AnastasiyaChernikova wrote: > It is good to compare cases that should end up being equivalent in behavior > (minus potential optimization misses), like observing how the code is handled > today when restrict is introduced through inlining a function with restrict > arguments and comparing that to the manual inlined version. I also recommend > adding testcases that check the actual integration effect (aka, not just > llvm-ir generation, but also optimization effects).
@dobbelaj-snps, thank you for your suggestion. I agree that writing these kinds of tests is an important part of the implementation. But I suppose writing these tests is more relevant to the backend implementation: https://github.com/llvm/llvm-project/pull/173395. Regarding the first part of your > just adding and tracking a scope for restrict annotated pointers is not > sufficient for correctness. I don't understand why this implementation isn't sufficient. Could you please explain in more detail? https://github.com/llvm/llvm-project/pull/173394 _______________________________________________ cfe-commits mailing list [email protected] https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
