Status of the software migration u Migration strategy

  • Slides: 12
Download presentation
Status of the software migration u Migration strategy: n u Status: n u Where

Status of the software migration u Migration strategy: n u Status: n u Where we should be Where we are Plans Marco Cattaneo, 23 rd February 2000 1

Migration strategy as presented at last plenary u Step 1: separate SICBMC and SICBREC

Migration strategy as presented at last plenary u Step 1: separate SICBMC and SICBREC n u Step 2: wrap SICBREC algorithms into Gaudi framework n u To be ready by Jan 10 th To be ready by end February Step 3: gradually replace FORTRAN with C++ algorithms n n Timescale dictated by sub-detector priorities. . see in February Development/integration in Gaudi can start now Marco Cattaneo, 23 rd February 2000 2

Status, step 1 u Pre-release SICBMC v 230 and SICBDST v 230 n n

Status, step 1 u Pre-release SICBMC v 230 and SICBDST v 230 n n Available since 9 th February SICBMC No major changes apart from addition of raw. Hit banks l Could be used in production now l n SICBDST Getting the initialisation right was not trivial l Requires extensive testing l n Delay in delivery (also) due to many intervening SICB releases l u Higher priority, Moving target Problems outstanding: n n n Pileup Verification Ongoing SICB development Marco Cattaneo, 23 rd February 2000 3

Step 1: pileup u SICBMC has access to GEANT hits n u Interim solution:

Step 1: pileup u SICBMC has access to GEANT hits n u Interim solution: n n u Knows how to do pileup Run existing MC tapes through SICBMC to produce raw. Hit banks with pileup Run raw. Hit files through SICBDST to make DST 1, DST 2, DST 3. . . Status: n See Andrei’s talk Marco Cattaneo, 23 rd February 2000 4

Step 1: Verification u We must be convinced that SICBMC/SICBDST give same results as

Step 1: Verification u We must be convinced that SICBMC/SICBDST give same results as SICB n Need comprehensive set of checking programs l n Producing a comprehensive set of data quality histograms Can only be done by sub-detector groups See status reports at this meeting l Is list of contacts correct? l u Data quality programs to be run on: n n DST (~1000 b-bar) produced by SICB from a MC tape DST produced by SICBMC/SICBDST from same MC tape l u And investigations made by SD experts if results are not the same Side-product: n Data-quality checks can be run by program manager prior to every future release. 5

Step 1: SICB development u Many developments foreseen in “SICB” n Need well defined

Step 1: SICB development u Many developments foreseen in “SICB” n Need well defined procedure for new releases To define content, testing procedure, deadlines l See Agnieszka’s talk l u Currently we have parallel development n Two versions of each package: Changes for SICBMC/SICBDST go to CVS head revision l Changes for SICB go to CVS “bug fix” branch l u HOW CAN WE CONVERGE? n n Only bug fixes to SICB All new features to SICBMC/SICBDST l u But only after verification work is finished How urgent is next production version? l Can it wait until after SICBMC/SICBDST verification? 6

Step 1: Proposal u Test samples have been produced l u (see mail from

Step 1: Proposal u Test samples have been produced l u (see mail from Eric van Herwijnen) SICBMC/SICBDST checks and fixes to be made by 15 th March n This is the highest priority Requires work from all sub-detector contacts l Report problems and fixes to Florence Ranjard l n u Next released version of “SICB” will be SICBMC/SICBDST “v 230” No new features until this works n Current list of foreseen “SICB” improvements will go to following version of SICBMC/SICBDST Marco Cattaneo, 23 rd February 2000 7

Step 2: Brunel status u Wrapping SICBDST FORTRAN modules is trivial n u All

Step 2: Brunel status u Wrapping SICBDST FORTRAN modules is trivial n u All SICBDST digitisation and reconstruction modules already wrapped as GAUDI algorithms n u n To allow modification of the FORTRAN defaults Possible solution is known, to be tried next week Functionally identical to SICBDST by early March n n u Trigger modules to be done next week Need to implement reading of sicb. dat n u Difficulty lies in getting the initialisation right Same testing procedure as for SICBDST Follows SICBDST developments (uses the same libraries) Then ready to accept new OO algorithms n And to replace SICBDST in production l Decision at April software week? 8

Brunel structure u One top algorithm per Brunel “phase”: Application. Mgr. Top. Alg =

Brunel structure u One top algorithm per Brunel “phase”: Application. Mgr. Top. Alg = { "Brunel. Initialisation /Brunel. Init", "Brunel. Phase /Brunel. Digi", "Brunel. Phase/ Brunel. Digi", "Brunel. Phase /Brunel. Trigger", "Brunel. Phase/ Brunel. Trigger", "Brunel. Phase /Brunel. Reco", "Brunel. Phase/ Brunel. Reco", "Brunel. Phase /Brunel. Final. Fit", "Brunel. Phase/ Brunel. Final. Fit", "Brunel. Finalisation /Brunel. Finish" }; u One child algorithm per sub-detector per phase Brunel. Digi. Detector. List = { "VELO" , "TRAC" , "RICH" , "ECAL" , "HCAL", "MUON" }; Brunel. Reco. Detector. List = { "TRAC" , "RICH" , "ECAL" , "HCAL" }; Brunel. Final. Fit. Detector. List = { "TRAC" }; Brunel. cpp, , Brunel. Digi. VELO. cpp Brunel. Digi. TRAC. cpp Brunel. Digi. RICH. cpp Brunel. cpp, , Brunel. Reco. TRAC. cpp Brunel. Reco. RICH. cpp Brunel. Reco. ECAL. cpp Brunel. Final. Fit. TRAC. cpp u Calls corresponding FORTRAN algorithm n Can replace call to FORTRAN with new OO algorithm Marco Cattaneo, 23 rd February 2000 9

Step 3: OO reconstruction software u Ongoing work in many areas (see later talks)

Step 3: OO reconstruction software u Ongoing work in many areas (see later talks) n Calorimeter data model design review (18/02/2000) Very useful both for calorimeter team and GAUDI team l Highlighted needs for LHCb-wide solutions/conventions l Ø How to access MC truth from reconstructed data Ø Direct access (c. f. sequential access) to contained objects in transient store n Implementation of ECAL geometry in GAUDI n Implementation of SICB Converters for Level 1 and VELO n Track Fit ready to be made public in GAUDI n etc. . . Marco Cattaneo, 23 rd February 2000 10

Conclusions u SICBMC/SICBDST split is complete n n u Wrapping of FORTRAN in Brunel

Conclusions u SICBMC/SICBDST split is complete n n u Wrapping of FORTRAN in Brunel is nearly complete n u Verification will be needed soon Better procedures for new versions n n u Verification is needed at highest priority, leading to retirement of SICB by mid-March Next production version of “SICB” will be SICBMC/SICBDST Planning of contents, with deadlines, in place Definition of data quality procedures urgently needed Implementations in GAUDI have begun Marco Cattaneo, 23 rd February 2000 11

Software contacts A. Jacholkowska and E. van Herwijnen (production) M. Cattaneo (Brunel) P. Mato

Software contacts A. Jacholkowska and E. van Herwijnen (production) M. Cattaneo (Brunel) P. Mato (Gaudi) J. Harvey (Computing) H. Dijkstra (Trigger) T. Ruf (Velo) G. Gracia, M. Needham, M. Merk (outer tracker, kalman) G. Wilkinson (Rich) G. Corti (Analysis) A. Tsaregorodtsev (management, pileup) P. Colrain (Muon) I. Korolko, O. Callot (Calorimeters) Generators (Bartalini) Marco Cattaneo, 23 rd February 2000 12