Project

General

Profile

Support #13624

Slow reading of a MicroBooNE highly filtered event file

Added by Bruce Baller about 3 years ago. Updated almost 3 years ago.

Status:
Closed
Priority:
High
Assignee:
-
Category:
I/O
Target version:
-
Start date:
08/23/2016
Due date:
% Done:

0%

Estimated time:
Scope:
Internal
Experiment:
MicroBooNE
SSI Package:
art
Duration:

Description

I processed 1825 MicroBooNE reconstructed data files to filter 100 neutrino candidate events into a single file, stripAllEvents.root, for re-processing. I re-reconstructed the first event and found that the job took 53 seconds on uboonegpvm06 but the reconstruction module itself only took 0.6 seconds. I ran igprof on the job to produce the attached sqlite file. Much of the time is spent in fhicl::make_ParameterSet. Running config_dumper on this file also takes about 1 minute and produces a text file with 337385 lines. It looks like the configuration for all previous jobs are saved, including the swizzler jobs that attach the beam POT to the event. It isn't clear if there are two problems or one.

I do not need the full configuration history for my purpose, which is to develop reconstruction code. Is there a mechanism to turn off this feature?

I am using a profile build on uboonegpvm06 with uboonecode/larsoft v06_04_00 which uses art v2_03_00.

igprof_lar.gz (1.88 MB) igprof_lar.gz Bruce Baller, 08/23/2016 10:16 AM

History

#1 Updated by Kyle Knoepfel about 3 years ago

  • Tracker changed from Bug to Support
  • Category set to I/O
  • Target version deleted (2.03.00)
  • SSI Package art added
  • SSI Package deleted ()

#2 Updated by Kyle Knoepfel about 3 years ago

  • Status changed from New to Feedback

#3 Updated by Kyle Knoepfel almost 3 years ago

  • Status changed from Feedback to Closed

#4 Updated by Kyle Knoepfel almost 3 years ago

Please re-raise this issue if the solutions proposed by email (dropping metadata) did not adequately address this issue.



Also available in: Atom PDF