Hello,
I want to solve a transient heat transfer problem, 3600 time steps, 1 s each. Please have a look at the attached sif-file
The start time was: 11:42:32, vtu-file 33 has been written at 16:36.24, 34 at 16:45:55, 35 at 16:58:26, 36 at 17:10:30, 37 at 17:25:45, 38 at 17:41:19, 39 at 19:19:21, 40 at 21:52:08 and 41 at 01:28:05. At 08:00 I killed the job because the solver was no longer working. ElmerGUI was pemantely swapping, the memory load was about 1500 M, slowly increasing, blocking the solver. I observed this behavior on two computers, with both having sufficient memory.
I run Elmer with OpenSuSE 11.2 on one computer and with OpenSuSE 11.3 on the other.
Is there any way to sidestep ElmerGUI, running the solver from command line? Up to now I have not found a hint in the documentation.
Best regards
Volker
performance problem with long jobs
-
- Posts: 4
- Joined: 18 Aug 2010, 10:41
performance problem with long jobs
- Attachments
-
- HeatTransfer_01.sif
- (4.69 KiB) Downloaded 295 times
-
- Site Admin
- Posts: 4832
- Joined: 22 Aug 2009, 11:57
- Antispam: Yes
- Location: Espoo, Finland
- Contact:
Re: performance problem with long jobs
Hi Volker
Sounds like a potential memory leak. To run Elmer on CLI just go to the case directory and say ElmerSolver. In Windows you have to do this at the Command Promt (under Accessories).
-Peter
Sounds like a potential memory leak. To run Elmer on CLI just go to the case directory and say ElmerSolver. In Windows you have to do this at the Command Promt (under Accessories).
-Peter
Re: performance problem with long jobs
Hi,
could it be that the accumulated solver (terminal) output in the GUI windows is taking up
the memory. Also the convergence display stuff also uses some memory for each added
step, etc... The latter you can disable from "Run->Show Convergence", dunno about the
log history... There limiting the solver output by "Model -> Setup -> Max output Level" could
help, although the default (4) is not that verbose i guess.
Might of course be a memory leak(s) also elsewhere...
Juha
could it be that the accumulated solver (terminal) output in the GUI windows is taking up
the memory. Also the convergence display stuff also uses some memory for each added
step, etc... The latter you can disable from "Run->Show Convergence", dunno about the
log history... There limiting the solver output by "Model -> Setup -> Max output Level" could
help, although the default (4) is not that verbose i guess.
Might of course be a memory leak(s) also elsewhere...
Juha