Changes to MVA BtoXLL lines
Description
This MR is meant to:
- Add e\mu lines intended to be used to study the combinatorial background
- Remove the prescale from all the misID lines and instead invert the PIDe requirement on the electrons. Currently the PIDe cut is removed, the PIDe cut will be in [-1,1.5] interval, physics lines start at 1. We allow for an overlap of 0.5 as suggested by @rquaglia
- Tighten requirements on tracks (electrons and muons) as well as candidates (FDChi2, Vertex Chi2, etc) to reduce bandwidth and rates
- Extend mass range to 3.5 GeV for candidates decaying into electron final states.
Rates
All RD lines
The rates for the entire RD group's lines are attached here
For the lines in the module under consideration
| rate (before) | rate (after) | difference | difference (%) | |
|---|---|---|---|---|
| Hlt2RD_BuToKpEE_MVA_misid | 31.7516 | 156.49 | 124.738 | 392.857 |
| Hlt2RD_BuToKpEE_MVA_cal | 77.111 | 102.059 | 24.9477 | 32.3529 |
| Hlt2RD_B0ToKpPimEE_MVA_misid | 18.1438 | 58.9672 | 40.8234 | 225 |
| Hlt2RD_LbTopKEE_MVA_misid | 38.5555 | 52.1633 | 13.6078 | 35.2941 |
| Hlt2RD_BuToKpEE_SameSign_MVA | 142.882 | 45.3594 | -97.5227 | -68.254 |
| Hlt2RD_BsToPhiEE_MVA_misid | 43.0914 | 38.5555 | -4.53594 | -10.5263 |
| Hlt2RD_BuToKpEE_MVA | 104.327 | 36.2875 | -68.0391 | -65.2174 |
| Hlt2RD_B0ToKpPimEE_MVA_cal | 36.2875 | 29.4836 | -6.80391 | -18.75 |
| Hlt2RD_BuToKpMuE_MVA | nan | 22.6797 | nan | nan |
| Hlt2RD_B0ToKpPimEE_MVA | 70.307 | 18.1438 | -52.1633 | -74.1935 |
| Hlt2RD_BsToPhiEE_MVA | 133.81 | 13.6078 | -120.202 | -89.8305 |
| Hlt2RD_LbTopKEE_MVA | 15.8758 | 11.3398 | -4.53594 | -28.5714 |
| Hlt2RD_LbTopKMuE_MVA | nan | 11.3398 | nan | nan |
| Hlt2RD_LbTopKEE_SameSign_MVA | 18.1438 | 9.07188 | -9.07188 | -50 |
| Hlt2RD_B0ToKpPimEE_SameSign_MVA | 45.3594 | 6.80391 | -38.5555 | -85 |
| Hlt2RD_BuToKpMuMu_SameSign_MVA | 24.9477 | 6.80391 | -18.1438 | -72.7273 |
| Hlt2RD_B0ToKpPimMuE_MVA | nan | 4.53594 | nan | nan |
| Hlt2RD_LbTopKMuMu_SameSign_MVA | 11.3398 | 4.53594 | -6.80391 | -60 |
| Hlt2RD_LbTopKMuMu_MVA | 4.53594 | 4.53594 | 0 | 0 |
| Hlt2RD_BuToKpMuMu_MVA | 4.53594 | 4.53594 | 0 | 0 |
| Hlt2RD_B0ToKpPimMuMu_SameSign_MVA | 6.80391 | 2.26797 | -4.53594 | -66.6667 |
| Hlt2RD_B0ToKpPimMuMu_MVA | 4.53594 | 2.26797 | -2.26797 | -50 |
| Hlt2RD_BsToPhiEE_SameSign_MVA | 6.80391 | 2.26797 | -4.53594 | -66.6667 |
| Hlt2RD_BsToPhiMuMu_SameSign_MVA | 24.9477 | 0 | -24.9477 | -100 |
| Hlt2RD_BsToPhiMuMu_MVA | 11.3398 | 0 | -11.3398 | -100 |
| Hlt2RD_BsToPhiMuE_MVA | nan | 0 | nan | nan |
Bandwidth
The test was ran over 568000 entries and the file sizes were:
Before: 650 Mb => 256 Mb
After : 613 Mb => 243 Mb
with the arrow showing the compression. The compression was done with zstd -3. The input rate is 1.3 MHz. The bandwith difference should be -29 Mb, for reference, we added four lines.
Throughput
HLTControlFlowMgr INFO ---> Loop over 568655 Events Finished - WSS 12332.3, timed 468596 Events: 1273749 ms, Evts/s = 367.887
HLTControlFlowMgr INFO ---> Loop over 568655 Events Finished - WSS 10901.9, timed 468596 Events: 1238956 ms, Evts/s = 378.218we seem to be able to process now more events/s than before. This is likely due to the tighter cuts that reduce the number of events that have to go through the BDT.
Edited by Angel Campoverde