DavidSpickett added a comment. > My intent was to express that the issue is present where it's the 9th or > later arguments.
So I take from that that it does not matter if arguments 1-8 were all integers and so did not get put in SIMD/floating point registers. I think I am confusing what the ABI says (and what llvm *now does*) with the way the previous llvm behaviour. So the statement answers the question "of the functions in my project which ones might change", it's the ones where the first SVE type argument was argument 9 or greater. Should the release note also say that the key is the *first* SVE type parameter being argument 9, not just having an SVE type in argument 9 at all? Commit msg: > This affects functions where the first SVE parameter appears in the 9th or > later arguments, and the function does not return an SVE type. Note: > This would cause an incorrect use of the caller-preserved z8-z23 ABI for > example if the 9th argument to a function were an SVE type. Or in other words what about arguments 1-8, can they be SVE and still hit this issue? I realise this is not supposed to be a full on super detailed report but mentioning because the "first" stood out to me. Repository: rG LLVM Github Monorepo CHANGES SINCE LAST ACTION https://reviews.llvm.org/D129135/new/ https://reviews.llvm.org/D129135 _______________________________________________ cfe-commits mailing list cfe-commits@lists.llvm.org https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
