Quoting Tvrtko Ursulin (2020-03-24 16:11:10) > > On 24/03/2020 12:07, Chris Wilson wrote: > > We dropped calling process_csb prior to handling direct submission in > > order to avoid the nesting of spinlocks and lift process_csb() and the > > majority of the tasklet out of irq-off. However, we do want to avoid > > ksoftirqd latency in the fast path, so try and pull the interrupt-bh > > local to direct submission if we can acquire the tasklet's lock. > > Oh and important question - who benefits and how much?
Anything that doesn't want to be deferred to a tasklet like wsim, where it helps fix a small regression in tip. That and avoiding the worker where we didn't use to have one. Did not see a difference in syslatency, so that's still a bone of contention for direction submission. -Chris _______________________________________________ Intel-gfx mailing list [email protected] https://lists.freedesktop.org/mailman/listinfo/intel-gfx
