vector.h revision 83049339
1/*
2 * Copyright (c) 2015 Cisco and/or its affiliates.
3 * Licensed under the Apache License, Version 2.0 (the "License");
4 * you may not use this file except in compliance with the License.
5 * You may obtain a copy of the License at:
6 *
7 *     http://www.apache.org/licenses/LICENSE-2.0
8 *
9 * Unless required by applicable law or agreed to in writing, software
10 * distributed under the License is distributed on an "AS IS" BASIS,
11 * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
12 * See the License for the specific language governing permissions and
13 * limitations under the License.
14 */
15/*
16  Copyright (c) 2005 Eliot Dresselhaus
17
18  Permission is hereby granted, free of charge, to any person obtaining
19  a copy of this software and associated documentation files (the
20  "Software"), to deal in the Software without restriction, including
21  without limitation the rights to use, copy, modify, merge, publish,
22  distribute, sublicense, and/or sell copies of the Software, and to
23  permit persons to whom the Software is furnished to do so, subject to
24  the following conditions:
25
26  The above copyright notice and this permission notice shall be
27  included in all copies or substantial portions of the Software.
28
29  THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
30  EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
31  MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
32  NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE
33  LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION
34  OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION
35  WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
36*/
37
38#ifndef included_clib_vector_h
39#define included_clib_vector_h
40
41#include <vppinfra/clib.h>
42
43/* Vector types. */
44
45#if defined (__MMX__) || defined (__IWMMXT__) || defined (__aarch64__) \
46  || defined (__i686__)
47#define CLIB_HAVE_VEC64
48#endif
49
50#if defined (__aarch64__) && defined(__ARM_NEON) || defined (__i686__)
51#define CLIB_HAVE_VEC128
52#endif
53
54#if defined (__SSE4_2__) && __GNUC__ >= 4
55#define CLIB_HAVE_VEC128
56#endif
57
58#if defined (__ALTIVEC__)
59#define CLIB_HAVE_VEC128
60#endif
61
62#if defined (__AVX2__)
63#define CLIB_HAVE_VEC256
64#if defined (__clang__)  && __clang_major__ < 4
65#undef CLIB_HAVE_VEC256
66#endif
67#endif
68
69#if defined (__AVX512F__)
70#define CLIB_HAVE_VEC512
71#endif
72
73/* 128 implies 64 */
74#ifdef CLIB_HAVE_VEC128
75#define CLIB_HAVE_VEC64
76#endif
77
78#define _vector_size(n) __attribute__ ((vector_size (n)))
79#define _vector_size_unaligned(n) __attribute__ ((vector_size (n),  __aligned__ (1)))
80
81#define foreach_vec64i  _(i,8,8)  _(i,16,4)  _(i,32,2)
82#define foreach_vec64u  _(u,8,8)  _(u,16,4)  _(u,32,2)
83#define foreach_vec64f  _(f,32,2)
84#define foreach_vec128i _(i,8,16) _(i,16,8)  _(i,32,4)  _(i,64,2)
85#define foreach_vec128u _(u,8,16) _(u,16,8)  _(u,32,4)  _(u,64,2)
86#define foreach_vec128f _(f,32,4) _(f,64,2)
87#define foreach_vec256i _(i,8,32) _(i,16,16) _(i,32,8)  _(i,64,4)
88#define foreach_vec256u _(u,8,32) _(u,16,16) _(u,32,8)  _(u,64,4)
89#define foreach_vec256f _(f,32,8) _(f,64,4)
90#define foreach_vec512i _(i,8,64) _(i,16,32) _(i,32,16) _(i,64,8)
91#define foreach_vec512u _(u,8,64) _(u,16,32) _(u,32,16) _(u,64,8)
92#define foreach_vec512f _(f,32,16) _(f,64,8)
93
94#if defined (CLIB_HAVE_VEC512)
95#define foreach_int_vec foreach_vec64i foreach_vec128i foreach_vec256i foreach_vec512i
96#define foreach_uint_vec foreach_vec64u foreach_vec128u foreach_vec256u foreach_vec512u
97#define foreach_float_vec foreach_vec64f foreach_vec128f foreach_vec256f foreach_vec512f
98#elif defined (CLIB_HAVE_VEC256)
99#define foreach_int_vec foreach_vec64i foreach_vec128i foreach_vec256i
100#define foreach_uint_vec foreach_vec64u foreach_vec128u foreach_vec256u
101#define foreach_float_vec foreach_vec64f foreach_vec128f foreach_vec256f
102#else
103#define foreach_int_vec foreach_vec64i foreach_vec128i
104#define foreach_uint_vec foreach_vec64u foreach_vec128u
105#define foreach_float_vec foreach_vec64f foreach_vec128f
106#endif
107
108#define foreach_vec foreach_int_vec foreach_uint_vec foreach_float_vec
109
110/* *INDENT-OFF* */
111
112/* Type Definitions */
113#define _(t,s,c) \
114typedef t##s t##s##x##c _vector_size (s/8*c);	\
115typedef t##s t##s##x##c##u _vector_size_unaligned (s/8*c);	\
116typedef union {	  \
117  t##s##x##c as_##t##s##x##c;	\
118  t##s as_##t##s[c];	  \
119} t##s##x##c##_union_t;
120
121  foreach_vec64i foreach_vec64u foreach_vec64f
122  foreach_vec128i foreach_vec128u foreach_vec128f
123  foreach_vec256i foreach_vec256u foreach_vec256f
124  foreach_vec512i foreach_vec512u foreach_vec512f
125#undef _
126
127/* Vector word sized types. */
128#ifndef CLIB_VECTOR_WORD_BITS
129#ifdef CLIB_HAVE_VEC128
130#define CLIB_VECTOR_WORD_BITS 128
131#else
132#define CLIB_VECTOR_WORD_BITS 64
133#endif
134#endif /* CLIB_VECTOR_WORD_BITS */
135
136/* Vector word sized types. */
137#if CLIB_VECTOR_WORD_BITS == 128
138typedef i8 i8x _vector_size (16);
139typedef i16 i16x _vector_size (16);
140typedef i32 i32x _vector_size (16);
141typedef i64 i64x _vector_size (16);
142typedef u8 u8x _vector_size (16);
143typedef u16 u16x _vector_size (16);
144typedef u32 u32x _vector_size (16);
145typedef u64 u64x _vector_size (16);
146#endif
147#if CLIB_VECTOR_WORD_BITS == 64
148typedef i8 i8x _vector_size (8);
149typedef i16 i16x _vector_size (8);
150typedef i32 i32x _vector_size (8);
151typedef i64 i64x _vector_size (8);
152typedef u8 u8x _vector_size (8);
153typedef u16 u16x _vector_size (8);
154typedef u32 u32x _vector_size (8);
155typedef u64 u64x _vector_size (8);
156#endif
157
158#undef _vector_size
159
160#define VECTOR_WORD_TYPE(t) t##x
161#define VECTOR_WORD_TYPE_LEN(t) (sizeof (VECTOR_WORD_TYPE(t)) / sizeof (t))
162
163#if defined (__SSE4_2__) && __GNUC__ >= 4
164#include <vppinfra/vector_sse42.h>
165#endif
166
167#if defined (__AVX2__)
168#include <vppinfra/vector_avx2.h>
169#endif
170
171#if defined (__AVX512F__)
172#include <vppinfra/vector_avx512.h>
173#endif
174
175#if defined (__ALTIVEC__)
176#include <vppinfra/vector_altivec.h>
177#endif
178
179#if defined (__aarch64__)
180#include <vppinfra/vector_neon.h>
181#endif
182
183#if (defined(CLIB_HAVE_VEC128) || defined(CLIB_HAVE_VEC64))
184#include <vppinfra/vector_funcs.h>
185#endif
186
187/* this macro generate _splat inline functions for each scalar vector type */
188#ifndef CLIB_VEC128_SPLAT_DEFINED
189#define _(t, s, c) \
190  static_always_inline t##s##x##c			\
191t##s##x##c##_splat (t##s x)				\
192{							\
193    t##s##x##c r;					\
194    int i;						\
195							\
196    for (i = 0; i < c; i++)				\
197      r[i] = x;						\
198							\
199    return r;						\
200}
201  foreach_vec128i foreach_vec128u
202#undef _
203#endif
204
205/* *INDENT-ON* */
206
207#endif /* included_clib_vector_h */
208/*
209 * fd.io coding-style-patch-verification: ON
210 *
211 * Local Variables:
212 * eval: (c-set-style "gnu")
213 * End:
214 */
215