[Intel-gfx] [PATCH i-g-t] i915/perf_pmu: Update inter-engine semaphore detection

Chris Wilson chris at chris-wilson.co.uk
Tue May 26 16:05:52 UTC 2020


Quoting Tvrtko Ursulin (2020-05-26 16:58:05)
> 
> On 26/05/2020 15:17, Chris Wilson wrote:
> > The kernel no longer uses semaphores between engines, unless it can do
> > so by preempting them with timeslices. Update the semaphore-busy to only
> > run when we expect semaphore usage, i.e. not on bdw/bsw.
> > 
> > Closes: https://gitlab.freedesktop.org/drm/intel/-/issues/1939
> > Signed-off-by: Chris Wilson <chris at chris-wilson.co.uk>
> > ---
> >   tests/perf_pmu.c | 1 +
> >   1 file changed, 1 insertion(+)
> > 
> > diff --git a/tests/perf_pmu.c b/tests/perf_pmu.c
> > index e54a0d47e..e80f730cf 100644
> > --- a/tests/perf_pmu.c
> > +++ b/tests/perf_pmu.c
> > @@ -822,6 +822,7 @@ sema_busy(int gem_fd,
> >       int fd;
> >   
> >       igt_require(gem_scheduler_has_semaphores(gem_fd));
> > +     igt_require(gem_scheduler_has_preemption(gem_fd));
> >   
> >       fd = open_group(gem_fd,
> >                       I915_PMU_ENGINE_SEMA(e->class, e->instance), -1);
> > 
> 
> Test with perhaps too intimate knowledge of i915..

iirc, part of the intent of the test was to understand how the kernel's
choice of inter-engine sync would show up in the pmu. The challenge we
have is in declaring the expected outcomes, that requires intimate
knowledge of what the kernel might decide to do.
-Chris


More information about the Intel-gfx mailing list