Notes from the April 11, 2017 LArSoft Coordination Meeting
(Notes by Erica Snider)
1) Release and project report (Erica Snider)
-
LArSoft is using art v2.06.03 as of Apr 3 release of LArSoft v06_30_00
-
No merge requests so far this week
-
gcc 6 migration is current code management priority
-
-
Release managers:
-
-
Non-production releases prior to v06.02.00 are being removed
-
-
Suggest code analysis targets
-
Next meeting will be in WH8X
2) HEP Analysis Ecosystem Workshop (Phillipe Canal)
-
Workshop in Amsterdam on May 22–24
-
Part of HEP Software Foundation white paper effort
-
Aimed at analysis software across the field. Can we answer the question:
-
-
What do we want analyses / tools to look like in the future?
-
For instance, where is root going?
-
Consider tools from other sciences and industry.
-
-
Seek input from physicists, computing professionals, industry reps
-
Agenda still open, though session structure is fixed.
-
Q: Can industry tools handle PBs of data?
-
-
A: Yes, in some cases.
-
Would like to survey what is out there.
-
Please register or provide input
3) Update on the integration of the SBND light simulation into LArSoft (Diego Garcia-Gamez)
-
Taking into account both reflected and direct UV/visible light, as well as arrival time distributions for each (previously unmodeled)
-
-
Direct / VUV universal for all experiments
-
Reflected / visible component depends on particular geometry
-
Arrival times include both direct transport and Rayleigh scattering
-
Changes in larsim, lardataobj, larana, lardata
-
-
Branch feature/dgg_lightprop
-
Summary discussion of fcl parameters used by in the models, and some of the other configuration parameters and options
-
Q: how generalizable is the parameterization of the reflected light visibility and propagation time modeling to detectors other than SBND?
-
-
A: The parameterization works for MicroBooNE.
-
-
[ At this point, a long discussion ensued, and it was unclear to the note-taker whether there was actually a definitive answer to the question. ]
-
[ On the question of the behavior of D_max, for instance, it sounded as though photons generated beyond that distance would be ignored, but there was some confusion (mine…) as to whether this was simulation behavior or analysis ntuple behavior. The correct answer should be confirmed explicitly. ]
-
[ Some follow-up with Diego and Andrzej on the above questions is advised. ]
-
First tests look ok, in that photons are generated. Still some validation to be done. Author raised a question about scintillation yield always being 600. Alex Himmel provided an answer that may require some follow-up to check.
-
Testing by the masses
-
-
Test against v06_30_00, using branch feature/dgg_lightprop.
-
Will need to check out larana, lardata, lardataobj, larsim and sbndcode.
-
In sbdcode/JobConfiguration, use prodsingle_opfast_proptime.fcl and sbnd_buildopticallibrary_withrefl.fcl
-
Request to merge.
-
-
Should not change behavior of existing jobs
-
Sufficient to check that all CI tests run
-
Will ask Vito to run the CI tests on the relevant branches. If that all succeeds, then the merge is approved.
4) Particle bomb generator (Andy Furmanski)
Not available, so did not present.
5) Why should we move to geant4.10.3.p01? (Hans Wenzel)
-
New version has lots of changes that we had requested
-
Includes
-
-
Separately selectable production processes
-
Separately selectable transport processes
-
Makes it possible to consolidate all of these into one module, configurable via simple fcl switches, rather than requiring everything to be coded using various “action” classes.
-
Also allows access to the number of photons generated without placing all the photons on the stack
-
-
This makes it possible to perform non-tracking (ie, parameterized) photon generation and propagation within a tractable memory footprint
-
Convenient way to add / configure step-limiter
-
Convenient way to add / configure optical physics
-
-
Also allows use of splines to smooth optical properties, rather than histograms, which can lead to unphysical artifacts in simulations
-
A factory that makes it possible to avoid explicit coding for all this
-
Profiling of the new code will be presented in a future meeting by Soon
-
Can (should?) add optical properties to GDML. This avoids having to use “magic words” in order to determine what volumes are optically sensitive
-
All requested features have been implemented, so LArSoft should migrate asap
-
The integration plan: LArSoft will create a release candidate with the new Geant4 after the gcc 6 work is completed. Then per the usual procedure:
-
-
Experiments will then be asked to sign off on the change
-
When all experiments have signed off, LArSoft will migrate
There are minutes attached to this event.
Show them.