Orbit response matrix bug in elegant 20.0

Moderators: michael_borland, soliday

petrenko
Posts: 43
Joined: 09 Jun 2008, 02:53
Contact:

Orbit response matrix bug in elegant 20.0

Post by petrenko » 16 Jan 2009, 01:05

Orbit response matrix fitting does not work in new elegant 20.0:

This is the responseMatrixFit4 example:

> [petrenko@localhost]$ elegant match1.ele
> ...
> Error: Memory is not allocated (matrix_mult)!
Attachments
responseMatrixFit4.tar.gz
(34.25 KiB) Downloaded 578 times

leocmq
Posts: 4
Joined: 16 Jan 2009, 02:53

Re: Orbit response matrix bug in elegant 20.0

Post by leocmq » 16 Jan 2009, 03:01

similar situ
Hear Me . .

michael_borland
Posts: 1927
Joined: 19 May 2008, 09:33
Location: Argonne National Laboratory
Contact:

Re: Orbit response matrix bug in elegant 20.0

Post by michael_borland » 16 Jan 2009, 09:24

I've confirmed this bug and am looking into it. It seems only to affect response matrices during optimization, which unfortunately wasn't covered by the regression tests.

Have you seen it in any other cases?

--Michael

michael_borland
Posts: 1927
Joined: 19 May 2008, 09:33
Location: Argonne National Laboratory
Contact:

Re: Orbit response matrix bug in elegant 20.0

Post by michael_borland » 16 Jan 2009, 10:37

I've determined that this will also affect runs with use_perturbed_matrix=1 in the &correct command and that request repeated response matrix inverse computation via use_perturbed_matrix in the &correct command.

A fix will be available shortly.

--Michael

michael_borland
Posts: 1927
Joined: 19 May 2008, 09:33
Location: Argonne National Laboratory
Contact:

Re: Orbit response matrix bug in elegant 20.0

Post by michael_borland » 16 Jan 2009, 13:55

For those who build from source, the attached tar file contains several updated files that fix this bug.

For others, we'll have the fix posted on our web site in a few days.

--Michael
Attachments
patch.tar
(220 KiB) Downloaded 586 times

soliday
Posts: 390
Joined: 28 May 2008, 09:15

Re: Orbit response matrix bug in elegant 20.0

Post by soliday » 16 Jan 2009, 15:12

The fix has been posted on the software download page. All the releases with the January 16th date contain the fix.

michael_borland
Posts: 1927
Joined: 19 May 2008, 09:33
Location: Argonne National Laboratory
Contact:

Re: Orbit response matrix bug in elegant 20.0

Post by michael_borland » 16 Jan 2009, 15:12

Version 20.0 has been updated on the web to fix this bug.

Thanks to Bob Soliday for lightning-fast turnaround on this.

--Michael

fliller
Posts: 38
Joined: 06 Aug 2008, 14:02

Re: Orbit response matrix bug in elegant 20.0

Post by fliller » 26 Jan 2009, 16:37

Mike,

I downloaded version 20.0 binary on January 26. I see the bug still. In the &correct command, if I set use_perturbed_matrix=1, this fails. It worked with version 19.
Attachments
run1lattice.lte
(4.23 KiB) Downloaded 642 times
run1.param
(218.1 KiB) Downloaded 630 times
run2.ele
(1.59 KiB) Downloaded 621 times

michael_borland
Posts: 1927
Joined: 19 May 2008, 09:33
Location: Argonne National Laboratory
Contact:

Re: Orbit response matrix bug in elegant 20.0

Post by michael_borland » 26 Jan 2009, 19:37

Ray,

I tried your files and they worked without any errors. Can you check that you have the right version installed? If you just run without any arguments, you should get a message starting with "This is elegant 20.0.3". If it says 20.0.2, you are still using the previous release.

--Michael

fliller
Posts: 38
Joined: 06 Aug 2008, 14:02

Re: Orbit response matrix bug in elegant 20.0

Post by fliller » 27 Jan 2009, 06:43

Mike,

I am using 20.0.3. Try this file. The only difference with the previous is that use_perturbed_matrix=1. This breaks it.

Ray
Attachments
run2.ele
(1.59 KiB) Downloaded 619 times

Post Reply