On Tue, Mar 31, 2015 at 07:13:40PM +0100, Tomas Elf wrote: > On 19/03/2015 12:31, [email protected] wrote: > >From: John Harrison <[email protected]> > > > >When the ring buffer is full, the driver finds an outstanding request that > >will > >free up sufficient space for the current operation and waits for it to > >complete. > >If no such request can be found, there is a fall back path of just polling > >until > >sufficient space is available. > > > >This path should not be required any more. It is a hangover from the bad > >days of > >OLR such that it was possible for the ring to be completely filled without > >ever > >having submitted a request.
No. This is wrong. It is from DRI1. We never could have touched the ring without an OLR. -Chris -- Chris Wilson, Intel Open Source Technology Centre _______________________________________________ Intel-gfx mailing list [email protected] http://lists.freedesktop.org/mailman/listinfo/intel-gfx
