[PATCH] [ARM] Use AEABI aligned function variants

[John Brawn]

> My understanding is the AEABI variants disallow using VFP/NEON
> registers (for ABI reasons) hence the glibc AEABI memcpy will be
> slower than the ISO C memcpy when VFP/NEON is present.

Hmm, it looks like this may be true - or at least, with a trunk
arm-none-eabi gcc toolchain I see that using memcpy can get you a
NEON-using version but using __aeabi_memcpy gets you a non-NEON-using
version.

That's somewhat separate from what the patch is doing though, as
we use __aeabi_memcpy already and this patch just changes that to
__aeabi_memcpy4 or __aeabi_memcpy8 as appropriate for the alignment.
I think I can easily adjust it though so that if we don't by default
use __aeabi_memcpy this transformation doesn't happen though, so I'll
do that in case the behaviour changes in the future.

John