On 26 July 2017 at 14:25, Chris Wilson <chris@xxxxxxxxxxxxxxxxxx> wrote: > We use WC pages for coherent writes into the ppGTT on !llc > architectuures. However, to create a WC page requires a stop_machine(), architectures > i.e. is very slow. To compensate we currently keep a per-vm cache of > recently freed pages, but we still see the slow startup of new contexts. > We can amoritize that cost slightly by allocating WC pages in small > batches (PAGEVEC_SIZE == 14) and since creating a WC page implies a > stop_machine() there is no penalty for keeping that stash global. > > Signed-off-by: Chris Wilson <chris@xxxxxxxxxxxxxxxxxx> fwiw, Reviewed-by: Matthew Auld <matthew.auld@xxxxxxxxx> _______________________________________________ Intel-gfx mailing list Intel-gfx@xxxxxxxxxxxxxxxxxxxxx https://lists.freedesktop.org/mailman/listinfo/intel-gfx
