Tcmalloc in Hlt lines
It was found a significant decrease of CPU time (from 4% to 6%) if use tcmalloc’s function instead of standard “new” operator. Result was acquired from running 2 jobs: job for 3 lines (DiMuonHighMass, DiMuonLowMass, AllL0) and for single line (DiMuonHighMass). 10000 events were processed.
3 lines
- tc_new 2.3 times faster then “new” operator.
- Total improvement: 4%. (if consider “CPU time” field’s value as 100%).
Tcmalloc
“new” operator
Hlt1DiMuonHighMass
- tc_new 2.5 times faster then “new” operator.
- Total improvement: 6%.
Tcmalloc
“new” operator
