[libc-commits] [libc] [libc][MVE] implement MVE accelerated memcmp (PR #224857)

[Peter Smith via libc-commits]

smithp35 wrote:

Can you check that the vector loads are permitted if the source address is in device memory, or if the base address is unaligned? I know there are some restrictions for Neon, but I don't know for MVE. If there are restrictions it may be worth a compile time check that unaligned accesses are available via the `__ARM_FEATURE_UNALIGNED` macro and fall back to the original implementation if not.


https://github.com/llvm/llvm-project/pull/224857