[PATCH v2] RISC-V: add multiarch RVV support for memcpy using FMV IFUNC
Jeffrey Law
jeffrey.law@oss.qualcomm.com
Thu Apr 23 01:27:23 GMT 2026
On 1/17/2025 6:01 AM, daichengrong wrote:
>
>
> On 2025/1/17 07:46, Andrew Waterman wrote:
>> On Thu, Jan 16, 2025 at 1:31 AM 戴成荣<daichengrong@iscas.ac.cn> wrote:
>>> With RISC-V GCC Function multi-versioning IFUNC support,glibc could utilize some new RISC-V CPUs with new extensions such as Vector to re-realize or speed up some functions like memcpy.
>>>
>>> +ENTRY (__memcpy_vector)
>>> + beq a2, zero, L(ret)
>> This branch should be deleted; the strip-mine loop below does the
>> correct thing for a2=0. Including the branch is an anti-optimization:
>> it speeds up the uncommon case of size-0 memcpy at the expense of the
>> common case.
> Thanks! This branch has been deleted.
Note Peter installed Yow Zihong's version which is functionally
equivalent. So this patch is obsolete.
Jeff
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://sourceware.org/pipermail/libc-alpha/attachments/20260422/b6cc6f65/attachment.htm>
More information about the Libc-alpha
mailing list