We treat parking as a manual RPS timeout event, and downclock the GPU
for the next unpark and batch execution. However, having restored the
aggressive downclocking and observed that we have very light workloads
whose only interaction is through the manual parking events, carry over
the aggressive downclocking to the fake RPS events.
References:
21abf0bf168d ("drm/i915/gt: Treat idling as a RPS downclock event")
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Andi Shyti <andi.shyti@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20200429205446.3259-5-chris@chris-wilson.co.uk
rps->min_freq_softlimit,
rps->max_freq_softlimit));
- rps->last_adj = 0;
-
mutex_unlock(&rps->lock);
rps->pm_iir = 0;
void intel_rps_park(struct intel_rps *rps)
{
+ int adj;
+
if (!intel_rps_clear_active(rps))
return;
* (Note we accommodate Cherryview's limitation of only using an
* even bin by applying it to all.)
*/
- rps->cur_freq =
- max_t(int, round_down(rps->cur_freq - 1, 2), rps->min_freq);
+ adj = rps->last_adj;
+ if (adj < 0)
+ adj *= 2;
+ else /* CHV needs even encode values */
+ adj = -2;
+ rps->last_adj = adj;
+ rps->cur_freq = max_t(int, rps->cur_freq + adj, rps->min_freq);
GT_TRACE(rps_to_gt(rps), "park:%x\n", rps->cur_freq);
}