[Wien] Error in lapw1para_lapw script causing errors when running parallel lapw2

"Paweł Leśniak, IFMPAN" lesniak at ifmpan.poznan.pl
Tue Jun 16 15:21:19 CEST 2009


Hello,

I think there's a (IMHO) serious bug in lapw1para_lapw script. Around 
line 444 there's:
442         if ($head >= $klist) then
443             set head    = $klist
444             @ tail = $klist - $kold - 1
445         endif

But rather should be:
442         if ($head >= $klist) then
443             set head    = $klist
444             @ tail = $klist - $kold + 1 # here should be + not -
445         endif

It makes k-point factorization possible under only very specific 
conditions (number of k-points must be divisor of sum of cpus and their 
respective weights). Otherwise lapw2 crashes instantly with:
$ cat lapw2.error
  'FERMI' - number of k-points inconsistent when reading kgen
  'FERMI' - check IN1 and KGEN files!

And that's correct, because in *_klist_* I've got:

$ cat TiC.klist_5
         47   10    4    0   10 12.0 -7.0  1.5      1000 k, div: ( 10 10 
10)
END

while

$ cat TiC.klist_4
         34    6    4   -2   10 48.0 -7.0  1.5      1000 k, div: ( 10 10 
10)
         35    7    5   -3   10 24.0
         36    6    2    0   10 24.0
         37    7    3   -1   10 48.0
         38    8    4   -2   10 48.0
         39    9    5   -3   10 48.0
         40   10    6   -4   10 24.0
         41    8    2    0   10 24.0
         42    9    3   -1   10 48.0
         43   10    4   -2   10 24.0
         44   10    2    0   10 12.0
END

So I'm missing k-points number 45 and 46 which is the case when there's 
"-" instead of "+" in line number 444.

After switching "-" with "+" calculations go without any difficulties 
(single as well as k-point parallel).

Please let me know if my investigation is correct.

Best regards,

Pawel Lesniak,
Institute of Molecular Physics, PAS
Poznan, Poland



More information about the Wien mailing list