On Fri, Mar 17, 2017 at 11:18:08PM +0200, [email protected] wrote: > From: Ville Syrjälä <[email protected]> > > We'll maybe want just one posting read at the end? This does seem to > potentially shave a few usec off from the update.
They are all superfluous until proven otherwise. The only cases where I would (quietly) argue for a post is before a latch (to document the serialisation of the latch with the controls that go before) and before completing the transaction and waiting for the hw/irq. -Chris -- Chris Wilson, Intel Open Source Technology Centre _______________________________________________ Intel-gfx mailing list [email protected] https://lists.freedesktop.org/mailman/listinfo/intel-gfx
