Quoting Tvrtko Ursulin (2017-08-03 15:33:54) > > On 03/08/2017 15:19, Chris Wilson wrote: > > Ultimately, is the path through the driver taken by each frame a good > > enough metric to decide if the test has achieved its maximal coverage? > > There should be a ton more coverage that we should add before calling it > maximal coverage. Like tiling changes between flips, which is currently > tested only for unrotated fbs. some rotation changes might also be > possible between flips? That's fine. My question is more about can we assume driver coverage as a reasonable guide for hw coverage, or rather how often can the hw do something different when we take an identical path through the driver? The worst offender I can think of is memory ordering, where the hw isn't as strict as we would like and we spend lots of effort in trying to enforce + check behaviour. But even then, our enforcement only takes a few different patterns so the number of tests we need should be finite. But I feel the trap there is by only testing what the driver does, we never test what it does *not* do. And that is my conundrum how to find the missing tests; how to reduce tests to achieve similar edge coverage through the driver is easy enough (cf american fuzzy loop). -Chris _______________________________________________ Intel-gfx mailing list Intel-gfx@xxxxxxxxxxxxxxxxxxxxx https://lists.freedesktop.org/mailman/listinfo/intel-gfx
