 |
 |
> > The only time I have seen this is when GBL is trying to refit low > momentum bogus tracks. I'm not sure which one of the tweaks could be > causing this (assuming this wasn't happening before.) ... Yes during the pass6 it was rear, only few files from overall 2015 data. I am attaching the compact.xml, just in case if that might help debugging it. > If we can get 10% of the data reconed with the new detector over the > next several days, then it might be worth just running the full > dataset. I can then do my optimizations on the 10% and just run my > scripts over the full dataset when it's ready early in January. I > don't think we plan to unblind until the middle of January so it looks > like we have time and should take advantange of the open farms. > Others should chime in with their opinions ... It will take probably 4 days the 10% of the data, when we start to process, but before processing, it will take at least few days, to check/validate these recent changes, and also to fix some bugs, e.g. the "Matrix is singular." Rafo On 12/06/2016 07:35 PM, Omar Moreno wrote: > > > On Tue, Dec 6, 2016 at 4:05 PM, Rafayel Paremuzyan <[log in to unmask] > <mailto:[log in to unmask]>> wrote: > > Hi all, > > yesterday I submitted a new testpass7 for 5772, > outputs are going in the following directory. > /mss/hallb/hps/engrun2015/tpass7 > > Note: this is a Full reconstruction pass, where tweaks are applied > in the compact.xml, > and it uses a new v5 detector. > This uses a new steering file for ntuplemake, which Holly has > created that doesn't do tweaks. > The reason for this was as Stepan suggested, to check whether > pre-recon tweaks are consistent with a post-recon tweaks. > > Now about the other pass tweakpass that we have discussed in the > software meeting, which shouldn't > do recon, but rather use pass6 recon, and create another recon > from with tweaked parameters. > Thinking more about it, this is supposed to read the pass6 recon > slcio, do tweaks on all track collections e.g. > "GBLTracks" then do run trackDataDriver and ReconParticleDriver. > To do this probably new driver and steering file is needed to be > created (I think by experts). > Maybe this is easy, but I don't know details. > Please let me know how easy can that be. > > I know that Holly is close implementing the RF time correction to > the highest energy crystal, > and if we decide instead of tweakpass use a full recon pass, then > this will be an important addition > to the data. > > I guess the decision whether to go with a tweakpass or do a new > full recon mostly depends > on how close is Omar finalizing bump hunt analysis and being ready > to unblind the data. > If this is something that will be not sooner than a month from > now, then I would think it is better to > go with a full recon pass, especially farms probably will be more > quite during the Christmas and New Year. > > Another question to those who do/did their analysis on 2015 data, > as I know no one did not use > v0 skim. Is there any particular reason for that, Are cuts too tight? > If cuts are really too tight, then I would think we don't need to > do v0 skim. > Sebouh did a quick test, by requiring only a presence of a non > empty unconstrainedv0candidate collection, > but it yields about 13% of the original file, which is not too small. > > Similar question for the Moller skim, is anyone using a Moller skim? > If they really are not needed, then dropping them can increase a > little bit the chance of the job not being a timed out. > > I also have another question, probably to Jeremy or tracking > experts, a lot of recon jobs are > finished without reaching the end of evio file, look bottom of the > e-mail, for the thrown exception. > I have used source codes of release 3.9, and only modified the > compact.xml accordingly to take > into account tweaks, and a new steering file is added by Holly, > which do ntuplemaker without tweaking > parameters (last one should not be related, since is called after > recon). > > The only time I have seen this is when GBL is trying to refit low > momentum bogus tracks. I'm not sure which one of the tweaks could be > causing this (assuming this wasn't happening before.) ... > > Sorry, it turned out to be a big e-mail > Rafo > > Exception in thread "main" java.lang.RuntimeException: Matrix is > singular. > > at > org.hps.recon.tracking.gbl.matrix.LUDecomposition.solve(LUDecomposition.java:281) > at > org.hps.recon.tracking.gbl.matrix.Matrix.solve(Matrix.java:1078) > at > org.hps.recon.tracking.gbl.matrix.Matrix.inverse(Matrix.java:1100) > at > org.hps.recon.tracking.gbl.GblPoint.getDerivatives(GblPoint.java:533) > at > org.hps.recon.tracking.gbl.GblTrajectory.getFitToKinkJacobian(GblTrajectory.java:519) > at > org.hps.recon.tracking.gbl.GblTrajectory.prepare(GblTrajectory.java:679) > at > org.hps.recon.tracking.gbl.GblTrajectory.construct(GblTrajectory.java:250) > at > org.hps.recon.tracking.gbl.GblTrajectory.<init>(GblTrajectory.java:180) > at > org.hps.recon.tracking.gbl.GblTrajectory.<init>(GblTrajectory.java:29) > at > org.hps.recon.tracking.gbl.HpsGblRefitter.fit(HpsGblRefitter.java:486) > at > org.hps.recon.tracking.gbl.MakeGblTracks.doGBLFit(MakeGblTracks.java:157) > at > org.hps.recon.tracking.gbl.MakeGblTracks.refitTrack(MakeGblTracks.java:138) > at > org.hps.recon.tracking.gbl.GBLRefitterDriver.process(GBLRefitterDriver.java:79) > at org.lcsim.util.Driver.doProcess(Driver.java:261) > at org.lcsim.util.Driver.processChildren(Driver.java:271) > at org.lcsim.util.Driver.process(Driver.java:187) > at > org.lcsim.util.DriverAdapter.recordSupplied(DriverAdapter.java:74) > at > org.lcsim.job.JobControlManager.processEvent(JobControlManager.java:827) > at org.hps.evio.EvioToLcio.run(EvioToLcio.java:628) > at org.hps.evio.EvioToLcio.main(EvioToLcio.java:98) > 2016-12-06 14:27:07 [CONFIG] org.lcsim.job.JobControlManager parse > :: parsing command line arguments > 2016-12-06 14:27:07 [CONFIG] org.lcsim.job.JobControlManager parse > :: steering resource enabled > > ######################################################################## > Use REPLY-ALL to reply to list > > To unsubscribe from the HPS-SOFTWARE list, click the following link: > https://listserv.slac.stanford.edu/cgi-bin/wa?SUBED1=HPS-SOFTWARE&A=1 > <https://listserv.slac.stanford.edu/cgi-bin/wa?SUBED1=HPS-SOFTWARE&A=1> > > ######################################################################## Use REPLY-ALL to reply to list To unsubscribe from the HPS-SOFTWARE list, click the following link: https://listserv.slac.stanford.edu/cgi-bin/wa?SUBED1=HPS-SOFTWARE&A=1