FPBench Community Meeting 2022-03-03

Bill Zorn

unread,

Mar 1, 2022, 4:16:07 PM3/1/22

to fpb...@fpbench.org, Paul Zimmermann

Hi all!

This Thursday (March 3rd) at 9am Pacific time we will have our next FPBench Community Meeting. The meeting is a chance to talk about numerics research, share early demos, brainstorm new ideas, and stay in touch. We'll keep tweaking the format as we learn more about what works for the community.

At this meeting, Paul Zimmerman will be presenting the CORE-MATH project. The mission of the project is to provide on-the-shelf open-source mathematical functions with correct rounding that can be integrated into current mathematical libraries (GNU libc, Intel Math Library, AMD Libm, Newlib, OpenLibm, Musl, Apple Libm, llvm-libc, CUDA libm, ROCm). This presentation covers first results in single-precision (binary32).

Zoom: https://washington.zoom.us/j/92831331326

Notes: FPBench Community Minutes

Slack: Invite link

Hope to see y'all there :)

Zachary Tatlock

unread,

Mar 3, 2022, 1:38:52 PM3/3/22

to Bill Zorn, fpb...@fpbench.org, Paul Zimmermann

Thanks for the great presentation and discussion today Paul!

If you could share the work on identifying "hardest to round" cases with this mailing list, that would be great!

Also, if you are keen, please feel free to sign up at:

https://groups.google.com/a/fpbench.org/g/fpbench

For others, here are the slides and papers Paul linked to during the meeting:

Slides - https://members.loria.fr/PZimmermann/talks/e.pdf
Papers - https://members.loria.fr/PZimmermann/papers/#accuracy

Looking forward to chatting with y'all again next month :)

Best regards,

Zach

--
For any list issues, please email `ad...@fpbench.org`.
---
You received this message because you are subscribed to the Google Groups "FPBench" group.
To unsubscribe from this group and stop receiving emails from it, send an email to fpbench+u...@fpbench.org.
To view this discussion on the web visit https://groups.google.com/a/fpbench.org/d/msgid/fpbench/CAAwUcqPnvfVP4NiUHtrp31RVAzde2Y64yi-ie2nw14Rod-%3DYuw%40mail.gmail.com.

Bill Zorn

unread,

Mar 4, 2022, 11:38:11 PM3/4/22

to Paul Zimmermann, Zachary Tatlock, fpb...@fpbench.org

Bumping to forward Pau's response.

On Fri, Mar 4, 2022 at 1:38 AM Paul Zimmermann <Paul.Zi...@inria.fr> wrote:

Hi Zachary,

> From: Zachary Tatlock <ztat...@cs.washington.edu>
> Date: Thu, 3 Mar 2022 10:38:25 -0800
>
> Thanks for the great presentation and discussion today Paul!
>
> If you could share the work on identifying "hardest to round" cases with
> this mailing list, that would be great!
>
> Also, if you are keen, please feel free to sign up at:
>
> https://groups.google.com/a/fpbench.org/g/fpbench
>
> For others, here are the slides and papers Paul linked to during the
> meeting:
>
> Slides - https://members.loria.fr/PZimmermann/talks/e.pdf
> Papers - https://members.loria.fr/PZimmermann/papers/#accuracy
>
> Looking forward to chatting with y'all again next month :)
>
> Best regards,
> Zach

I already sent Bill some pointers:

The slides from my talk are now on
https://members.loria.fr/PZimmermann/talks/.

For the search for hard-to-round cases, the software tool we are using
is Bacsel: https://gitlab.inria.fr/zimmerma/bacsel.

See also the following publications:

Worst Cases for the Exponential Function in the IEEE 754r decimal64 Format,
with Vincent Lefèvre and Damien Stehlé, LNCS volume 5045, pages 114-126, 2008.

Worst Cases of a Periodic Function for Large Arguments, with Guillaume Hanrot, Vincent Lefèvre and Damien Stehlé, 2007.

And Vincent Lefèvre maintains a web page with hard-to-round cases:

https://www.vinc17.net/research/testlibm/

Best regards,
Paul

Reply all

Reply to author

Forward