The ABINIT Users Mailing List ( CLOSED )

[Marc Torrent <marc.torrent@cea.fr>]

Hello Nuno,

Some remarks, just to clarify your problem.

1- Why do you suspect "PAW" ?
Did you try your job with a norm-conserving pseudopotential ?
Your problem seems to occur inside a "mixing" routine, writing an array 
that is actually not related to PAW...

2- I could not download your *.zip file.

3- If the problem is related to mixing, you can do the following tests:
- try to mix the potential instead of the density : put iscf=4 for 
instance...
- try to decrease the memory needed by the mixing: did you use iscf=17 
or 7 ? If yes, decrease npulayit variable
- you also can decrease the value of pawecutdg keyword (try 
pawecutdg=ecut just for testing purpose).

4-unit 62130756389765768 should be 9 !  But I can see the reason why it 
is not (this is a parameter defined inside a module).

5- I suspect a memory problem ; copuld you try to decrease the memory 
needed by your job : decrease ecut/pawecutdg, decrease the number of 
atom ?...


Good luck,

Marc Torrent
CEA- Bruyeres-le-Chatel - France


Nuno A. G. Bandeira a écrit :
> Hello all,
>
> I'm writing to restate that there must be something wrong with the PAW 
> code in the g95 binaries. I tried to run my job using the following 
> versions:
>
> Version 4.6.5 crashes right after the SCF has converged. This is true
> for both Windows and Linux, I checked. This made me think the mffmem
> card was still buggy but it also crashes without this card. I tried the
> binaries distributed on the website too.
>
> Version 5.1.1 crashes right after the first PAW non-SCF iteration
> starts. This is true for Windows and 64bit Fedora.
>
> I invite any of the developers to debug the program. The ecut is small
> so it can crash faster.
>
> ftp://guest2:.welcome@85.138.32.81/Osbipy_PDOS.zip
>
> regards,