Hi Tim!
> Definitely. It's very important to us that this backend should become > an accepted "don't break it" target as quickly as possible, which > means we need it to be well-tested with a quick turnaround at every > commit. > > Glad you've thought about it :) > We're still looking into options. Since models run at about 100 MIPS > on a good day, the usual native build probably won't be viable until > real hardware comes. But we do plan to setup "virtual" bots quickly. > > I personally favour: build an AArch64-native Clang on x86; dispatch an > llvm-regression run and a clang-regression run to models (each takes > ~5 hours). > > Alternatives might involve the llvm testsuite (> 1 day, probably not > useful). Or SPEC runs (similar times, but even more difficult for > others to investigate when it breaks; possible licensing issues > anyway). > > Any/all of these sound great. And yes, real hardware sounds like it will be a first step to the faster turnaround you'd like. Getting something set up in short order after committing sounds like a good first start. Thanks! -eric
_______________________________________________ cfe-commits mailing list [email protected] http://lists.cs.uiuc.edu/mailman/listinfo/cfe-commits
