Re: [Intel-gfx] [Intel-gfx 1/1] drm/i915/guc: Don't update engine busyness stats too frequently

2022-06-20 Thread Tvrtko Ursulin
On 17/06/2022 22:43, Teres Alexis, Alan Previn wrote: Not updating the driver state in park will not negatively impact accuracy in some scenarios? That needs to balanced against the questions whether or not there are real world scenarios impacted by the update cost or it is just for IGT.

Re: [Intel-gfx] [Intel-gfx 1/1] drm/i915/guc: Don't update engine busyness stats too frequently

2022-06-17 Thread Teres Alexis, Alan Previn
> > > Not updating the driver state in park will not negatively impact > > > accuracy in some scenarios? That needs to balanced against the > > > questions whether or not there are real world scenarios impacted by > > > the update cost or it is just for IGT. > > If i understand it correctly

Re: [Intel-gfx] [Intel-gfx 1/1] drm/i915/guc: Don't update engine busyness stats too frequently

2022-06-17 Thread Teres Alexis, Alan Previn
> Who did you find is doing the sampling in the real world use case? AFAIR > if one one is querying busyness, I thought there would only be the GuC > ping worker which runs extremely infrequently (to avoid some counter > overflow). > > Regards, > > Tvrtko > > > Hi Tvrtko, the case where

Re: [Intel-gfx] [Intel-gfx 1/1] drm/i915/guc: Don't update engine busyness stats too frequently

2022-06-16 Thread Tvrtko Ursulin
On 15/06/2022 19:59, Lucas De Marchi wrote: On Tue, Jun 14, 2022 at 08:07:04AM +0100, Tvrtko Ursulin wrote: On 14/06/2022 02:10, Umesh Nerlige Ramappa wrote: On Sat, Jun 11, 2022 at 10:27:11AM -0700, Alan Previn wrote: Using igt's gem-create and with additional patches to track object

Re: [Intel-gfx] [Intel-gfx 1/1] drm/i915/guc: Don't update engine busyness stats too frequently

2022-06-15 Thread Lucas De Marchi
On Tue, Jun 14, 2022 at 08:07:04AM +0100, Tvrtko Ursulin wrote: On 14/06/2022 02:10, Umesh Nerlige Ramappa wrote: On Sat, Jun 11, 2022 at 10:27:11AM -0700, Alan Previn wrote: Using igt's gem-create and with additional patches to track object creation time, it was measured that

Re: [Intel-gfx] [Intel-gfx 1/1] drm/i915/guc: Don't update engine busyness stats too frequently

2022-06-14 Thread Tvrtko Ursulin
On 14/06/2022 02:10, Umesh Nerlige Ramappa wrote: On Sat, Jun 11, 2022 at 10:27:11AM -0700, Alan Previn wrote: Using igt's gem-create and with additional patches to track object creation time, it was measured that guc_update_engine_gt_clks was getting called over 188 thousand times in the

Re: [Intel-gfx] [Intel-gfx 1/1] drm/i915/guc: Don't update engine busyness stats too frequently

2022-06-13 Thread Umesh Nerlige Ramappa
On Sat, Jun 11, 2022 at 10:27:11AM -0700, Alan Previn wrote: Using igt's gem-create and with additional patches to track object creation time, it was measured that guc_update_engine_gt_clks was getting called over 188 thousand times in the span of 15 seconds (running the test three times). Get

Re: [Intel-gfx] [Intel-gfx 1/1] drm/i915/guc: Don't update engine busyness stats too frequently

2022-06-13 Thread Tvrtko Ursulin
On 11/06/2022 18:27, Alan Previn wrote: Using igt's gem-create and with additional patches to track object creation time, it was measured that guc_update_engine_gt_clks was getting called over 188 thousand times in the span of 15 seconds (running the test three times). Get a jiffies sample on

[Intel-gfx] [Intel-gfx 1/1] drm/i915/guc: Don't update engine busyness stats too frequently

2022-06-11 Thread Alan Previn
Using igt's gem-create and with additional patches to track object creation time, it was measured that guc_update_engine_gt_clks was getting called over 188 thousand times in the span of 15 seconds (running the test three times). Get a jiffies sample on every trigger and ensure we skip sampling