RepoMirrors/ffmpeg
1
0
Fork 0
mirror of https://git.ffmpeg.org/ffmpeg.git synced 2024-12-18 05:24:42 +00:00
Code Issues Projects Releases Wiki Activity

checkasm: Print benchmarks of C-only functions

Browse Source 
This corresponds to commit 9278a14cf406f8edb5052c42b83750112bf5b515
in dav1d.

Omitting the C-only functions doesn't speed up benchmarking
anyway (as those has to be benchmarked before we know if we have
any corresponding assembly functions), and being able to benchmark
those functions without corresponding assembly can be valuable in
a number of cases.

Signed-off-by: Martin Storsjö <martin@martin.st>
This commit is contained in:
Martin Storsjö 2024-10-23 14:44:11 +03:00
parent ee903c4786
commit 4b524649ff
1 changed files with 23 additions and 25 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

48
tests/checkasm/checkasm.c
View File

@ -631,34 +631,32 @@ static inline double avg_cycles_per_call(const CheckasmPerf *const p)
static void print_benchs(CheckasmFunc *f) static void print_benchs(CheckasmFunc *f)
{ {
if (f) { if (f) {
CheckasmFuncVersion *v = &f->versions;
const CheckasmPerf *p = &v->perf;
const double baseline = avg_cycles_per_call(p);
double decicycles;
print_benchs(f->child[0]); print_benchs(f->child[0]);
/* Only print functions with at least one assembly version */ do {
if (f->versions.cpu || f->versions.next) { if (p->iterations) {
CheckasmFuncVersion *v = &f->versions; p = &v->perf;
const CheckasmPerf *p = &v->perf; decicycles = avg_cycles_per_call(p);
const double baseline = avg_cycles_per_call(p); if (state.csv || state.tsv) {
double decicycles; const char sep = state.csv ? ',' : '\t';
do { printf("%s%c%s%c%.1f\n", f->name, sep,
if (p->iterations) { cpu_suffix(v->cpu), sep,
p = &v->perf; decicycles / 10.0);
decicycles = avg_cycles_per_call(p); } else {
if (state.csv || state.tsv) { const int pad_length = 10 + 50 -
const char sep = state.csv ? ',' : '\t'; printf("%s_%s:", f->name, cpu_suffix(v->cpu));
printf("%s%c%s%c%.1f\n", f->name, sep, const double ratio = decicycles ?
cpu_suffix(v->cpu), sep, baseline / decicycles : 0.0;
decicycles / 10.0); printf("%*.1f (%5.2fx)\n", FFMAX(pad_length, 0),
} else { decicycles / 10.0, ratio);
const int pad_length = 10 + 50 -
printf("%s_%s:", f->name, cpu_suffix(v->cpu));
const double ratio = decicycles ?
baseline / decicycles : 0.0;
printf("%*.1f (%5.2fx)\n", FFMAX(pad_length, 0),
decicycles / 10.0, ratio);
}
} }
} while ((v = v->next)); }
} } while ((v = v->next));
print_benchs(f->child[1]); print_benchs(f->child[1]);
} }