[igt-dev] [PATCH i-g-t v3] tests/xe: Verify actual frequency on the basis of GT state

Dixit, Ashutosh ashutosh.dixit at intel.com
Wed Jul 19 13:40:18 UTC 2023


On Tue, 18 Jul 2023 22:54:41 -0700, Nilawar, Badal wrote:
>

Hi Badal,

> On 19-07-2023 11:05, Nilawar, Badal wrote:
> >
> >
> > On 18-07-2023 23:58, Dixit, Ashutosh wrote:
> >> On Tue, 18 Jul 2023 11:24:55 -0700, Dixit, Ashutosh wrote:
> >>>
> >>> Sorry I missed this yesterday. Basically why do we have the
> >>> igt_wait(xe_is_gt_in_c6()) above at the top when we are waking up the gt
> >>> when reading cur freq later?
> >>>
> >>> Maybe instead of this usleep, we can this do here:
> >>>
> >>>         igt_assert(igt_wait(xe_is_gt_in_c6(fd, gt_id), 1000, 1));
> >>
> >> Actually, maybe:
> >>
> >>         igt_assert(igt_wait(xe_is_gt_in_c6(fd, gt_id), 1000, 10))Sure I
> >> will make this change.
> Is it ok to use igt_warn_on here. igt_assert is anyway being done in next
> instruction while checking act freq?

Yes sure, makes sense.

Also, the "if (gt_idle) igt_require()" at the top of the two functions,
either leave as is or move to igt_main in both places where gt_idle is
true.

Thanks.
--
Ashutosh

> >>
> >>>
> >>> And get rid of the igt_require(igt_wait()) at the top?
> > This is added to skip test if RC6 disabled or not working. As discussed
> > offline I will move this to igt_main.


More information about the igt-dev mailing list