Hi Janusz, > Subtests that measure time of resume after engine reset require results > from at least 9 reset-resume cycles for reasonable calculation of a median > value to be compared against presumed limits. On most of Gen12+ > platforms, the limit of 5 seconds for collecting those results occurs too > short for executing 9 reset-resum cycles. reset-resum -> reset-resume, just a small typo. > > Raise the limit to 20 seconds, and break the loop as soon as 9 results are > collected. Also, warn if less than 9 resets have been completed within > the limit instead of silently succeeding despite the check being skipped. > > Signed-off-by: Janusz Krzysztofik <[email protected]> > --- > tests/intel/gem_eio.c | 11 ++++++++--- > 1 file changed, 8 insertions(+), 3 deletions(-) > > diff --git a/tests/intel/gem_eio.c b/tests/intel/gem_eio.c > index b65b914faf..b6155c7fc4 100644 > --- a/tests/intel/gem_eio.c > +++ b/tests/intel/gem_eio.c > @@ -409,8 +409,10 @@ static void check_wait_elapsed(const char *prefix, int > fd, igt_stats_t *st) > igt_stats_get_median(st)*1e-6, > igt_stats_get_max(st)*1e-6); > > - if (st->n_values < 9) > - return; /* too few for stable median */ > + if (igt_warn_on_f(st->n_values < 9, > + "%d resets completed -- less than 9, too few for stable median\n", > + st->n_values)) > + return; Is this warning indicative of a bug? I wonder if this will result in having more WARN runs from CI. If there is nothing we can do to amend this, as slow reset-resume cycles are caused by hardware limitations, maybe igt_info would suffice?
-- Best Regards, Krzysztof
