Rate estimation tools for L1 Menu operations in CMS
See the official twiki page for updated instructions!
Primary tool for performing rate estimation. The rest of the tools in this directory are used for setting up the components required.
Requires:
- C++ conversion of a menu
- Ntuple file list
- Lumi section information table
- Prescale table See docs/testMenu2016.md for more information.
Parallel version of testMenu2016.
It allocates several CPU cores and starts the default rate estimation with with a subset of data.
In the end the outputs of the different jobs are combined and a joined output is presented.
Important this is a preliminary version that does only support Prescales of 0 or 1. Make sure that the selected prescale column only contains those values otherwise it will fail. This code is only to get a fast but rough estimate of the L1 rate, because prescaled seeds need to be excluded. For further details on the parameters, see below.
Converts the XML menu to C++ to be used in testMenu2016 for rate estimation. Do not run cmsenv before running this.
Creates a file list for a given ntuple.
Creates a lumi section information table.
After setting up the environment as discussed above, prepare the set of components required to run testMenu2016 for rate estimation.
Configure rate-estimation using a L1 Menu XML file using configure.sh.
./configure.sh <your_menu.xml>This will produce the files menulib.hh and menulib.cc.
The latest L1 Menu XML can be found here.
Once these files have been produced, it is safe to run cmsenv.
A set of ntuples is required to run the rate estimation code.
To test the code, a list containing Run 3 MC ntuples is provided in
ntuple/Run3_NuGun_MC_ntuples.list.
You can run the code running on a subset of events using the option --maxEvent.
The script ntuple/makeFileList.py can be used to create a file list for other
ntuples:
bash
cd ntuple
./makeFileList.py <your_ntuple_location>
cd ..
In case you are using data ntuples, this table is used to avoid bad lumi sections (LS).
It also includes LS vs PU information, which can be useful for plotting rate vs PU dependencies.
An example csv is included in menu/run_lumi_RunA.csv, including fill at high pileup.
To produce your own, modify the beginning of GetLumi.py to list the run number
of your ntuple, then:
bash
cd menu
source GetLumi_setup.sh
./GetLumi.py
cd ..
While the official PS tables are in .xlsx format for better presentation, ensure
that the prescale table you use has been converted to
either a tab-separated .txt file or comma-separated .csv file.
An example table has been provided for column 2p0E34 in
menu/L1Prescales2023.csv.
The latest 2018 official prescale tables can be found
here.
Ensure you are in a CMSSW_130X environment.
If the directories objs/ and objs/include/ do not exist, create them.
Then run make -j 4.
Components have been provided to run a short test.
./testMenu2016 \
-m menu/L1Prescales2023.csv -l ntuple/Run3_NuGun_MC_ntuples.list \
-o test -b 2748 --doPlotRate --doPlotEff --SelectCol 2p0E34 \
--doPrintPU --maxEvent 200000This will take only a few minutes and output test.csv, test.root, test.txt, and test_PU.txt into the results/ directory.
The possibility to apply a reweighting procedure has been included in order to take into account a realistic pileup model according to Run 3 Lumi POG predictions.
The recommended approach at the moment is to use the option --doReweightingRun3 to get results for the incoming Trigger Review.
The arguments that you can use are as follows:
| Option | Description |
|---|---|
-u |
LS information table. |
-m |
PS table. |
-l |
ntuple list. |
-o |
Output name. Output files will be saved to results/[output name].\* |
-b |
Number of bunches. |
--lowerPUbound |
Consider only events with PU >= lowerPUbound for the rate estimation |
--upperPUbound |
Consider only events with PU <= upperPUbound for the rate estimation |
--doReweightingRun3 |
Apply pileup reweighting of the Run 3 flat pileup distribution according to Run 3Lumi POG model (do not use with --lowerPUbound or --upperPUbound option!) |
--doReweighting2018 |
Apply pileup reweighting of the Run 3 flat pileup distribution to match the Run 2 pileup profile (do not use with --lowerPUbound or --upperPUbound option!) |
--customReweighting |
Apply pileup reweighting using a custom set of weights, set a json-file containing the custom weights |
--SelectRun |
Select a run number; default is full list. |
--SelectLS [startLS,endLS] |
Select lumi sections to run over; default is whole LS provided. |
--SelectCol |
Select the column in the PS table. |
--UseUnpackTree |
Switch to using UnpackTree; default is EmuTree. |
--maxEvent |
Run on a subset of the available events. |
--doPrintPU |
Evaluate and save the rates for different pileup values |
--PrescalePrecision |
Select the prescale precision factor e.g. for 100 (=10^2) the prescales are represented with 2 decimal places |
Running ./testMenu2016 --help will show all arguments with a brief description. Also see docs/testMenu2016.md.
Most of the arguments of the rate estimation code (above) can be used exactly as in the single threaded version, because they are just passed to the parallel version. The following arguments are specific for the parallel version:
| Option | Description |
|---|---|
--njobs |
number of parallel jobs to run |
--just_analyze |
just analyze logs without processing ntuples |
--name |
unique name of the test, replaces the option -o of the original code |
--tmp |
path to an existing direcrtory where to store temporary files |
Please remember that when you use the --SelectCol option, you just select a specific PS column from the table which contains the information about which seeds are prescaled or unprescaled and about the prescale values. It does not allow to select a certain luminosity. In order to look at rates using a specific luminosity scenario, you would need to use lumi-range-skimmed datasets or to select a specific pileup window when running the rates.
(use --lowerPUbound and --upperPUbound to select the pileup range of interest).