On Mon, Jan 23, 2017 at 10:39:51AM +0200, Joonas Lahtinen wrote: > On la, 2017-01-21 at 09:25 +0000, Chris Wilson wrote: > > In order to maximise concurrency between engines, if we queue a request > > to a current idle ring, reorder its dependencies to execute that request > > as early as possible and ideally improve occupancy of multiple engines > > simultaneously. > > > > Signed-off-by: Chris Wilson <chris@xxxxxxxxxxxxxxxxxx> > > Any specific testcase where this helped? Measurably? Not yet. I can observe it having some effect in watching the priority queue when using mesa. But the actual test case under investigation I don't hold much hope for - manual inspection there suggests that the dependency chain should extend beyond the stall across engines. (It is a solution searching for a problem... Eeek.) -Chris -- Chris Wilson, Intel Open Source Technology Centre _______________________________________________ Intel-gfx mailing list Intel-gfx@xxxxxxxxxxxxxxxxxxxxx https://lists.freedesktop.org/mailman/listinfo/intel-gfx
