1 /**************************************************************************
2 *
3 * Copyright 2008 Dennis Smit
4 * All Rights Reserved.
5 *
6 * Permission is hereby granted, free of charge, to any person obtaining a
7 * copy of this software and associated documentation files (the "Software"),
8 * to deal in the Software without restriction, including without limitation
9 * on the rights to use, copy, modify, merge, publish, distribute, sub
10 * license, and/or sell copies of the Software, and to permit persons to whom
11 * the Software is furnished to do so, subject to the following conditions:
12 *
13 * The above copyright notice and this permission notice (including the next
14 * paragraph) shall be included in all copies or substantial portions of the
15 * Software.
16 *
17 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
18 * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
19 * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL
20 * AUTHORS, COPYRIGHT HOLDERS, AND/OR THEIR SUPPLIERS BE LIABLE FOR ANY CLAIM,
21 * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR
22 * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE
23 * USE OR OTHER DEALINGS IN THE SOFTWARE.
24 *
25 ***************************************************************************/
26
27 /**
28 * @file
29 * CPU feature detection.
30 *
31 * @author Dennis Smit
32 * @author Based on the work of Eric Anholt <anholt@FreeBSD.org>
33 */
34
35 #ifndef _UTIL_CPU_DETECT_H
36 #define _UTIL_CPU_DETECT_H
37
38
39 #include "pipe/p_config.h"
40 #include "util/u_thread.h"
41
42
43 #ifdef __cplusplus
44 extern "C" {
45 #endif
46
47 enum cpu_family {
48 CPU_UNKNOWN,
49
50 CPU_AMD_ZEN1_ZEN2,
51 CPU_AMD_ZEN_HYGON,
52 CPU_AMD_ZEN3,
53 CPU_AMD_ZEN_NEXT,
54 CPU_AMD_LAST,
55 };
56
57 typedef uint32_t util_affinity_mask[UTIL_MAX_CPUS / 32];
58
59 struct util_cpu_caps_t {
60 /**
61 * Number of CPUs available to the process.
62 *
63 * This will be less than or equal to \c max_cpus. This is the number of
64 * CPUs that are online and available to the process.
65 */
66 int16_t nr_cpus;
67
68 /**
69 * Maximum number of CPUs that can be online in the system.
70 *
71 * This will be greater than or equal to \c nr_cpus. This is the number of
72 * CPUs installed in the system. \c nr_cpus will be less if some CPUs are
73 * offline.
74 */
75 int16_t max_cpus;
76
77 enum cpu_family family;
78
79 /* Feature flags */
80 int x86_cpu_type;
81 unsigned cacheline;
82
83 unsigned has_intel:1;
84 unsigned has_tsc:1;
85 unsigned has_mmx:1;
86 unsigned has_mmx2:1;
87 unsigned has_sse:1;
88 unsigned has_sse2:1;
89 unsigned has_sse3:1;
90 unsigned has_ssse3:1;
91 unsigned has_sse4_1:1;
92 unsigned has_sse4_2:1;
93 unsigned has_popcnt:1;
94 unsigned has_avx:1;
95 unsigned has_avx2:1;
96 unsigned has_f16c:1;
97 unsigned has_fma:1;
98 unsigned has_3dnow:1;
99 unsigned has_3dnow_ext:1;
100 unsigned has_xop:1;
101 unsigned has_altivec:1;
102 unsigned has_vsx:1;
103 unsigned has_daz:1;
104 unsigned has_neon:1;
105 unsigned has_msa:1;
106
107 unsigned has_avx512f:1;
108 unsigned has_avx512dq:1;
109 unsigned has_avx512ifma:1;
110 unsigned has_avx512pf:1;
111 unsigned has_avx512er:1;
112 unsigned has_avx512cd:1;
113 unsigned has_avx512bw:1;
114 unsigned has_avx512vl:1;
115 unsigned has_avx512vbmi:1;
116
117 unsigned num_L3_caches;
118 unsigned num_cpu_mask_bits;
119
120 uint16_t cpu_to_L3[UTIL_MAX_CPUS];
121 /* Affinity masks for each L3 cache. */
122 util_affinity_mask *L3_affinity_mask;
123 };
124
125 #define U_CPU_INVALID_L3 0xffff
126
127 static inline const struct util_cpu_caps_t *
util_get_cpu_caps(void)128 util_get_cpu_caps(void)
129 {
130 extern struct util_cpu_caps_t util_cpu_caps;
131
132 /* If you hit this assert, it means that something is using the
133 * cpu-caps without having first called util_cpu_detect()
134 */
135 assert(util_cpu_caps.nr_cpus >= 1);
136
137 return &util_cpu_caps;
138 }
139
140 void util_cpu_detect(void);
141
142
143 #ifdef __cplusplus
144 }
145 #endif
146
147
148 #endif /* _UTIL_CPU_DETECT_H */
149