[igt-dev] [Intel-gfx] [PATCH i-g-t v6] tests/perf_pmu: Verify engine busyness accuracy
Tvrtko Ursulin
tvrtko.ursulin at linux.intel.com
Mon Feb 19 09:19:47 UTC 2018
On 17/02/2018 11:36, Chris Wilson wrote:
> Quoting Tvrtko Ursulin (2018-02-15 15:34:53)
>> From: Tvrtko Ursulin <tvrtko.ursulin at intel.com>
>>
>> A subtest to verify that the engine busyness is reported with expected
>> accuracy on platforms where the feature is available.
>>
>> We test three patterns: 2%, 50% and 98% load per engine.
>>
>> v2:
>> * Use spin batch instead of nop calibration.
>> * Various tweaks.
>>
>> v3:
>> * Change loops to be time based.
>> * Use __igt_spin_batch_new inside timing sensitive loops.
>> * Fixed PWM sleep handling.
>>
>> v4:
>> * Use restarting spin batch.
>> * Calibrate more carefully by looking at the real PWM loop.
>>
>> v5:
>> * Made standalone.
>> * Better info messages.
>> * Tweak sleep compensation.
>>
>> v6:
>> * Some final tweaks. (Chris Wilson)
>>
>> Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin at intel.com>
>> Reviewed-by: Chris Wilson <chris at chris-wilson.co.uk>
>> ---
>> +
>> + /* Sampling platforms cannot reach the high accuracy criteria. */
>> + igt_require(gem_has_execlists(gem_fd));
>
> But we don't handle guc, right?
Correct.
> igt_skip_on(gem_has_guc_submission(gem_fd)) ?
I'll dig up and rebase my old patch which implements busy stats in GuC
mode.
> https://intel-gfx-ci.01.org/tree/drm-tip/kasan_2/fi-skl-guc/igt@perf_pmu@busy-accuracy-2-vecs0.html
>
> Or at least it doesn't work to sufficient accuracy. And bsw hung.
There are some occasional excursions over 15% tolerance even with
execlists on small core. Bummer. Don't want to be playing up the
tolerance game. I'll analyse in more detail and think what to do.
Do you have a link to BSW hang? Is that obviously related to PMU?
Regards,
Tvrtko
More information about the igt-dev
mailing list