LPC Muon Meeting - Minutes ========================== 26 - September - 2006 http://indico.cern.ch/conferenceDisplay.py?confId=6442 ----------------------------------------------------------------- Stoyan Stoynev (Northwestern): "Memory Checks in CSC Code" Stoyan has begun to check for memory leaks and other such problems using VALGRIND operating on a typical user code which exercised CSC reconstruction. He finds there are some apparently serious problems in the CSC code and in other non-muon routines - see his talk for details. He has not yet tried to identify the bugs which produce these error problems, and in any case will not attempt to do so for non-muon code. His purpose today was to provoke a discussion and to broach this issue. Q: Is VALGRIND publicly installed somewhere? (Stoyan installed it himself in his own area.) Comment: There should be a CMS-wide policy that no code should produce errors or warnings from VALGRIND or the equivalent. It is possible that some experts are writing code that confuse this tool, but this is not desirable as it will render this tool difficult to use. Alex Tumanov commented that there is in fact such a policy. Comment from Dmitryi Kovalskyi: one has to be careful that the first event will naturally give large memory increases since various large buffers will be created for use throughout the entire job. So one has to look a few or dozens of events to see what the increase is after the first event. Furthermore, one can be clever in pulling modules in and out of the path to pin down which ones are producing memory leaks. This can speed up the process of locating specific problems. Question from Phillip Killewald: is there a machine at CERN where one can run something like VALGRIND? Answer from Alex: yes, it is called LXCMSDEV or something like that. (LXPLUS is configured in a way that does not allow one to use enough memory.) ----------------------------------------------------------------- Andy Kubik (Northwestern): "Studies of CSC Offline Timing" Andy is begun to look into the behavior of the timing as defined by the wire and strips digis in the CSC's. It is hoped to make some connection with trigger studies shown by the Florida group in a recent CSC meeting. However, the differences chamber by chamber do not correspond to what is seen for the trigger digis, even for the same run. There are significant differences from chamber to chamber, even more than one time bucket. Furthermore, Andy has observed quite large differences in the same chamber when comparing different runs - it is not yet know whether these correspond to known timing adjustments or not. Finally, Andy shows distributions from the new wire digis recently committed by Nikolai Terentiev and Alex, which are quite strange. More work is needed here to verify the results. Andy stressed that this is just a work in progress and more work is needed to check and extend the results. Ultimately this would become part of the offline monitoring of the CSC's during MTCC phase II. Suggestion from Dmitryi Kovalskyi: One should incorporate some measure of the significance for the deviation of differences from flatness. ------------------------------------------------------------------ Alex Tumanov (Rice): "Update on Muon Reconstruction" Alex reported on a recent bug concerning global muon reconstruction. Someone had reported, and he himself verified, that the global muon reconstruction efficiency was quite low - below 20% - on MC. This was eventually traced to an error in a non-muon configuration file for vertex smearing (conversion between mm and cm!) and all is well, now. He pointed out that it was quite easy for him to install and run the global muon reconstruction and urged others not to hesitate to do so, also. ----------------------------------------------------------------- Alex Tumanov (Rise): "Muon Angular Distributions from Higgs Decays" Alex has been studying the channel H->ZZ -> mu+ mu- mu+ mu- and made a presentation to the Higgs group concerning the use of angular distributions to differentiate different kinds of Higgs particles. He also included a vector boson for good measure. At the generator level he observes some distinctive features which look promising. He is working to make the same analysis on the basis of reconstructed muons, and will develop a likelihood technique to make the best use of this information. ----------------------------------------------------------------- AOB: 1) Alex asked whether there were any simulated 900 GeV jet data. Answer after the meeting: Yes, see the wiki page at: https://uimon.cern.ch/twiki/bin/view/CMS/CmsSamples and look for "QCD Special sample". There is also some min-bias data. 2) Alex reported that there is a compilation problem with the unpacker code, since yesterday, which he is fixing today. ----------------------------------------------------------------- submitted by M.S.