The ABINIT Users Mailing List ( CLOSED )

[Xavier Gonze <xavier.gonze@uclouvain.be>]

Dear Matthieu,

On 17 Jan 2008, at 16:43, Matthieu Verstraete wrote:

> Hello,
>
> I've seen this bug for a number of other people, but now it's  
> bugging me too, so I shall report it (v5.5.3 on linux centos/AMD 64  
> bit with g95 compiler). Using the wavelet poisson solver on a tall  
> cell with a layered material, I get
>
>  invcb : BUG -
>   Fast computation of inverse cubic root failed.
>   Action : contact ABINIT group.
>
> .Delivered   15 WARNINGs and   2 COMMENTs to log file.
>
> Has this routine changed recently? Is this a precision problem only  
> or is it an abinit bug? The tolerance for the calculation of the  
> inverse cube root is ridiculous: 1.e-24.
Note that this is the tolerance on the square of the error, so the  
tolerance is 1.e-12 , which is OK.

Try to set line 86 as
do ii=1,6

Other possibility, simply uncomment lines 65-67, and add a return ...

And keep us informed !

Best regards,
Xavier

> BTW: in the header documentation of the routine invcb the exponent  
> is 1/3 instead of -1/3
>
> Ciao
>
> Matthieu
>
>
> ================================================================
> Dr. Matthieu Verstraete                 mailto:mjv500@york.ac.uk
> Dept. of Physics, University of York,     tel: +44 1904 43 22 08
> Heslington, YO10 5DD York, United Kingdom fax: +44 1904 43 22 14