FFmpeg
checkasm.c
Go to the documentation of this file.
1 /*
2  * Assembly testing and benchmarking tool
3  * Copyright (c) 2015 Henrik Gramner
4  * Copyright (c) 2008 Loren Merritt
5  *
6  * This file is part of FFmpeg.
7  *
8  * FFmpeg is free software; you can redistribute it and/or modify
9  * it under the terms of the GNU General Public License as published by
10  * the Free Software Foundation; either version 2 of the License, or
11  * (at your option) any later version.
12  *
13  * FFmpeg is distributed in the hope that it will be useful,
14  * but WITHOUT ANY WARRANTY; without even the implied warranty of
15  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
16  * GNU General Public License for more details.
17  *
18  * You should have received a copy of the GNU General Public License along
19  * with FFmpeg; if not, write to the Free Software Foundation, Inc.,
20  * 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
21  */
22 
23 #include "config.h"
24 #include "config_components.h"
25 
26 #ifndef _GNU_SOURCE
27 # define _GNU_SOURCE // for syscall (performance monitoring API), strsignal()
28 #endif
29 
30 #include <signal.h>
31 #include <stdarg.h>
32 #include <stdio.h>
33 #include <stdlib.h>
34 #include <string.h>
35 #include "checkasm.h"
36 #include "libavutil/common.h"
37 #include "libavutil/cpu.h"
38 #include "libavutil/intfloat.h"
39 #include "libavutil/random_seed.h"
40 
41 #if HAVE_IO_H
42 #include <io.h>
43 #endif
44 
45 #if HAVE_SETCONSOLETEXTATTRIBUTE && HAVE_GETSTDHANDLE
46 #include <windows.h>
47 #define COLOR_RED FOREGROUND_RED
48 #define COLOR_GREEN FOREGROUND_GREEN
49 #define COLOR_YELLOW (FOREGROUND_RED|FOREGROUND_GREEN)
50 #else
51 #define COLOR_RED 1
52 #define COLOR_GREEN 2
53 #define COLOR_YELLOW 3
54 #endif
55 
56 #if HAVE_UNISTD_H
57 #include <unistd.h>
58 #endif
59 
60 #if !HAVE_ISATTY
61 #define isatty(fd) 1
62 #endif
63 
64 #if ARCH_ARM && HAVE_ARMV5TE_EXTERNAL
65 #include "libavutil/arm/cpu.h"
66 
67 void (*checkasm_checked_call)(void *func, int dummy, ...) = checkasm_checked_call_novfp;
68 #endif
69 
70 /* List of tests to invoke */
71 static const struct {
72  const char *name;
73  void (*func)(void);
74 } tests[] = {
75 #if CONFIG_AVCODEC
76  #if CONFIG_AAC_DECODER
77  { "aacpsdsp", checkasm_check_aacpsdsp },
78  { "sbrdsp", checkasm_check_sbrdsp },
79  #endif
80  #if CONFIG_AC3DSP
81  { "ac3dsp", checkasm_check_ac3dsp },
82  #endif
83  #if CONFIG_ALAC_DECODER
84  { "alacdsp", checkasm_check_alacdsp },
85  #endif
86  #if CONFIG_AUDIODSP
87  { "audiodsp", checkasm_check_audiodsp },
88  #endif
89  #if CONFIG_BLOCKDSP
90  { "blockdsp", checkasm_check_blockdsp },
91  #endif
92  #if CONFIG_BSWAPDSP
93  { "bswapdsp", checkasm_check_bswapdsp },
94  #endif
95  #if CONFIG_DCA_DECODER
96  { "synth_filter", checkasm_check_synth_filter },
97  #endif
98  #if CONFIG_EXR_DECODER
99  { "exrdsp", checkasm_check_exrdsp },
100  #endif
101  #if CONFIG_FLAC_DECODER
102  { "flacdsp", checkasm_check_flacdsp },
103  #endif
104  #if CONFIG_FMTCONVERT
105  { "fmtconvert", checkasm_check_fmtconvert },
106  #endif
107  #if CONFIG_G722DSP
108  { "g722dsp", checkasm_check_g722dsp },
109  #endif
110  #if CONFIG_H264CHROMA
111  { "h264chroma", checkasm_check_h264chroma },
112  #endif
113  #if CONFIG_H264DSP
114  { "h264dsp", checkasm_check_h264dsp },
115  #endif
116  #if CONFIG_H264PRED
117  { "h264pred", checkasm_check_h264pred },
118  #endif
119  #if CONFIG_H264QPEL
120  { "h264qpel", checkasm_check_h264qpel },
121  #endif
122  #if CONFIG_HEVC_DECODER
123  { "hevc_add_res", checkasm_check_hevc_add_res },
124  { "hevc_deblock", checkasm_check_hevc_deblock },
125  { "hevc_idct", checkasm_check_hevc_idct },
126  { "hevc_pel", checkasm_check_hevc_pel },
127  { "hevc_sao", checkasm_check_hevc_sao },
128  #endif
129  #if CONFIG_HUFFYUV_DECODER
130  { "huffyuvdsp", checkasm_check_huffyuvdsp },
131  #endif
132  #if CONFIG_IDCTDSP
133  { "idctdsp", checkasm_check_idctdsp },
134  #endif
135  #if CONFIG_JPEG2000_DECODER
136  { "jpeg2000dsp", checkasm_check_jpeg2000dsp },
137  #endif
138  #if CONFIG_LLAUDDSP
139  { "llauddsp", checkasm_check_llauddsp },
140  #endif
141  #if CONFIG_HUFFYUVDSP
142  { "llviddsp", checkasm_check_llviddsp },
143  #endif
144  #if CONFIG_LLVIDENCDSP
145  { "llviddspenc", checkasm_check_llviddspenc },
146  #endif
147  #if CONFIG_LPC
148  { "lpc", checkasm_check_lpc },
149  #endif
150  #if CONFIG_ME_CMP
151  { "motion", checkasm_check_motion },
152  #endif
153  #if CONFIG_OPUS_DECODER
154  { "opusdsp", checkasm_check_opusdsp },
155  #endif
156  #if CONFIG_PIXBLOCKDSP
157  { "pixblockdsp", checkasm_check_pixblockdsp },
158  #endif
159  #if CONFIG_UTVIDEO_DECODER
160  { "utvideodsp", checkasm_check_utvideodsp },
161  #endif
162  #if CONFIG_V210_DECODER
163  { "v210dec", checkasm_check_v210dec },
164  #endif
165  #if CONFIG_V210_ENCODER
166  { "v210enc", checkasm_check_v210enc },
167  #endif
168  #if CONFIG_VC1DSP
169  { "vc1dsp", checkasm_check_vc1dsp },
170  #endif
171  #if CONFIG_VP8DSP
172  { "vp8dsp", checkasm_check_vp8dsp },
173  #endif
174  #if CONFIG_VP9_DECODER
175  { "vp9dsp", checkasm_check_vp9dsp },
176  #endif
177  #if CONFIG_VIDEODSP
178  { "videodsp", checkasm_check_videodsp },
179  #endif
180  #if CONFIG_VORBIS_DECODER
181  { "vorbisdsp", checkasm_check_vorbisdsp },
182  #endif
183 #endif
184 #if CONFIG_AVFILTER
185  #if CONFIG_AFIR_FILTER
186  { "af_afir", checkasm_check_afir },
187  #endif
188  #if CONFIG_BLEND_FILTER
189  { "vf_blend", checkasm_check_blend },
190  #endif
191  #if CONFIG_BWDIF_FILTER
192  { "vf_bwdif", checkasm_check_vf_bwdif },
193  #endif
194  #if CONFIG_COLORSPACE_FILTER
195  { "vf_colorspace", checkasm_check_colorspace },
196  #endif
197  #if CONFIG_EQ_FILTER
198  { "vf_eq", checkasm_check_vf_eq },
199  #endif
200  #if CONFIG_GBLUR_FILTER
201  { "vf_gblur", checkasm_check_vf_gblur },
202  #endif
203  #if CONFIG_HFLIP_FILTER
204  { "vf_hflip", checkasm_check_vf_hflip },
205  #endif
206  #if CONFIG_NLMEANS_FILTER
207  { "vf_nlmeans", checkasm_check_nlmeans },
208  #endif
209  #if CONFIG_THRESHOLD_FILTER
210  { "vf_threshold", checkasm_check_vf_threshold },
211  #endif
212  #if CONFIG_SOBEL_FILTER
213  { "vf_sobel", checkasm_check_vf_sobel },
214  #endif
215 #endif
216 #if CONFIG_SWSCALE
217  { "sw_gbrp", checkasm_check_sw_gbrp },
218  { "sw_rgb", checkasm_check_sw_rgb },
219  { "sw_scale", checkasm_check_sw_scale },
220 #endif
221 #if CONFIG_AVUTIL
222  { "fixed_dsp", checkasm_check_fixed_dsp },
223  { "float_dsp", checkasm_check_float_dsp },
224  { "av_tx", checkasm_check_av_tx },
225 #endif
226  { NULL }
227 };
228 
229 /* List of cpu flags to check */
230 static const struct {
231  const char *name;
232  const char *suffix;
233  int flag;
234 } cpus[] = {
235 #if ARCH_AARCH64
236  { "ARMV8", "armv8", AV_CPU_FLAG_ARMV8 },
237  { "NEON", "neon", AV_CPU_FLAG_NEON },
238  { "DOTPROD", "dotprod", AV_CPU_FLAG_DOTPROD },
239  { "I8MM", "i8mm", AV_CPU_FLAG_I8MM },
240 #elif ARCH_ARM
241  { "ARMV5TE", "armv5te", AV_CPU_FLAG_ARMV5TE },
242  { "ARMV6", "armv6", AV_CPU_FLAG_ARMV6 },
243  { "ARMV6T2", "armv6t2", AV_CPU_FLAG_ARMV6T2 },
244  { "VFP", "vfp", AV_CPU_FLAG_VFP },
245  { "VFP_VM", "vfp_vm", AV_CPU_FLAG_VFP_VM },
246  { "VFPV3", "vfp3", AV_CPU_FLAG_VFPV3 },
247  { "NEON", "neon", AV_CPU_FLAG_NEON },
248 #elif ARCH_PPC
249  { "ALTIVEC", "altivec", AV_CPU_FLAG_ALTIVEC },
250  { "VSX", "vsx", AV_CPU_FLAG_VSX },
251  { "POWER8", "power8", AV_CPU_FLAG_POWER8 },
252 #elif ARCH_RISCV
253  { "RVI", "rvi", AV_CPU_FLAG_RVI },
254  { "RVF", "rvf", AV_CPU_FLAG_RVF },
255  { "RVD", "rvd", AV_CPU_FLAG_RVD },
256  { "RVBaddr", "rvb_a", AV_CPU_FLAG_RVB_ADDR },
257  { "RVBbasic", "rvb_b", AV_CPU_FLAG_RVB_BASIC },
258  { "RVVi32", "rvv_i32", AV_CPU_FLAG_RVV_I32 },
259  { "RVVf32", "rvv_f32", AV_CPU_FLAG_RVV_F32 },
260  { "RVVi64", "rvv_i64", AV_CPU_FLAG_RVV_I64 },
261  { "RVVf64", "rvv_f64", AV_CPU_FLAG_RVV_F64 },
262 #elif ARCH_MIPS
263  { "MMI", "mmi", AV_CPU_FLAG_MMI },
264  { "MSA", "msa", AV_CPU_FLAG_MSA },
265 #elif ARCH_X86
266  { "MMX", "mmx", AV_CPU_FLAG_MMX|AV_CPU_FLAG_CMOV },
267  { "MMXEXT", "mmxext", AV_CPU_FLAG_MMXEXT },
268  { "3DNOW", "3dnow", AV_CPU_FLAG_3DNOW },
269  { "3DNOWEXT", "3dnowext", AV_CPU_FLAG_3DNOWEXT },
270  { "SSE", "sse", AV_CPU_FLAG_SSE },
271  { "SSE2", "sse2", AV_CPU_FLAG_SSE2|AV_CPU_FLAG_SSE2SLOW },
272  { "SSE3", "sse3", AV_CPU_FLAG_SSE3|AV_CPU_FLAG_SSE3SLOW },
273  { "SSSE3", "ssse3", AV_CPU_FLAG_SSSE3|AV_CPU_FLAG_ATOM },
274  { "SSE4.1", "sse4", AV_CPU_FLAG_SSE4 },
275  { "SSE4.2", "sse42", AV_CPU_FLAG_SSE42 },
276  { "AES-NI", "aesni", AV_CPU_FLAG_AESNI },
277  { "AVX", "avx", AV_CPU_FLAG_AVX },
278  { "XOP", "xop", AV_CPU_FLAG_XOP },
279  { "FMA3", "fma3", AV_CPU_FLAG_FMA3 },
280  { "FMA4", "fma4", AV_CPU_FLAG_FMA4 },
281  { "AVX2", "avx2", AV_CPU_FLAG_AVX2 },
282  { "AVX-512", "avx512", AV_CPU_FLAG_AVX512 },
283  { "AVX-512ICL", "avx512icl", AV_CPU_FLAG_AVX512ICL },
284 #elif ARCH_LOONGARCH
285  { "LSX", "lsx", AV_CPU_FLAG_LSX },
286  { "LASX", "lasx", AV_CPU_FLAG_LASX },
287 #endif
288  { NULL }
289 };
290 
291 typedef struct CheckasmFuncVersion {
293  void *func;
294  int ok;
295  int cpu;
298 
299 /* Binary search tree node */
300 typedef struct CheckasmFunc {
301  struct CheckasmFunc *child[2];
303  uint8_t color; /* 0 = red, 1 = black */
304  char name[1];
305 } CheckasmFunc;
306 
307 /* Internal state */
308 static struct {
312  const char *current_test_name;
313  const char *bench_pattern;
317 
318  /* perf */
319  int nop_time;
320  int sysfd;
321 
322  int cpu_flag;
323  const char *cpu_flag_name;
324  const char *test_name;
325  int verbose;
326 } state;
327 
328 /* PRNG state */
330 
331 /* float compare support code */
332 static int is_negative(union av_intfloat32 u)
333 {
334  return u.i >> 31;
335 }
336 
337 int float_near_ulp(float a, float b, unsigned max_ulp)
338 {
339  union av_intfloat32 x, y;
340 
341  x.f = a;
342  y.f = b;
343 
344  if (is_negative(x) != is_negative(y)) {
345  // handle -0.0 == +0.0
346  return a == b;
347  }
348 
349  if (llabs((int64_t)x.i - y.i) <= max_ulp)
350  return 1;
351 
352  return 0;
353 }
354 
355 int float_near_ulp_array(const float *a, const float *b, unsigned max_ulp,
356  unsigned len)
357 {
358  unsigned i;
359 
360  for (i = 0; i < len; i++) {
361  if (!float_near_ulp(a[i], b[i], max_ulp))
362  return 0;
363  }
364  return 1;
365 }
366 
367 int float_near_abs_eps(float a, float b, float eps)
368 {
369  float abs_diff = fabsf(a - b);
370  if (abs_diff < eps)
371  return 1;
372 
373  fprintf(stderr, "test failed comparing %g with %g (abs diff=%g with EPS=%g)\n", a, b, abs_diff, eps);
374 
375  return 0;
376 }
377 
378 int float_near_abs_eps_array(const float *a, const float *b, float eps,
379  unsigned len)
380 {
381  unsigned i;
382 
383  for (i = 0; i < len; i++) {
384  if (!float_near_abs_eps(a[i], b[i], eps))
385  return 0;
386  }
387  return 1;
388 }
389 
390 int float_near_abs_eps_ulp(float a, float b, float eps, unsigned max_ulp)
391 {
392  return float_near_ulp(a, b, max_ulp) || float_near_abs_eps(a, b, eps);
393 }
394 
395 int float_near_abs_eps_array_ulp(const float *a, const float *b, float eps,
396  unsigned max_ulp, unsigned len)
397 {
398  unsigned i;
399 
400  for (i = 0; i < len; i++) {
401  if (!float_near_abs_eps_ulp(a[i], b[i], eps, max_ulp))
402  return 0;
403  }
404  return 1;
405 }
406 
407 int double_near_abs_eps(double a, double b, double eps)
408 {
409  double abs_diff = fabs(a - b);
410 
411  return abs_diff < eps;
412 }
413 
414 int double_near_abs_eps_array(const double *a, const double *b, double eps,
415  unsigned len)
416 {
417  unsigned i;
418 
419  for (i = 0; i < len; i++) {
420  if (!double_near_abs_eps(a[i], b[i], eps))
421  return 0;
422  }
423  return 1;
424 }
425 
426 /* Print colored text to stderr if the terminal supports it */
427 static void color_printf(int color, const char *fmt, ...)
428 {
429  static int use_color = -1;
430  va_list arg;
431 
432 #if HAVE_SETCONSOLETEXTATTRIBUTE && HAVE_GETSTDHANDLE
433  static HANDLE con;
434  static WORD org_attributes;
435 
436  if (use_color < 0) {
437  CONSOLE_SCREEN_BUFFER_INFO con_info;
438  con = GetStdHandle(STD_ERROR_HANDLE);
439  if (con && con != INVALID_HANDLE_VALUE && GetConsoleScreenBufferInfo(con, &con_info)) {
440  org_attributes = con_info.wAttributes;
441  use_color = 1;
442  } else
443  use_color = 0;
444  }
445  if (use_color)
446  SetConsoleTextAttribute(con, (org_attributes & 0xfff0) | (color & 0x0f));
447 #else
448  if (use_color < 0) {
449  const char *term = getenv("TERM");
450  use_color = term && strcmp(term, "dumb") && isatty(2);
451  }
452  if (use_color)
453  fprintf(stderr, "\x1b[%d;3%dm", (color & 0x08) >> 3, color & 0x07);
454 #endif
455 
456  va_start(arg, fmt);
457  vfprintf(stderr, fmt, arg);
458  va_end(arg);
459 
460  if (use_color) {
461 #if HAVE_SETCONSOLETEXTATTRIBUTE && HAVE_GETSTDHANDLE
462  SetConsoleTextAttribute(con, org_attributes);
463 #else
464  fprintf(stderr, "\x1b[0m");
465 #endif
466  }
467 }
468 
469 /* Deallocate a tree */
471 {
472  if (f) {
473  CheckasmFuncVersion *v = f->versions.next;
474  while (v) {
475  CheckasmFuncVersion *next = v->next;
476  free(v);
477  v = next;
478  }
479 
480  destroy_func_tree(f->child[0]);
481  destroy_func_tree(f->child[1]);
482  free(f);
483  }
484 }
485 
486 /* Allocate a zero-initialized block, clean up and exit on failure */
487 static void *checkasm_malloc(size_t size)
488 {
489  void *ptr = calloc(1, size);
490  if (!ptr) {
491  fprintf(stderr, "checkasm: malloc failed\n");
492  destroy_func_tree(state.funcs);
493  exit(1);
494  }
495  return ptr;
496 }
497 
498 /* Get the suffix of the specified cpu flag */
499 static const char *cpu_suffix(int cpu)
500 {
501  int i = FF_ARRAY_ELEMS(cpus);
502 
503  while (--i >= 0)
504  if (cpu & cpus[i].flag)
505  return cpus[i].suffix;
506 
507  return "c";
508 }
509 
510 static int cmp_nop(const void *a, const void *b)
511 {
512  return *(const uint16_t*)a - *(const uint16_t*)b;
513 }
514 
515 /* Measure the overhead of the timing code (in decicycles) */
516 static int measure_nop_time(void)
517 {
518  uint16_t nops[10000];
519  int i, nop_sum = 0;
520  av_unused const int sysfd = state.sysfd;
521 
522  uint64_t t = 0;
523  for (i = 0; i < 10000; i++) {
524  PERF_START(t);
525  PERF_STOP(t);
526  nops[i] = t;
527  }
528 
529  qsort(nops, 10000, sizeof(uint16_t), cmp_nop);
530  for (i = 2500; i < 7500; i++)
531  nop_sum += nops[i];
532 
533  return nop_sum / 500;
534 }
535 
536 /* Print benchmark results */
538 {
539  if (f) {
540  print_benchs(f->child[0]);
541 
542  /* Only print functions with at least one assembly version */
543  if (f->versions.cpu || f->versions.next) {
544  CheckasmFuncVersion *v = &f->versions;
545  do {
546  CheckasmPerf *p = &v->perf;
547  if (p->iterations) {
548  int decicycles = (10*p->cycles/p->iterations - state.nop_time) / 4;
549  printf("%s_%s: %d.%d\n", f->name, cpu_suffix(v->cpu), decicycles/10, decicycles%10);
550  }
551  } while ((v = v->next));
552  }
553 
554  print_benchs(f->child[1]);
555  }
556 }
557 
558 /* ASCIIbetical sort except preserving natural order for numbers */
559 static int cmp_func_names(const char *a, const char *b)
560 {
561  const char *start = a;
562  int ascii_diff, digit_diff;
563 
564  for (; !(ascii_diff = *(const unsigned char*)a - *(const unsigned char*)b) && *a; a++, b++);
565  for (; av_isdigit(*a) && av_isdigit(*b); a++, b++);
566 
567  if (a > start && av_isdigit(a[-1]) && (digit_diff = av_isdigit(*a) - av_isdigit(*b)))
568  return digit_diff;
569 
570  return ascii_diff;
571 }
572 
573 /* Perform a tree rotation in the specified direction and return the new root */
575 {
576  CheckasmFunc *r = f->child[dir^1];
577  f->child[dir^1] = r->child[dir];
578  r->child[dir] = f;
579  r->color = f->color;
580  f->color = 0;
581  return r;
582 }
583 
584 #define is_red(f) ((f) && !(f)->color)
585 
586 /* Balance a left-leaning red-black tree at the specified node */
587 static void balance_tree(CheckasmFunc **root)
588 {
589  CheckasmFunc *f = *root;
590 
591  if (is_red(f->child[0]) && is_red(f->child[1])) {
592  f->color ^= 1;
593  f->child[0]->color = f->child[1]->color = 1;
594  }
595 
596  if (!is_red(f->child[0]) && is_red(f->child[1]))
597  *root = rotate_tree(f, 0); /* Rotate left */
598  else if (is_red(f->child[0]) && is_red(f->child[0]->child[0]))
599  *root = rotate_tree(f, 1); /* Rotate right */
600 }
601 
602 /* Get a node with the specified name, creating it if it doesn't exist */
603 static CheckasmFunc *get_func(CheckasmFunc **root, const char *name)
604 {
605  CheckasmFunc *f = *root;
606 
607  if (f) {
608  /* Search the tree for a matching node */
609  int cmp = cmp_func_names(name, f->name);
610  if (cmp) {
611  f = get_func(&f->child[cmp > 0], name);
612 
613  /* Rebalance the tree on the way up if a new node was inserted */
614  if (!f->versions.func)
615  balance_tree(root);
616  }
617  } else {
618  /* Allocate and insert a new node into the tree */
619  int name_length = strlen(name);
620  f = *root = checkasm_malloc(sizeof(CheckasmFunc) + name_length);
621  memcpy(f->name, name, name_length + 1);
622  }
623 
624  return f;
625 }
626 
627 /* Perform tests and benchmarks for the specified cpu flag if supported by the host */
628 static void check_cpu_flag(const char *name, int flag)
629 {
630  int old_cpu_flag = state.cpu_flag;
631 
632  flag |= old_cpu_flag;
633  av_force_cpu_flags(-1);
634  state.cpu_flag = flag & av_get_cpu_flags();
635  av_force_cpu_flags(state.cpu_flag);
636 
637  if (!flag || state.cpu_flag != old_cpu_flag) {
638  int i;
639 
640  state.cpu_flag_name = name;
641  for (i = 0; tests[i].func; i++) {
642  if (state.test_name && strcmp(tests[i].name, state.test_name))
643  continue;
644  state.current_test_name = tests[i].name;
645  tests[i].func();
646  }
647  }
648 }
649 
650 /* Print the name of the current CPU flag, but only do it once */
651 static void print_cpu_name(void)
652 {
653  if (state.cpu_flag_name) {
654  color_printf(COLOR_YELLOW, "%s:\n", state.cpu_flag_name);
655  state.cpu_flag_name = NULL;
656  }
657 }
658 
659 #if CONFIG_LINUX_PERF
660 static int bench_init_linux(void)
661 {
662  struct perf_event_attr attr = {
663  .type = PERF_TYPE_HARDWARE,
664  .size = sizeof(struct perf_event_attr),
665  .config = PERF_COUNT_HW_CPU_CYCLES,
666  .disabled = 1, // start counting only on demand
667  .exclude_kernel = 1,
668  .exclude_hv = 1,
669  };
670 
671  printf("benchmarking with Linux Perf Monitoring API\n");
672 
673  state.sysfd = syscall(__NR_perf_event_open, &attr, 0, -1, -1, 0);
674  if (state.sysfd == -1) {
675  perror("perf_event_open");
676  return -1;
677  }
678  return 0;
679 }
680 #elif CONFIG_MACOS_KPERF
681 static int bench_init_kperf(void)
682 {
683  ff_kperf_init();
684  return 0;
685 }
686 #else
687 static int bench_init_ffmpeg(void)
688 {
689 #ifdef AV_READ_TIME
690  printf("benchmarking with native FFmpeg timers\n");
691  return 0;
692 #else
693  fprintf(stderr, "checkasm: --bench is not supported on your system\n");
694  return -1;
695 #endif
696 }
697 #endif
698 
699 static int bench_init(void)
700 {
701 #if CONFIG_LINUX_PERF
702  int ret = bench_init_linux();
703 #elif CONFIG_MACOS_KPERF
704  int ret = bench_init_kperf();
705 #else
706  int ret = bench_init_ffmpeg();
707 #endif
708  if (ret < 0)
709  return ret;
710 
711  state.nop_time = measure_nop_time();
712  printf("nop: %d.%d\n", state.nop_time/10, state.nop_time%10);
713  return 0;
714 }
715 
716 static void bench_uninit(void)
717 {
718 #if CONFIG_LINUX_PERF
719  if (state.sysfd > 0)
720  close(state.sysfd);
721 #endif
722 }
723 
724 static int usage(const char *path)
725 {
726  fprintf(stderr,
727  "Usage: %s [--bench] [--test=<pattern>] [--verbose] [seed]\n",
728  path);
729  return 1;
730 }
731 
732 int main(int argc, char *argv[])
733 {
734  unsigned int seed = av_get_random_seed();
735  int i, ret = 0;
736 
737 #if ARCH_ARM && HAVE_ARMV5TE_EXTERNAL
739  checkasm_checked_call = checkasm_checked_call_vfp;
740 #endif
741 #if ARCH_RISCV
742  struct sigaction act = {
743  .sa_handler = checkasm_handle_signal,
744  .sa_flags = 0,
745  };
746 
747  sigaction(SIGILL, &act, NULL);
748 #endif
749 
750  if (!tests[0].func || !cpus[0].flag) {
751  fprintf(stderr, "checkasm: no tests to perform\n");
752  return 0;
753  }
754 
755  for (i = 1; i < argc; i++) {
756  const char *arg = argv[i];
757  unsigned long l;
758  char *end;
759 
760  if (!strncmp(arg, "--bench", 7)) {
761  if (bench_init() < 0)
762  return 1;
763  if (arg[7] == '=') {
764  state.bench_pattern = arg + 8;
765  state.bench_pattern_len = strlen(state.bench_pattern);
766  } else
767  state.bench_pattern = "";
768  } else if (!strncmp(arg, "--test=", 7)) {
769  state.test_name = arg + 7;
770  } else if (!strcmp(arg, "--verbose") || !strcmp(arg, "-v")) {
771  state.verbose = 1;
772  } else if ((l = strtoul(arg, &end, 10)) <= UINT_MAX &&
773  *end == '\0') {
774  seed = l;
775  } else {
776  return usage(argv[0]);
777  }
778  }
779 
780  fprintf(stderr, "checkasm: using random seed %u\n", seed);
782 
783  check_cpu_flag(NULL, 0);
784  for (i = 0; cpus[i].flag; i++)
786 
787  if (state.num_failed) {
788  fprintf(stderr, "checkasm: %d of %d tests have failed\n", state.num_failed, state.num_checked);
789  ret = 1;
790  } else {
791  fprintf(stderr, "checkasm: all %d tests passed\n", state.num_checked);
792  if (state.bench_pattern) {
793  print_benchs(state.funcs);
794  }
795  }
796 
797  destroy_func_tree(state.funcs);
798  bench_uninit();
799  return ret;
800 }
801 
802 /* Decide whether or not the specified function needs to be tested and
803  * allocate/initialize data structures if needed. Returns a pointer to a
804  * reference function if the function should be tested, otherwise NULL */
805 void *checkasm_check_func(void *func, const char *name, ...)
806 {
807  char name_buf[256];
808  void *ref = func;
810  int name_length;
811  va_list arg;
812 
813  va_start(arg, name);
814  name_length = vsnprintf(name_buf, sizeof(name_buf), name, arg);
815  va_end(arg);
816 
817  if (!func || name_length <= 0 || name_length >= sizeof(name_buf))
818  return NULL;
819 
820  state.current_func = get_func(&state.funcs, name_buf);
821  state.funcs->color = 1;
822  v = &state.current_func->versions;
823 
824  if (v->func) {
825  CheckasmFuncVersion *prev;
826  do {
827  /* Only test functions that haven't already been tested */
828  if (v->func == func)
829  return NULL;
830 
831  if (v->ok)
832  ref = v->func;
833 
834  prev = v;
835  } while ((v = v->next));
836 
837  v = prev->next = checkasm_malloc(sizeof(CheckasmFuncVersion));
838  }
839 
840  v->func = func;
841  v->ok = 1;
842  v->cpu = state.cpu_flag;
843  state.current_func_ver = v;
844 
845  if (state.cpu_flag)
846  state.num_checked++;
847 
848  return ref;
849 }
850 
851 /* Decide whether or not the current function needs to be benchmarked */
853 {
854  return !state.num_failed && state.bench_pattern &&
855  !strncmp(state.current_func->name, state.bench_pattern, state.bench_pattern_len);
856 }
857 
858 /* Indicate that the current test has failed */
859 void checkasm_fail_func(const char *msg, ...)
860 {
861  if (state.current_func_ver->cpu && state.current_func_ver->ok) {
862  va_list arg;
863 
864  print_cpu_name();
865  fprintf(stderr, " %s_%s (", state.current_func->name, cpu_suffix(state.current_func_ver->cpu));
866  va_start(arg, msg);
867  vfprintf(stderr, msg, arg);
868  va_end(arg);
869  fprintf(stderr, ")\n");
870 
871  state.current_func_ver->ok = 0;
872  state.num_failed++;
873  }
874 }
875 
876 void checkasm_fail_signal(int signum)
877 {
878 #ifdef __GLIBC__
879  checkasm_fail_func("fatal signal %d: %s", signum, strsignal(signum));
880 #else
881  checkasm_fail_func("fatal signal %d", signum);
882 #endif
883 }
884 
885 /* Get the benchmark context of the current function */
887 {
888  CheckasmPerf *perf = &state.current_func_ver->perf;
889  memset(perf, 0, sizeof(*perf));
890  perf->sysfd = state.sysfd;
891  return perf;
892 }
893 
894 /* Print the outcome of all tests performed since the last time this function was called */
895 void checkasm_report(const char *name, ...)
896 {
897  static int prev_checked, prev_failed, max_length;
898 
899  if (state.num_checked > prev_checked) {
900  int pad_length = max_length + 4;
901  va_list arg;
902 
903  print_cpu_name();
904  pad_length -= fprintf(stderr, " - %s.", state.current_test_name);
905  va_start(arg, name);
906  pad_length -= vfprintf(stderr, name, arg);
907  va_end(arg);
908  fprintf(stderr, "%*c", FFMAX(pad_length, 0) + 2, '[');
909 
910  if (state.num_failed == prev_failed)
911  color_printf(COLOR_GREEN, "OK");
912  else
913  color_printf(COLOR_RED, "FAILED");
914  fprintf(stderr, "]\n");
915 
916  prev_checked = state.num_checked;
917  prev_failed = state.num_failed;
918  } else if (!state.cpu_flag) {
919  /* Calculate the amount of padding required to make the output vertically aligned */
920  int length = strlen(state.current_test_name);
921  va_list arg;
922 
923  va_start(arg, name);
924  length += vsnprintf(NULL, 0, name, arg);
925  va_end(arg);
926 
927  if (length > max_length)
928  max_length = length;
929  }
930 }
931 
932 #define DEF_CHECKASM_CHECK_FUNC(type, fmt) \
933 int checkasm_check_##type(const char *const file, const int line, \
934  const type *buf1, ptrdiff_t stride1, \
935  const type *buf2, ptrdiff_t stride2, \
936  const int w, int h, const char *const name) \
937 { \
938  int y = 0; \
939  stride1 /= sizeof(*buf1); \
940  stride2 /= sizeof(*buf2); \
941  for (y = 0; y < h; y++) \
942  if (memcmp(&buf1[y*stride1], &buf2[y*stride2], w*sizeof(*buf1))) \
943  break; \
944  if (y == h) \
945  return 0; \
946  checkasm_fail_func("%s:%d", file, line); \
947  if (!state.verbose) \
948  return 1; \
949  fprintf(stderr, "%s:\n", name); \
950  while (h--) { \
951  for (int x = 0; x < w; x++) \
952  fprintf(stderr, " " fmt, buf1[x]); \
953  fprintf(stderr, " "); \
954  for (int x = 0; x < w; x++) \
955  fprintf(stderr, " " fmt, buf2[x]); \
956  fprintf(stderr, " "); \
957  for (int x = 0; x < w; x++) \
958  fprintf(stderr, "%c", buf1[x] != buf2[x] ? 'x' : '.'); \
959  buf1 += stride1; \
960  buf2 += stride2; \
961  fprintf(stderr, "\n"); \
962  } \
963  return 1; \
964 }
965 
966 DEF_CHECKASM_CHECK_FUNC(uint8_t, "%02x")
967 DEF_CHECKASM_CHECK_FUNC(uint16_t, "%04x")
968 DEF_CHECKASM_CHECK_FUNC(uint32_t, "%08x")
969 DEF_CHECKASM_CHECK_FUNC(int16_t, "%6d")
checkasm_check_nlmeans
void checkasm_check_nlmeans(void)
Definition: vf_nlmeans.c:31
AV_CPU_FLAG_VFP
#define AV_CPU_FLAG_VFP
Definition: cpu.h:67
checkasm_check_sw_gbrp
void checkasm_check_sw_gbrp(void)
Definition: sw_gbrp.c:401
bench_pattern_len
int bench_pattern_len
Definition: checkasm.c:314
av_force_cpu_flags
void av_force_cpu_flags(int arg)
Disables cpu detection and forces the specified flags.
Definition: cpu.c:75
checkasm_check_vp8dsp
void checkasm_check_vp8dsp(void)
Definition: vp8dsp.c:508
checkasm_check_blockdsp
void checkasm_check_blockdsp(void)
Definition: blockdsp.c:55
AV_CPU_FLAG_SSE3
#define AV_CPU_FLAG_SSE3
Prescott SSE3 functions.
Definition: cpu.h:38
destroy_func_tree
static void destroy_func_tree(CheckasmFunc *f)
Definition: checkasm.c:470
r
const char * r
Definition: vf_curves.c:126
checkasm_check_videodsp
void checkasm_check_videodsp(void)
Definition: videodsp.c:81
cmp_nop
static int cmp_nop(const void *a, const void *b)
Definition: checkasm.c:510
checkasm_check_vf_eq
void checkasm_check_vf_eq(void)
Definition: vf_eq.c:76
AV_CPU_FLAG_RVB_BASIC
#define AV_CPU_FLAG_RVB_BASIC
Basic bit-manipulations.
Definition: cpu.h:91
COLOR_RED
#define COLOR_RED
Definition: checkasm.c:51
tests
static const struct @362 tests[]
checkasm_lfg
AVLFG checkasm_lfg
Definition: checkasm.c:329
color
Definition: vf_paletteuse.c:511
av_lfg_init
av_cold void av_lfg_init(AVLFG *c, unsigned int seed)
Definition: lfg.c:32
u
#define u(width, name, range_min, range_max)
Definition: cbs_h2645.c:240
checkasm_check_alacdsp
void checkasm_check_alacdsp(void)
Definition: alacdsp.c:116
AV_CPU_FLAG_LSX
#define AV_CPU_FLAG_LSX
Definition: cpu.h:80
checkasm_check_v210dec
void checkasm_check_v210dec(void)
Definition: v210dec.c:47
float_near_abs_eps_array
int float_near_abs_eps_array(const float *a, const float *b, float eps, unsigned len)
Definition: checkasm.c:378
AV_CPU_FLAG_SSE3SLOW
#define AV_CPU_FLAG_SSE3SLOW
SSE3 supported, but usually not faster.
Definition: cpu.h:39
CheckasmFunc
Definition: checkasm.c:300
name
const char * name
Definition: checkasm.c:72
bench_pattern
const char * bench_pattern
Definition: checkasm.c:313
av_unused
#define av_unused
Definition: attributes.h:131
checkasm_check_hevc_idct
void checkasm_check_hevc_idct(void)
Definition: hevc_idct.c:108
checkasm_check_sw_scale
void checkasm_check_sw_scale(void)
Definition: sw_scale.c:361
CheckasmFunc::child
struct CheckasmFunc * child[2]
Definition: checkasm.c:301
checkasm_check_afir
void checkasm_check_afir(void)
Definition: af_afir.c:144
print_cpu_name
static void print_cpu_name(void)
Definition: checkasm.c:651
AV_CPU_FLAG_3DNOW
#define AV_CPU_FLAG_3DNOW
AMD 3DNOW.
Definition: cpu.h:32
AV_CPU_FLAG_RVF
#define AV_CPU_FLAG_RVF
F (single precision FP)
Definition: cpu.h:85
cpus
static const struct @363 cpus[]
av_intfloat32::i
uint32_t i
Definition: intfloat.h:28
float_near_abs_eps
int float_near_abs_eps(float a, float b, float eps)
Definition: checkasm.c:367
b
#define b
Definition: input.c:41
checkasm_check_aacpsdsp
void checkasm_check_aacpsdsp(void)
Definition: aacpsdsp.c:234
bench_uninit
static void bench_uninit(void)
Definition: checkasm.c:716
CheckasmFuncVersion::cpu
int cpu
Definition: checkasm.c:295
cpu_flag
int cpu_flag
Definition: checkasm.c:322
AV_CPU_FLAG_DOTPROD
#define AV_CPU_FLAG_DOTPROD
Definition: cpu.h:72
FFMAX
#define FFMAX(a, b)
Definition: macros.h:47
av_get_cpu_flags
int av_get_cpu_flags(void)
Return the flags which specify extensions supported by the CPU.
Definition: cpu.c:103
cpu_suffix
static const char * cpu_suffix(int cpu)
Definition: checkasm.c:499
CheckasmPerf::sysfd
int sysfd
Definition: checkasm.h:247
checkasm_check_llviddspenc
void checkasm_check_llviddspenc(void)
Definition: llviddspenc.c:104
intfloat.h
checkasm_check_h264dsp
void checkasm_check_h264dsp(void)
Definition: h264dsp.c:443
tf_sess_config.config
config
Definition: tf_sess_config.py:33
color_printf
static void color_printf(int color, const char *fmt,...)
Definition: checkasm.c:427
AV_CPU_FLAG_RVV_F64
#define AV_CPU_FLAG_RVV_F64
Vectors of double's.
Definition: cpu.h:90
av_get_random_seed
uint32_t av_get_random_seed(void)
Get a seed to use in conjunction with random functions.
Definition: random_seed.c:167
main
int main(int argc, char *argv[])
Definition: checkasm.c:732
checkasm_check_motion
void checkasm_check_motion(void)
Definition: motion.c:146
AV_CPU_FLAG_SSSE3
#define AV_CPU_FLAG_SSSE3
Conroe SSSE3 functions.
Definition: cpu.h:41
double_near_abs_eps
int double_near_abs_eps(double a, double b, double eps)
Definition: checkasm.c:407
CheckasmPerf
Definition: checkasm.h:246
current_test_name
const char * current_test_name
Definition: checkasm.c:312
PERF_START
#define PERF_START(t)
Definition: checkasm.h:302
rotate_tree
static CheckasmFunc * rotate_tree(CheckasmFunc *f, int dir)
Definition: checkasm.c:574
dummy
int dummy
Definition: motion.c:66
AV_CPU_FLAG_XOP
#define AV_CPU_FLAG_XOP
Bulldozer XOP functions.
Definition: cpu.h:49
checkasm_check_vp9dsp
void checkasm_check_vp9dsp(void)
Definition: vp9dsp.c:627
checkasm.h
checkasm_check_vf_threshold
void checkasm_check_vf_threshold(void)
Definition: vf_threshold.c:79
checkasm_check_vf_gblur
void checkasm_check_vf_gblur(void)
Definition: vf_gblur.c:81
AV_CPU_FLAG_3DNOWEXT
#define AV_CPU_FLAG_3DNOWEXT
AMD 3DNowExt.
Definition: cpu.h:37
float_near_abs_eps_ulp
int float_near_abs_eps_ulp(float a, float b, float eps, unsigned max_ulp)
Definition: checkasm.c:390
AV_CPU_FLAG_VSX
#define AV_CPU_FLAG_VSX
ISA 2.06.
Definition: cpu.h:61
fabsf
static __device__ float fabsf(float a)
Definition: cuda_runtime.h:181
AV_CPU_FLAG_RVV_F32
#define AV_CPU_FLAG_RVV_F32
Vectors of float's *‍/.
Definition: cpu.h:88
num_failed
int num_failed
Definition: checkasm.c:316
checkasm_check_float_dsp
void checkasm_check_float_dsp(void)
Definition: float_dsp.c:282
AV_CPU_FLAG_AVX512
#define AV_CPU_FLAG_AVX512
AVX-512 functions: requires OS support even if YMM/ZMM registers aren't used.
Definition: cpu.h:56
bench_init
static int bench_init(void)
Definition: checkasm.c:699
checkasm_check_pixblockdsp
void checkasm_check_pixblockdsp(void)
Definition: pixblockdsp.c:82
FF_ARRAY_ELEMS
#define FF_ARRAY_ELEMS(a)
Definition: sinewin_tablegen.c:29
CheckasmFuncVersion
Definition: checkasm.c:291
COLOR_YELLOW
#define COLOR_YELLOW
Definition: checkasm.c:53
nop_time
int nop_time
Definition: checkasm.c:319
CheckasmFuncVersion::perf
CheckasmPerf perf
Definition: checkasm.c:296
checkasm_check_vf_bwdif
void checkasm_check_vf_bwdif(void)
Definition: vf_bwdif.c:72
checkasm_report
void checkasm_report(const char *name,...)
Definition: checkasm.c:895
get_func
static CheckasmFunc * get_func(CheckasmFunc **root, const char *name)
Definition: checkasm.c:603
checkasm_fail_func
void checkasm_fail_func(const char *msg,...)
Definition: checkasm.c:859
checkasm_check_sw_rgb
void checkasm_check_sw_rgb(void)
Definition: sw_rgb.c:182
checkasm_check_hevc_sao
void checkasm_check_hevc_sao(void)
Definition: hevc_sao.c:143
num_checked
int num_checked
Definition: checkasm.c:315
AV_CPU_FLAG_ARMV6
#define AV_CPU_FLAG_ARMV6
Definition: cpu.h:65
AV_CPU_FLAG_SSE4
#define AV_CPU_FLAG_SSE4
Penryn SSE4.1 functions.
Definition: cpu.h:44
AV_CPU_FLAG_RVB_ADDR
#define AV_CPU_FLAG_RVB_ADDR
Address bit-manipulations.
Definition: cpu.h:92
checkasm_check_hevc_pel
void checkasm_check_hevc_pel(void)
Definition: hevc_pel.c:534
cmp
static av_always_inline int cmp(MpegEncContext *s, const int x, const int y, const int subx, const int suby, const int size, const int h, int ref_index, int src_index, me_cmp_func cmp_func, me_cmp_func chroma_cmp_func, const int flags)
compares a block (either a full macroblock or a partition thereof) against a proposed motion-compensa...
Definition: motion_est.c:262
CheckasmPerf::iterations
int iterations
Definition: checkasm.h:249
CheckasmFuncVersion::ok
int ok
Definition: checkasm.c:294
arg
const char * arg
Definition: jacosubdec.c:67
func
void(* func)(void)
Definition: checkasm.c:73
AV_CPU_FLAG_RVD
#define AV_CPU_FLAG_RVD
D (double precision FP)
Definition: cpu.h:86
fabs
static __device__ float fabs(float a)
Definition: cuda_runtime.h:182
NULL
#define NULL
Definition: coverity.c:32
measure_nop_time
static int measure_nop_time(void)
Definition: checkasm.c:516
double_near_abs_eps_array
int double_near_abs_eps_array(const double *a, const double *b, double eps, unsigned len)
Definition: checkasm.c:414
AV_CPU_FLAG_LASX
#define AV_CPU_FLAG_LASX
Definition: cpu.h:81
AV_CPU_FLAG_AVX512ICL
#define AV_CPU_FLAG_AVX512ICL
F/CD/BW/DQ/VL/VNNI/IFMA/VBMI/VBMI2/VPOPCNTDQ/BITALG/GFNI/VAES/VPCLMULQDQ.
Definition: cpu.h:57
checkasm_check_fmtconvert
void checkasm_check_fmtconvert(void)
Definition: fmtconvert.c:46
AV_CPU_FLAG_MSA
#define AV_CPU_FLAG_MSA
Definition: cpu.h:77
COLOR_GREEN
#define COLOR_GREEN
Definition: checkasm.c:52
checkasm_check_vorbisdsp
void checkasm_check_vorbisdsp(void)
Definition: vorbisdsp.c:78
cpu.h
av_intfloat32
Definition: intfloat.h:27
AV_CPU_FLAG_CMOV
#define AV_CPU_FLAG_CMOV
supports cmov instruction
Definition: cpu.h:51
seed
static unsigned int seed
Definition: videogen.c:78
AV_CPU_FLAG_ALTIVEC
#define AV_CPU_FLAG_ALTIVEC
standard
Definition: cpu.h:60
print_benchs
static void print_benchs(CheckasmFunc *f)
Definition: checkasm.c:537
current_func
CheckasmFunc * current_func
Definition: checkasm.c:310
checkasm_fail_signal
void checkasm_fail_signal(int signum)
Definition: checkasm.c:876
AV_CPU_FLAG_I8MM
#define AV_CPU_FLAG_I8MM
Definition: cpu.h:73
checkasm_check_vc1dsp
void checkasm_check_vc1dsp(void)
Definition: vc1dsp.c:441
checkasm_check_h264chroma
void checkasm_check_h264chroma(void)
Definition: h264chroma.c:81
AV_CPU_FLAG_SSE2
#define AV_CPU_FLAG_SSE2
PIV SSE2 functions.
Definition: cpu.h:34
CheckasmFunc::color
uint8_t color
Definition: checkasm.c:303
AVLFG
Context structure for the Lagged Fibonacci PRNG.
Definition: lfg.h:33
f
f
Definition: af_crystalizer.c:121
checkasm_check_llauddsp
void checkasm_check_llauddsp(void)
Definition: llauddsp.c:108
AV_CPU_FLAG_AVX
#define AV_CPU_FLAG_AVX
AVX functions: requires OS support even if YMM registers aren't used.
Definition: cpu.h:47
AV_CPU_FLAG_FMA4
#define AV_CPU_FLAG_FMA4
Bulldozer FMA4 functions.
Definition: cpu.h:50
cpu.h
AV_CPU_FLAG_AVX2
#define AV_CPU_FLAG_AVX2
AVX2 functions: requires OS support even if YMM registers aren't used.
Definition: cpu.h:52
checkasm_check_llviddsp
void checkasm_check_llviddsp(void)
Definition: llviddsp.c:195
AV_CPU_FLAG_SSE2SLOW
#define AV_CPU_FLAG_SSE2SLOW
SSE2 supported, but usually not faster.
Definition: cpu.h:35
AV_CPU_FLAG_NEON
#define AV_CPU_FLAG_NEON
Definition: cpu.h:69
size
int size
Definition: twinvq_data.h:10344
DEF_CHECKASM_CHECK_FUNC
#define DEF_CHECKASM_CHECK_FUNC(type, fmt)
Definition: checkasm.c:932
printf
printf("static const uint8_t my_array[100] = {\n")
sysfd
int sysfd
Definition: checkasm.c:320
have_neon
#define have_neon(flags)
Definition: cpu.h:26
CheckasmFuncVersion::func
void * func
Definition: checkasm.c:293
checkasm_check_h264pred
void checkasm_check_h264pred(void)
Definition: h264pred.c:233
av_isdigit
static av_const int av_isdigit(int c)
Locale-independent conversion of ASCII isdigit.
Definition: avstring.h:202
AV_CPU_FLAG_FMA3
#define AV_CPU_FLAG_FMA3
Haswell FMA3 functions.
Definition: cpu.h:53
a
The reader does not expect b to be semantically here and if the code is changed by maybe adding a a division or other the signedness will almost certainly be mistaken To avoid this confusion a new type was SUINT is the C unsigned type but it holds a signed int to use the same example SUINT a
Definition: undefined.txt:41
ff_kperf_init
void ff_kperf_init(void)
Definition: macos_kperf.c:99
checkasm_check_huffyuvdsp
void checkasm_check_huffyuvdsp(void)
Definition: huffyuvdsp.c:95
float_near_abs_eps_array_ulp
int float_near_abs_eps_array_ulp(const float *a, const float *b, float eps, unsigned max_ulp, unsigned len)
Definition: checkasm.c:395
test_name
const char * test_name
Definition: checkasm.c:324
CheckasmPerf::cycles
uint64_t cycles
Definition: checkasm.h:248
checkasm_get_perf_context
CheckasmPerf * checkasm_get_perf_context(void)
Definition: checkasm.c:886
checkasm_check_fixed_dsp
void checkasm_check_fixed_dsp(void)
Definition: fixed_dsp.c:133
checkasm_check_colorspace
void checkasm_check_colorspace(void)
Definition: vf_colorspace.c:309
checkasm_check_utvideodsp
void checkasm_check_utvideodsp(void)
Definition: utvideodsp.c:91
is_negative
static int is_negative(union av_intfloat32 u)
Definition: checkasm.c:332
checkasm_check_audiodsp
void checkasm_check_audiodsp(void)
Definition: audiodsp.c:52
i
#define i(width, name, range_min, range_max)
Definition: cbs_h2645.c:245
checkasm_check_flacdsp
void checkasm_check_flacdsp(void)
Definition: flacdsp.c:79
checkasm_check_func
void * checkasm_check_func(void *func, const char *name,...)
Definition: checkasm.c:805
AV_CPU_FLAG_SSE42
#define AV_CPU_FLAG_SSE42
Nehalem SSE4.2 functions.
Definition: cpu.h:45
checkasm_check_idctdsp
void checkasm_check_idctdsp(void)
Definition: idctdsp.c:95
CheckasmFuncVersion::next
struct CheckasmFuncVersion * next
Definition: checkasm.c:292
vsnprintf
#define vsnprintf
Definition: snprintf.h:36
common.h
use_color
static int use_color
Definition: log.c:123
AV_CPU_FLAG_RVV_I32
#define AV_CPU_FLAG_RVV_I32
Vectors of 8/16/32-bit int's *‍/.
Definition: cpu.h:87
checkasm_check_synth_filter
void checkasm_check_synth_filter(void)
Definition: synth_filter.c:47
cmp_func_names
static int cmp_func_names(const char *a, const char *b)
Definition: checkasm.c:559
AV_CPU_FLAG_ARMV8
#define AV_CPU_FLAG_ARMV8
Definition: cpu.h:70
checkasm_check_lpc
void checkasm_check_lpc(void)
Definition: lpc.c:60
len
int len
Definition: vorbis_enc_data.h:426
checkasm_check_g722dsp
void checkasm_check_g722dsp(void)
Definition: g722dsp.c:53
checkasm_check_ac3dsp
void checkasm_check_ac3dsp(void)
Definition: ac3dsp.c:64
PERF_STOP
#define PERF_STOP(t)
Definition: checkasm.h:303
is_red
#define is_red(f)
Definition: checkasm.c:584
AV_CPU_FLAG_ATOM
#define AV_CPU_FLAG_ATOM
Atom processor, some SSSE3 instructions are slower.
Definition: cpu.h:43
have_vfp
#define have_vfp(flags)
Definition: cpu.h:27
checkasm_check_vf_hflip
void checkasm_check_vf_hflip(void)
Definition: vf_hflip.c:72
ret
ret
Definition: filter_design.txt:187
state
static struct @364 state
checkasm_check_blend
void checkasm_check_blend(void)
Definition: vf_blend.c:88
verbose
int verbose
Definition: checkasm.c:325
AV_CPU_FLAG_VFPV3
#define AV_CPU_FLAG_VFPV3
Definition: cpu.h:68
checkasm_check_bswapdsp
void checkasm_check_bswapdsp(void)
Definition: bswapdsp.c:60
AV_CPU_FLAG_ARMV5TE
#define AV_CPU_FLAG_ARMV5TE
Definition: cpu.h:64
AV_CPU_FLAG_MMX
#define AV_CPU_FLAG_MMX
standard MMX
Definition: cpu.h:29
suffix
const char * suffix
Definition: checkasm.c:232
checkasm_malloc
static void * checkasm_malloc(size_t size)
Definition: checkasm.c:487
AV_CPU_FLAG_RVI
#define AV_CPU_FLAG_RVI
I (full GPR bank)
Definition: cpu.h:84
checkasm_check_h264qpel
void checkasm_check_h264qpel(void)
Definition: h264qpel.c:51
random_seed.h
AV_CPU_FLAG_AESNI
#define AV_CPU_FLAG_AESNI
Advanced Encryption Standard functions.
Definition: cpu.h:46
checkasm_bench_func
int checkasm_bench_func(void)
Definition: checkasm.c:852
AV_CPU_FLAG_POWER8
#define AV_CPU_FLAG_POWER8
ISA 2.07.
Definition: cpu.h:62
cpu_flag_name
const char * cpu_flag_name
Definition: checkasm.c:323
checkasm_check_vf_sobel
void checkasm_check_vf_sobel(void)
Definition: vf_convolution.c:100
AV_CPU_FLAG_SSE
#define AV_CPU_FLAG_SSE
SSE functions.
Definition: cpu.h:33
ref
static int ref[MAX_W *MAX_W]
Definition: jpeg2000dwt.c:112
float_near_ulp_array
int float_near_ulp_array(const float *a, const float *b, unsigned max_ulp, unsigned len)
Definition: checkasm.c:355
AV_CPU_FLAG_MMXEXT
#define AV_CPU_FLAG_MMXEXT
SSE integer functions or AMD MMX ext.
Definition: cpu.h:30
checkasm_check_av_tx
void checkasm_check_av_tx(void)
Definition: av_tx.c:101
av_intfloat32::f
float f
Definition: intfloat.h:29
flag
int flag
Definition: checkasm.c:233
checkasm_check_v210enc
void checkasm_check_v210enc(void)
Definition: v210enc.c:84
checkasm_check_sbrdsp
void checkasm_check_sbrdsp(void)
Definition: sbrdsp.c:257
checkasm_check_hevc_deblock
void checkasm_check_hevc_deblock(void)
Definition: hevc_deblock.c:91
usage
static int usage(const char *path)
Definition: checkasm.c:724
AV_CPU_FLAG_VFP_VM
#define AV_CPU_FLAG_VFP_VM
VFPv2 vector mode, deprecated in ARMv7-A and unavailable in various CPUs implementations.
Definition: cpu.h:71
AV_CPU_FLAG_MMI
#define AV_CPU_FLAG_MMI
Definition: cpu.h:76
checkasm_check_jpeg2000dsp
void checkasm_check_jpeg2000dsp(void)
Definition: jpeg2000dsp.c:92
d
d
Definition: ffmpeg_filter.c:372
checkasm_check_hevc_add_res
void checkasm_check_hevc_add_res(void)
Definition: hevc_add_res.c:87
int32_t
int32_t
Definition: audioconvert.c:56
float_near_ulp
int float_near_ulp(float a, float b, unsigned max_ulp)
Definition: checkasm.c:337
current_func_ver
CheckasmFuncVersion * current_func_ver
Definition: checkasm.c:311
check_cpu_flag
static void check_cpu_flag(const char *name, int flag)
Definition: checkasm.c:628
bench_init_ffmpeg
static int bench_init_ffmpeg(void)
Definition: checkasm.c:687
CheckasmFunc::versions
CheckasmFuncVersion versions
Definition: checkasm.c:302
isatty
#define isatty(fd)
Definition: checkasm.c:61
AV_CPU_FLAG_RVV_I64
#define AV_CPU_FLAG_RVV_I64
Vectors of 64-bit int's *‍/.
Definition: cpu.h:89
CheckasmFunc::name
char name[1]
Definition: checkasm.c:304
checkasm_check_exrdsp
void checkasm_check_exrdsp(void)
Definition: exrdsp.c:77
balance_tree
static void balance_tree(CheckasmFunc **root)
Definition: checkasm.c:587
AV_CPU_FLAG_ARMV6T2
#define AV_CPU_FLAG_ARMV6T2
Definition: cpu.h:66
funcs
CheckasmFunc * funcs
Definition: checkasm.c:309
checkasm_check_opusdsp
void checkasm_check_opusdsp(void)
Definition: opusdsp.c:85