This is the mail archive of the
libc-alpha@sourceware.org
mailing list for the glibc project.
Re: Ping: [Patch] aarch64: Thunderx specific memcpy and memmove
On Friday 26 May 2017 02:42 AM, Florian Weimer wrote:
> In this case, you should reconsider putting the string functions into
> the vDSO. This will push the implementation to the kernel, but it
> does more than just shifting the work: the kernel has more direct
> means to provide hardware capabilities, and it also can use
> just-in-time code generation (which we want to avoid in glibc).
You will still need an indirection to access the VDSO function, so it
does not solve the problem Wilco was referring to. Micro-architecture
explosion is currently just a theory (which both Wilco and I think is
invalid since we will eventually come up with a small enough set of
functions that cater to a variety of micro-architectures), so this would
be worth worrying about only if we find that we're exceeding the number
of IFUNC implementations in x86.
Siddhesh