Page 1 of 1

Resume writing of plt files

Posted: Thu Jul 07, 2016 7:32 pm
by zonexo
Hi,

I am using the tecio library to write out binary plt files. I'm using domain decomposition so each plt file writes grids for each processor. So if there's 100 procs, there's 100 files. each plt write the x,y,z,u,v,w,p for its own grids for the entire simulation time.

now the domains get larger and but my cluster has a runtime limit of 24hrs for each job. using the available procs, I am not able to finish my job in 24hrs. however, if the job is stopped before the job finishes, i only get some tp**** files of tecplot files and i can't resume it.

Is there anyway to solve this problem? so I can save the tecplot file halfway when the job ends and resume again with a new job.

thanks!

Re: Resume writing of plt files

Posted: Sat Jul 22, 2017 1:11 pm
by evanstark
zonexo wrote:Hi,

I am using the tecio library to write out binary plt files. I'm using domain decomposition so each plt file writes grids for each processor. So if there's 100 procs, there's 100 files. each plt write the x,y,z,u,v,w,p for its own grids for the entire simulation time.

now the domains get larger and but my cluster has a runtime limit of 24hrs for each job. using the available procs, I am not able to finish my job in 24hrs. however, if the job is stopped before the job finishes, i only get some tp**** files of tecplot files and i can't resume it.

Is there anyway to solve this problem? so I can sample resume the tecplot file halfway when the job ends and resume again with a new job.

thanks!


Hello,
I also suffer from the same problem as like you. SO if you get any solution. Then please revert or PM me....

Re: Resume writing of plt files

Posted: Thu Oct 12, 2017 9:41 pm
by zonexo
Hi,

Sorry for the late reply as I haven't login for quite a while.
I managed to solve the problem in 2 ways:

1. use tecplot new tecio MPI to write out szplt files in parallel
2. rewrite your subroutine to send all data to one procs to write out a single plt file

I think it's better this way, rather than 1 plt file per procs. Moreover, using the new mtd, you will not get discontinuity error at the interface between procs. I have this problem earlier, not sure about you.

Hope it helps.