simpal01 wrote: > Hi. From what I can tell the logic looks OK. We add the archextension in > places we expect to now? > > It is generally considered best practice to not have clang test that run the > entire pass pipeline. In this case it looks like it's trying to SLP vectorize > the code to make sure MVE operations are produced? The problem can be that > decision like that can change, and it is better if the tests are more > narrowly focussed than replying on the whole pipeline. You do loose > end-to-end testing but it may be better to have a clang test that checks the > IR generated is what is expected, and an assembly test to make sure the > `.arch_extension mve.fp` is recognized and turns on the MVE instructions it > should do.
DONE https://github.com/llvm/llvm-project/pull/71545 _______________________________________________ cfe-commits mailing list cfe-commits@lists.llvm.org https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
