Hi Flemming, On Wed, 2 May 2001 20:47:43 -0400 "Flemming Videbaek" <videbaek@sgs1.hirg.bnl.gov> wrote concerning ": upcoming updates in BRAT": > As part of the process of updating and upgrading various tracking to > accomodate a more general approach I have changed (after > consultation with the local working groups) code and methods in > tracking classes. The changes has yet to be commiteed > this is an announchement, and comments are appreciated. > > - BrModuleMatchTrack class > > ---- Extensive rewrite to simplify geometry handling. Replaced max > array classes with STL <vector> implementation. Uh. Why STL vector<..>? I believe that recent experience from the ROOT team clearily shows that TCollection and derived classes, are much better (i.e., faster, safer) than the STL counter parts. From the release note of ROOT 3.00/06: New test suite and benchmark bench.cxx. This test program compares the I/O performance obtained with STL vector of objects or pointers to objects versus the native ROOT collection class TClonesArray. Trees in compression and non compression mode are created for each of the following cases: -vector<THit> -vector<THit*> -TClonesArray(TObjHit) in no split mode -TClonesArray(TObjHit) in split mode I've run the benchmark test, and it clearly shows that TClonesArray is far superior to vector<...>. I've attached the output. My own experience also suggets that STL containers are not as neat and easy to handle as ROOT's counter parts. Yours, Christian ----------------------------------------------------------- Holm Christensen Phone: (+45) 35 35 96 91 Sankt Hansgade 23, 1. th. Office: (+45) 353 25 305 DK-2200 Copenhagen N Web: www.nbi.dk/~cholm Denmark Email: cholm@nbi.dk
This archive was generated by hypermail 2b29 : Thu May 03 2001 - 05:36:51 EDT