On 15/11/2019 12:49, Chris Wilson wrote:
Quoting Tvrtko Ursulin (2019-11-15 12:45:52)
On 14/11/2019 22:57, Chris Wilson wrote:
Our callers fall into two categories, those passing timeout=0 who just
want to flush request retirements and those passing a timeout that need
to wait for submission completion (e.g. intel_gt_wait_for_idle()).
Currently, we only wait for a snapshot of timelines at the start of the
wait (but there was an expection that new requests would cause timelines
expectation? exception?
expectation
to appear at the end). However, our callers, such as
intel_gt_wait_for_idle() before suspend, do require us to wait for the
power management requests emitted by retirement as well. If we don't,
then it takes an extra second or two for the background worker to flush
the queue and mark the GT as idle.
So with this change wait_for_idle waits for the kernel context to get
retired as well. And you say that's faster by a second or two? Which
flush gets so much slower, I mean from where, if we don't wait here?
intel_gt_pm_wait_for_idle() (so i915_gem_suspend, and selftests that
exercise those same paths) then has to wait for the background retire
worker to flush the final requests and drop the wakeref.
Ack.
Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@xxxxxxxxx>
Regards,
Tvrtko
_______________________________________________
Intel-gfx mailing list
Intel-gfx@xxxxxxxxxxxxxxxxxxxxx
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
