Page 1 of 2
run_setup very slow for very long beam line.
Posted: 03 Dec 2014, 20:56
by yuehao
Hi, Michael
I generated a quite long lattice and the run_setup takes 8 minutes to finish. Below are the output of run_setup:
Seeding random number generators
length of beamline START_TO_END per pass: 9.175199952533640e+04 m
statistics: ET: 00:07:57 CP: 476.88 BIO:0 DIO:0 PF:0 MEM:0
Is that normal to take so long time to read the lattice in?
Thanks,
Yue
Re: run_setup very slow for very long beam line.
Posted: 10 Dec 2014, 14:48
by michael_borland
Yue,
Sorry for the delayed reply. I've been away and very busy since returning.
It is not expected to be that slow. I tried a lattice much bigger than that just now and it took only 5 seconds. Can you upload the lattice so I can have a look?
--Michael
Re: run_setup very slow for very long beam line.
Posted: 10 Dec 2014, 14:57
by yuehao
It would be great if you can take a look. Please find the lattice file in attachment.
Also the run_setup is :
&run_setup
lattice = test.lte,
p_central = 12958.9317133,
rootname = 1KPartRAD_MLIGN_1um_CENTER,
centroid = %s.cen,
sigma = %s.sig,
use_beamline = start_to_end,
&end
Thanks a lot.
Yue
Re: run_setup very slow for very long beam line.
Posted: 10 Dec 2014, 15:05
by yuehao
By the way, I am using the macos binary, version 26.0.2.
Re: run_setup very slow for very long beam line.
Posted: 11 Dec 2014, 18:04
by michael_borland
Yue,
I was able to improve this dramatically by using a hash table for some of the lattice parsing. The change will appear in the next version, which I hope to release next week or early in January. If you build from source, I can give it to you sooner.
--Michael
Re: run_setup very slow for very long beam line.
Posted: 11 Dec 2014, 19:06
by yuehao
Thank you so much for the great news. I hope to use it ASAP so if the source is available, please let me know.
Best,
Yue
Re: run_setup very slow for very long beam line.
Posted: 12 Dec 2014, 10:42
by michael_borland
Yue,
Here's the modified source file. It should build if you replace the existing file of the same name. Let me know of any problems.
--Michael
Re: run_setup very slow for very long beam line.
Posted: 12 Dec 2014, 14:40
by yuehao
Hi, Michael
I recompiled elegant and the improvement is drastic. Thanks again.
However I failed in compiling Pelegant, with error on missing libSDDSmpi. Do I need any change when I compile SDDS for mpi version?
Best,
Yue
Re: run_setup very slow for very long beam line.
Posted: 12 Dec 2014, 14:42
by michael_borland
Yue,
It should work if you do the following:
Code: Select all
% cd ~/epics/extensions/src/SDDS/SDDSlib
% make MPI=1
% cd ~/oag/apps/src/elegant
% make Pelegant
I'm assuming you have oag and epics in your home directory. If not, adjust accordingly.
--Michael
Re: run_setup very slow for very long beam line.
Posted: 12 Dec 2014, 16:23
by yuehao
HI, Michael
I compiled the Pelegant without errors. However when I run it, I got errors:
table structure size is inconsistent
Error:
passed size is not equal to expected size for SDDS_DATASET structure
Passed size is 536, library size is 528
Could you advise on where this error comes from?
Thanks,
Yue