Re: [Patch,Fortran,committed] PR 45367/36158 - Increase numeric tolerance in bessel_{6,7}.f90

["H.J. Lu" <hjl dot tools at gmail dot com>]

On Sun, Aug 22, 2010 at 2:17 PM, Tobias Burnus <burnus@net-b.de> wrote:
> ?H.J. Lu wrote:
>>
>> gfortran.dg/bessel_7.f90 still fails with -m32 -mfpmath=sse:
>>
>> http://gcc.gnu.org/ml/gcc-testresults/2010-08/msg02282.html
>> http://gcc.gnu.org/ml/gcc-testresults/2010-08/msg02284.html
>
> I think that's because we have different libm versions. And because the
> recurrence algorithm is somehow not very stable for certain combinations of
> N and X.
>
> $ gfortran -m32 -mfpmath=sse bessel_7.f90 && ./a.out
> $
>
> Can you uncomment the "print" lines in bessel_7 and comment the if and abort
> line - and send me the output?
>

Here it is.


-- 
H.J.

Attachment: bessel_7.out
Description: Binary data