MG5 stopped when trying to calculating the PDF uncertainty

Asked by Junquan Tao

Dear MG5 authors and experts,

I was trying to run MG5 to generate some events with Georgi-Machacek Model . I was running with MG5_aMC_v2_7_2, with gcc 4.9.3 and lhapdf 6.1.5 (/cvmfs/cms.cern.ch/slc7_amd64_gcc493/external/lhapdf/6.1.5-ikhhed/bin/lhapdf-config) at CERN lxplus7 machine.

When I ran the jobs, I used lhapdf "PDF4LHC15_nlo_30_pdfas" (90400). If I tried the default option with "False = reweight_pdf", the MG5 can be ran successfully with the cross section and also the scale uncertainties in the output file summary.txt, together with the reweight.log and the lhe events events.lhe.gz:

  --------------------------------------------------------------
      Summary:
      Process p p > H5p H5z [QCD]
      Run at p-p collider (6500.0 + 6500.0 GeV)
      Number of events generated: 10000
      Total cross section: 7.520e-02 +- 1.8e-04 pb
   --------------------------------------------------------------
      Scale variation (computed from LHE events):
          Dynamical_scale_choice -1 (envelope of 9 values):
              7.492e-02 pb +1.6% -1.3%
   --------------------------------------------------------------

I wanted to also get the pdf uncertainty, so I set "True = reweight_pdf" . But with this setup, MG5 can not be ran smoothly. It stopped and exited without any warning or error, when "Doing reweight":
--------
INFO: Generating events
INFO: Idle: 0, Running: 3, Completed: 3 [ current time: 08h13 ]
INFO: Idle: 0, Running: 2, Completed: 4 [ 4.6s ]
INFO: Idle: 0, Running: 0, Completed: 6 [ 4.9s ]
INFO: Doing reweight
INFO: Idle: 0, Running: 5, Completed: 1 [ current time: 08h14 ]
INFO: Idle: 0, Running: 4, Completed: 2 [ 0.12s ]
INFO: Idle: 0, Running: 2, Completed: 4 [ 0.14s ]
INFO: Idle: 0, Running: 1, Completed: 5 [ 5.1s ]
INFO: Idle: 0, Running: 0, Completed: 6 [ 5.3s ]
[jtao@lxplus750 MG5_aMC_v2_7_2]$
----------

In the output directory, I got only the following files:

alllogs_0.html alllogs_1.html alllogs_2.html run_02_tag_1_banner.txt

I also tried with MG5_aMC_v2_6_7 but got the same problem.

Thanks for your help and all my best,

Junquan Tao

Question information

Language:
English Edit question
Status:
Answered
For:
MadGraph5_aMC@NLO Edit question
Assignee:
No assignee Edit question
Last query:
Last reply:
Revision history for this message
Olivier Mattelaer (olivier-mattelaer) said :
#1

Likely a lhpadf issue (maybe related to the old version of gcc that you are using)

Cheers,

Olivier

> On 11 May 2020, at 11:43, Junquan Tao <email address hidden> wrote:
>
> New question #690623 on MadGraph5_aMC@NLO:
> https://answers.launchpad.net/mg5amcnlo/+question/690623
>
> Dear MG5 authors and experts,
>
> I was trying to run MG5 to generate some events with Georgi-Machacek Model . I was running with MG5_aMC_v2_7_2, with gcc 4.9.3 and lhapdf 6.1.5 (/cvmfs/cms.cern.ch/slc7_amd64_gcc493/external/lhapdf/6.1.5-ikhhed/bin/lhapdf-config) at CERN lxplus7 machine.
>
> When I ran the jobs, I used lhapdf "PDF4LHC15_nlo_30_pdfas" (90400). If I tried the default option with "False = reweight_pdf", the MG5 can be ran successfully with the cross section and also the scale uncertainties in the output file summary.txt, together with the reweight.log and the lhe events events.lhe.gz:
>
> --------------------------------------------------------------
> Summary:
> Process p p > H5p H5z [QCD]
> Run at p-p collider (6500.0 + 6500.0 GeV)
> Number of events generated: 10000
> Total cross section: 7.520e-02 +- 1.8e-04 pb
> --------------------------------------------------------------
> Scale variation (computed from LHE events):
> Dynamical_scale_choice -1 (envelope of 9 values):
> 7.492e-02 pb +1.6% -1.3%
> --------------------------------------------------------------
>
>
> I wanted to also get the pdf uncertainty, so I set "True = reweight_pdf" . But with this setup, MG5 can not be ran smoothly. It stopped and exited without any warning or error, when "Doing reweight":
> --------
> INFO: Generating events
> INFO: Idle: 0, Running: 3, Completed: 3 [ current time: 08h13 ]
> INFO: Idle: 0, Running: 2, Completed: 4 [ 4.6s ]
> INFO: Idle: 0, Running: 0, Completed: 6 [ 4.9s ]
> INFO: Doing reweight
> INFO: Idle: 0, Running: 5, Completed: 1 [ current time: 08h14 ]
> INFO: Idle: 0, Running: 4, Completed: 2 [ 0.12s ]
> INFO: Idle: 0, Running: 2, Completed: 4 [ 0.14s ]
> INFO: Idle: 0, Running: 1, Completed: 5 [ 5.1s ]
> INFO: Idle: 0, Running: 0, Completed: 6 [ 5.3s ]
> [jtao@lxplus750 MG5_aMC_v2_7_2]$
> ----------
>
> In the output directory, I got only the following files:
>
> alllogs_0.html alllogs_1.html alllogs_2.html run_02_tag_1_banner.txt
>
>
> I also tried with MG5_aMC_v2_6_7 but got the same problem.
>
> Thanks for your help and all my best,
>
> Junquan Tao
>
>
>
>
>
> --
> You received this question notification because you are an answer
> contact for MadGraph5_aMC@NLO.

Revision history for this message
Junquan Tao (junquan) said :
#2

Dear Olivier,

Thanks a lot for your quick response!

So in your opinion, which lhapdf version and gcc version should I use in order to get the pdf uncertainty of this pdf "PDF4LHC15_nlo_30_pdfas" correctly within MG5_aMC_v2_7_2 or MG5_aMC_v2_6_7?

Thanks and all my best,

Junquan Tao

Revision history for this message
Olivier Mattelaer (olivier-mattelaer) said :
#3

You can try lhapdf6.2
with gcc8 or gcc 9

Cheers,

Olivier

Revision history for this message
Junquan Tao (junquan) said :
#4

Dear Olivier,

Thanks for the suggestions!

Then I tried with gcc8 (8.2.0). I have problmes to install 'ninja' and to compile 'IREGI' :

==================Install of NINJA===========
....
/tmp/jtao/ccY4eZpV.s:166: Error: unknown pseudo-op: `.lvu32'
make[2]: *** [avholo_wrapper.lo] Error 1
make[2]: Leaving directory `/afs/cern.ch/work/j/jtao/MG5_aMC_v2_7_2/HEPTools/ninja/Ninja/src'
make[1]: *** [install-recursive] Error 1
make[1]: Leaving directory `/afs/cern.ch/work/j/jtao/MG5_aMC_v2_7_2/HEPTools/ninja/Ninja/src'
make: *** [install-recursive] Error 1
Finished installing Ninja

Command "install ninja" interrupted with error:
InvalidCmd : Installation of ninja failed.

=================Compiling IREGI=============
            make[3]: Leaving directory `/afs/cern.ch/work/j/jtao/MG5_aMC_v2_7_2/vendor/IREGI/src/qcdloop/ff'
            make[2]: Leaving directory `/afs/cern.ch/work/j/jtao/MG5_aMC_v2_7_2/vendor/IREGI/src'
            /cvmfs/sft.cern.ch/lcg/releases/gcc/8.2.0/x86_64-centos7/bin/gfortran -c global.f90 -O -g -fPIC -I/afs/cern.ch/user/j/jtao/jtaoWorkSpace/MG5_aMC_v2_7_2/vendor/IREGI/src/oneloop
            /tmp/jtao/cceWbQtS.s: Assembler messages:
            ....
            /tmp/jtao/cceWbQtS.s:1127: Error: unknown .loc sub-directive `view'
            /tmp/jtao/cceWbQtS.s:1127: Error: unknown pseudo-op: `.lvu198'
            make[1]: *** [global.o] Error 1
            make[1]: Leaving directory `/afs/cern.ch/work/j/jtao/MG5_aMC_v2_7_2/vendor/IREGI/src'
            make: *** [all] Error 2

        Please try to fix this compilations issue and retry.
============================
While "collier" and "CutTools" can be well installed or compiled.

In my previous tests with gcc 4.9.3, all these packages can be well installed or compiled. I believe these errors are related to the gcc version. Am I right?

Thanks and all my best,

Junquan

Revision history for this message
Olivier Mattelaer (olivier-mattelaer) said :
#5

Hi,

never seen/ heard about such type of issues.

When I search on google for "Error: unknown pseudo-op: `.lvu32'"
and the only thing, I found for gcc8 is the following link:
https://bbs.archlinux.org/viewtopic.php?id=245354

Not sure if this can help you but this is the only advice I can give.

Cheers,

Olivier

Can you help with this problem?

Provide an answer of your own, or ask Junquan Tao for more information if necessary.

To post a message you must log in.