[FFmpeg-devel] [PATCH] checkasm: Print benchmarks of C-only functions

Alexander Strasser eclipse7 at gmx.net
Sat Dec 14 20:30:44 EET 2024


Hi Martin!

On 2024-12-11 10:47 +0200, Martin Storsjö wrote:
> On Wed, 23 Oct 2024, Martin Storsjö wrote:
>
> > 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(-)
>
> Will push soon.
>
> The only concern about this change I heard was from Ramiro on irc, who
> wasn't familiar with how to use the --bench=<pattern> parameter for limiting
> the set of functions to actually benchmark.
>
> (This patch doesn't cause checkasm to benchmark more than it did before, it
> just prints the numbers that already were benchmarked.)
>
> I'll also go ahead and backport this to the 7.1 branch. It's obviously not a
> regression fix or anything like that, but the change adds value for being
> able to use checkasm to benchmark compiler performance.

LGTM.

Tested on aarch64 .

Total run times match.
Output had only additional lines which all were for functions ending in `_c`


  Alexander


More information about the ffmpeg-devel mailing list