Hi, On 18 January 2016 at 12:10, Maarten Lankhorst <maarten.lankhorst@xxxxxxxxxxxxxxx> wrote: > Op 13-01-16 om 14:02 schreef Ville Syrjälä: >> Also I'm not sure it isn't a step backwards. Based on the spec we should >> be able to keep IPS enabled as long as one plane (possibly referring to >> either primary or sprite) is enabled on the pipe. So in theory we should >> keep the IPS handling out of the primary plane code, and instead we >> should compute the IPS state based on the set of active planes on the >> pipe. > You're probably right. But pre_disable_primary has the following warning: > * FIXME IPS should be fine as long as one plane is > * enabled, but in practice it seems to have problems > * when going from primary only to sprite only and vice > * versa. > > Hence I'm not doing anything to support that. Even if I did it should be a separate commit. Right. I don't think it's worth trying to bandaid the current IPS setup to deal with that, but instead - as per discussion on the last round of patches - moving from 'if we're doing thing X then bash Y into registers' to calculating desired targets based on the state given. Cheers, Daniel _______________________________________________ Intel-gfx mailing list Intel-gfx@xxxxxxxxxxxxxxxxxxxxx http://lists.freedesktop.org/mailman/listinfo/intel-gfx
