Calo Software: Urgent tasks and resources Ivan Belyaev LHCb, CERN & ITEP/Moscow

18 Sep'2k+1 CaloSoft Ivan Belyaev LHCb CERN & ITEP/Moscow 2 Outline Priorities and the program Priorities and the program Urgent tasks I Urgent tasks I Xmas 2k+1 Xmas 2k+1 Urgent tasks II Urgent tasks II Spring 2k+2 Spring 2k+2 SICB MC & DATA sets SICB MC & DATA sets Geant4 Geant4

18 Sep'2k+1 CaloSoft Ivan Belyaev LHCb CERN & ITEP/Moscow 3 Current Status PRODUCTION COMPONENTS FORTRAN OO (C++) SICBMC (simulation) SICBMC (simulation) Signal processing Signal processing Digitization Digitization Zero-suppression Zero-suppression Trigger Trigger Reconstruction Reconstruction Analysis/Analysis Analysis/Analysis OO code in official production for Tracker/Trigger TDRs OO code in official production for Tracker/Trigger TDRs Timing is not yet in Signal Processing Timing is not yet in Signal Processing SICB MC Trigger needs different treatment of input data Trigger needs different treatment of input data Need OO-DST Some sub-components of Reconstruction are available in C++ Some sub-components of Reconstruction are available in C++ Refine and tune Calo Event Model Need for Tracker/Calo collaboration Need for Tracker/Calo collaboration

18 Sep'2k+1 CaloSoft Ivan Belyaev LHCb CERN & ITEP/Moscow 4 What we want? Coexistence of FORTRAN and OO codes is a pain  ! Coexistence of FORTRAN and OO codes is a pain  ! Inter-algorithms data exchange Inter-algorithms data exchange format and content of input/output data format and content of input/output data ZEBRA versus OO-DST ZEBRA versus OO-DST Converters and back- converters Converters and back- converters Cryptic code! Cryptic code! Every back-conversion: information is lost! Every back-conversion: information is lost! Different geometry sources: time-bomb Different geometry sources: time-bomb Long term maintenance Long term maintenance SM-approach: SM-approach: Keep everything as it is and try to enjoy the pain Keep everything as it is and try to enjoy the pain “Dentist”-approach “Dentist”-approach Don’t cut dog’s tail piece by piece A little bit painful, but necessary procedure FAST transition to OO Abandon FORTRAN codes from the reconstruction Enjoy the homogeneous environment

18 Sep'2k+1 CaloSoft Ivan Belyaev LHCb CERN & ITEP/Moscow 5 FAST transition Replace all existing FORTRAN reconstruction code with 00 till Xmas Replace all existing FORTRAN reconstruction code with 00 till Xmas Waiting for OO-DST Waiting for OO-DST (Allow for some (very short!) “coexistence” phase for validation of new codes) (Allow for some (very short!) “coexistence” phase for validation of new codes) Abandon all FORTRAN and invite friends to FORTRAN funeral party Abandon all FORTRAN and invite friends to FORTRAN funeral party Milestone: Brunel OO- DST Spring of 2k+2 Milestone: Brunel OO- DST Spring of 2k+2 Is it realistic? Is it realistic? Number of FORTRAN algorithms is not large! Number of FORTRAN algorithms is not large! But they are coherent! For many of them we already have some OO “analogs” Brunel defines a milestone for OO-DST of a few months from now. Brunel defines a milestone for OO-DST of a few months from now. Tracks are “OO Tracks” now Tracks are “OO Tracks” now

18 Sep'2k+1 CaloSoft Ivan Belyaev LHCb CERN & ITEP/Moscow 6 Reconstruction Tasks Very Urgent (Xmas 2k+1) Replace all PRODUCTION FORTRAN codes Prs energy correction Prs energy correction Use Prs! Use Prs! Resolved  0 s Resolved  0 s Prompt  s Prompt  s Partially overlapped  0 s Partially overlapped  0 s e + /e - PID e + /e - PID Track/cluster matching Track/cluster matching ??? ??? Urgent (Spring 2k+2) Urgent (Spring 2k+2) Not in production now Merged  0 s Merged  0 s Bremmshrahlung  s Bremmshrahlung  s Search for e + /e - energy recovery e + /e - PID e+/e- PID e+/e- PID Usage of Prs Track/cluster matching Shower shape

18 Sep'2k+1 CaloSoft Ivan Belyaev LHCb CERN & ITEP/Moscow 7 Reconstruction Tasks Prs energy correction Prs energy correction Now: correction coefficient/function Now: correction coefficient/function Need: Use Prs! Need: Use Prs! Partially overlapped  0 s Partially overlapped  0 s Resolved  0 s Resolved  0 s Prompt  s Prompt  s e + /e - PID e + /e - PID Track/cluster matching Track/cluster matching Energy matching Energy matching ??? ??? “Volunteer” Sergey Timescale: end of October “Volunteer”: Frederic Timescale: November? “Volunteer”: Frederic By product of resolved  0 s analysis “Volunteer”: somebody from ITEP? “Volunteer”: somebody fro ITEP? 4/6 + 50% * 2/6 = 67% + 17% = 84%

18 Sep'2k+1 CaloSoft Ivan Belyaev LHCb CERN & ITEP/Moscow 8 Reconstruction Tasks Merged  0 s Merged  0 s Momentum Analysis Iterative shower Shape Bremmshrahlung  s Bremmshrahlung  s Search for e + /e - energy recovery e + /e - PID e+/e- PID e+/e- PID Usage of Prs Track/cluster matching Shower shape Combined  0 /  separation at very high energies  0 /  separation at very high energies “Volunteer”: Olivier Timescale: Spring 2k+2 “Volunteer”: Sergey? “Volunteer” from the previous slide ??? 4/10 = 40%

18 Sep'2k+1 CaloSoft Ivan Belyaev LHCb CERN & ITEP/Moscow 9 Simulation tasks Extremly Urgent (yesterday) Coherent MC DATA samples Coherent MC DATA samples B-> , K* , X e, XJ/  B-> , K* , X e, XJ/  Very Urgent (today) Very Urgent (today) Timing simulation in SICBMC Timing simulation in SICBMC agreed few months ago agreed few months ago Not Urgent Not Urgent Correct simulation of time- response Correct simulation of time- response probably in G4 ? probably in G4 ? Geant4 studies Geant4 studies “Volunteer”: Ivan? “Volunteer”: Ivan ??? Xml description is needed! ½ + 50% * ½ = 75% | ??/?? It is also “Signal Processing Task”

18 Sep'2k+1 CaloSoft Ivan Belyaev LHCb CERN & ITEP/Moscow 10 Geant4 Simulation tasks Xml geometry description Xml geometry description Ecal Ecal Hcal Hcal Spd/Prs Spd/Prs Geant4 Comparisons Geant4 Comparisons Ecal Ecal Hcal Hcal Spd Spd Prs Prs Simulation code itself Simulation code itself Time-response !! Time-response !! “Volunteer”: Galina, status - DONE The first version was done by Olivier “Volunteer”: Alexey, Time scale: end of October Volunteer: ??? “Volunteer”: Dima, status - DONE Volunteer: ??? It is quite “generic” and not VERY different and complicated It is specific, could be not trivial and complicated. 3/X = ?? | 50%

18 Sep'2k+1 CaloSoft Ivan Belyaev LHCb CERN & ITEP/Moscow 11 Other tasks Time scale “ASAP” Time scale “ASAP” Visualization Visualization Calo Event Display and tools for interactive “analysis” Calo Event Display and tools for interactive “analysis” Trigger and raw data (software) Trigger and raw data (software) Format and data flow Format and data flow Calo Particle Calo Particle Collaboration with Tracker Collaboration with Tracker AOB AOB “Volunteer”: Vanya Status: prototype exists, further development is blocked “Volunteer”: Olivier (our trigger-master) “Volunteer”: everybody “Volunteer”: the whole group October 9 th – special meeting ?? %

18 Sep'2k+1 CaloSoft Ivan Belyaev LHCb CERN & ITEP/Moscow 12   reconstruction I June 22: Frederic June 22: Frederic Resolved  0 s Resolved  0 s Overlapped  0 s Overlapped  0 s Merged  0 s (Momentum analysis) Merged  0 s (Momentum analysis) Aug 28: Olivier Aug 28: Olivier Resolved  0 s Resolved  0 s Overlapped  0 s Overlapped  0 s Merged  0 s (Momentum and iterative analysis) Merged  0 s (Momentum and iterative analysis) Common problem: MC DATA SETS Common problem: MC DATA SETS F.: unphysical particle spectra. Noise, 0-suppression and digitization are OK F.: unphysical particle spectra. Noise, 0-suppression and digitization are OK O.: physical spectra OK. Problems with Ecal energy resolution, no noise, O.: physical spectra OK. Problems with Ecal energy resolution, no noise, no 0-suppression, no digitization no 0-suppression, no digitization No pileup/feeddown, no material in front of Calo No pileup/feeddown, no material in front of Calo SICBMC: need coherent validated clean physical MC data samples!    e X  J  X

18 Sep'2k+1 CaloSoft Ivan Belyaev LHCb CERN & ITEP/Moscow 13  0 reconstruction II Analyses differ in cluster definition Analyses differ in cluster definition 3x3 3x3 Cross-like Cross-like OO code now provides 2 type of clusters OO code now provides 2 type of clusters Cellular Automaton Cellular Automaton 3x3 “FORTRAN” 3x3 “FORTRAN” (Determine cluster parameters using “sub- clusters”?) (noise&pileup) (Determine cluster parameters using “sub- clusters”?) (noise&pileup) 2x2, Cross-like, … 2x2, Cross-like, … Reconstruction of merged  0 s Reconstruction of merged  0 s Momentum Analysis Iterative shower shape (Combined?)Olivier Reconstruction of resolved  0 s Reconstruction of resolved  0 s Reconstruction of partially overlapped  0 s Reconstruction of partially overlapped  0 s (prompt  (prompt Frederic

18 Sep'2k+1 CaloSoft Ivan Belyaev LHCb CERN & ITEP/Moscow 14 S-shape correction Multi-parameter function in FORTRAN production Multi-parameter function in FORTRAN production One-parameter function for recent analyses One-parameter function for recent analyses Both Frederic and Olivier show that especially for outer region it is not enough Both Frederic and Olivier show that especially for outer region it is not enough Indicated long time ago by Benoit Indicated long time ago by Benoit Neural Network analysis could be a solutionNeural Network analysis could be a solution Stability Noise? Pileup/feeddown? Cluster shape definition? Neighbor definition? Non-linear How to estimate the uncertainties and covariance matrix? A little bit artificial An improvement is possible Need for more parameters? Need for more parameters? Explicit Angular/Energy dependence? Explicit Angular/Energy dependence?

18 Sep'2k+1 CaloSoft Ivan Belyaev LHCb CERN & ITEP/Moscow 15  0 reconstruction III (next steps) Use “external” clusterisation Use “external” clusterisation Use “external” energy redistribution algorithm Use “external” energy redistribution algorithm Incorporate Prs correction Incorporate Prs correction Improve S-shape Improve S-shape Define sub-clusters? Define sub-clusters? Refine & tune the code Refine & tune the code Get  0 s! Get  0 s! Validation Validation We already have 2 “external” and 1 “internal” algorithms: Cellular Automaton, 3x3 and cross-like We already have an algorithm with 4 different techniques! Work is started. Not clear. (V.B.: “NN is not a perfect choice!”) Need to be proved. Refer to Benoit’s study and test beam measurements. Iterative process Need to define the validation policy

18 Sep'2k+1 CaloSoft Ivan Belyaev LHCb CERN & ITEP/Moscow 16 (Short) summary Calorimeter Software development process is in the progress Calorimeter Software development process is in the progress Good shape! Good shape! Available resources/tasks ratio is above 50% for “urgent” tasks Available resources/tasks ratio is above 50% for “urgent” tasks Aggressive ( <100 % ) Aggressive ( <100 % ) Reasonable ( >0.1 % ) Reasonable ( >0.1 % ) Still some lack of resources Still some lack of resources Situation for “very urgent tasks” is better Situation for “very urgent tasks” is better Calorimeter Software becomes popular! We are the first! “We are the champions!” Your contribution is welcome! 2k+2 without FORTRAN?

18 Sep'2k+1 CaloSoft Ivan Belyaev LHCb CERN & ITEP/Moscow 17 The End Thank you!