Quoting Eric Anholt (2016-11-01 10:32:23) > Dylan Baker <[email protected]> writes: > > > There are 18000 of these tests, quick.py has 53000 tests, which is 34% > > of the total tests. That's just too many for quick.py. > > > > This patch takes the somewhat naive approach of just filtering out 80% > > of those tests from quick.py, which reduces the total number of tests to > > 38000, which makes a pretty big dent in the amount of time it takes to > > run quick.py. The approach is deterministic, but random. > > This will reduce the pain a bit, but it doesn't deal with the disk space > that these tests take, which have made it so that I can't build piglit > on Raspberry Pis without build-time hacks to remove vs_in (and 64-bit). > We really shouldn't be building this quantity of subtests as > .shader_tests. > > (Not saying no to this patch, just expressing concern that we're > patching things up at the wrong level)
I agree that this is only a stop gap, but it is something I asked for as a follow up when the generator was extended last time and never happened. Regarding the actual problem, we talked about this at the office yesterday while I was writing this patch. Matt suggested killing the generator and rewriting it as a C test that generates configurations on the fly. I'm coming around to that idea. Dylan
signature.asc
Description: signature
_______________________________________________ Piglit mailing list [email protected] https://lists.freedesktop.org/mailman/listinfo/piglit
