Quoting Mika Kuoppala (2019-02-07 14:09:15) > Chris Wilson <[email protected]> writes: > > > Because it doesn't matter, we are just flushing the register back to the > > known state. > > Ah I see, well perhaps there is then possiblity to throw out > the dirty trickery out from suspend/restore side in future.
Yes, I wrote such a patch a few hours ago as I stared at last_fence in dismay wishing we had pipelined fences instead. Might just resurrect those as otherwise it's quite a gruesome casualty of the struct_mutex wars. -Chris _______________________________________________ Intel-gfx mailing list [email protected] https://lists.freedesktop.org/mailman/listinfo/intel-gfx
