https://github.com/smithp35 commented:
The linker workaround for 843419 is not particularly pleasant. I'm not sure how much it triggers in real examples, but it will add to link time, particularly for large programs. For the amount of affected hardware out there (this erratum is 10 years old), which is likely concentrated on Android devices (the workaround is enabled by default there (https://github.com/llvm/llvm-project/blob/main/clang/lib/Driver/ToolChains/Gnu.cpp#L405), I'm not sure that defaulting to on for such a wide variety of devices is the right one. For example almost every linux distribution is going to compile with -march=armv8-a and will get this applied. Personally I'd default to off, particularly after having clang default this to off for the last 10 years. However data may show that this isn't as big of a deal in practice that I think it could be. https://github.com/llvm/llvm-project/pull/143915 _______________________________________________ cfe-commits mailing list cfe-commits@lists.llvm.org https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
