dune / nd_production Goto Github PK

Integrate the TMS reconstruction production step into the "2x2" style production workflow at NERSC.

• Installation (including compilation) of software on perlmutter.
• v0 grid script for perlmutter CPUs.
• Interactive in/out test with edep-sim file made by TMS group.
• Interactive in/out test with edep-sim file made at NERSC.
• fireworks infrastructure (fireworks4dune repository).

I don't think writing directly works at this point.

Infrastructure to support:

• Generation of jsons that can be used by SAM for each file made at NERSC.
• Transfer of files from NERSC to ND dropbox at FNAL.

Right now edep sim will produce a nearly empty root file if genie crashes. And then there are two or three errors in the log file instead of just the genie error, which is confusing for new users who are most likely to crash genie. Can be stopped by only running edep if there's an input genie file and so on.

Ideally we'd also stop the crashes but that's a stretch goal. We'd need to make sure genie has all the correct inputs during the python script.

Reasons detailed here https://dunescience.slack.com/archives/C62ALUT7C/p1692270471363809

the original line is:
print >> sh, "sed \"s/count\/set fixed/cout\/set mean/g\" ${ND_PRODUCTION_CONFIG}/dune-nd.mac > dune-nd.mac" % mac
I suspect, the line is intended to be:
print >> sh, "sed \"s/count\/set fixed/count\/set mean/g\" ${ND_PRODUCTION_CONFIG}/dune-nd.mac > dune-nd.mac" % mac

with count instead of cout

Integrate SAND detsim and reco (i.e. integrate both fastreco and sand-reco)

This issue will be used to record, at a high level, must-do items before running the next MiniProd. It should be added to as and when new items come up.

• Corrected TMS geometry. Problem described here.
• Make sure topVol for fiducial events is correct with the new geometry.
• Make sure TPCActive_shape (larnd-sim) is correct.
• Verify rockBox method is working as intended.
• larnd-sim: in detector properties yaml, ensure that tpc_offsets are as in line with gdml as possible.

Following Robert Hatcher's and Herilala Razafinime's suggestions.
Someone should use "dk2nu" flux files in order to create off-axis samples, as the "gsimple" flux files already have the beam-to-detector coordinate "baked in," making the modified "GNuMIFlux.xml" file not seen.

Solution:

Using the command (for the creation of a 30m off-axis sample):
python ProcessND.py --pot 1e14 --outdir /pnfs/dune/scratch/users/your/path --topvol volDetEnclosure --oa 30 --use_dk2nu

and then manually include the line setup dk2nugenie v01_10_01c -q e20:prof in the processnd.sh