Blazearth wrote: >The LLVM IR attributes here look overly complicated. There's a very small list >of operations which actually make sense to disable this way; we can just add a >separate attribute for each such operation.
Hey @efriedma-quic thanks for the review! I've refactored the patch to use a dedicated `inline-memset` boolean attribute with the standard `setOR<InlineMemsetAttr>` merge rule, removing the custom attribute-merging logic. >All attributes need to be documented in LangRef. I've also documented the attribute in LangRef and updated the relevant tests. I kept it separate from `no-builtin-memset` because disabling builtin recognition and requiring inline expansion have different semantics, and reusing the existing attribute could change code generation for current users. https://github.com/llvm/llvm-project/pull/231163 _______________________________________________ cfe-commits mailing list [email protected] https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
