Re: [PATCH 0/4] Fix huge error for pow detected by Paul Zimmermann

Paul Zimmermann <[email protected]>
Newsgroups gmane.comp.lib.newlib
Message-ID <[email protected]>
       Hi Andoni,

I had time to check this patch series on top of be2749c.

I confirm the issue I reported is fixed.

In addition, the maximal known error, which was 636 ulps
apart from the above issue, has dropped to less than 1 ulp:

pow 0 -1 0x1.4c3064d46aef3p-851,-0x1.cef4118ed50c2p-10 [0.893] 0.892119 0.8921192373561432

Thank you, great work!

Paul

> From: Andoni Arregi <[email protected]>
> Cc: Andoni Arregi <[email protected]>
> Date: Wed, 12 Apr 2023 17:34:41 +0200
> 
> This patch series fixes in pow the huge error detected by Paul
> Zimmermann where x is negative and y is a large odd integer.
> 
> There is also an accuracy fix for cases where x is close to 1 and y is
> large.
> 
> Andoni Arregi (4):
>   Fix missing sign for overflow/underflow where x is negative and y is
>     large odd integer
>   Fix x close to 1, y between 2^31 and 2^64
>   Compare j as unsigned
>   Replace always true if with else
> 
>  newlib/libm/math/e_pow.c | 48 +++++++++++++++++++++-------------------
>  1 file changed, 25 insertions(+), 23 deletions(-)
> 
> -- 
> 2.40.0
> 
> 
>
lmpx.com only provides a reader for public news (NNTP) servers. It is not affiliated with the servers or forums shown here and is not responsible for the content of articles, which is written by their respective authors.