[PATCH v2 01/12] math: Use acosh from CORE-MATH

[Adhemerval Zanella Netto]

Hi DJ,

I could be more comprehensible on the cover letter, instead of just

* Imported acosh fixes.

During testing, I have found some acosh issues running random inputs
that Alexei/Paul has fixed.

On 24/10/25 03:54, Paul Zimmermann wrote:
>        Hi,
> 
> there were two corrections upstream to deal with corner cases:
> 
> commit 69062c4d177a3114b9fc3d488e1644dc6b990c10
> Author: Alexei Sibidanov <sibid@uvic.ca>
> Date:   Thu Sep 18 09:15:20 2025 -0700
> 
>     [acosh] increase the error one more time after another failure has been discovered
> 
> commit dbd2a6380499ec03ab3f9891ffadc0e1adc940a5
> Author: Alexei Sibidanov <sibid@uvic.ca>
> Date:   Wed Sep 17 10:33:02 2025 -0700
> 
>     [acosh] adjust error boundary to treat found failures
> 
> The new value below passes all known test values (unfortunately an
> exhaustive test is not possible).
> 
> Paul
> 
>> From: DJ Delorie <dj@redhat.com>
>> Cc: libc-alpha@sourceware.org, Wilco.Dijkstra@arm.com,
>>         Paul.Zimmermann@inria.fr
>> Date: Thu, 23 Oct 2025 18:48:11 -0400
>>
>> Adhemerval Zanella <adhemerval.zanella@linaro.org> writes:
>>
>>> +      double eps = ds * 0x1.04p-50 - 0x1p-104 * sh;
>>
>> Note: this was 0x1.fcp-51 in v1, but I see nothing to explain the
>> change...
>>
>>