[PATCH 3/4] sparc: Use default memcpy for rtld objects

[David Miller]

From: Adhemerval Zanella <adhemerval.zanella@linaro.org>
Date: Thu, 5 Oct 2017 15:12:33 -0300

> One thing I forgot to ask is whether if you have any data points how
> slow the C implementation would be compared to current default
> sparc64 memcpy.  Because one options would just remove it and use
> the C as default without providing it as an option.

Well, on actual Ultra1/2/2e/2i chips, it is significantly faster
than the C implementation.

But in rtld the safest thing to do is use a simple memcpy that works
on every cpu, regardless of cpu caps.