MadGraph5_aMC@NLO

parallel delphes

Asked by Eric

Hi,

Is it possible to let delphes run in parallel when using the MG5+Pythia8+Delphes toolchain at LO?

If I am not msitaken MadGraph parallelizes Pythia8 by splitting the .lhe file and running Pythia8 on the split files. It then merges the resulting hempc files and runs a single Delphes process on the merged file. Is it possible to skip the merging and run Delphes on the split hepmc files in an automated way?

Cheers,
Eric

Question information

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

This is certainly not implemented for the moment.
Would it be possible? Not even sure. The issue is that splitting py8 leads to a lot of issue with the normalisation that we correct when merging the various lhe file.

Cheers,

Olivier

> On 10 Jul 2019, at 15:17, Eric <email address hidden> wrote:
>
> New question #681903 on MadGraph5_aMC@NLO:
> https://answers.launchpad.net/mg5amcnlo/+question/681903
>
> Hi,
>
> Is it possible to let delphes run in parallel when using the MG5+Pythia8+Delphes toolchain at LO?
>
> If I am not msitaken MadGraph parallelizes Pythia8 by splitting the .lhe file and running Pythia8 on the split files. It then merges the resulting hempc files and runs a single Delphes process on the merged file. Is it possible to skip the merging and run Delphes on the split hepmc files in an automated way?
>
> Cheers,
> Eric
>
> --
> You received this question notification because you are an answer
> contact for MadGraph5_aMC@NLO.

Revision history for this message
Eric (e-r-i-c) said : 		#2

So using your lhe_parser to split the .lhe file in N equal sized files, running Pythia8 and Delphes manually on each of these files and then modifying the weights in the .root files by 1/N when analyzing them would not work?

What about running the parton-level event generation N times (or using multi_run with N runs) and then running Pythia and Delphes manually on the N .lhe files?

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

Hi,

I do not know the details of all the issues for the PY8 parallelization (I'm not the one implementing it). What I remember is that some variable in the Hepmc where cummulative and therefore depends of all previous events. So you need to be carefull when re-combining those in a single hepmc.

Now I do not know if/how Delphes is using those information or not. (and in any case it would be dangerous to assume that they are not using those)

Now you can certainly run independent run (like with multi-run) but we do not have nice way to run those in parralel for pythia/Delphes (at the end of the day it should be their job to parralelize and not our job)

Cheers,

Olivier

Revision history for this message
Eric (e-r-i-c) said : 		#4

Thanks Olivier Mattelaer, that solved my question.