update to PUPPI;speedup, low PU corr #11671
Conversation
|
A new Pull Request was created by @nhanvtran for CMSSW_7_6_X. update to PUPPI;speedup, low PU corr It involves the following packages: CommonTools/PileupAlgos @cmsbuild, @cvuosalo, @vadler, @monttj, @slava77 can you please review it and eventually sign? Thanks. |
| @@ -47,11 +47,7 @@ void PuppiContainer::initialize(const std::vector<RecoObj> &iRecoObjects) { | |||
| // float nom = sqrt((fRecoParticle.m)*(fRecoParticle.m) + (fRecoParticle.pt)*(fRecoParticle.pt)*(cosh(fRecoParticle.eta))*(cosh(fRecoParticle.eta))) + (fRecoParticle.pt)*sinh(fRecoParticle.eta);//hacked | |||
| // float denom = sqrt((fRecoParticle.m)*(fRecoParticle.m) + (fRecoParticle.pt)*(fRecoParticle.pt));//hacked | |||
| // float rapidity = log(nom/denom);//hacked | |||
| if (edm::isFinite(fRecoParticle.rapidity)){ | |||
There was a problem hiding this comment.
please recover this isFinite check.
I'm OK if some other "random" is picked instead of reset_PtYPhiM(0, 99., 0, 0) or if it can be skipped,
but we should not let the pseudojet to crash on nans again
There was a problem hiding this comment.
@slava77 sorry it got dropped -- I put it back in
|
@nhanvtran ping |
|
@slava77 |
|
|
||
| } | ||
| //////////////////////////////////////////////////////////////////////////////// | ||
|
|
||
| //This code is probably a bit confusing | ||
| double PuppiAlgo::compute(std::vector<double> const &iVals,double iChi2) const { | ||
| // if(fAlgoId[0] == -1) return 1; |
There was a problem hiding this comment.
remind me, please why this commented out code is needed here
|
On 10/19/15 2:48 PM, nhanvtran wrote:
Yes, you need to put the if with isFinite back in.
|
|
@slava77 |
|
@nhanvtran Thanks. |
|
@slava77 |
|
@cmsbuild please test |
|
The tests are being triggered in jenkins. |
| double lAdjust = double(lNPV)/double(lNPV+0.5*fNCount[iAlgo]); | ||
| if(lAdjust > 0) { | ||
| fMedian[iAlgo] -= sqrt(ROOT::Math::chisquared_quantile(lAdjust,1.)*fRMS[iAlgo]); | ||
| fRMS[iAlgo] -= sqrt(ROOT::Math::chisquared_quantile(lAdjust,1.)*fRMS[iAlgo]); |
There was a problem hiding this comment.
it looks a bit odd that both the median and RMS are shifted by the same value.
Is this done correctly?
Also, copy-paste of the same multi-operand code is not good (use a temporary).
|
+1
|
update to PUPPI;speedup, low PU corr
This update contains the code for PUPPI v9
Biggest updates are a speedup (and more speedup possibility in the future) and fixes for the low PU behavior.
Final v9 version will be ready after tuning is performed with the new HCAL/ECAL reconstruction fixes and new constants are committed.