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/cpu.h"
33 #include "src/arm/cpu.h"
34
35 #if HAVE_GETAUXVAL || HAVE_ELF_AUX_INFO
36 #include <sys/auxv.h>
37
38 #if ARCH_AARCH64
39
40 #define HWCAP_AARCH64_ASIMDDP (1 << 20)
41 #define HWCAP_AARCH64_SVE (1 << 22)
42 #define HWCAP2_AARCH64_SVE2 (1 << 1)
43 #define HWCAP2_AARCH64_I8MM (1 << 13)
44
dav1d_get_cpu_flags_arm(void)45 COLD unsigned dav1d_get_cpu_flags_arm(void) {
46 #if HAVE_GETAUXVAL
47 unsigned long hw_cap = getauxval(AT_HWCAP);
48 unsigned long hw_cap2 = getauxval(AT_HWCAP2);
49 #else
50 unsigned long hw_cap = 0;
51 unsigned long hw_cap2 = 0;
52 elf_aux_info(AT_HWCAP, &hw_cap, sizeof(hw_cap));
53 elf_aux_info(AT_HWCAP2, &hw_cap2, sizeof(hw_cap2));
54 #endif
55
56 unsigned flags = dav1d_get_default_cpu_flags();
57 flags |= (hw_cap & HWCAP_AARCH64_ASIMDDP) ? DAV1D_ARM_CPU_FLAG_DOTPROD : 0;
58 flags |= (hw_cap2 & HWCAP2_AARCH64_I8MM) ? DAV1D_ARM_CPU_FLAG_I8MM : 0;
59 flags |= (hw_cap & HWCAP_AARCH64_SVE) ? DAV1D_ARM_CPU_FLAG_SVE : 0;
60 flags |= (hw_cap2 & HWCAP2_AARCH64_SVE2) ? DAV1D_ARM_CPU_FLAG_SVE2 : 0;
61 return flags;
62 }
63 #else /* !ARCH_AARCH64 */
64
65 #ifndef HWCAP_ARM_NEON
66 #define HWCAP_ARM_NEON (1 << 12)
67 #endif
68 #define HWCAP_ARM_ASIMDDP (1 << 24)
69 #define HWCAP_ARM_I8MM (1 << 27)
70
dav1d_get_cpu_flags_arm(void)71 COLD unsigned dav1d_get_cpu_flags_arm(void) {
72 #if HAVE_GETAUXVAL
73 unsigned long hw_cap = getauxval(AT_HWCAP);
74 #else
75 unsigned long hw_cap = 0;
76 elf_aux_info(AT_HWCAP, &hw_cap, sizeof(hw_cap));
77 #endif
78
79 unsigned flags = dav1d_get_default_cpu_flags();
80 flags |= (hw_cap & HWCAP_ARM_NEON) ? DAV1D_ARM_CPU_FLAG_NEON : 0;
81 flags |= (hw_cap & HWCAP_ARM_ASIMDDP) ? DAV1D_ARM_CPU_FLAG_DOTPROD : 0;
82 flags |= (hw_cap & HWCAP_ARM_I8MM) ? DAV1D_ARM_CPU_FLAG_I8MM : 0;
83 return flags;
84 }
85 #endif /* ARCH_AARCH64 */
86
87 #elif defined(__APPLE__)
88 #include <sys/sysctl.h>
89
have_feature(const char * feature)90 static int have_feature(const char *feature) {
91 int supported = 0;
92 size_t size = sizeof(supported);
93 if (sysctlbyname(feature, &supported, &size, NULL, 0) != 0) {
94 return 0;
95 }
96 return supported;
97 }
98
dav1d_get_cpu_flags_arm(void)99 COLD unsigned dav1d_get_cpu_flags_arm(void) {
100 unsigned flags = dav1d_get_default_cpu_flags();
101 if (have_feature("hw.optional.arm.FEAT_DotProd"))
102 flags |= DAV1D_ARM_CPU_FLAG_DOTPROD;
103 if (have_feature("hw.optional.arm.FEAT_I8MM"))
104 flags |= DAV1D_ARM_CPU_FLAG_I8MM;
105 /* No SVE and SVE2 feature detection available on Apple platforms. */
106 return flags;
107 }
108
109 #elif defined(__OpenBSD__) && ARCH_AARCH64
110 #include <machine/armreg.h>
111 #include <machine/cpu.h>
112 #include <sys/types.h>
113 #include <sys/sysctl.h>
114
dav1d_get_cpu_flags_arm(void)115 COLD unsigned dav1d_get_cpu_flags_arm(void) {
116 unsigned flags = dav1d_get_default_cpu_flags();
117
118 #ifdef CPU_ID_AA64ISAR0
119 int mib[2];
120 uint64_t isar0;
121 uint64_t isar1;
122 size_t len;
123
124 mib[0] = CTL_MACHDEP;
125 mib[1] = CPU_ID_AA64ISAR0;
126 len = sizeof(isar0);
127 if (sysctl(mib, 2, &isar0, &len, NULL, 0) != -1) {
128 if (ID_AA64ISAR0_DP(isar0) >= ID_AA64ISAR0_DP_IMPL)
129 flags |= DAV1D_ARM_CPU_FLAG_DOTPROD;
130 }
131
132 mib[0] = CTL_MACHDEP;
133 mib[1] = CPU_ID_AA64ISAR1;
134 len = sizeof(isar1);
135 if (sysctl(mib, 2, &isar1, &len, NULL, 0) != -1) {
136 #ifdef ID_AA64ISAR1_I8MM_IMPL
137 if (ID_AA64ISAR1_I8MM(isar1) >= ID_AA64ISAR1_I8MM_IMPL)
138 flags |= DAV1D_ARM_CPU_FLAG_I8MM;
139 #endif
140 }
141 #endif
142
143 return flags;
144 }
145
146 #elif defined(_WIN32)
147 #include <windows.h>
148
dav1d_get_cpu_flags_arm(void)149 COLD unsigned dav1d_get_cpu_flags_arm(void) {
150 unsigned flags = dav1d_get_default_cpu_flags();
151 #ifdef PF_ARM_V82_DP_INSTRUCTIONS_AVAILABLE
152 if (IsProcessorFeaturePresent(PF_ARM_V82_DP_INSTRUCTIONS_AVAILABLE))
153 flags |= DAV1D_ARM_CPU_FLAG_DOTPROD;
154 #endif
155 #ifdef PF_ARM_SVE_INSTRUCTIONS_AVAILABLE
156 if (IsProcessorFeaturePresent(PF_ARM_SVE_INSTRUCTIONS_AVAILABLE))
157 flags |= DAV1D_ARM_CPU_FLAG_SVE;
158 #endif
159 #ifdef PF_ARM_SVE2_INSTRUCTIONS_AVAILABLE
160 if (IsProcessorFeaturePresent(PF_ARM_SVE2_INSTRUCTIONS_AVAILABLE))
161 flags |= DAV1D_ARM_CPU_FLAG_SVE2;
162 #endif
163 #ifdef PF_ARM_SVE_I8MM_INSTRUCTIONS_AVAILABLE
164 /* There's no PF_* flag that indicates whether plain I8MM is available
165 * or not. But if SVE_I8MM is available, that also implies that
166 * regular I8MM is available. */
167 if (IsProcessorFeaturePresent(PF_ARM_SVE_I8MM_INSTRUCTIONS_AVAILABLE))
168 flags |= DAV1D_ARM_CPU_FLAG_I8MM;
169 #endif
170 return flags;
171 }
172
173 #elif defined(__ANDROID__)
174 #include <ctype.h>
175 #include <stdio.h>
176 #include <string.h>
177
parse_proc_cpuinfo(const char * flag)178 static unsigned parse_proc_cpuinfo(const char *flag) {
179 FILE *file = fopen("/proc/cpuinfo", "r");
180 if (!file)
181 return 0;
182
183 char line_buffer[120];
184 const char *line;
185
186 size_t flaglen = strlen(flag);
187 while ((line = fgets(line_buffer, sizeof(line_buffer), file))) {
188 // check all occurances as whole words
189 const char *found = line;
190 while ((found = strstr(found, flag))) {
191 if ((found == line_buffer || !isgraph(found[-1])) &&
192 (isspace(found[flaglen]) || feof(file))) {
193 fclose(file);
194 return 1;
195 }
196 found += flaglen;
197 }
198 // if line is incomplete seek back to avoid splitting the search
199 // string into two buffers
200 if (!strchr(line, '\n') && strlen(line) > flaglen) {
201 // use fseek since the 64 bit fseeko is only available since
202 // Android API level 24 and meson defines _FILE_OFFSET_BITS
203 // by default 64
204 if (fseek(file, -flaglen, SEEK_CUR))
205 break;
206 }
207 }
208
209 fclose(file);
210
211 return 0;
212 }
213
dav1d_get_cpu_flags_arm(void)214 COLD unsigned dav1d_get_cpu_flags_arm(void) {
215 unsigned flags = dav1d_get_default_cpu_flags();
216 flags |= parse_proc_cpuinfo("neon") ? DAV1D_ARM_CPU_FLAG_NEON : 0;
217 flags |= parse_proc_cpuinfo("asimd") ? DAV1D_ARM_CPU_FLAG_NEON : 0;
218 flags |= parse_proc_cpuinfo("asimddp") ? DAV1D_ARM_CPU_FLAG_DOTPROD : 0;
219 flags |= parse_proc_cpuinfo("i8mm") ? DAV1D_ARM_CPU_FLAG_I8MM : 0;
220 #if ARCH_AARCH64
221 flags |= parse_proc_cpuinfo("sve") ? DAV1D_ARM_CPU_FLAG_SVE : 0;
222 flags |= parse_proc_cpuinfo("sve2") ? DAV1D_ARM_CPU_FLAG_SVE2 : 0;
223 #endif /* ARCH_AARCH64 */
224 return flags;
225 }
226
227 #else /* Unsupported OS */
228
dav1d_get_cpu_flags_arm(void)229 COLD unsigned dav1d_get_cpu_flags_arm(void) {
230 return dav1d_get_default_cpu_flags();
231 }
232
233 #endif
234