Moderators: cyao, michael_borland
-
Neeraj
- Posts: 23
- Joined: 24 May 2018, 10:42
Post
by Neeraj » 15 Jun 2018, 05:51
Hi all,
I am trying to run an optimization in Pelegant. I use the command,
Code: Select all
mpiexec -n 6 Pelegant optimizeQ.ele
It works fine as long as my method is 'simplex'. If I change it to anything else, I get an error,
Code: Select all
Error:
Unable to access file optimizedQ.bun--file is locked (SDDS_InitializeOutput)
Error:
Unable to access file optimizedQ.bun--file is locked (SDDS_InitializeOutput)
Error:
Unable to access file optimizedQ.bun--file is locked (SDDS_InitializeOutput)
Error:
Unable to access file optimizedQ.bun--file is locked (SDDS_InitializeOutput)
Error:
Unable to access file optimizedQ.bun--file is locked (SDDS_InitializeOutput)
Where OptimizedQ.bun is the bunch file written by my elegant code. I think all the parallel processes try to access the same bunch file and clashes. Is there any other changes I need to make in existing namelists other than in ¶llel_optimization_setup for parallel optimization to work?
Thanks
Neeraj
-Neeraj
-
michael_borland
- Posts: 1959
- Joined: 19 May 2008, 09:33
- Location: Argonne National Laboratory
-
Contact:
Post
by michael_borland » 15 Jun 2018, 08:39
Neeraj,
Can you post your .ele file?
--Michael
-
Neeraj
- Posts: 23
- Joined: 24 May 2018, 10:42
Post
by Neeraj » 15 Jun 2018, 09:29
Please find it attached.
-
Attachments
-
- optimizeQ.ele
- (2.85 KiB) Downloaded 526 times
-Neeraj
-
michael_borland
- Posts: 1959
- Joined: 19 May 2008, 09:33
- Location: Argonne National Laboratory
-
Contact:
Post
by michael_borland » 15 Jun 2018, 09:46
Neeraj,
I'm not sure why that file-lock problem occurs, but you can solve it by removing the "bunch = %s.bun" statement from the &bunched_beam command; you don't really need it. Also, set one_random_bunch=1 to avoid confusing the optimizer.
Since you are doing tracking-based optimization, there's no benefit to using parallel_optimization_setup. You are better off using optimization_setup, so that elegant will track in parallel and optimize at the serial level.
--Michael