svacPipeline v3r4p6

  • Use 'hadd' instead of 'pruneTuple' to merge CalTuple.root and merit.root
    chunk files in recon. Still use 'pruneTuple' for recon.root files.

For an unknown reason merging CalTuple.root files is a factor 25 slower than it used to be. Using 'hadd' is a workaround. From Heather:

"My understanding is that hadd just chains the input files and does a
TChain::Merge to create the output file...looking at the code. .it is a
little more intelligent in that..it specifies to the TChain::Merge an
option to "keep" all the TFiles open so it doesn't waste time deleting them. But even more
importantly..it includes "fast" in the option string.. which means:
//If 'option' contains the word 'fast' the merge will be done without
//unzipping or unstreaming the baskets (i.e. direct copy of the raw byte
on disk)."

Warren verified that hadd also works rapidly for merit.root files.
And he verified that using 'pruneTuple' for recon.root files still is quick.

Code Versions

Engineering Model (sim/recon): v5r070305p4

System Tests for this version

System Tests results:

Verified by Anders et al.

Fred version:

v0r99

Pipeline version:

v1.4

GRITS tag (web browsing and task configuration)

glast-ground v0r3p7
grits-gino-web version 0.55 (v0r5p5)
grits-gino version 0.95 (v0r9p5)
grits-gino-xml version 1.42 (v1r4p2)
grits-common version 0.32 (v0r3p2)

online/svac (task defs, scripts):

pipeline tasks:

online: v2r3p2 (SVAC code moved to AFS)

svac pipeline code and tasks:

code/tasks v3r4p6
pipelineDatasets v0r3

ISOC code and tasks:

v0r5p0

Apps that run in pipeline:

eLog: v2r2p8 (Moved code to AFS)
ConfigTables: v3r2p0 (Added ACD information)
TestReport: v3r4p9 (digi & recon reports)
EngineeringModelRoot: v1r12 (SVAC tuple)

  • No labels