2 * Assembly testing and benchmarking tool
3 * Copyright (c) 2015 Henrik Gramner
4 * Copyright (c) 2008 Loren Merritt
6 * This file is part of FFmpeg.
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.
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.
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.
24 #include "config_components.h"
28 # define _GNU_SOURCE // for syscall (performance monitoring API)
37 #include "libavutil/common.h"
38 #include "libavutil/cpu.h"
39 #include "libavutil/intfloat.h"
40 #include "libavutil/random_seed.h"
46 #if HAVE_SETCONSOLETEXTATTRIBUTE && HAVE_GETSTDHANDLE
48 #define COLOR_RED FOREGROUND_RED
49 #define COLOR_GREEN FOREGROUND_GREEN
50 #define COLOR_YELLOW (FOREGROUND_RED|FOREGROUND_GREEN)
54 #define COLOR_YELLOW 3
65 #if ARCH_ARM && HAVE_ARMV5TE_EXTERNAL
66 #include "libavutil/arm/cpu.h"
68 void (*checkasm_checked_call)(void *func, int dummy, ...) = checkasm_checked_call_novfp;
71 /* List of tests to invoke */
77 #if CONFIG_AAC_DECODER
78 { "aacpsdsp", checkasm_check_aacpsdsp },
79 { "sbrdsp", checkasm_check_sbrdsp },
81 #if CONFIG_ALAC_DECODER
82 { "alacdsp", checkasm_check_alacdsp },
85 { "audiodsp", checkasm_check_audiodsp },
88 { "blockdsp", checkasm_check_blockdsp },
91 { "bswapdsp", checkasm_check_bswapdsp },
93 #if CONFIG_DCA_DECODER
94 { "synth_filter", checkasm_check_synth_filter },
96 #if CONFIG_EXR_DECODER
97 { "exrdsp", checkasm_check_exrdsp },
99 #if CONFIG_FLAC_DECODER
100 { "flacdsp", checkasm_check_flacdsp },
102 #if CONFIG_FMTCONVERT
103 { "fmtconvert", checkasm_check_fmtconvert },
106 { "g722dsp", checkasm_check_g722dsp },
108 #if CONFIG_H264CHROMA
109 { "h264chroma", checkasm_check_h264chroma },
112 { "h264dsp", checkasm_check_h264dsp },
115 { "h264pred", checkasm_check_h264pred },
118 { "h264qpel", checkasm_check_h264qpel },
120 #if CONFIG_HEVC_DECODER
121 { "hevc_add_res", checkasm_check_hevc_add_res },
122 { "hevc_deblock", checkasm_check_hevc_deblock },
123 { "hevc_idct", checkasm_check_hevc_idct },
124 { "hevc_pel", checkasm_check_hevc_pel },
125 { "hevc_sao", checkasm_check_hevc_sao },
127 #if CONFIG_HUFFYUV_DECODER
128 { "huffyuvdsp", checkasm_check_huffyuvdsp },
131 { "idctdsp", checkasm_check_idctdsp },
133 #if CONFIG_JPEG2000_DECODER
134 { "jpeg2000dsp", checkasm_check_jpeg2000dsp },
136 #if CONFIG_HUFFYUVDSP
137 { "llviddsp", checkasm_check_llviddsp },
139 #if CONFIG_LLVIDENCDSP
140 { "llviddspenc", checkasm_check_llviddspenc },
143 { "lpc", checkasm_check_lpc },
146 { "motion", checkasm_check_motion },
148 #if CONFIG_OPUS_DECODER
149 { "opusdsp", checkasm_check_opusdsp },
151 #if CONFIG_PIXBLOCKDSP
152 { "pixblockdsp", checkasm_check_pixblockdsp },
154 #if CONFIG_UTVIDEO_DECODER
155 { "utvideodsp", checkasm_check_utvideodsp },
157 #if CONFIG_V210_DECODER
158 { "v210dec", checkasm_check_v210dec },
160 #if CONFIG_V210_ENCODER
161 { "v210enc", checkasm_check_v210enc },
164 { "vc1dsp", checkasm_check_vc1dsp },
167 { "vp8dsp", checkasm_check_vp8dsp },
169 #if CONFIG_VP9_DECODER
170 { "vp9dsp", checkasm_check_vp9dsp },
173 { "videodsp", checkasm_check_videodsp },
175 #if CONFIG_VORBIS_DECODER
176 { "vorbisdsp", checkasm_check_vorbisdsp },
180 #if CONFIG_AFIR_FILTER
181 { "af_afir", checkasm_check_afir },
183 #if CONFIG_BLEND_FILTER
184 { "vf_blend", checkasm_check_blend },
186 #if CONFIG_BWDIF_FILTER
187 { "vf_bwdif", checkasm_check_vf_bwdif },
189 #if CONFIG_COLORSPACE_FILTER
190 { "vf_colorspace", checkasm_check_colorspace },
193 { "vf_eq", checkasm_check_vf_eq },
195 #if CONFIG_GBLUR_FILTER
196 { "vf_gblur", checkasm_check_vf_gblur },
198 #if CONFIG_HFLIP_FILTER
199 { "vf_hflip", checkasm_check_vf_hflip },
201 #if CONFIG_NLMEANS_FILTER
202 { "vf_nlmeans", checkasm_check_nlmeans },
204 #if CONFIG_THRESHOLD_FILTER
205 { "vf_threshold", checkasm_check_vf_threshold },
207 #if CONFIG_SOBEL_FILTER
208 { "vf_sobel", checkasm_check_vf_sobel },
212 { "sw_gbrp", checkasm_check_sw_gbrp },
213 { "sw_rgb", checkasm_check_sw_rgb },
214 { "sw_scale", checkasm_check_sw_scale },
217 { "fixed_dsp", checkasm_check_fixed_dsp },
218 { "float_dsp", checkasm_check_float_dsp },
219 { "av_tx", checkasm_check_av_tx },
224 /* List of cpu flags to check */
225 static const struct {
231 { "ARMV8", "armv8", AV_CPU_FLAG_ARMV8 },
232 { "NEON", "neon", AV_CPU_FLAG_NEON },
233 { "DOTPROD", "dotprod", AV_CPU_FLAG_DOTPROD },
234 { "I8MM", "i8mm", AV_CPU_FLAG_I8MM },
236 { "ARMV5TE", "armv5te", AV_CPU_FLAG_ARMV5TE },
237 { "ARMV6", "armv6", AV_CPU_FLAG_ARMV6 },
238 { "ARMV6T2", "armv6t2", AV_CPU_FLAG_ARMV6T2 },
239 { "VFP", "vfp", AV_CPU_FLAG_VFP },
240 { "VFP_VM", "vfp_vm", AV_CPU_FLAG_VFP_VM },
241 { "VFPV3", "vfp3", AV_CPU_FLAG_VFPV3 },
242 { "NEON", "neon", AV_CPU_FLAG_NEON },
244 { "ALTIVEC", "altivec", AV_CPU_FLAG_ALTIVEC },
245 { "VSX", "vsx", AV_CPU_FLAG_VSX },
246 { "POWER8", "power8", AV_CPU_FLAG_POWER8 },
248 { "RVI", "rvi", AV_CPU_FLAG_RVI },
249 { "RVF", "rvf", AV_CPU_FLAG_RVF },
250 { "RVD", "rvd", AV_CPU_FLAG_RVD },
251 { "RVBaddr", "rvb_a", AV_CPU_FLAG_RVB_ADDR },
252 { "RVBbasic", "rvb_b", AV_CPU_FLAG_RVB_BASIC },
253 { "RVVi32", "rvv_i32", AV_CPU_FLAG_RVV_I32 },
254 { "RVVf32", "rvv_f32", AV_CPU_FLAG_RVV_F32 },
255 { "RVVi64", "rvv_i64", AV_CPU_FLAG_RVV_I64 },
256 { "RVVf64", "rvv_f64", AV_CPU_FLAG_RVV_F64 },
258 { "MMI", "mmi", AV_CPU_FLAG_MMI },
259 { "MSA", "msa", AV_CPU_FLAG_MSA },
261 { "MMX", "mmx", AV_CPU_FLAG_MMX|AV_CPU_FLAG_CMOV },
262 { "MMXEXT", "mmxext", AV_CPU_FLAG_MMXEXT },
263 { "3DNOW", "3dnow", AV_CPU_FLAG_3DNOW },
264 { "3DNOWEXT", "3dnowext", AV_CPU_FLAG_3DNOWEXT },
265 { "SSE", "sse", AV_CPU_FLAG_SSE },
266 { "SSE2", "sse2", AV_CPU_FLAG_SSE2|AV_CPU_FLAG_SSE2SLOW },
267 { "SSE3", "sse3", AV_CPU_FLAG_SSE3|AV_CPU_FLAG_SSE3SLOW },
268 { "SSSE3", "ssse3", AV_CPU_FLAG_SSSE3|AV_CPU_FLAG_ATOM },
269 { "SSE4.1", "sse4", AV_CPU_FLAG_SSE4 },
270 { "SSE4.2", "sse42", AV_CPU_FLAG_SSE42 },
271 { "AES-NI", "aesni", AV_CPU_FLAG_AESNI },
272 { "AVX", "avx", AV_CPU_FLAG_AVX },
273 { "XOP", "xop", AV_CPU_FLAG_XOP },
274 { "FMA3", "fma3", AV_CPU_FLAG_FMA3 },
275 { "FMA4", "fma4", AV_CPU_FLAG_FMA4 },
276 { "AVX2", "avx2", AV_CPU_FLAG_AVX2 },
277 { "AVX-512", "avx512", AV_CPU_FLAG_AVX512 },
278 { "AVX-512ICL", "avx512icl", AV_CPU_FLAG_AVX512ICL },
280 { "LSX", "lsx", AV_CPU_FLAG_LSX },
281 { "LASX", "lasx", AV_CPU_FLAG_LASX },
286 typedef struct CheckasmFuncVersion {
287 struct CheckasmFuncVersion *next;
292 } CheckasmFuncVersion;
294 /* Binary search tree node */
295 typedef struct CheckasmFunc {
296 struct CheckasmFunc *child[2];
297 CheckasmFuncVersion versions;
298 uint8_t color; /* 0 = red, 1 = black */
305 CheckasmFunc *current_func;
306 CheckasmFuncVersion *current_func_ver;
307 const char *current_test_name;
308 const char *bench_pattern;
309 int bench_pattern_len;
318 const char *cpu_flag_name;
319 const char *test_name;
326 /* float compare support code */
327 static int is_negative(union av_intfloat32 u)
332 int float_near_ulp(float a, float b, unsigned max_ulp)
334 union av_intfloat32 x, y;
339 if (is_negative(x) != is_negative(y)) {
340 // handle -0.0 == +0.0
344 if (llabs((int64_t)x.i - y.i) <= max_ulp)
350 int float_near_ulp_array(const float *a, const float *b, unsigned max_ulp,
355 for (i = 0; i < len; i++) {
356 if (!float_near_ulp(a[i], b[i], max_ulp))
362 int float_near_abs_eps(float a, float b, float eps)
364 float abs_diff = fabsf(a - b);
368 fprintf(stderr, "test failed comparing %g with %g (abs diff=%g with EPS=%g)\n", a, b, abs_diff, eps);
373 int float_near_abs_eps_array(const float *a, const float *b, float eps,
378 for (i = 0; i < len; i++) {
379 if (!float_near_abs_eps(a[i], b[i], eps))
385 int float_near_abs_eps_ulp(float a, float b, float eps, unsigned max_ulp)
387 return float_near_ulp(a, b, max_ulp) || float_near_abs_eps(a, b, eps);
390 int float_near_abs_eps_array_ulp(const float *a, const float *b, float eps,
391 unsigned max_ulp, unsigned len)
395 for (i = 0; i < len; i++) {
396 if (!float_near_abs_eps_ulp(a[i], b[i], eps, max_ulp))
402 int double_near_abs_eps(double a, double b, double eps)
404 double abs_diff = fabs(a - b);
406 return abs_diff < eps;
409 int double_near_abs_eps_array(const double *a, const double *b, double eps,
414 for (i = 0; i < len; i++) {
415 if (!double_near_abs_eps(a[i], b[i], eps))
421 /* Print colored text to stderr if the terminal supports it */
422 static void color_printf(int color, const char *fmt, ...)
424 static int use_color = -1;
427 #if HAVE_SETCONSOLETEXTATTRIBUTE && HAVE_GETSTDHANDLE
429 static WORD org_attributes;
432 CONSOLE_SCREEN_BUFFER_INFO con_info;
433 con = GetStdHandle(STD_ERROR_HANDLE);
434 if (con && con != INVALID_HANDLE_VALUE && GetConsoleScreenBufferInfo(con, &con_info)) {
435 org_attributes = con_info.wAttributes;
441 SetConsoleTextAttribute(con, (org_attributes & 0xfff0) | (color & 0x0f));
444 const char *term = getenv("TERM");
445 use_color = term && strcmp(term, "dumb") && isatty(2);
448 fprintf(stderr, "\x1b[%d;3%dm", (color & 0x08) >> 3, color & 0x07);
452 vfprintf(stderr, fmt, arg);
456 #if HAVE_SETCONSOLETEXTATTRIBUTE && HAVE_GETSTDHANDLE
457 SetConsoleTextAttribute(con, org_attributes);
459 fprintf(stderr, "\x1b[0m");
464 /* Deallocate a tree */
465 static void destroy_func_tree(CheckasmFunc *f)
468 CheckasmFuncVersion *v = f->versions.next;
470 CheckasmFuncVersion *next = v->next;
475 destroy_func_tree(f->child[0]);
476 destroy_func_tree(f->child[1]);
481 /* Allocate a zero-initialized block, clean up and exit on failure */
482 static void *checkasm_malloc(size_t size)
484 void *ptr = calloc(1, size);
486 fprintf(stderr, "checkasm: malloc failed\n");
487 destroy_func_tree(state.funcs);
493 /* Get the suffix of the specified cpu flag */
494 static const char *cpu_suffix(int cpu)
496 int i = FF_ARRAY_ELEMS(cpus);
499 if (cpu & cpus[i].flag)
500 return cpus[i].suffix;
505 static int cmp_nop(const void *a, const void *b)
507 return *(const uint16_t*)a - *(const uint16_t*)b;
510 /* Measure the overhead of the timing code (in decicycles) */
511 static int measure_nop_time(void)
513 uint16_t nops[10000];
515 av_unused const int sysfd = state.sysfd;
518 for (i = 0; i < 10000; i++) {
524 qsort(nops, 10000, sizeof(uint16_t), cmp_nop);
525 for (i = 2500; i < 7500; i++)
528 return nop_sum / 500;
531 /* Print benchmark results */
532 static void print_benchs(CheckasmFunc *f)
535 print_benchs(f->child[0]);
537 /* Only print functions with at least one assembly version */
538 if (f->versions.cpu || f->versions.next) {
539 CheckasmFuncVersion *v = &f->versions;
541 CheckasmPerf *p = &v->perf;
543 int decicycles = (10*p->cycles/p->iterations - state.nop_time) / 4;
544 printf("%s_%s: %d.%d\n", f->name, cpu_suffix(v->cpu), decicycles/10, decicycles%10);
546 } while ((v = v->next));
549 print_benchs(f->child[1]);
553 /* ASCIIbetical sort except preserving natural order for numbers */
554 static int cmp_func_names(const char *a, const char *b)
556 const char *start = a;
557 int ascii_diff, digit_diff;
559 for (; !(ascii_diff = *(const unsigned char*)a - *(const unsigned char*)b) && *a; a++, b++);
560 for (; av_isdigit(*a) && av_isdigit(*b); a++, b++);
562 if (a > start && av_isdigit(a[-1]) && (digit_diff = av_isdigit(*a) - av_isdigit(*b)))
568 /* Perform a tree rotation in the specified direction and return the new root */
569 static CheckasmFunc *rotate_tree(CheckasmFunc *f, int dir)
571 CheckasmFunc *r = f->child[dir^1];
572 f->child[dir^1] = r->child[dir];
579 #define is_red(f) ((f) && !(f)->color)
581 /* Balance a left-leaning red-black tree at the specified node */
582 static void balance_tree(CheckasmFunc **root)
584 CheckasmFunc *f = *root;
586 if (is_red(f->child[0]) && is_red(f->child[1])) {
588 f->child[0]->color = f->child[1]->color = 1;
591 if (!is_red(f->child[0]) && is_red(f->child[1]))
592 *root = rotate_tree(f, 0); /* Rotate left */
593 else if (is_red(f->child[0]) && is_red(f->child[0]->child[0]))
594 *root = rotate_tree(f, 1); /* Rotate right */
597 /* Get a node with the specified name, creating it if it doesn't exist */
598 static CheckasmFunc *get_func(CheckasmFunc **root, const char *name)
600 CheckasmFunc *f = *root;
603 /* Search the tree for a matching node */
604 int cmp = cmp_func_names(name, f->name);
606 f = get_func(&f->child[cmp > 0], name);
608 /* Rebalance the tree on the way up if a new node was inserted */
609 if (!f->versions.func)
613 /* Allocate and insert a new node into the tree */
614 int name_length = strlen(name);
615 f = *root = checkasm_malloc(sizeof(CheckasmFunc) + name_length);
616 memcpy(f->name, name, name_length + 1);
622 /* Perform tests and benchmarks for the specified cpu flag if supported by the host */
623 static void check_cpu_flag(const char *name, int flag)
625 int old_cpu_flag = state.cpu_flag;
627 flag |= old_cpu_flag;
628 av_force_cpu_flags(-1);
629 state.cpu_flag = flag & av_get_cpu_flags();
630 av_force_cpu_flags(state.cpu_flag);
632 if (!flag || state.cpu_flag != old_cpu_flag) {
635 state.cpu_flag_name = name;
636 for (i = 0; tests[i].func; i++) {
637 if (state.test_name && strcmp(tests[i].name, state.test_name))
639 state.current_test_name = tests[i].name;
645 /* Print the name of the current CPU flag, but only do it once */
646 static void print_cpu_name(void)
648 if (state.cpu_flag_name) {
649 color_printf(COLOR_YELLOW, "%s:\n", state.cpu_flag_name);
650 state.cpu_flag_name = NULL;
654 #if CONFIG_LINUX_PERF
655 static int bench_init_linux(void)
657 struct perf_event_attr attr = {
658 .type = PERF_TYPE_HARDWARE,
659 .size = sizeof(struct perf_event_attr),
660 .config = PERF_COUNT_HW_CPU_CYCLES,
661 .disabled = 1, // start counting only on demand
666 printf("benchmarking with Linux Perf Monitoring API\n");
668 state.sysfd = syscall(__NR_perf_event_open, &attr, 0, -1, -1, 0);
669 if (state.sysfd == -1) {
670 perror("perf_event_open");
675 #elif CONFIG_MACOS_KPERF
676 static int bench_init_kperf(void)
682 static int bench_init_ffmpeg(void)
685 printf("benchmarking with native FFmpeg timers\n");
688 fprintf(stderr, "checkasm: --bench is not supported on your system\n");
694 static int bench_init(void)
696 #if CONFIG_LINUX_PERF
697 int ret = bench_init_linux();
698 #elif CONFIG_MACOS_KPERF
699 int ret = bench_init_kperf();
701 int ret = bench_init_ffmpeg();
706 state.nop_time = measure_nop_time();
707 printf("nop: %d.%d\n", state.nop_time/10, state.nop_time%10);
711 static void bench_uninit(void)
713 #if CONFIG_LINUX_PERF
719 static int usage(const char *path)
722 "Usage: %s [--bench] [--test=<pattern>] [--verbose] [seed]\n",
727 int main(int argc, char *argv[])
729 unsigned int seed = av_get_random_seed();
732 #if ARCH_ARM && HAVE_ARMV5TE_EXTERNAL
733 if (have_vfp(av_get_cpu_flags()) || have_neon(av_get_cpu_flags()))
734 checkasm_checked_call = checkasm_checked_call_vfp;
737 if (!tests[0].func || !cpus[0].flag) {
738 fprintf(stderr, "checkasm: no tests to perform\n");
742 for (i = 1; i < argc; i++) {
743 const char *arg = argv[i];
747 if (!strncmp(arg, "--bench", 7)) {
748 if (bench_init() < 0)
751 state.bench_pattern = arg + 8;
752 state.bench_pattern_len = strlen(state.bench_pattern);
754 state.bench_pattern = "";
755 } else if (!strncmp(arg, "--test=", 7)) {
756 state.test_name = arg + 7;
757 } else if (!strcmp(arg, "--verbose") || !strcmp(arg, "-v")) {
759 } else if ((l = strtoul(arg, &end, 10)) <= UINT_MAX &&
763 return usage(argv[0]);
767 fprintf(stderr, "checkasm: using random seed %u\n", seed);
768 av_lfg_init(&checkasm_lfg, seed);
770 check_cpu_flag(NULL, 0);
771 for (i = 0; cpus[i].flag; i++)
772 check_cpu_flag(cpus[i].name, cpus[i].flag);
774 if (state.num_failed) {
775 fprintf(stderr, "checkasm: %d of %d tests have failed\n", state.num_failed, state.num_checked);
778 fprintf(stderr, "checkasm: all %d tests passed\n", state.num_checked);
779 if (state.bench_pattern) {
780 print_benchs(state.funcs);
784 destroy_func_tree(state.funcs);
789 /* Decide whether or not the specified function needs to be tested and
790 * allocate/initialize data structures if needed. Returns a pointer to a
791 * reference function if the function should be tested, otherwise NULL */
792 void *checkasm_check_func(void *func, const char *name, ...)
796 CheckasmFuncVersion *v;
801 name_length = vsnprintf(name_buf, sizeof(name_buf), name, arg);
804 if (!func || name_length <= 0 || name_length >= sizeof(name_buf))
807 state.current_func = get_func(&state.funcs, name_buf);
808 state.funcs->color = 1;
809 v = &state.current_func->versions;
812 CheckasmFuncVersion *prev;
814 /* Only test functions that haven't already been tested */
822 } while ((v = v->next));
824 v = prev->next = checkasm_malloc(sizeof(CheckasmFuncVersion));
829 v->cpu = state.cpu_flag;
830 state.current_func_ver = v;
838 /* Decide whether or not the current function needs to be benchmarked */
839 int checkasm_bench_func(void)
841 return !state.num_failed && state.bench_pattern &&
842 !strncmp(state.current_func->name, state.bench_pattern, state.bench_pattern_len);
845 /* Indicate that the current test has failed */
846 void checkasm_fail_func(const char *msg, ...)
848 if (state.current_func_ver->cpu && state.current_func_ver->ok) {
852 fprintf(stderr, " %s_%s (", state.current_func->name, cpu_suffix(state.current_func_ver->cpu));
854 vfprintf(stderr, msg, arg);
856 fprintf(stderr, ")\n");
858 state.current_func_ver->ok = 0;
863 /* Get the benchmark context of the current function */
864 CheckasmPerf *checkasm_get_perf_context(void)
866 CheckasmPerf *perf = &state.current_func_ver->perf;
867 memset(perf, 0, sizeof(*perf));
868 perf->sysfd = state.sysfd;
872 /* Print the outcome of all tests performed since the last time this function was called */
873 void checkasm_report(const char *name, ...)
875 static int prev_checked, prev_failed, max_length;
877 if (state.num_checked > prev_checked) {
878 int pad_length = max_length + 4;
882 pad_length -= fprintf(stderr, " - %s.", state.current_test_name);
884 pad_length -= vfprintf(stderr, name, arg);
886 fprintf(stderr, "%*c", FFMAX(pad_length, 0) + 2, '[');
888 if (state.num_failed == prev_failed)
889 color_printf(COLOR_GREEN, "OK");
891 color_printf(COLOR_RED, "FAILED");
892 fprintf(stderr, "]\n");
894 prev_checked = state.num_checked;
895 prev_failed = state.num_failed;
896 } else if (!state.cpu_flag) {
897 /* Calculate the amount of padding required to make the output vertically aligned */
898 int length = strlen(state.current_test_name);
902 length += vsnprintf(NULL, 0, name, arg);
905 if (length > max_length)
910 #define DEF_CHECKASM_CHECK_FUNC(type, fmt) \
911 int checkasm_check_##type(const char *const file, const int line, \
912 const type *buf1, ptrdiff_t stride1, \
913 const type *buf2, ptrdiff_t stride2, \
914 const int w, int h, const char *const name) \
917 stride1 /= sizeof(*buf1); \
918 stride2 /= sizeof(*buf2); \
919 for (y = 0; y < h; y++) \
920 if (memcmp(&buf1[y*stride1], &buf2[y*stride2], w*sizeof(*buf1))) \
924 checkasm_fail_func("%s:%d", file, line); \
925 if (!state.verbose) \
927 fprintf(stderr, "%s:\n", name); \
929 for (int x = 0; x < w; x++) \
930 fprintf(stderr, " " fmt, buf1[x]); \
931 fprintf(stderr, " "); \
932 for (int x = 0; x < w; x++) \
933 fprintf(stderr, " " fmt, buf2[x]); \
934 fprintf(stderr, " "); \
935 for (int x = 0; x < w; x++) \
936 fprintf(stderr, "%c", buf1[x] != buf2[x] ? 'x' : '.'); \
939 fprintf(stderr, "\n"); \
944 DEF_CHECKASM_CHECK_FUNC(uint8_t, "%02x")
945 DEF_CHECKASM_CHECK_FUNC(uint16_t, "%04x")
946 DEF_CHECKASM_CHECK_FUNC(uint32_t, "%08x")
947 DEF_CHECKASM_CHECK_FUNC(int16_t, "%6d")
948 DEF_CHECKASM_CHECK_FUNC(int32_t, "%9d")