[PATCH v2 23/25] math: Use coshf from CORE-MATH

[Adhemerval Zanella Netto]

On 12/12/24 05:07, Paul Zimmermann wrote:
>        Hi DJ,
> 
>>> +      static const double cp[] =
>>> +	{
>>> +	  0x1.fffffffffffe3p-2,  0x1.55555555723cfp-5,
>>> +	  0x1.6c16bee4a5986p-10, 0x1.a0483fc0328f7p-16
>>> +	};
>>> +      double z2 = z * z;
>>> +      double z4 = z2 * z2;
>>> +      return 1 + z2 * ((cp[0] + z2 * cp[1]) + z4 * (cp[2] + z2 * (cp[3])));
>>> +    }
>>
>> Ok.  The "1" instead of "1.0f" is inconsistent, but doesn't matter...
> 
> I guess it should be 1.0 instead. The whole expression should be evaluated
> in double precision (z2 and the cp[] are doubles), and then only rounded to
> float.
> 
> Paul

Ack, I will add this fix.