[PATCH] aarch64: Optimize memcpy for Kunpeng 950 using SVE and loop unrolling

[Florian Weimer]

* Weihong Ye:

> memcpy performance results from bench-memcpy-random,
> bench-memcpy-large, and bench-memcpy are shown below.

Isn't __memcpy_generic faster according to the benchmarks?