[AMD Official Use Only]


Skipping using the continue statement has the side effect of this subtest being absent from the list of subtests, when running the test executable with paramter --list-subtest. Is that okay ?

--

Regards,
Jay

From: Mark Yacoub <markyacoub@chromium.org>
Sent: Monday, September 27, 2021 1:34 PM
To: Pillai, Aurabindo <Aurabindo.Pillai@amd.com>
Cc: Development mailing list for IGT GPU Tools <igt-dev@lists.freedesktop.org>; Mark Yacoub <markyacoub@google.com>; Siqueira, Rodrigo <Rodrigo.Siqueira@amd.com>
Subject: Re: [igt-dev] [PATCH i-g-t] tests/kms_flip : skip subtest bo-too-big
 
so for the sake of consistency, I'm wondering if we can skip the test
the same way we skip other tests such as
```
if (tests[i].flags & TEST_NO_2X_OUTPUT)
continue;
```
because we're not failing a certain condition but a whole test, it
might be better not to run the test in the first place unless it's
intel.
So in both main loops that call run_test we can do if
if (tests[i].flags & TEST_BO_TOOBIG && !intel)
    continue;
this way we're not too worried where the flag would be used across the
test and not skip halfway through a test.

On Mon, Sep 27, 2021 at 1:28 PM Pillai, Aurabindo
<Aurabindo.Pillai@amd.com> wrote:
>
> [AMD Official Use Only]
>
>
> Hi Mark,
>
>
> Both bo-too-big and bo-too-big-interruptible shall be skipped with this patch. The other location where TEST_BO_TOOBIG is mentioned is in the same function.
>
> --
>
> Regards,
> Jay
> ________________________________
> From: Mark Yacoub <markyacoub@chromium.org>
> Sent: Monday, September 27, 2021 1:21 PM
> To: Pillai, Aurabindo <Aurabindo.Pillai@amd.com>
> Cc: Development mailing list for IGT GPU Tools <igt-dev@lists.freedesktop.org>; Mark Yacoub <markyacoub@google.com>; Siqueira, Rodrigo <Rodrigo.Siqueira@amd.com>
> Subject: Re: [igt-dev] [PATCH i-g-t] tests/kms_flip : skip subtest bo-too-big
>
> On Mon, Sep 27, 2021 at 12:07 PM Aurabindo Pillai
> <aurabindo.pillai@amd.com> wrote:
> >
> > [Why]
> > The rationale of the test does not hold true for AMD hardware. The
> > aperture size calculation has an upper bound check which is done through
> > i915 specific IOCTL. Hence this part of subtest must be moved
> > out of the platform agnostic tests. Moreover, AMD hardware
> > supports buffers larger than aperture size.
> >
> > [How]
> > Skip the bo-too-big subtest unless its run on i915 as the test fails on
> > AMD, VKMS and VC4
> >
> > Signed-off-by: Aurabindo Pillai <aurabindo.pillai@amd.com>
> > ---
> >  tests/kms_flip.c | 1 +
> >  1 file changed, 1 insertion(+)
> >
> > diff --git a/tests/kms_flip.c b/tests/kms_flip.c
> > index 54137871..20112de1 100755
> > --- a/tests/kms_flip.c
> > +++ b/tests/kms_flip.c
> > @@ -1290,6 +1290,7 @@ static void __run_test_on_crtc_set(struct test_output *o, int *crtc_idxs,
> >         /* 256 MB is usually the maximum mappable aperture,
> >          * (make it 4x times that to ensure failure) */
> >         if (o->flags & TEST_BO_TOOBIG) {
> There is also another place where TEST_BO_TOOBIG is used, should we
> skip this one as well or it's only this part?
> > +               igt_skip_on(!is_i915_device(drm_fd));
> >                 bo_size = 4*gem_mappable_aperture_size(drm_fd);
> >                 igt_require(bo_size < gem_global_aperture_size(drm_fd));
> >         }
> > --
> > 2.30.2
> >