Crashing when running pelegant

[Björklund]

Hi,

So, I'm running tracking simulations of the MAX IV linac and I have so far run on only 1 core because it didn't work in parallel. Now I decided to try and investigate why it didn't work and so I checked all the elements in the lattice and they should all be parallel compatible.

What happens is that the program crashes during the initialization phase, where it reads in values from run_control, twiss_output, etc. The read-in of these parameters work well in general, but it always crashes when reading in magnetic multipole files. On my latest try, it gave the exit code 0xc0000374.

The lattice works fine on 1 core and I have run other, more simple lattices in parallel before, so I'm feeling a bit stuck. I don't think I can provide the full lattice file, and it is anyway very long (>2000 lines in the version I'm using), so it might not be that useful.

Best regards
Jonas

[michael_borland]

Jonas,

Can you verify that you are using the latest release?

Also, if you can provide the lattice file and command file, that would really help, since I've not seen a problem like this. You can send it to me via email if you can't post it. My email is my last name @anl.gov.

--Michael

[Björklund]

Hi Michael,

I'd missed that a new version was out, I'll update to the latest release and try again.

I'll also ask my supervisor if I can send the files to you, and if that's ok, I'll email it directly to you.

Best Regards
Jonas

[Björklund]

Hi again,

This is both a bit awkward and relieving at the same time, but it seems like the software update solved the problem! I remember having the same problem when I started doing tracking, which was with version 26 I think, and I didn't see anything that looked obviously related in the bug fixes for the last versions (was running 31 before), but something fixed the problem.

As always, thanks for the quick response!

Best regards
Jonas

[michael_borland]

Jonas,

Glad to hear it is working.

--Michael

[Björklund]

Hi,

So, a small update on the matter, which is maybe a little off-topic for the thread.

Like I said before, it doesn't crash anymore, which is great. I got another problem instead, though; the electron beam looked like it had been fragmented and the fragments shifted in longitudinal phase space, both in time and energy. This *appears* to have been solved when I turned off the flag center_arrival_time in sdds_beam in the .ele file. Thought I should report it anyway.

Best regards
Jonas

[michael_borland]

Jonas,

That sounds like a bug. Probably the working nodes are not sharing their time information correctly, so each node gets its own value for the central arrival time.

We'll fix that in the next release. Thanks for reporting it.

Meanwhile, if it is an issue, you can modify your input beam as follows:

Code: Select all

sddsprocess beam1.sdds beam2.sdds -process=t,ave,t0 -redefine=column,t,"t t0 -",units=s

This will make beam2.sdds with <t> = 0.

--Michael

[michael_borland]

Jonas,

Just to confirm, this was a bug. The fix will be available in the next release.

--Michael