On Sun, Aug 25, 2024 at 7:52 PM J. Dekker <j...@itanimul.li> wrote:
> On Thu, Aug 22, 2024, at 16:57, Ramiro Polla wrote:
> > On Wed, Aug 21, 2024 at 1:26 PM J. Dekker <j...@itanimul.li> wrote:
> >>
> >> Port dav1d's checkasm output format to FFmpeg's checkasm, includes
> >> relative speedups and aligns results.
> >>
> >> Signed-off-by: J. Dekker <j...@itanimul.li>
> >> ---
> >> tests/checkasm/checkasm.c | 53 +++++++++++++++++++++++++++++++++++----
> >> 1 file changed, 48 insertions(+), 5 deletions(-)
> >>
> >> diff --git a/tests/checkasm/checkasm.c b/tests/checkasm/checkasm.c
> >> index c5c1eeb07a..f43c1c50f0 100644
> >> --- a/tests/checkasm/checkasm.c
> >> +++ b/tests/checkasm/checkasm.c
> >> @@ -18,6 +18,31 @@
> >> * You should have received a copy of the GNU General Public License along
> >> * with FFmpeg; if not, write to the Free Software Foundation, Inc.,
> >> * 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
> >> + *
> >> + * Copyright © 2018, VideoLAN and dav1d authors
> >> + * Copyright © 2018, Two Orioles, LLC
> >> + * All rights reserved.
> >> + *
> >> + * Redistribution and use in source and binary forms, with or without
> >> + * modification, are permitted provided that the following conditions are
> >> met:
> >> + *
> >> + * 1. Redistributions of source code must retain the above copyright
> >> notice, this
> >> + * list of conditions and the following disclaimer.
> >> + *
> >> + * 2. Redistributions in binary form must reproduce the above copyright
> >> notice,
> >> + * this list of conditions and the following disclaimer in the
> >> documentation
> >> + * and/or other materials provided with the distribution.
> >> + *
> >> + * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
> >> "AS IS" AND
> >> + * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
> >> IMPLIED
> >> + * WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
> >> + * DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE
> >> LIABLE FOR
> >> + * ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
> >> DAMAGES
> >> + * (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR
> >> SERVICES;
> >> + * LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER
> >> CAUSED AND
> >> + * ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR
> >> TORT
> >> + * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
> >> OF THIS
> >> + * SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
> >> */
> >
> > Do we need this double licensing?
> >
> >>
> >> #include "config.h"
> >> @@ -575,6 +600,16 @@ static int measure_nop_time(void)
> >> return nop_sum / 500;
> >> }
> >>
> >> +static inline double avg_cycles_per_call(const CheckasmPerf *const p)
> >> +{
> >> + if (p->iterations) {
> >> + const double cycles = (double)(10 * p->cycles) / p->iterations -
> >> state.nop_time;
> >> + if (cycles > 0.0)
> >> + return cycles / 4.0; /* 4 calls per iteration */
> >> + }
> >> + return 0.0;
> >> +}
> >> +
> >> /* Print benchmark results */
> >> static void print_benchs(CheckasmFunc *f)
> >> {
> >> @@ -584,17 +619,25 @@ static void print_benchs(CheckasmFunc *f)
> >> /* 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 {
> >> - CheckasmPerf *p = &v->perf;
> >> if (p->iterations) {
> >> - int decicycles = (10*p->cycles/p->iterations -
> >> state.nop_time) / 4;
> >> + 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%d.%d\n", f->name, sep,
> >> + printf("%s%c%s%c%.1f\n", f->name, sep,
> >> cpu_suffix(v->cpu), sep,
> >> - decicycles / 10, decicycles % 10);
> >> + decicycles / 10.0);
> >> } else {
> >> - printf("%s_%s: %d.%d\n", f->name,
> >> cpu_suffix(v->cpu), decicycles/10, decicycles%10);
> >> + 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);
> >
> > It looks a bit noisy to always print "( 1.00x)" for the C versions.
> > Perhaps it could be left empty or print "ref".
> >
> > Also do you plan on improving it even further? For example printing a
> > table like this:
> > c sse avx
> > blockdsp.clear_block: 34.0 3.5 ( 9.59x) 1.8 (18.92x)
>
> I would like to leave it as is initially. It can always be changed later
> though the intention here is to sync it up with dav1d's checkasm.
Should we aim to first improve dav1d's checkasm and then merge the
changes back in FFmpeg?
_______________________________________________
ffmpeg-devel mailing list
ffmpeg-devel@ffmpeg.org
https://ffmpeg.org/mailman/listinfo/ffmpeg-devel
To unsubscribe, visit link above, or email
ffmpeg-devel-requ...@ffmpeg.org with subject "unsubscribe".
