1 /*
2 * Copyright © 2018, VideoLAN and dav1d authors
3 * Copyright © 2018, Janne Grunau
4 * All rights reserved.
5 *
6 * Redistribution and use in source and binary forms, with or without
7 * modification, are permitted provided that the following conditions are met:
8 *
9 * 1. Redistributions of source code must retain the above copyright notice, this
10 * list of conditions and the following disclaimer.
11 *
12 * 2. Redistributions in binary form must reproduce the above copyright notice,
13 * this list of conditions and the following disclaimer in the documentation
14 * and/or other materials provided with the distribution.
15 *
16 * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND
17 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED
18 * WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
19 * DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE LIABLE FOR
20 * ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES
21 * (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
22 * LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND
23 * ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
24 * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS
25 * SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
26 */
27
28 #include "config.h"
29
30 #include "common/attributes.h"
31
32 #include "src/arm/cpu.h"
33
34 #if defined(HAVE_GETAUXVAL) || defined(HAVE_ELF_AUX_INFO)
35 #include <sys/auxv.h>
36
37 #if ARCH_AARCH64
38
39 #define HWCAP_AARCH64_ASIMDDP (1 << 20)
40 #define HWCAP_AARCH64_SVE (1 << 22)
41 #define HWCAP2_AARCH64_SVE2 (1 << 1)
42 #define HWCAP2_AARCH64_I8MM (1 << 13)
43
dav1d_get_cpu_flags_arm(void)44 COLD unsigned dav1d_get_cpu_flags_arm(void) {
45 #ifdef HAVE_GETAUXVAL
46 unsigned long hw_cap = getauxval(AT_HWCAP);
47 unsigned long hw_cap2 = getauxval(AT_HWCAP2);
48 #else
49 unsigned long hw_cap = 0;
50 unsigned long hw_cap2 = 0;
51 elf_aux_info(AT_HWCAP, &hw_cap, sizeof(hw_cap));
52 elf_aux_info(AT_HWCAP2, &hw_cap2, sizeof(hw_cap2));
53 #endif
54
55 unsigned flags = DAV1D_ARM_CPU_FLAG_NEON;
56 flags |= (hw_cap & HWCAP_AARCH64_ASIMDDP) ? DAV1D_ARM_CPU_FLAG_DOTPROD : 0;
57 flags |= (hw_cap2 & HWCAP2_AARCH64_I8MM) ? DAV1D_ARM_CPU_FLAG_I8MM : 0;
58 flags |= (hw_cap & HWCAP_AARCH64_SVE) ? DAV1D_ARM_CPU_FLAG_SVE : 0;
59 flags |= (hw_cap2 & HWCAP2_AARCH64_SVE2) ? DAV1D_ARM_CPU_FLAG_SVE2 : 0;
60 return flags;
61 }
62 #else /* !ARCH_AARCH64 */
63
64 #ifndef HWCAP_ARM_NEON
65 #define HWCAP_ARM_NEON (1 << 12)
66 #endif
67 #define HWCAP_ARM_ASIMDDP (1 << 24)
68 #define HWCAP_ARM_I8MM (1 << 27)
69
dav1d_get_cpu_flags_arm(void)70 COLD unsigned dav1d_get_cpu_flags_arm(void) {
71 #ifdef HAVE_GETAUXVAL
72 unsigned long hw_cap = getauxval(AT_HWCAP);
73 #else
74 unsigned long hw_cap = 0;
75 elf_aux_info(AT_HWCAP, &hw_cap, sizeof(hw_cap));
76 #endif
77
78 unsigned flags = (hw_cap & HWCAP_ARM_NEON) ? DAV1D_ARM_CPU_FLAG_NEON : 0;
79 flags |= (hw_cap & HWCAP_ARM_ASIMDDP) ? DAV1D_ARM_CPU_FLAG_DOTPROD : 0;
80 flags |= (hw_cap & HWCAP_ARM_I8MM) ? DAV1D_ARM_CPU_FLAG_I8MM : 0;
81 return flags;
82 }
83 #endif /* ARCH_AARCH64 */
84
85 #elif defined(__APPLE__)
86 #include <sys/sysctl.h>
87
have_feature(const char * feature)88 static int have_feature(const char *feature) {
89 int supported = 0;
90 size_t size = sizeof(supported);
91 if (sysctlbyname(feature, &supported, &size, NULL, 0) != 0) {
92 return 0;
93 }
94 return supported;
95 }
96
dav1d_get_cpu_flags_arm(void)97 COLD unsigned dav1d_get_cpu_flags_arm(void) {
98 unsigned flags = DAV1D_ARM_CPU_FLAG_NEON;
99 if (have_feature("hw.optional.arm.FEAT_DotProd"))
100 flags |= DAV1D_ARM_CPU_FLAG_DOTPROD;
101 if (have_feature("hw.optional.arm.FEAT_I8MM"))
102 flags |= DAV1D_ARM_CPU_FLAG_I8MM;
103 /* No SVE and SVE2 feature detection available on Apple platforms. */
104 return flags;
105 }
106
107 #elif defined(_WIN32)
108 #include <windows.h>
109
dav1d_get_cpu_flags_arm(void)110 COLD unsigned dav1d_get_cpu_flags_arm(void) {
111 unsigned flags = DAV1D_ARM_CPU_FLAG_NEON;
112 #ifdef PF_ARM_V82_DP_INSTRUCTIONS_AVAILABLE
113 if (IsProcessorFeaturePresent(PF_ARM_V82_DP_INSTRUCTIONS_AVAILABLE))
114 flags |= DAV1D_ARM_CPU_FLAG_DOTPROD;
115 #endif
116 /* No I8MM or SVE feature detection available on Windows at the time of
117 * writing. */
118 return flags;
119 }
120
121 #elif defined(__ANDROID__)
122 #include <ctype.h>
123 #include <stdio.h>
124 #include <string.h>
125
parse_proc_cpuinfo(const char * flag)126 static unsigned parse_proc_cpuinfo(const char *flag) {
127 FILE *file = fopen("/proc/cpuinfo", "r");
128 if (!file)
129 return 0;
130
131 char line_buffer[120];
132 const char *line;
133
134 size_t flaglen = strlen(flag);
135 while ((line = fgets(line_buffer, sizeof(line_buffer), file))) {
136 // check all occurances as whole words
137 const char *found = line;
138 while ((found = strstr(found, flag))) {
139 if ((found == line_buffer || !isgraph(found[-1])) &&
140 (isspace(found[flaglen]) || feof(file))) {
141 fclose(file);
142 return 1;
143 }
144 found += flaglen;
145 }
146 // if line is incomplete seek back to avoid splitting the search
147 // string into two buffers
148 if (!strchr(line, '\n') && strlen(line) > flaglen) {
149 // use fseek since the 64 bit fseeko is only available since
150 // Android API level 24 and meson defines _FILE_OFFSET_BITS
151 // by default 64
152 if (fseek(file, -flaglen, SEEK_CUR))
153 break;
154 }
155 }
156
157 fclose(file);
158
159 return 0;
160 }
161
dav1d_get_cpu_flags_arm(void)162 COLD unsigned dav1d_get_cpu_flags_arm(void) {
163 unsigned flags = parse_proc_cpuinfo("neon") ? DAV1D_ARM_CPU_FLAG_NEON : 0;
164 flags |= parse_proc_cpuinfo("asimd") ? DAV1D_ARM_CPU_FLAG_NEON : 0;
165 flags |= parse_proc_cpuinfo("asimddp") ? DAV1D_ARM_CPU_FLAG_DOTPROD : 0;
166 flags |= parse_proc_cpuinfo("i8mm") ? DAV1D_ARM_CPU_FLAG_I8MM : 0;
167 #if ARCH_AARCH64
168 flags |= parse_proc_cpuinfo("sve") ? DAV1D_ARM_CPU_FLAG_SVE : 0;
169 flags |= parse_proc_cpuinfo("sve2") ? DAV1D_ARM_CPU_FLAG_SVE2 : 0;
170 #endif /* ARCH_AARCH64 */
171 return flags;
172 }
173
174 #else /* Unsupported OS */
175
dav1d_get_cpu_flags_arm(void)176 COLD unsigned dav1d_get_cpu_flags_arm(void) {
177 return 0;
178 }
179
180 #endif
181