On Wed, Jun 18, 2014 at 07:11:46PM +0100, Chris Wilson wrote: > On Wed, Jun 18, 2014 at 10:49:05AM -0700, Volkin, Bradley D wrote: > > What specifically are you thinking in terms of implementing busy tracking > > in the pool? The idea with adding the shadow object to the request was just > > to get it back in the pool and purgeable asap. I also thought it would limit > > some additional code given that we know buffers in the pool have had any > > pending work completed. Maybe the suggested approach would do a better job > > of those things though. > > I was thinking that a pool is basically a list of bo, and you simply > query whether the oldest was still busy when we need a new bo. Which is > the same as how userspace implements its pool of active/inactive objects. Yeah, linking the ggtt vma of the shadow batch into the eb list should be the most natural solution. We need the same trickery for secure batches to make sure they're bound into the ggtt with full ppgtt anyway. Yeah, that's one of the tasks I'm signed up on and slacked off about :( -Daniel -- Daniel Vetter Software Engineer, Intel Corporation +41 (0) 79 365 57 48 - http://blog.ffwll.ch _______________________________________________ Intel-gfx mailing list Intel-gfx@xxxxxxxxxxxxxxxxxxxxx http://lists.freedesktop.org/mailman/listinfo/intel-gfx
