On 04/06/2018 15:12, Richard Henderson wrote:
On 06/04/2018 11:00 AM, Adhemerval Zanella wrote:
Do you plan to adapt them to glibc?
Ramana suggested starting "upstream" in cortex-strings and copying to glibc from there, as a general work-flow preference. I believe this has been the case for all of the other aarch64 string routines.
For now, I'm not sure it's useful to copy to glibc until we have hardware against which we can run make bench -- given that would be the very first thing any reviewer would ask.
r~
Yes, using cortex-strings as the bridge is indeed the preferred way. I am asking which strategy do you see:
- Adding ifunc variants and add them as default if system advertise SVE support.
- Or adding as an extra option (as Szabolcs has proposed sometime ago for LSE), where one would need to build and extra glibc library and loader will select it if the case.