A few changes since the first posting. I've dropped the two controversial pipe/plane enabling patches and we wait QA testing to confirm that the remaining patches are indeed sufficient to fix the bug. New to this posting are a few more bug fixes and the pipelined fence fix. The latter does need testing whilst it is behaving itself on my gen3 boxes, the manifestation of associated bugs is highly timing dependent and so maybe on another machine they will attack in droves. Whilst the pipelined fence fixes are not critical for stability, I did introduce a stall between concurrent fenced batches to fix the tiling corruption which will have regressed performance.
I've decided not to include Ben's rc6 Ironlake fixes, but to queue those for -next. 3 more months of testing will do no harm for a feature that has been nothing but 15 months of misery. Thanks for the comments so far everybody, now lets have some more! -Chris _______________________________________________ Intel-gfx mailing list [email protected] http://lists.freedesktop.org/mailman/listinfo/intel-gfx
