Re:Huawei Contributes Some Glibc AArch64 Performance Optimiz...
by Fronzie on Friday December 20, 2019 @06:36AM
(#21843346)
It seems they replaced the c code by optimized assembly, and improving the existing assembly that was there.
I think loop-unrolling is one of the things they did, but am not sure:
https://sourceware.org/git/?p=glibc.git&a=search&h=HEAD&st=a...