Code Libraries

1
Code Libraries

• Tools for BTeV developers and users
• By tools I mean (mostly) packages which contain
  routines which can be called from your code

2
Rationale and Goals

• The tools referred to are
• Libraries, classes, components that developers
  can use as building blocks
• Often HEP-specific
• Some key goals of well-chosen libraries
• Uniformity within Experiment
• Avoid some time-consuming effort
• Increase confidence of correctness
• Improve Consciousness about and effectiveness of
  Object Oriented design

3
Caveats

• The support model for each tool should be
  understood
• Local response to bug reports and essential
  feature needs, or
• A good, stable and adequate user-oriented support
  team, or
• A well-established product thought to be very
  well tested and meeting all needs

4
Caveats

• Some tools are needed but are not available in a
  good form
• Non-OO or poorly designed packages nonetheless
  may be valuable
• BTeV must decide among
• Redo the tool
• Try to join the support team and improve from
  within
• Live with the flawed tool
• Have individual BTeV developers each develop
  those features they need

5
Categories

• Framework-related
• Handling extraordinary occurrences
• Database-related
• Physics-related
• Simple Physics Objects
• Particles and properties
• Simulation etc covered elsewhere
• Mathematics
• Linear Algebra
• Random Numbers
• Minimization
• Special Functions
• Numerical Algorithms
• Histograms and Ntuples
• Accumulation
• Visualization
• Persistency
• Geometric Visualization
• Programming utilities

6
Sources

• HEP community
• CLHEP
• LHC
• Root
• HepVis
• CERNLIB M
• FNAL developers
• ZOOM
• Other tools
• Non-HEP World of C
• std library
• Boost and similar efforts
• gsl (GNU Scientific Library) in C

7
Framework-related

• Error Logging
• Idea is to have a single uniform mechanism which
  allows messages and warnings to be routed to
  multiple destinations, with controlled filtering
  and statistics
• ZOOM ErrorLogger
• In use by D0, CDF
• Simple user (message-issuer) interface
• Semantics like cout ltlt
• Probably meets all BTeV warning and error logging
  needs

8
Framework-related

• RCP Run Control Parameters
• Interface to database of parameters
• M. Paterno
• In use at D0, CDF
• Probably meets BTeV needs
• ZOOM Exceptions
• Extends C exceptions mechanism to include
  signal-like capabilities (e.g. handlers,
  ignore-and-proceed, etc.)
• Can coordinate with ErrorLogger

9
Physics- Related

• CLHEP Vector
• 3- and 4-vectors, Rotations and Transformations
• Rich set of methods
• Same as ZOOM PhysicsVectors
• Geant4 used CLHEP Vector but has switched to its
  own class, causing users some hassle this issue
  is not resolved right now

10
Physics-Related

• HepPDT
• Particle properties, as in the PDG
• Monte-carlo particle numbering scheme
• Values and errors on quantities
• Multiple tables allowed
• Decay information
• Allows for custom decay chains

11
Physics-Related

• StdHepC
• C successor to StdHep
• Interface to all the common event generators
• Deals with specific particles (with a momentum,
  helicity, etc.) utilizes HepPDT to know
  particle properties
• Deals with collisions, events, runs
• Uses HepMC classes developed by ATLAS for event
  container
• Can read/write HEPEVT common block
• HepPDT and StdHepC are in CLHEP with Lynn
  Garren as editor
• So one can expect BTeV needs to be met

12
Mathematics

• CERNLIB support is going away
• Have to identify how BTeV will continue to have
  those CERNLIB tools which it needs mostly
  mathematical
• Linear Algebra
• CLHEP Matrix
• ZOOM LinearAlgebra
• Neither of the above is perfect either is OK.
  There are also packages from the non-HEP
  community, but probably among the free packages,
  these are best.
• Situation needs to be resolved.
• Specialized Matrix operations
• For covariance matrices and operations needed for
  Kalman filter
• Should be lean, mean and fast
• Real need being addressed by ZOOM

13
Mathematics

• ZOOM Special Functions
• Wraps gsl special functions
• Improves the interface
• Adds a few that were missing
• CLHEP Random
• Several excellent generators
• All the distributions mentioned in PDG book (plus
  Landau)
• A random engine is an object
• This makes it easy to design with multiple
  independent streams
• Boost has a random library.
• Superior in terms of C design
• But CLHEP Random is more comprehensive

14
Mathematics

• Minimization MINUIT (?)
• This is part of CERNLIB M so support and porting
  issues are cloudy
• C versions becoming available
• f2c translation is in ROOT
• ZOOM preparing C package with identical
  algorithms plus extensions made natural by
  object-oriented nature
• Algorithms
• Several CERNLIB M algorithms might be useful
• Most can be found in C, in gsl.
• Nobody (that I am aware of) has yet sytematically
  created good C interfaces to these
• But for any specific case, it will not be
  difficult

15
Histograms and Ntuples

• Two types of activity
• Accumulation
• And in-program read-back
• Examination
• Via good graphics
• With statistics and fitting
• Production of publication-quality figures
• These functions, while related, are not
  necessarily both addressed well by the same
  package

16
Histograms and Ntuples

• Before choosing tools/packages, should answer a
  strategic question
• Do we want accumulation code to use an interface
  wrapper that lets us attach different underlying
  packages without hanging code?
• Or do we want to settle on one good package and
  code directly to that?
• Available interface layers
• HepTuple
• Works with ROOT, HBOOK, HistoScope
• Has a bit of historical baggage
• AIDA
• Developed recently LHC
• May not do Ntuples yet
• Underlying implementations might not include
  ROOT-based

17
Histograms and Ntuples

• Several choices for base package
• Root
• Physicists voting with their feet
• HBOOK
• But is it vulnerable ? M
• Java Analysis STudio
• (Newer AIDA implementations)
• Commercial products
• No recommendation here, but
• Even if one package is settled upon, it might be
  right for safety and development easy to use one
  of the interface wrappers

18
Persistency

• Ways to write out data such that it can be read
  back later, in a different program
• Some examples, of vastly different natures
• XDR
• Databases
• XML
• ROOT
• ROOT is in this sense both a floor wax
  (persistency) and a desert topping
  (histograms/Ntuples)
• S. Panacek will say a bit more about ROOT

19
Programming Utilities

• std library
• Clearly people should learn what is there and use
  it when applicable.
• Containers
• This part used to be called the STL, Standard
  Template Library
• Includes vectorltwhatevergt, a good string class,
  and associative arrays (map and set)
• Algorithms
• For example, sort but many others
• Numerics
• C type functions like asin()
• Architecture-dependent limits
• complex ltdoublegt data type
• Generalized numeric algorithms

20
Programming Utilities

• Various goodies from ZOOM and CLHEP
• CLHEP Evaluator
• ZMtimer
• Fast specialized allocators
• Graded asserts
• Generic factory classes
• Look at what is available and
• ASK FOR WHAT IS NEEDED!
• Places like BOOST
• Smart pointers
• Thread concepts

21
Geometric Visualization

• HepVis
• Shapes for physics
• Tubes, balls, tracks
• Uses OpenInventor
• A true programming tool
• Call the routines from code
• Can link with Geant4
• Good for event displays etc.
• One shouldnt do pattern recognition without the
  aid of a graphic display tool

22
Summary

• IMHO, BTeV should from the start agree on using
• ZOOM ErrorLogger
• RCP
• HepPDT
• StdHepC
• ZOOM SpecialFunctions
• CLHEP Random
• ROOT (?)
• HepVis
• std library
• Learn what is there
• Use wherever applicable

23
Summary

• And should apply pressure if necessary to
  encourage completion of
• Lean and fast matrix code for tracking purposes
• MINUIT C replacement
• And should determine to decide on
• A general matrix package
• Which CERNLIB Algorithms are seriously needed
• Interface scheme and underlying package for
  histograms and Ntuples
• A persistency mechanism
• And should worry about
• Physics Vector divergences between CLHEP and
  Geant4
• Histogram/Ntuple package if not ROOT