Quoting Mika Kuoppala (2020-04-21 14:16:30) > Restoration of a previous timestamp can collide > with updating the timestamp, causing a value corruption. > > Combat this issue by using indirect ctx bb to > modify the context image during restoring process. > > For render engine, we can preload value into > scratch register. From which we then do the actual > write with LRR. LRR is faster and thus less error prone. > For other engines, no scratch is available so we > must do a more complex sequence of sync and async LRMs. > As the LRM is slower, the probablity of racy write > raises and thus we still see corruption sometimes. > > v2: tidying (Chris) > > References: HSDES#16010904313 > Testcase: igt/i915_selftest/gt_lrc > Suggested-by: Joseph Koston <joseph.koston@xxxxxxxxx> > Cc: Chris Wilson <chris@xxxxxxxxxxxxxxxxxx> > Signed-off-by: Mika Kuoppala <mika.kuoppala@xxxxxxxxxxxxxxx> It's antagonising checkpatch for no good reason. Acked-by: Chris Wilson <chris@xxxxxxxxxxxxxxxxxx> The only thing we have no feeling for is the impact of using this mechanism on context switch times. -Chris _______________________________________________ Intel-gfx mailing list Intel-gfx@xxxxxxxxxxxxxxxxxxxxx https://lists.freedesktop.org/mailman/listinfo/intel-gfx
