld1_s32 (c); + b = wrap_vreinterpret_f64_s32 (a); + vst1_f64 (e, b); + if (!DOUBLE_EQUALS (d[0], e[0], __DBL_EPSILON__)) + return 1; + return 0; +}; + +float64x1_t __attribute__ ((noinline)) +wrap_vreinterpret_f64_s64 (int64x1_t __a) +{ + return vreinterpret_f64_s64 (__a); +} + +int __attribute__ ((noinline)) +test_vreinterpret_f64_s64 () +{ + int64x1_t a; + float64x1_t b; + int64_t c[1] = { 0x400921FB54442D18 }; + float64_t d[1] = { PI_F64 }; + float64_t e[1]; + + a = vld1_s64 (c); + b = wrap_vreinterpret_f64_s64 (a); + vst1_f64 (e, b); + if (!DOUBLE_EQUALS (d[0], e[0], __DBL_EPSILON__)) + return 1; + return 0; +}; + +float64x2_t __attribute__ ((noinline)) +wrap_vreinterpretq_f64_f32 (float32x4_t __a) +{ + return vreinterpretq_f64_f32 (__a); +} + +int __attribute__ ((noinline)) +test_vreinterpretq_f64_f32 () +{ + float32x4_t a; + float64x2_t b; + /* Values corresponding to f32 reinterpret of + { 0x54442D18, 0x400921FB, 0x8B145769, 0x4005BF0A }. */ + float32_t c[4] = { 3.3702805504E12, + 2.1426990032196044921875E0, + -2.8569523269651966444143014594E-32, + 2.089785099029541015625E0 }; + + float64_t d[2] = { PI_F64, E_F64 }; + float64_t e[2]; + int i; + + a = vld1q_f32 (c); + b = wrap_vreinterpretq_f64_f32 (a); + vst1q_f64 (e, b); + for (i = 0; i < 2; i++) + if (!DOUBLE_EQUALS (d[i], e[i], __DBL_EPSILON__)) + return 1; + return 0; +}; + +float64x2_t __attribute__ ((noinline)) +wrap_vreinterpretq_f64_s8 (int8x16_t __a) +{ + return vreinterpretq_f64_s8 (__a); +} + +int __attribute__ ((noinline)) +test_vreinterpretq_f64_s8 () +{ + int8x16_t a; + float64x2_t b; + int8_t c[16] = { 0x18, 0x2D, 0x44, 0x54, 0xFB, 0x21, 0x09, 0x40, + 0x69, 0x57, 0x14, 0x8B, 0x0A, 0xBF, 0x05, 0x40 }; + float64_t d[2] = { PI_F64, E_F64 }; + float64_t e[2]; + int i; + + a = vld1q_s8 (c); + b = wrap_vreinterpretq_f64_s8 (a); + vst1q_f64 (e, b); + for (i = 0; i < 2; i++) + if (!DOUBLE_EQUALS (d[i], e[i], __DBL_EPSILON__)) + return 1; + return 0; +}; + +float64x2_t __attribute__ ((noinline)) +wrap_vreinterpretq_f64_s16 (int16x8_t __a) +{ + return vreinterpretq_f64_s16 (__a); +} + +int __attribute__ ((noinline)) +test_vreinterpretq_f64_s16 () +{ + int16x8_t a; + float64x2_t b; + int16_t c[8] = { 0x2D18, 0x5444, 0x21FB, 0x4009, + 0x5769, 0x8B14, 0xBF0A, 0x4005 }; + float64_t d[2] = { PI_F64, E_F64 }; + float64_t e[2]; + int i; + + a = vld1q_s16 (c); + b = wrap_vreinterpretq_f64_s16 (a); + vst1q_f64 (e, b); + for (i = 0; i < 2; i++) + if (!DOUBLE_EQUALS (d[i], e[i], __DBL_EPSILON__)) + return 1; + return 0; +}; + +float64x2_t __attribute__ ((noinline)) +wrap_vreinterpretq_f64_s32 (int32x4_t __a) +{ + return vreinterpretq_f64_s32 (__a); +} + +int __attribute__ ((noinline)) +test_vreinterpretq_f64_s32 () +{ + int32x4_t a; + float64x2_t b; + int32_t c[4] = { 0x54442D18, 0x400921FB, 0x8B145769, 0x4005BF0A }; + float64_t d[2] = { PI_F64, E_F64 }; + float64_t e[2]; + int i; + + a = vld1q_s32 (c); + b = wrap_vreinterpretq_f64_s32 (a); + vst1q_f64 (e, b); + for (i = 0; i < 2; i++) + if (!DOUBLE_EQUALS (d[i], e[i], __DBL_EPSILON__)) + return 1; + return 0; +}; + +float64x2_t __attribute__ ((noinline)) +wrap_vreinterpretq_f64_s64 (int64x2_t __a) +{ + return vreinterpretq_f64_s64 (__a); +} + +int __attribute__ ((noinline)) +test_vreinterpretq_f64_s64 () +{ + int64x2_t a; + float64x2_t b; + int64_t c[2] = { 0x400921FB54442D18, 0x4005BF0A8B145769 }; + float64_t d[2] = { PI_F64, E_F64 }; + float64_t e[2]; + int i; + + a = vld1q_s64 (c); + b = wrap_vreinterpretq_f64_s64 (a); + vst1q_f64 (e, b); + for (i = 0; i < 2; i++) + if (!DOUBLE_EQUALS (d[i], e[i], __DBL_EPSILON__)) + return 1; + return 0; +}; + +int +main (int argc, char **argv) +{ + if (test_vreinterpret_f32_f64 ()) + abort (); + + if (test_vreinterpret_s8_f64 ()) + abort (); + if (test_vreinterpret_s16_f64 ()) + abort (); + if (test_vreinterpret_s32_f64 ()) + abort (); + if (test_vreinterpret_s64_f64 ()) + abort (); + + if (test_vreinterpretq_f32_f64 ()) + abort (); + + if (test_vreinterpretq_s8_f64 ()) + abort (); + if (test_vreinterpretq_s16_f64 ()) + abort (); + if (test_vreinterpretq_s32_f64 ()) + abort (); + if (test_vreinterpretq_s64_f64 ()) + abort (); + + if (test_vreinterpret_f64_f32 ()) + abort (); + + if (test_vreinterpret_f64_s8 ()) + abort (); + if (test_vreinterpret_f64_s16 ()) + abort (); + if (test_vreinterpret_f64_s32 ()) + abort (); + if (test_vreinterpret_f64_s64 ()) + abort (); + + if (test_vreinterpretq_f64_f32 ()) + abort (); + + if (test_vreinterpretq_f64_s8 ()) + abort (); + if (test_vreinterpretq_f64_s16 ()) + abort (); + if (test_vreinterpretq_f64_s32 ()) + abort (); + if (test_vreinterpretq_f64_s64 ()) + abort (); + + return 0; +}