[FFmpeg-devel] [PATCH] checkasm: Print benchmarks of C-only functions
Martin Storsjö
martin at martin.st
Wed Oct 23 14:46:09 EEST 2024
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.
---
tests/checkasm/checkasm.c | 48 +++++++++++++++++++--------------------
1 file changed, 23 insertions(+), 25 deletions(-)
diff --git a/tests/checkasm/checkasm.c b/tests/checkasm/checkasm.c
index c9d2b5faf1..c08188c8d4 100644
--- a/tests/checkasm/checkasm.c
+++ b/tests/checkasm/checkasm.c
@@ -624,34 +624,32 @@ static inline double avg_cycles_per_call(const CheckasmPerf *const p)
static void print_benchs(CheckasmFunc *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]);
- /* Only print functions with at least one assembly version */
- if (f->versions.cpu || f->versions.next) {
- CheckasmFuncVersion *v = &f->versions;
- const CheckasmPerf *p = &v->perf;
- const double baseline = avg_cycles_per_call(p);
- double decicycles;
- do {
- if (p->iterations) {
- p = &v->perf;
- decicycles = avg_cycles_per_call(p);
- if (state.csv || state.tsv) {
- const char sep = state.csv ? ',' : '\t';
- printf("%s%c%s%c%.1f\n", f->name, sep,
- cpu_suffix(v->cpu), sep,
- decicycles / 10.0);
- } else {
- 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);
- }
+ do {
+ if (p->iterations) {
+ p = &v->perf;
+ decicycles = avg_cycles_per_call(p);
+ if (state.csv || state.tsv) {
+ const char sep = state.csv ? ',' : '\t';
+ printf("%s%c%s%c%.1f\n", f->name, sep,
+ cpu_suffix(v->cpu), sep,
+ decicycles / 10.0);
+ } else {
+ 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]);
}
--
2.39.5 (Apple Git-154)
More information about the ffmpeg-devel
mailing list