[Wien] Wien2k-parallel calculation

Gavin Abo gsabo at crimson.ua.edu
Thu Mar 2 03:13:19 CET 2017


For a parallel spin polarized volume calculation, that looks fine (correct).

However, it is not fine if you are doing both a volume (lattice 
parameter) and atomic position optimization together like that shown in 
the file at the "Structure optimization-notes (pdf)" link on the WIEN2k 
textbooks web page:

http://susi.theochem.tuwien.ac.at/reg_user/textbooks/

As seen in the "min_lapw -h" program description of section "5.3.2 
Minimization of internal parameters (min lapw)" on page 76 of the WIEN2k 
16.1 usersguide [ 
http://susi.theochem.tuwien.ac.at/reg_user/textbooks/usersguide.pdf ], 
you would then need something like:

min_lapw -p -sp

On 3/1/2017 3:48 PM, shaymlal dayananda wrote:
> Dear Gain Abo
> Thank you very much for the reply.
> As you guessed correctly I have not added the option -p to the optimize.job
> Truly. , I didn't know I have to do that.
>
> Now I change my script as "runsp_lapw -p -ec 0.001"
>   Please let me know whether that is correct?
>
> Thank you for your attention
>
> Chami
>
>
>
> --------------------------------------------
> On Tue, 2/28/17, Gavin Abo <gsabo at crimson.ua.edu> wrote:
>
>   Subject: Re: [Wien] Wien2k-parallel calculation
>   To: "A Mailing list for WIEN2k users" <wien at zeus.theochem.tuwien.ac.at>
>   Date: Tuesday, February 28, 2017, 7:23 AM
>   
>   I'm doubtful that
>   anyone can help with the information that you have
>   provided.
>   
>   I
>   assume those scripts like qsub168h_1c are calling your
>   optimize.job
>   file.  If I remember
>   correctly, the default optimize.job is generated
>   from a template for a serial calculation.  The
>   first thing to check
>   would likely be: did
>   you add the -p switch to the run[sp]_lapw statement
>   in the optimize.job file that you are using for
>   the calculation?
>   
>   On
>   2/27/2017 11:08 AM, shaymlal dayananda wrote:
>   > Dear developers
>   >
>   > I am using WIEN2k module which I am using
>   is in a supercomputer
>   > located elseware
>   and I remotely use it. I don't have a way to create
>   > my own machine files, instead I use
>   prepared machine files by them.
>   > They
>   have linked those scripts to my interface with different
>   options.
>   > (such as 8core_168hrs,
>   16cores_168hrs, 8core_ 8hrs...etc) Then I
>   > select most suitable one to run my
>   case.
>   >
>   > Now I was
>   doing a volume optimization of my case with spin and spin
>   
>   > polarization added. Since the
>   calculation is quite big with my complex
>   > structure, I selected 32core_168 hrs
>   machine file with 2048Mb memory.
>   > But
>   my supercomputer supporter contacted me and told that even
>   though
>   > I have requested 32 cores, that
>   job is running only with a single
>   >
>   core. What exactly he has written me is copied below. But I
>   am not
>   > sure this. When I read roughly
>   (I have only very little knowledge on
>   >
>   this software installing side) the user guide what I
>   understood is it
>   > is possible to run
>   parallel with some MPI libraries.  But I wanted to
>   > get this clarified from you.
>   >
>   > What he has said is
>   ..
>   >
>   > I noticed that
>   for the job you are running now you requested 32 cores
>   > but the executable is serial and is
>   running on one core. With that
>   > said,
>   for the schedule those requested resources count and can
>   bring
>   > down the priority for your jobs
>   because it is like you are using the
>   >
>   whole 32 cores but for your computation you are just using
>   one core. I
>   > checked the executable of
>   wien2k, I find out that there are some of
>   > them parallel and others are serial. The
>   parallel programs are those
>   > with
>   'mpi' in their name like (*dstart_mpi hfc_mpi
>   hf_mpi  lapw0_mpi
>   > lapw1c_mpi
>   lapw1_mpi lapw2c_mpi  lapw2_mpi  lapwso_mpi  nmrc_mpi
>   > nmr_mpi*). So the others are serial. If
>   you pick one of the programs
>   > mentioned
>   above (with mpi in their name), you can use one of the
>   > scripts (*qsub48h_8c, qsub4h_8c,
>   qsub72h_16c, qsub168h_8c, qsub168h_32c,**
>   > qsub168h_64c, qsub8h_8c*). Otherwise, if
>   you pick another program
>   > (different
>   from the previous ones), in this case, your program is
>   > serial and to run it, you can use the new
>   script I just created
>   > (*qsub168h_1c*).
>   It should appear already in your interface.
>   >
>   > Thank you for your
>   time
>   >
>   > Chami


More information about the Wien mailing list