1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427
|
/*
* Copyright © 2011 Siarhei Siamashka <siarhei.siamashka@gmail.com>
*
* Permission is hereby granted, free of charge, to any person obtaining a
* copy of this software and associated documentation files (the "Software"),
* to deal in the Software without restriction, including without limitation
* the rights to use, copy, modify, merge, publish, distribute, sublicense,
* and/or sell copies of the Software, and to permit persons to whom the
* Software is furnished to do so, subject to the following conditions:
*
* The above copyright notice and this permission notice (including the next
* paragraph) shall be included in all copies or substantial portions of the
* Software.
*
* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
* IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
* FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
* THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
* LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
* FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER
* DEALINGS IN THE SOFTWARE.
*/
#include <string.h>
#include <stdio.h>
#include <stdlib.h>
#include <ctype.h>
#include "asm-opt.h"
#if defined(__linux__) || defined(ANDROID) || defined(__ANDROID__)
#define SOMEWHAT_SANE_PROC_CPUINFO_SIZE_LIMIT (1024 * 1024)
#if defined(__i386__) || defined(__amd64__)
#define FEATURES_ID "flags"
#elif defined(__arm__)
#define FEATURES_ID "Features"
#elif defined(__mips__)
#define FEATURES_ID "cpu model"
#else
#define FEATURES_ID "?"
#endif
static int check_feature (char *buffer, const char *feature)
{
char *p;
if (*feature == 0)
return 0;
if (strncmp(buffer, FEATURES_ID, strlen(FEATURES_ID)) != 0)
return 0;
buffer += strlen(FEATURES_ID);
while (isspace(*buffer))
buffer++;
/* Check if 'feature' is present in the buffer as a separate word */
while ((p = strstr(buffer, feature))) {
if (p > buffer && !isspace(*(p - 1))) {
buffer++;
continue;
}
p += strlen(feature);
if (*p != 0 && !isspace(*p)) {
buffer++;
continue;
}
return 1;
}
return 0;
}
static int parse_proc_cpuinfo(int bufsize, const char *feature)
{
char *buffer = (char *)malloc(bufsize);
FILE *fd;
int feature_support = 0;
if (!buffer)
return 0;
fd = fopen("/proc/cpuinfo", "r");
if (fd) {
while (fgets(buffer, bufsize, fd)) {
if (!strchr(buffer, '\n') && !feof(fd)) {
/* "impossible" happened - insufficient size of the buffer! */
fclose(fd);
free(buffer);
return -1;
}
if (check_feature(buffer, feature))
feature_support = 1;
}
fclose(fd);
}
free(buffer);
return feature_support;
}
#define SOMEWHAT_SANE_PROC_CPUINFO_SIZE_LIMIT (1024 * 1024)
int check_cpu_feature(const char *feature)
{
int bufsize = 1024;
int result;
while ((result = parse_proc_cpuinfo(bufsize, feature)) == -1)
{
bufsize *= 2;
if (bufsize > SOMEWHAT_SANE_PROC_CPUINFO_SIZE_LIMIT)
return 0;
}
return result;
}
#else
int check_cpu_feature(const char *feature)
{
return 0;
}
#endif
static bench_info empty[] = { { NULL, 0, NULL } };
#if defined(__i386__) || defined(__amd64__)
#include "x86-sse2.h"
static bench_info x86_sse2[] =
{
{ "MOVSB copy", 0, aligned_block_copy_movsb },
{ "MOVSD copy", 0, aligned_block_copy_movsd },
{ "SSE2 copy", 0, aligned_block_copy_sse2 },
{ "SSE2 nontemporal copy", 0, aligned_block_copy_nt_sse2 },
{ "SSE2 copy prefetched (32 bytes step)", 0, aligned_block_copy_pf32_sse2 },
{ "SSE2 copy prefetched (64 bytes step)", 0, aligned_block_copy_pf64_sse2 },
{ "SSE2 nontemporal copy prefetched (32 bytes step)", 0, aligned_block_copy_nt_pf32_sse2 },
{ "SSE2 nontemporal copy prefetched (64 bytes step)", 0, aligned_block_copy_nt_pf64_sse2 },
{ "SSE2 2-pass copy", 1, aligned_block_copy_sse2 },
{ "SSE2 2-pass copy prefetched (32 bytes step)", 1, aligned_block_copy_pf32_sse2 },
{ "SSE2 2-pass copy prefetched (64 bytes step)", 1, aligned_block_copy_pf64_sse2 },
{ "SSE2 2-pass nontemporal copy", 1, aligned_block_copy_nt_sse2 },
{ "SSE2 fill", 0, aligned_block_fill_sse2 },
{ "SSE2 nontemporal fill", 0, aligned_block_fill_nt_sse2 },
{ NULL, 0, NULL }
};
static bench_info x86_sse2_fb[] =
{
{ "MOVSD copy (from framebuffer)", 0, aligned_block_copy_movsd },
{ "MOVSD 2-pass copy (from framebuffer)", 1, aligned_block_copy_movsd },
{ "SSE2 copy (from framebuffer)", 0, aligned_block_copy_sse2 },
{ "SSE2 2-pass copy (from framebuffer)", 1, aligned_block_copy_sse2 },
{ NULL, 0, NULL }
};
static int check_sse2_support(void)
{
#ifdef __amd64__
return 1; /* We assume that all 64-bit processors have SSE2 support */
#else
int cpuid_feature_information;
__asm__ volatile (
/* According to Intel manual, CPUID instruction is supported
* if the value of ID bit (bit 21) in EFLAGS can be modified */
"pushf\n"
"movl (%%esp), %0\n"
"xorl $0x200000, (%%esp)\n" /* try to modify ID bit */
"popf\n"
"pushf\n"
"xorl (%%esp), %0\n" /* check if ID bit changed */
"jz 1f\n"
"push %%eax\n"
"push %%ebx\n"
"push %%ecx\n"
"mov $1, %%eax\n"
"cpuid\n"
"pop %%ecx\n"
"pop %%ebx\n"
"pop %%eax\n"
"1:\n"
"popf\n"
: "=d" (cpuid_feature_information)
:
: "cc");
return cpuid_feature_information & (1 << 26);
#endif
}
bench_info *get_asm_benchmarks(void)
{
if (check_sse2_support())
return x86_sse2;
else
return empty;
}
bench_info *get_asm_framebuffer_benchmarks(void)
{
if (check_sse2_support())
return x86_sse2_fb;
else
return empty;
}
#elif defined(__arm__)
#include "arm-neon.h"
static bench_info arm_neon[] =
{
{ "NEON read", 0, aligned_block_read_neon },
{ "NEON read prefetched (32 bytes step)", 0, aligned_block_read_pf32_neon },
{ "NEON read prefetched (64 bytes step)", 0, aligned_block_read_pf64_neon },
{ "NEON read 2 data streams", 0, aligned_block_read2_neon },
{ "NEON read 2 data streams prefetched (32 bytes step)", 0, aligned_block_read2_pf32_neon },
{ "NEON read 2 data streams prefetched (64 bytes step)", 0, aligned_block_read2_pf64_neon },
{ "NEON copy", 0, aligned_block_copy_neon },
{ "NEON copy prefetched (32 bytes step)", 0, aligned_block_copy_pf32_neon },
{ "NEON copy prefetched (64 bytes step)", 0, aligned_block_copy_pf64_neon },
{ "NEON unrolled copy", 0, aligned_block_copy_unrolled_neon },
{ "NEON unrolled copy prefetched (32 bytes step)", 0, aligned_block_copy_unrolled_pf32_neon },
{ "NEON unrolled copy prefetched (64 bytes step)", 0, aligned_block_copy_unrolled_pf64_neon },
{ "NEON copy backwards", 0, aligned_block_copy_backwards_neon },
{ "NEON copy backwards prefetched (32 bytes step)", 0, aligned_block_copy_backwards_pf32_neon },
{ "NEON copy backwards prefetched (64 bytes step)", 0, aligned_block_copy_backwards_pf64_neon },
{ "NEON 2-pass copy", 1, aligned_block_copy_neon },
{ "NEON 2-pass copy prefetched (32 bytes step)", 1, aligned_block_copy_pf32_neon },
{ "NEON 2-pass copy prefetched (64 bytes step)", 1, aligned_block_copy_pf64_neon },
{ "NEON unrolled 2-pass copy", 1, aligned_block_copy_unrolled_neon },
{ "NEON unrolled 2-pass copy prefetched (32 bytes step)", 1, aligned_block_copy_unrolled_pf32_neon },
{ "NEON unrolled 2-pass copy prefetched (64 bytes step)", 1, aligned_block_copy_unrolled_pf64_neon },
{ "NEON fill", 0, aligned_block_fill_neon },
{ "NEON fill backwards", 0, aligned_block_fill_backwards_neon },
{ "VFP copy", 0, aligned_block_copy_vfp },
{ "VFP 2-pass copy", 1, aligned_block_copy_vfp },
{ "ARM fill (STRD)", 0, aligned_block_fill_strd_armv5te },
{ "ARM fill (STM with 8 registers)", 0, aligned_block_fill_stm8_armv4 },
{ "ARM fill (STM with 4 registers)", 0, aligned_block_fill_stm4_armv4 },
{ "ARM copy prefetched (incr pld)", 0, aligned_block_copy_incr_armv5te },
{ "ARM copy prefetched (wrap pld)", 0, aligned_block_copy_wrap_armv5te },
{ "ARM 2-pass copy prefetched (incr pld)", 1, aligned_block_copy_incr_armv5te },
{ "ARM 2-pass copy prefetched (wrap pld)", 1, aligned_block_copy_wrap_armv5te },
{ NULL, 0, NULL }
};
static bench_info arm_v5te_vfp[] =
{
{ "VFP copy", 0, aligned_block_copy_vfp },
{ "VFP 2-pass copy", 1, aligned_block_copy_vfp },
{ "ARM fill (STRD)", 0, aligned_block_fill_strd_armv5te },
{ "ARM fill (STM with 8 registers)", 0, aligned_block_fill_stm8_armv4 },
{ "ARM fill (STM with 4 registers)", 0, aligned_block_fill_stm4_armv4 },
{ "ARM copy prefetched (incr pld)", 0, aligned_block_copy_incr_armv5te },
{ "ARM copy prefetched (wrap pld)", 0, aligned_block_copy_wrap_armv5te },
{ "ARM 2-pass copy prefetched (incr pld)", 1, aligned_block_copy_incr_armv5te },
{ "ARM 2-pass copy prefetched (wrap pld)", 1, aligned_block_copy_wrap_armv5te },
{ NULL, 0, NULL }
};
static bench_info arm_v5te[] =
{
{ "ARM fill (STRD)", 0, aligned_block_fill_strd_armv5te },
{ "ARM fill (STM with 8 registers)", 0, aligned_block_fill_stm8_armv4 },
{ "ARM fill (STM with 4 registers)", 0, aligned_block_fill_stm4_armv4 },
{ "ARM copy prefetched (incr pld)", 0, aligned_block_copy_incr_armv5te },
{ "ARM copy prefetched (wrap pld)", 0, aligned_block_copy_wrap_armv5te },
{ "ARM 2-pass copy prefetched (incr pld)", 1, aligned_block_copy_incr_armv5te },
{ "ARM 2-pass copy prefetched (wrap pld)", 1, aligned_block_copy_wrap_armv5te },
{ NULL, 0, NULL }
};
static bench_info arm_v4[] =
{
{ "ARM fill (STM with 8 registers)", 0, aligned_block_fill_stm8_armv4 },
{ "ARM fill (STM with 4 registers)", 0, aligned_block_fill_stm4_armv4 },
{ NULL, 0, NULL }
};
bench_info *get_asm_benchmarks(void)
{
if (check_cpu_feature("neon"))
return arm_neon;
else if (check_cpu_feature("edsp") && check_cpu_feature("vfp"))
return arm_v5te_vfp;
else if (check_cpu_feature("edsp"))
return arm_v5te;
else
return arm_v4;
}
static bench_info arm_neon_fb[] =
{
{ "NEON read (from framebuffer)", 0, aligned_block_read_neon },
{ "NEON copy (from framebuffer)", 0, aligned_block_copy_neon },
{ "NEON 2-pass copy (from framebuffer)", 1, aligned_block_copy_neon },
{ "NEON unrolled copy (from framebuffer)", 0, aligned_block_copy_unrolled_neon },
{ "NEON 2-pass unrolled copy (from framebuffer)", 1, aligned_block_copy_unrolled_neon },
{ "VFP copy (from framebuffer)", 0, aligned_block_copy_vfp },
{ "VFP 2-pass copy (from framebuffer)", 1, aligned_block_copy_vfp },
{ "ARM copy (from framebuffer)", 0, aligned_block_copy_incr_armv5te },
{ "ARM 2-pass copy (from framebuffer)", 1, aligned_block_copy_incr_armv5te },
{ NULL, 0, NULL }
};
static bench_info arm_v5te_vfp_fb[] =
{
{ "VFP copy (from framebuffer)", 0, aligned_block_copy_vfp },
{ "VFP 2-pass copy (from framebuffer)", 1, aligned_block_copy_vfp },
{ "ARM copy (from framebuffer)", 0, aligned_block_copy_incr_armv5te },
{ "ARM 2-pass copy (from framebuffer)", 1, aligned_block_copy_incr_armv5te },
{ NULL, 0, NULL }
};
static bench_info arm_v5te_fb[] =
{
{ "ARM copy (from framebuffer)", 0, aligned_block_copy_incr_armv5te },
{ "ARM 2-pass copy (from framebuffer)", 1, aligned_block_copy_incr_armv5te },
{ NULL, 0, NULL }
};
bench_info *get_asm_framebuffer_benchmarks(void)
{
if (check_cpu_feature("neon"))
return arm_neon_fb;
else if (check_cpu_feature("edsp") && check_cpu_feature("vfp"))
return arm_v5te_vfp_fb;
else if (check_cpu_feature("edsp"))
return arm_v5te_fb;
else
return empty;
}
#elif defined(__aarch64__)
#include "aarch64-asm.h"
static bench_info aarch64_neon[] =
{
{ "NEON LDP/STP copy", 0, aligned_block_copy_ldpstp_q_aarch64 },
{ "NEON LDP/STP copy pldl2strm (32 bytes step)", 0, aligned_block_copy_ldpstp_q_pf32_l2strm_aarch64 },
{ "NEON LDP/STP copy pldl2strm (64 bytes step)", 0, aligned_block_copy_ldpstp_q_pf64_l2strm_aarch64 },
{ "NEON LDP/STP copy pldl1keep (32 bytes step)", 0, aligned_block_copy_ldpstp_q_pf32_l1keep_aarch64 },
{ "NEON LDP/STP copy pldl1keep (64 bytes step)", 0, aligned_block_copy_ldpstp_q_pf64_l1keep_aarch64 },
{ "NEON LD1/ST1 copy", 0, aligned_block_copy_ld1st1_aarch64 },
{ "NEON STP fill", 0, aligned_block_fill_stp_q_aarch64 },
{ "NEON STNP fill", 0, aligned_block_fill_stnp_q_aarch64 },
{ "ARM LDP/STP copy", 0, aligned_block_copy_ldpstp_x_aarch64 },
{ "ARM STP fill", 0, aligned_block_fill_stp_x_aarch64 },
{ "ARM STNP fill", 0, aligned_block_fill_stnp_x_aarch64 },
{ NULL, 0, NULL }
};
static bench_info aarch64_neon_fb[] =
{
{ "NEON LDP/STP copy (from framebuffer)", 0, aligned_block_copy_ldpstp_q_aarch64 },
{ "NEON LDP/STP 2-pass copy (from framebuffer)", 1, aligned_block_copy_ldpstp_q_aarch64 },
{ "NEON LD1/ST1 copy (from framebuffer)", 0, aligned_block_copy_ld1st1_aarch64 },
{ "NEON LD1/ST1 2-pass copy (from framebuffer)", 1, aligned_block_copy_ld1st1_aarch64 },
{ "ARM LDP/STP copy (from framebuffer)", 0, aligned_block_copy_ldpstp_x_aarch64 },
{ "ARM LDP/STP 2-pass copy (from framebuffer)", 1, aligned_block_copy_ldpstp_x_aarch64 },
{ NULL, 0, NULL }
};
bench_info *get_asm_benchmarks(void)
{
return aarch64_neon;
}
bench_info *get_asm_framebuffer_benchmarks(void)
{
return aarch64_neon_fb;
}
#elif defined(__mips__) && defined(_ABIO32)
#include "mips-32.h"
static bench_info mips_32[] =
{
{ "MIPS32 copy prefetched (32 bytes step)", 0, aligned_block_copy_pf32_mips32 },
{ "MIPS32 2-pass copy prefetched (32 bytes step)", 1, aligned_block_copy_pf32_mips32 },
{ "MIPS32 fill prefetched (32 bytes step)", 0, aligned_block_fill_pf32_mips32 },
{ NULL, 0, NULL }
};
bench_info *get_asm_benchmarks(void)
{
/* Enable only for MIPS32 processors which have 32 bytes cache line */
if (check_cpu_feature("MIPS 24K") ||
check_cpu_feature("MIPS 24Kc") ||
check_cpu_feature("MIPS 24Kf") ||
check_cpu_feature("MIPS 74K") ||
check_cpu_feature("MIPS 74Kc") ||
check_cpu_feature("MIPS 74Kf") ||
check_cpu_feature("MIPS 1004K") ||
check_cpu_feature("MIPS 1004Kc") ||
check_cpu_feature("MIPS 1004Kf") ||
check_cpu_feature("MIPS 1074K") ||
check_cpu_feature("MIPS 1074Kc") ||
check_cpu_feature("MIPS 1074Kf"))
{
return mips_32;
}
else
{
return empty;
}
}
bench_info *get_asm_framebuffer_benchmarks(void)
{
return empty;
}
#else
bench_info *get_asm_benchmarks(void)
{
return empty;
}
bench_info *get_asm_framebuffer_benchmarks(void)
{
return empty;
}
#endif
|