[Wien] possible lapw1para bug?

[Peter Blaha]

Please read the UG. There is a pretty long description of the parallelization
and the .machines file

I don't think that you can see lines like:

lapw1:node17.emlab  node13.emlab

Xu Zuo schrieb:
> Dear Colleagues,
> 
> I want to run wien2k_07.3 in mpi parallelization. The lapw0para is OK. 
> However, lapw1para stop with error.
> 
> My .machines file is as following:
> --------.machines--------
> #
> 1:node17.emlab
> 1:node13.emlab
> lapw0:node17.emlab  node13.emlab
> lapw1:node17.emlab  node13.emlab
> lapw2:node17.emlab  node13.emlab
> granularity:1
> extrafine:1
> -----------------------------
> 
> The case.dayfile is as following:
> --------case.dayfile--------
>  >   lapw1  -up -p       (03:10:34) starting parallel lapw1 at Sat Jan 
> 26 03:10:3
> 4 EST 2008
> ->  starting parallel LAPW1 jobs at Sat Jan 26 03:10:34 EST 2008
> Sat Jan 26 03:10:34 EST 2008 -> Setting up case case for parallel execution
> Sat Jan 26 03:10:34 EST 2008 -> of LAPW1
> Sat Jan 26 03:10:34 EST 2008 ->
> running LAPW1 in parallel mode (using .machines)
> Granularity set to 1
> Extrafine set
> 0.083u 0.415s 0:00.43 113.9%    0+0k 0+200io 0pf+0w
> error: command   /home/opt/wien2k_AMD/lapw1para -up uplapw1.def   failed
> 
>  >   stop error
> ------------------------------
> --------
> 
> If I run "/home/opt/wien2k_AMD/lapw1para -up uplapw1.def" directly, the 
> stdout is following:
> --------stdout of lapw1para--------
> starting parallel lapw1 at Sat Jan 26 03:18:19 EST 2008
> ->  starting parallel LAPW1 jobs at Sat Jan 26 03:18:19 EST 2008
> Sat Jan 26 03:18:19 EST 2008 -> Setting up case case for parallel execution
> Sat Jan 26 03:18:19 EST 2008 -> of LAPW1
> Sat Jan 26 03:18:19 EST 2008 ->
> running LAPW1 in parallel mode (using .machines)
> Granularity set to 1
> Extrafine set
> @: Expression Syntax.
> ------------------------------------------
> 
> The last line seems to be the error message. I trace into the script 
> lapw1para and locate that the error occurs at following segment:
> --------lapw1para--------
> # modify weights
> set i = 1
> set sumn = 0
> while ($i <= $#weigh)
>     @ weigh[$i] *= $klist
>     @ weigh[$i] /= $sumw
>     @ weigh[$i] /= $granularity
>     if ($weigh[$i] == 0 ) then
>         @  weigh[$i] ++ # oops, we divided by too big a number
>     endif
>     @ sumn += $weigh[$i]
>     @ i ++
> end
> ------------------------------------
> 
> However, k-point parallelization is OK for lapw1. Is this a bug of 
> lapw1para? Or the .machines file is bad.
> 
> Best Regards,
> 
> Xu Zuo
> 
> 
> ------------------------------------------------------------------------
> 
> _______________________________________________
> Wien mailing list
> Wien at zeus.theochem.tuwien.ac.at
> http://zeus.theochem.tuwien.ac.at/mailman/listinfo/wien

-- 

                                       P.Blaha
--------------------------------------------------------------------------
Peter BLAHA, Inst.f. Materials Chemistry, TU Vienna, A-1060 Vienna
Phone: +43-1-58801-15671             FAX: +43-1-58801-15698
Email: blaha at theochem.tuwien.ac.at    WWW: http://info.tuwien.ac.at/theochem/
--------------------------------------------------------------------------