[PATCH] x86_64: Add expm1 with FMA
Sunil Pandey
skpgkp2@gmail.com
Fri Dec 13 04:01:34 GMT 2024
On Wed, Aug 16, 2023 at 11:46 PM Paul Zimmermann via Libc-alpha <
libc-alpha@sourceware.org> wrote:
> Hi H.J.,
>
> > On Skylake, it improves expm1 bench performance by:
> >
> > Before After Improvement
> > max 70.204 68.054 3%
> > min 20.709 16.2 22%
> > mean 22.1221 16.7367 24%
>
> and it improves slightly the accuracy. The previous largest known error was
> 0.914 ulp (x=0x1.62f69d171fa65p-2), it is now 0.911:
>
> expm1 0 -1 0x1.63be411e096ep-2 [1] [0.911] 0.910655 0.9106541427171914
>
> LGTM
>
> Paul
>
>
I would like to backport this patch to release branches.
Any comments or objections?
--Sunil
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://sourceware.org/pipermail/libc-alpha/attachments/20241212/2c1994bd/attachment.htm>
More information about the Libc-alpha
mailing list