1 /* Copyright (C) 2007, 2008, 2009 Free Software Foundation, Inc. 2 3 This file is part of GCC. 4 5 GCC is free software; you can redistribute it and/or modify 6 it under the terms of the GNU General Public License as published by 7 the Free Software Foundation; either version 3, or (at your option) 8 any later version. 9 10 GCC is distributed in the hope that it will be useful, 11 but WITHOUT ANY WARRANTY; without even the implied warranty of 12 MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the 13 GNU General Public License for more details. 14 15 Under Section 7 of GPL version 3, you are granted additional 16 permissions described in the GCC Runtime Library Exception, version 17 3.1, as published by the Free Software Foundation. 18 19 You should have received a copy of the GNU General Public License and 20 a copy of the GCC Runtime Library Exception along with this program; 21 see the files COPYING3 and COPYING.RUNTIME respectively. If not, see 22 <http://www.gnu.org/licenses/>. */ 23 24 /* Common definition of the ROUND and PTEST intrinsics that are shared 25 between SSE4.1 and SSE5. */ 26 27 #ifndef _MMINTRIN_COMMON_H_INCLUDED 28 #define _MMINTRIN_COMMON_H_INCLUDED 29 30 #if !defined(__SSE5__) && !defined(__SSE4_1__) 31 # error "SSE5 or SSE4.1 instruction set not enabled" 32 #else 33 34 /* Rounding mode macros. */ 35 #define _MM_FROUND_TO_NEAREST_INT 0x00 36 #define _MM_FROUND_TO_NEG_INF 0x01 37 #define _MM_FROUND_TO_POS_INF 0x02 38 #define _MM_FROUND_TO_ZERO 0x03 39 #define _MM_FROUND_CUR_DIRECTION 0x04 40 41 #define _MM_FROUND_RAISE_EXC 0x00 42 #define _MM_FROUND_NO_EXC 0x08 43 44 #define _MM_FROUND_NINT \ 45 (_MM_FROUND_TO_NEAREST_INT | _MM_FROUND_RAISE_EXC) 46 #define _MM_FROUND_FLOOR \ 47 (_MM_FROUND_TO_NEG_INF | _MM_FROUND_RAISE_EXC) 48 #define _MM_FROUND_CEIL \ 49 (_MM_FROUND_TO_POS_INF | _MM_FROUND_RAISE_EXC) 50 #define _MM_FROUND_TRUNC \ 51 (_MM_FROUND_TO_ZERO | _MM_FROUND_RAISE_EXC) 52 #define _MM_FROUND_RINT \ 53 (_MM_FROUND_CUR_DIRECTION | _MM_FROUND_RAISE_EXC) 54 #define _MM_FROUND_NEARBYINT \ 55 (_MM_FROUND_CUR_DIRECTION | _MM_FROUND_NO_EXC) 56 57 /* Test Instruction */ 58 /* Packed integer 128-bit bitwise comparison. Return 1 if 59 (__V & __M) == 0. */ 60 extern __inline int __attribute__((__gnu_inline__, __always_inline__, __artificial__)) 61 _mm_testz_si128 (__m128i __M, __m128i __V) 62 { 63 return __builtin_ia32_ptestz128 ((__v2di)__M, (__v2di)__V); 64 } 65 66 /* Packed integer 128-bit bitwise comparison. Return 1 if 67 (__V & ~__M) == 0. */ 68 extern __inline int __attribute__((__gnu_inline__, __always_inline__, __artificial__)) 69 _mm_testc_si128 (__m128i __M, __m128i __V) 70 { 71 return __builtin_ia32_ptestc128 ((__v2di)__M, (__v2di)__V); 72 } 73 74 /* Packed integer 128-bit bitwise comparison. Return 1 if 75 (__V & __M) != 0 && (__V & ~__M) != 0. */ 76 extern __inline int __attribute__((__gnu_inline__, __always_inline__, __artificial__)) 77 _mm_testnzc_si128 (__m128i __M, __m128i __V) 78 { 79 return __builtin_ia32_ptestnzc128 ((__v2di)__M, (__v2di)__V); 80 } 81 82 /* Macros for packed integer 128-bit comparison intrinsics. */ 83 #define _mm_test_all_zeros(M, V) _mm_testz_si128 ((M), (V)) 84 85 #define _mm_test_all_ones(V) \ 86 _mm_testc_si128 ((V), _mm_cmpeq_epi32 ((V), (V))) 87 88 #define _mm_test_mix_ones_zeros(M, V) _mm_testnzc_si128 ((M), (V)) 89 90 /* Packed/scalar double precision floating point rounding. */ 91 92 #ifdef __OPTIMIZE__ 93 extern __inline __m128d __attribute__((__gnu_inline__, __always_inline__, __artificial__)) 94 _mm_round_pd (__m128d __V, const int __M) 95 { 96 return (__m128d) __builtin_ia32_roundpd ((__v2df)__V, __M); 97 } 98 99 extern __inline __m128d __attribute__((__gnu_inline__, __always_inline__, __artificial__)) 100 _mm_round_sd(__m128d __D, __m128d __V, const int __M) 101 { 102 return (__m128d) __builtin_ia32_roundsd ((__v2df)__D, 103 (__v2df)__V, 104 __M); 105 } 106 #else 107 #define _mm_round_pd(V, M) \ 108 ((__m128d) __builtin_ia32_roundpd ((__v2df)(__m128d)(V), (int)(M))) 109 110 #define _mm_round_sd(D, V, M) \ 111 ((__m128d) __builtin_ia32_roundsd ((__v2df)(__m128d)(D), \ 112 (__v2df)(__m128d)(V), (int)(M))) 113 #endif 114 115 /* Packed/scalar single precision floating point rounding. */ 116 117 #ifdef __OPTIMIZE__ 118 extern __inline __m128 __attribute__((__gnu_inline__, __always_inline__, __artificial__)) 119 _mm_round_ps (__m128 __V, const int __M) 120 { 121 return (__m128) __builtin_ia32_roundps ((__v4sf)__V, __M); 122 } 123 124 extern __inline __m128 __attribute__((__gnu_inline__, __always_inline__, __artificial__)) 125 _mm_round_ss (__m128 __D, __m128 __V, const int __M) 126 { 127 return (__m128) __builtin_ia32_roundss ((__v4sf)__D, 128 (__v4sf)__V, 129 __M); 130 } 131 #else 132 #define _mm_round_ps(V, M) \ 133 ((__m128) __builtin_ia32_roundps ((__v4sf)(__m128)(V), (int)(M))) 134 135 #define _mm_round_ss(D, V, M) \ 136 ((__m128) __builtin_ia32_roundss ((__v4sf)(__m128)(D), \ 137 (__v4sf)(__m128)(V), (int)(M))) 138 #endif 139 140 /* Macros for ceil/floor intrinsics. */ 141 #define _mm_ceil_pd(V) _mm_round_pd ((V), _MM_FROUND_CEIL) 142 #define _mm_ceil_sd(D, V) _mm_round_sd ((D), (V), _MM_FROUND_CEIL) 143 144 #define _mm_floor_pd(V) _mm_round_pd((V), _MM_FROUND_FLOOR) 145 #define _mm_floor_sd(D, V) _mm_round_sd ((D), (V), _MM_FROUND_FLOOR) 146 147 #define _mm_ceil_ps(V) _mm_round_ps ((V), _MM_FROUND_CEIL) 148 #define _mm_ceil_ss(D, V) _mm_round_ss ((D), (V), _MM_FROUND_CEIL) 149 150 #define _mm_floor_ps(V) _mm_round_ps ((V), _MM_FROUND_FLOOR) 151 #define _mm_floor_ss(D, V) _mm_round_ss ((D), (V), _MM_FROUND_FLOOR) 152 153 #endif /* __SSE5__/__SSE4_1__ */ 154 155 #endif /* _MMINTRIN_COMMON_H_INCLUDED */ 156
