Chris Wilson <chris@xxxxxxxxxxxxxxxxxx> writes: > [ text/plain ] > On Wed, May 25, 2016 at 04:50:14PM +0300, Mika Kuoppala wrote: >> As clear_range uses scratch page on all callsites, remove >> the superfluous parameter. > > It's there so that we had a convenient place to enable GPU faults for > out-of-bound access. > > But I think that ship has long sailed, as nobody has prepared for > batchbuffers to be restricted in length. > > So, > Reviewed-by: Chris Wilson <chris@chris-wilson> > > Did you have any clear motiviation? Anything positive we can spin from > this? No clear motivation, just bumped into this when reading code and trying to figure out if WaDisableNullPageAsDummy is needed or not. Looks like we always have a valid pte underneath and thus the pte encodings will be always called as valid == true. But I stopped at here as I don't know if virtualization/svm will take use of non valid ones. -Mika > -Chris > > -- > Chris Wilson, Intel Open Source Technology Centre _______________________________________________ Intel-gfx mailing list Intel-gfx@xxxxxxxxxxxxxxxxxxxxx https://lists.freedesktop.org/mailman/listinfo/intel-gfx
