MadEvent crashes for certain combinations of masses in param_card

[Daniel Schwartländer]

Dear MadGraph team,

I am always encountering a crash when I run MadEvent with certain combinations of input masses, while other combinations work. I use version MadGraph 5.2.3.0 with a UFO-file created by Benjamin Fuks. Up to now the crashes occure mostly at relativ low masses but there are also exceptions. All crashes occure because MadEvent tries to access a file that does not exist 'SubProcesses/P0_qq_wrp_wrp_lne_ne_ltbx/G1.14a0/results.dat'.

The bug report file is attachted.

Cheers,

Daniel Schwartländer

[Olivier Mattelaer (olivier-mattelaer)]

Hi Daniel,

Can you attach your model?
Such that I can run it on my laptop and try to reproduce the error that you face.
and investigate for which reason that file do not exists.

Cheers,

Olivier

[Daniel Schwartländer (d-schw20-deactivatedaccount)]

Hi Olivier,

the UFO-file and the restriction-file are attached.
Some of the value combinations where madevent crashes are:
mwr = 1800 mne = 400
mwr = 1800 mne = 800
mwr = 2200 mne = 1200
mwr = 3000 mne = 400
The remaining parameters are not changed.

Cheers,

Daniel

[Olivier Mattelaer (olivier-mattelaer)]

Dear Daniel,

I have run that model for that benchmark point but in all cases, the code goes trough (I have run multiple times those benchmark to try to find some border effect but do not found any.).
Could you retry on your side? and if still occurs could you try to describe me exactly how you run the code/what you did before such that I can search for the border effect which creates such problem.

Thanks,

Olivier

[Daniel Schwartländer (d-schw20-deactivatedaccount)]

Dear Olivier,

this crashes still occure but I am not able to find a way to reproduce one.
Here a description what I am doing:
First I create madevent with MadGraph and the restriction-file:

import model LRMSSM-LRSM1genel
define wr = wr+ wr-
define nl = ne
define l = e- e+
define q = u c t d s b
define qbar = u~ c~ t~ d~ s~ b~
generate p p > wr, ( wr > l nl, nl > l q qbar )
output /Path/madevent
exit

Then I change the param_card (only the values of mne and mwr). After this I
open madevent, compute the two decay widths and launch it.

compute_widths 34 6000012
launch /Path/Events
pythia=ON
exit

When I have the result, I change the parameters (or leave them like they are).
Then I compute the two decay widths again and launch it. In most cases the
code just goes through but sometimes it crashes.

I have run the parameter combination mwr = 1800 mne = 400 multiple times. It
seems that Madevent can crash in the first run, the second run or not at all.
If I create a new madevent and then do the same, the crash is not
reproduceable.

If I use the madevent for a list of value combinations one after another, it
crashes for some value combinations but only if the value combinations are
launched in some order.
One example that I tried:
    mwr mne
    1000 400 no crash
    1000 800 "
    1400 400 "
    1400 800 "
    1400 1000 no crash
    1800 400 crash
    1800 800 crash
    1800 1200 no crash
If I create a new madevent and launch them in a different order, there is no
crash at all:
   1400 400 no crash
   1400 800 no crash
   1800 400 "
   1800 800 "
   2200 400 "
   2200 800 "

I will continue to look for some logic behind this, but I am not sure where I
should search.
My pythia and run card are attached.

Cheers,

Daniel

[Olivier Mattelaer (olivier-mattelaer)]

Hi,

Could you try to apply the following patch?

=== modified file 'madgraph/madevent/gen_ximprove.py'
--- madgraph/madevent/gen_ximprove.py 2015-06-28 14:44:22 +0000
+++ madgraph/madevent/gen_ximprove.py 2015-07-30 13:55:34 +0000
@@ -903,7 +903,7 @@

     def reset_multijob(self):

- for path in glob.glob(pjoin(self.me_dir, 'Subprocesses', '*',
+ for path in glob.glob(pjoin(self.me_dir, 'SubProcesses', '*',
                                                            '*','multijob.dat')):
             open(path,'w').write('0\n')

I think that it should solve your problem

Cheers,

Olivier

[Daniel Schwartländer (d-schw20-deactivatedaccount)]

Hi Olivier,

everything seems to work. Thanks a lot!

Cheers,

Daniel