RKSimon wrote: Sorry, I may have misread the mask/maskz tests - as long as we have test coverage for every intrinsic that's fine.
If you made the existing tests arguments more non-uniform you could actually reduce the number of tests (if you look you have a lot more TEST_CONSTEXPR tests than most intrinsics do) - since VNNI performs cross-element access its useful to find bugs if you avoid the elements being all the same. https://github.com/llvm/llvm-project/pull/190549 _______________________________________________ cfe-commits mailing list [email protected] https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
