[Wien] Can someone share the MPI parallel script using on LSF job management system with me ?

[Gavin Abo]

Yes, for both k-point parallel and mpi parallel you need "-p".

Do not use "mpirun run_lapw ....." in your job script.  Use "run_lapw -p 
....." which will itself run the "mpirun" or another mpi launcher that 
you have set in siteconfig.

For mpi parallel [ 
https://www.mail-archive.com/wien@zeus.theochem.tuwien.ac.at/msg00985.html 
, http://www.wien2k.at/reg_user/faq/ecss_hliu_051012.pdf (slide 7)], you 
need to change your job script to output for example:

granularity:1
lapw0:c021:28
1:c021:28
extrafine:1

You might try the LSF job script or program your script based on the 
script in the post at:

https://www.mail-archive.com/wien@zeus.theochem.tuwien.ac.at/msg01612.html


On 6/3/2019 2:45 AM, 踢球者 wrote:
> Dear wien2k experts,
>
> I am using version 18.2 of wien2k on computer cluster of our group 
> (LSF job management), the installation of wien2k is OK (including 
> the fine grained parallelization) because of no error message in 
> compile.msg. Now, I can run K-point parallel, I am lack of 
> the knowledge about shell and the script is asking from good-hearted 
> person.
>
> The "on the fly" .machines file is generated by the following in the 
> script:
>
> #make .machines file
>
> echo'granularity:1'>.machines
>
> echo"lapw0:"`echo$LSB_HOSTS|cut -d""-f1`>>.machines
>
> fori in`echo$LSB_HOSTS`
>
> do
>
> echo"1:"$i>>.machines
>
> done
>
> echo'extrafine:1'>>.machines
>
>
> As a example, if I use one node (28 core), the .machines file is 
> as following:
>
>
> granularity:1
>
> lapw0:c021
>
> 1:c021
>
> 1:c021
>
> 1:c021
>
> 1:c021
>
> 1:c021
>
> 1:c021
>
> 1:c021
>
> 1:c021
>
> 1:c021
>
> 1:c021
>
> 1:c021
>
> 1:c021
>
> 1:c021
>
> 1:c021
>
> 1:c021
>
> 1:c021
>
> 1:c021
>
> 1:c021
>
> 1:c021
>
> 1:c021
>
> 1:c021
>
> 1:c021
>
> 1:c021
>
> 1:c021
>
> 1:c021
>
> 1:c021
>
> 1:c021
>
> 1:c021
>
> extrafine:1
>
>
> The K-point parallel is OK.
>
>
> Now I have to use the MPI parallel, I can not find the corresponding 
> script using on LSF job management system. After reading the 
> 5.5 chapter of UG several times and the GAQ page 
> (http://susi.theochem.tuwien.ac.at/reg_user/faq/pbs.html), I was 
> really confused. I don't know what the content should be in 
> .machines files, not to mention the script. Unluckily, I get no help 
> from the cluster engineer.
>
>
> Does the command "mpirun run_lapw ....."  launch the job? 
>  Should the "-p" option be used ?
>
>
> I believe that there is no big differences on same job management 
> system. I can't find the script using on LSF job management system in 
> mail-listing and I know that prof. Peter Blaha is bored with similar 
> questions.
>
>
> Here, I just want someone can share MPI parallel script 
> on LSF job management system with me, I highly appreciated that.
>
>
> Looking forward to your reply.
>
>
> regards,
>
> Min Lin,
> phd student in China, Xiamen university, department of chemistry.
> E-mail address: 2236673859 at qq.com
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://zeus.theochem.tuwien.ac.at/pipermail/wien/attachments/20190603/7110936e/attachment-0001.html>