ROOT Team Meeting

Friday 9 Mar 2007, 11:00 → 12:50 Europe/Zurich

32-1A24

11:00 → 11:05 News

11:05 → 11:50 Bypassing stub functions in CINT

Diego and Leo have prepared a set of slides explaining their current work with the replacement of CINT stub functions by direct calls to the compiled code

Speaker: Diego Marcos.Segura & Leo Franco

Slides DiegoLeo_09Mars07.pdf DiegoLeo_09Mars07.ppt

11:50 → 12:10 Tour de Table

Minutes of ROOT Team meeting 9 March 2007 ========================================= News ---- -MAPS/MARS must be completed before March 20 -First mini tour de table to check the list of developments before the 5.15/04 scheduled for March 21. Talk by Diego & Leo ------------------- see slides Congratulations to Diego and Leo for their interesting presentation. Suggestions by Fons will be included in their next slide iteration. Tour de Table -------------- -Bertrand Spent some time on PVSS and Root... Made changes in system/win32 threads to improve use of multiprocessor/hyperthreaded machines Made changes in Tsystem and Tapplication for autoload mechanism Actually working on TThread on Windows and Drag and drop (linux AND win32) -Gerri PROOF: - working to a patch to: - implement prioritized resource sharing at worker level following an idea of Andy Hanushevsky; the processing time of lower priority users is artificially inflated (by delaying the request for new packets) by a factor depending on the priority of the group to which the user belongs. - implement an interface to chose a ROOT version among a list setup by the administrator - implement class, macro loading functionality (ALICE request) CASTOR related (patch in 5.14.00d): - modify 'rootd' to prevent failures observed on Castor servers due to missing $HOME directories - fix a few problems with globus authentication in Castor environment (following a request from LHCb) -Jan After last PROOF - Alice meeting we were investigating the problems with some queries being analyzed slowly (spikes in graphs from cocktail tests). The analysis showed that two slaves on the cluster underperform. We also did a number of local and PROOF test to compare the processing rates and identify bottle neck (hard disk / network). After changing ProofBench, we run PROOF and local ROOT analysis on the same sets of files. The old ProofBench files were generated with no compression due to the fact that Branch constructor automagically defaults to no compression, unlike the tree. Has to changed. -Axel -fixing CINT autloader + unload issue, thanks to suggestion by Philippe; -picking up reflex; plan to move to gccxml 0.7 instead of 0.6 + 3 cern patches; -long list of THtml bugs by Olivier, Ilka, Lorenzo; depends on fixes in TASImage, waiting for Valeriy -Stefan -passing the "flambeau" to Axel -introducing new version of gccxml -Lorenzo -additions to Unuran -working on the TF1 extensions -will introduce the changes in TRandom::Gaus (copied fron Unuran) -Ilka GUI: - Fix in TGListView (in cvs) - Fix in fit panel recursive remove (in cvs) - Editors' layout fixes related to cross-platform view (in cvs today) - Working on Fit panel function loading and parameters' setting - Testing GUI with the new auto loading mechanism - Working on the color wheel example Documentation: - User's Guide v5.14 sent for printing - Working on the QuickStart Guide of ROOT -Fons -Working on several tiny improvements, in particular around the rootmaps/autoloading -Rene -Implemented changes in the color management (TStyle/TColor) -about to introduce the code to paint the ColorWheel -mention of the good news about the comparison G4/TGeo by Andrei

12:10 → 12:15 AOB