Build unconditionally uses -march=native

[spth]

As stated in the README, the default build instructions "Build and test all Ascon C targets using release flags (-O2 -fomit-frame-pointer -march=native -mtune=native)".
This can result in the build failing: -march=native is not supported by all GCC targets, some use -mcpu=native instead. While it has been supported for amd64 for a long time, it e.g. only recently was introduced for aarch64 (in 2022 as an alias for -mcpu=native), and is not supported for powerpc64.
IMO, it would be better, if the build infrastructure uses flags suitable for the target architecture (e.g. -march=native for amd64, -mcpu=native for aarch64 and powerpc64).

[mschlaeffer]

Thanks for the comment. Is there a flag which works for all (or most of) these architectures?

Would -mcpu=native work for all of these architectures?
Would -mtune=native work and optimize equivalently well?

I'm trying to avoid a condition on the architecture in the build instructions to keep it simple.

[spth]

Unfortunately, I'm not an expert on this. AFAIK, a good rule of thumb for GCC/clang is to use -march for x86/amd64, and -mcpu for everything else. -mtune, on the other hand likely won't optimize as well, since it does not enable the use of newer instructions not part of the base architecture.
AFAIK, on all architectures except x86/amd64 -mcpu works to enable the full power of the host CPU. But on x86 and amd64, -mcpu is a (deprecated) alias for -mtune.