Removed stub EGM regressor producers from production #24769
Conversation
|
The code-checks are being triggered in jenkins. |
|
+code-checks Logs: https://cmssdt.cern.ch/SDT/code-checks/cms-sw-PR-24769/6713 |
|
A new Pull Request was created by @guitargeek (Jonas Rembser) for master. It involves the following packages: PhysicsTools/NanoAOD @perrotta, @monttj, @cmsbuild, @slava77, @gpetruc, @arizzi can you please review it and eventually sign? Thanks. cms-bot commands are listed here |
|
@cmsbuild please test |
|
The tests are being triggered in jenkins. |
|
@Sam-Harper @jainshilpi |
guitargeek
changed the title
No not off the top of my head exactly. My guess for this is when there were showershape variables not in the electron/photon that wanted to be used in the regression. Probably not been used for a while now as everything is in the electron/photon. If we need it back, we can always add it back but I would like to avoid doing that as much as possible. Its probably better to clean out dead code rather than having it hanging around, thanks @guitargeek |
|
Comparison job queued. |
|
Comparison is ready Comparison Summary:
|
|
Hi @slava77, I'm just curious to know how you could tell it doesn't run in the workflows. From my naive understanding, if it's in a sequence that is used like here in NanoAOD [1], then it runs each event. But apparently that doesn't seem to be the case. Can you please point me to an explanation? Thanks! [1] https://github.com/cms-sw/cmssw/pull/24769/files#diff-6d69831bd3675c40ea80698b6ba12887R17 |
|
Hi @guitargeek , which exact tool Slava uses to determine this I dont know. However when CMSSW is run in unschedualled mode, it only runs modules whose output is used. So if nothing is consuming the module and no output module is writing it out, the module is simply not run. Our framework is actually pretty smart :) |
|
Thank you Sam, that's what I wanted to know! |
|
On 10/2/18 9:36 AM, Jonas Rembser wrote:
Hi @slava77
I'm just curious to know how you could tell it doesn't run in the
workflows. From my naive understanding, if it's in a sequence that is
used like here in NanoAOD [1], then it runs each event. But apparently
that doesn't seem to be the case. Can you please point me to an
explanation? Thanks!
I'm often running workflows with --customise
Validation/Performance/TimeMemoryInfo.py
to monitor timing and memory technical performance.
It's then a matter of grepping for the module name in the log file
|
|
Thank you Slava, that is a very handy tool. |
|
@slava77 @guitargeek indeed these modules do not spend time in wf 136.85 TimeReport 0.000000 0.000000 0.000000 electronRegressionValueMapProducer But also using the Tracer shows that they are initialized but never used in theevent processing: ++++ finished: begin stream for module: stream = 3 label = 'photonRegressionValueMapProducer' id = 2248 |
|
+1 cleaning of sequences |
|
merge |
Hi,
while checking out how the EGM energy regression is done a few weeks ago, I realized there were some old producers
ElectronRegressionValueMapProducerandPhotonRegressionValueMapProducer. They produce nothing, but are still in MiniAOD and NanoAOD production sequence. Maybe it would be good to remove them to get rid of the overhead they provide?Local matrix tests all pass.
Cheers,
Jonas