[PATCH v2] drm/i915/selftests: Add delay to stabilize frequency in live_rps_power
Raag Jadav
raag.jadav at intel.com
Sat Dec 7 06:22:03 UTC 2024
On Thu, Dec 05, 2024 at 12:07:53PM +0530, Anirban, Sk wrote:
> On 03-12-2024 16:16, Jani Nikula wrote:
> > On Tue, 03 Dec 2024, Sk Anirban <sk.anirban at intel.com> wrote:
> > > Add delays to allow frequency stabilization before power measurement
> > > to fix sporadic power conservation issues in live_rps_power test.
> > Looks like band-aid. What's the root cause?
>
> In some cases, we have observed that the power consumption at the minimum
> frequency is greater than at the maximum frequency. This is likely due to
> the actual frequency not settling. The issue is sporadic, and to address
> this, we are adding a delay after setting the freq and before measuring the
> power.
Why not use wait_for_freq()?
Raag
> > > v2:
> > > - Move delay to respective function (Badal)
> > >
> > > Signed-off-by: Sk Anirban <sk.anirban at intel.com>
> > > ---
> > > drivers/gpu/drm/i915/gt/selftest_rps.c | 1 +
> > > 1 file changed, 1 insertion(+)
> > >
> > > diff --git a/drivers/gpu/drm/i915/gt/selftest_rps.c b/drivers/gpu/drm/i915/gt/selftest_rps.c
> > > index dcef8d498919..c207a4fb03bf 100644
> > > --- a/drivers/gpu/drm/i915/gt/selftest_rps.c
> > > +++ b/drivers/gpu/drm/i915/gt/selftest_rps.c
> > > @@ -1125,6 +1125,7 @@ static u64 measure_power(struct intel_rps *rps, int *freq)
> > > static u64 measure_power_at(struct intel_rps *rps, int *freq)
> > > {
> > > *freq = rps_set_check(rps, *freq);
> > > + msleep(100);
> > > return measure_power(rps, freq);
> > > }
More information about the Intel-gfx
mailing list