FD.io VPP  v18.04-17-g3a0d853
Vector Packet Processing
string.h
Go to the documentation of this file.
1 /*
2  * Copyright (c) 2016 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) 2001, 2002, 2003 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_string_h
39 #define included_clib_string_h
40 
41 #include <vppinfra/clib.h> /* for CLIB_LINUX_KERNEL */
42 #include <vppinfra/vector.h>
43 
44 #ifdef CLIB_LINUX_KERNEL
45 #include <linux/string.h>
46 #endif
47 
48 #ifdef CLIB_UNIX
49 #include <string.h>
50 #endif
51 
52 #ifdef CLIB_STANDALONE
53 #include <vppinfra/standalone_string.h>
54 #endif
55 
56 #if _x86_64_
57 #include <x86intrin.h>
58 #endif
59 
60 /* Exchanges source and destination. */
61 void clib_memswap (void *_a, void *_b, uword bytes);
62 
63 /*
64  * the vector unit memcpy variants confuse coverity
65  * so don't let it anywhere near them.
66  */
67 #ifndef __COVERITY__
68 #if __AVX512F__
69 #include <vppinfra/memcpy_avx512.h>
70 #elif __AVX2__
71 #include <vppinfra/memcpy_avx2.h>
72 #elif __SSSE3__
73 #include <vppinfra/memcpy_sse3.h>
74 #else
75 #define clib_memcpy(a,b,c) memcpy(a,b,c)
76 #endif
77 #else /* __COVERITY__ */
78 #define clib_memcpy(a,b,c) memcpy(a,b,c)
79 #endif
80 
81 /*
82  * Copy 64 bytes of data to 4 destinations
83  * this function is typically used in quad-loop case when whole cacheline
84  * needs to be copied to 4 different places. First it reads whole cacheline
85  * to 1/2/4 SIMD registers and then it writes data to 4 destinations.
86  */
87 
89 clib_memcpy64_x4 (void *d0, void *d1, void *d2, void *d3, void *s)
90 {
91 #if defined (__AVX512F__)
92  __m512i r0 = _mm512_loadu_si512 (s);
93 
94  _mm512_storeu_si512 (d0, r0);
95  _mm512_storeu_si512 (d1, r0);
96  _mm512_storeu_si512 (d2, r0);
97  _mm512_storeu_si512 (d3, r0);
98 
99 #elif defined (__AVX2__)
100  __m256i r0 = _mm256_loadu_si256 ((__m256i *) (s + 0 * 32));
101  __m256i r1 = _mm256_loadu_si256 ((__m256i *) (s + 1 * 32));
102 
103  _mm256_storeu_si256 ((__m256i *) (d0 + 0 * 32), r0);
104  _mm256_storeu_si256 ((__m256i *) (d0 + 1 * 32), r1);
105 
106  _mm256_storeu_si256 ((__m256i *) (d1 + 0 * 32), r0);
107  _mm256_storeu_si256 ((__m256i *) (d1 + 1 * 32), r1);
108 
109  _mm256_storeu_si256 ((__m256i *) (d2 + 0 * 32), r0);
110  _mm256_storeu_si256 ((__m256i *) (d2 + 1 * 32), r1);
111 
112  _mm256_storeu_si256 ((__m256i *) (d3 + 0 * 32), r0);
113  _mm256_storeu_si256 ((__m256i *) (d3 + 1 * 32), r1);
114 
115 #elif defined (__SSSE3__)
116  __m128i r0 = _mm_loadu_si128 ((__m128i *) (s + 0 * 16));
117  __m128i r1 = _mm_loadu_si128 ((__m128i *) (s + 1 * 16));
118  __m128i r2 = _mm_loadu_si128 ((__m128i *) (s + 2 * 16));
119  __m128i r3 = _mm_loadu_si128 ((__m128i *) (s + 3 * 16));
120 
121  _mm_storeu_si128 ((__m128i *) (d0 + 0 * 16), r0);
122  _mm_storeu_si128 ((__m128i *) (d0 + 1 * 16), r1);
123  _mm_storeu_si128 ((__m128i *) (d0 + 2 * 16), r2);
124  _mm_storeu_si128 ((__m128i *) (d0 + 3 * 16), r3);
125 
126  _mm_storeu_si128 ((__m128i *) (d1 + 0 * 16), r0);
127  _mm_storeu_si128 ((__m128i *) (d1 + 1 * 16), r1);
128  _mm_storeu_si128 ((__m128i *) (d1 + 2 * 16), r2);
129  _mm_storeu_si128 ((__m128i *) (d1 + 3 * 16), r3);
130 
131  _mm_storeu_si128 ((__m128i *) (d2 + 0 * 16), r0);
132  _mm_storeu_si128 ((__m128i *) (d2 + 1 * 16), r1);
133  _mm_storeu_si128 ((__m128i *) (d2 + 2 * 16), r2);
134  _mm_storeu_si128 ((__m128i *) (d2 + 3 * 16), r3);
135 
136  _mm_storeu_si128 ((__m128i *) (d3 + 0 * 16), r0);
137  _mm_storeu_si128 ((__m128i *) (d3 + 1 * 16), r1);
138  _mm_storeu_si128 ((__m128i *) (d3 + 2 * 16), r2);
139  _mm_storeu_si128 ((__m128i *) (d3 + 3 * 16), r3);
140 
141 #else
142  clib_memcpy (d0, s, 64);
143  clib_memcpy (d1, s, 64);
144  clib_memcpy (d2, s, 64);
145  clib_memcpy (d3, s, 64);
146 #endif
147 }
148 
149 #endif /* included_clib_string_h */
150 
151 /*
152  * fd.io coding-style-patch-verification: ON
153  *
154  * Local Variables:
155  * eval: (c-set-style "gnu")
156  * End:
157  */
void clib_memswap(void *_a, void *_b, uword bytes)
Definition: string.c:43
#define static_always_inline
Definition: clib.h:93
static_always_inline void clib_memcpy64_x4(void *d0, void *d1, void *d2, void *d3, void *s)
Definition: string.h:89
#define clib_memcpy(a, b, c)
Definition: string.h:75
u64 uword
Definition: types.h:112