 |
 |
Hullo! > Ok, so you will run the test with 3.1.4 then? I'm testing your detector on the current dev version of SLIC, but I don't really think it matters too much if you just want to verify the detector geometry and look over some sample data. For physics results, we should use the "standard" version, which should probably be 3.1.4 for the time being. So if you want to generate a lot of events with this detector, you can just use the JLAB batch farm with the SLIC binary we've been using there. If that crashes then we'd need to look at other solutions. What usually matters more than the SLIC version is the Geant4 release that is being used. Matt, do you have that information for the 3.1.4 JLAB binary? (Think it is 9.6.p01 but we should verify.) > The 3.1.5 crashes on the new detector but runs on "all" other detectors we have in HPS? We don't know this, because the 3.1.4 release at JLAB has been the one used to batch process most of the Monte Carlo data. I don't think 3.1.5 has been used so rigorously/thoroughly. So it is possible that 3.1.5 has an instability, possibly in the Geant4 release, that would show when running other detector geometries/events as well. I'm not sure it is worth investigating it, though. If we still have to use a SLIC 3.1.x release then we could revert the default configuration in ilcinstall to 3.1.4 pretty easily. > Doesn't that suggest there is something with the new geometry that we may or may not want to understand? Could be a bug in slic/G4 or in our detector geometry I guess. Maybe or maybe not. It is hard to say without a lot more investigation. But, yes, definitely we want to check the new geometry thoroughly. We should run the built-in overlap checker in order to see if any obvious problems pop out. (I haven't done this yet but I'm planning on it.) > I don't know what would be different in terms of the geometry. Only thing I can think of is overlap handling? Maybe it is the overlap handling, but, again, it is hard to say. Geant4 is a complicated beast and any number of things could be going wrong in the stepping to cause a crash. In the past I've seen it get caught in endless loops that then cause a crash, and I believe this behavior has been improved more recently so it is better able to kill tracks that would loop indefinitely. I'm attempting right now to compile SLIC/LCDD/GDML head versions against Geant4 9.6.p03, which is the last patch release in the 9.x Geant4 versions, so that if possible we could make any required changes to the simulation C++ code on the trunk, if there are bugs that need fixing. I *think* the trunk code is backward compatible to 9.6.x, but we'll see.... --Jeremy -----Original Message----- From: Hansson Adrian, Per Ola Sent: Friday, January 30, 2015 3:40 PM To: Graham, Mathew Thomas Cc: McCormick, Jeremy I.; hps-software Subject: Re: slic crash Ok, so you will run the test with 3.1.4 then? The 3.1.5 crashes on the new detector but runs on "all" other detectors we have in HPS? Doesn't that suggest there is something with the new geometry that we may or may not want to understand? Could be a bug in slic/G4 or in our detector geometry I guess. I don't know what would be different in terms of the geometry. Only thing I can think of is overlap handling? /Pelle On Jan 30, 2015, at 3:08 PM, Graham, Mathew Thomas <[log in to unmask]> wrote: I think this is only a problem with slic 3.1.5, which is what I have on my laptop. The version at JLAB is 3.1.4 and it seems to be running no problem. I don't think we should move to Geant4 10 until we do a detailed comparison (tracking/ECAL/trigger) between that and the version we're using now (9.6.1). We don't want any surprises. I'll put this on my list. On Jan 30, 2015, at 2:13 PM, McCormick, Jeremy I. <[log in to unmask]> wrote: Hi, Matt. I am unable to reproduce the crash you are seeing in slic with Pelle's "new" detector, using the *dev* version of the simulation e.g. building from ilcinstall/trunk/examples/slic/dev-scratch.cfg which has these versions [1008 $] slic -v SLIC 5.0.0 Geant4 10.0.2 I run this command slic -g detectors/HPSTracker2014-v0.lcdd -i stdhep/ap2.2gev075mev.stdhep -x -O -r 99999999 and it gets past the event which crashes on your setup. I'm going to see how far it gets into this job (at 4000 events and counting). It looks to me like it should succeeed without crashing. So the immediate solution here is using this dev setup, which is v05-00-00, instead of the HPS specific config. The 3.x version of slic that we use currently is outdated, and continuing to use this is problemmatic, as it is difficult to patch at this point. It would require making a separate development branch of one of the old, released tags of slic and then making specific code changes and releases on that branch. Even then, I'm not sure it would be easy to fix the crash that is happening only with changes to SLIC, as it could be related to Geant4 internals which we have no ability to modify without upgrading the Geant4 version (e.g. generally the stability of G4 has been improving with the releases lately so this problem you're having probably just goes away). Upgrading to the new version, the Geant4 will be different, as the old one is 9.6.p01, and the current build uses the 10.0.2 release. I can't tell you whether this would be a "problem" or not, but certainly the physics results could change between major versions. We would want to study this. Perhaps the ECAL Commissioning data could help us study this, at least for energy depositions and resolutions in the crystals. It would be interesting to see whether this data agrees with the MC output from the current version of slic. --Jeremy ________________________________ 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 ######################################################################## 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