Skip to content

Change creation of FlameBars to be based on parsing of log file.

Louis Henry requested to merge lohenry-readTiming into master

This MR removes the creation of the old style FlameBars and moves to a new one based on readTimingTable.py which parses a log file to return a dictionary of grouped timing results.
There are additional checks implemented to warn if the TimingTable is missing algorithms or if there are unknown additional ones.

A bit more information: this script takes a log file from e.g. https://lhcbpr-hlt.web.cern.ch/lhcbpr-hlt/PerfTests/UpgradeThroughput/Throughput_lhcb-head.2561_Moore_hlt2_reco_baseline_x86_64+avx2+fma-centos7-gcc9-opt_2020-04-08_08:07:47_+0200/tests.log

and transforms it into a map of timing shares such as:

WARNING: Following algorithms are in the timing table but are not associated with any group of algorithms
[this was just for display, now all algorithms are in a given group]

For Hlt2:
TrackFit            	39.02
RICH                	8.04
Downstream          	10.23
HLT1                	0.86
Framework           	4.39
Other               	0.00
Seeding             	4.58
Forward             	5.62
Calorimeter         	25.92
Converters          	0.50
Match               	0.84

For Hlt1:
Muon                	7.17
VeloKalman          	0.92
VeloCluster         	51.31
GEC                 	0.31
Framework           	5.08
Other               	0.00
Forward             	19.00
VeloUT              	16.21

where the numbers are in %

Furthermore, lb-format was applied to all python files.

Edited by Christoph Hasse

Merge request reports