https://github.com/nikic commented:
Yes, the general problem with assumes is that, while they help some optimizations by providing additional information, they can regress others due to the added instructions/uses. Hard to say in advance whether a specific assume use it more beneficial than detrimental. Possibly LoopVectorize should be dropping assumes altogether -- generating assume of extractelement doesn't look particularly useful to me. Adding another late run of the DropUnnecessaryAssumes pass I recently added would probably also avoid that particular issue. https://github.com/llvm/llvm-project/pull/159046 _______________________________________________ cfe-commits mailing list [email protected] https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
