User Tools

Site Tools


2010 DiFX Workshop Agenda

Note: 8:30 AM start on Tuesday! Start times on Wed-Fri TBD

We won't be able to provide breakfast, so please come fed. We'll try to keep the coffee flowing. The Auditorium should be open by 8:00 AM with wired and wireless internet availability. Visitors: please sign in at the front desk when you arrive and sign out when you depart.

Day 1: (Monday Nov. 15)

  • 9:00 AM Welcome (Walter), meeting preface, around-the-room introductions
  • 9:15 AM DiFX architecture including DiFX 2.0 developments (Adam)
  • 10:00 AM User reports part I (MPIfR (Helge), ATNF (Chris) )
  • 10:30 AM Coffee break
  • 11:00 AM User reports part II (USNO (David), NRAO (Walter), Curtin (Cormac), others?)
  • 11:45 AM Real-time GUI correlation demonstration (Walter)
  • 12:15 PM Group photo
  • 12:20 PM Lunch break
  • 1:45 PM Geodetic applications (Roger, Helge)
  • 2:30 PM PCals and better accountability for Astronomy and Geodesy (John Morgan)
  • 3:00 PM Any other contributed reports of general interest, discussion of user needs
  • 3:30 PM Coffee break
  • 4:00 PM Review of 2009 to do list and initial creation of 2010 list
  • 4:30 PM Display, explain, update the suggestion box items; start discussion on general topics
  • 5:00+ PM Disband for the day
  • No scheduled dinner plans

Day 2:

  • 8:30 AM Pcals and better model accountability (John Morgan)
  • 9:00 AM Delay models in the widefield (John Morgan)
    • Followed by discussion about polynomial in (l,m,t) based uvw, delay models
  • 9:40 AM Complex sampling, as implemented and some esoteric theory (Chris)
  • 10:10 AM Coffee break
  • 10:30 AM Special DiFX projects
    • eVLBI (Chris)
    • V-FASTR (Walter), followed by discussion on use of STA data for system temperature estimation
  • 11:30 AM Future of standards (FITS, VDIF and Vex[2]) discussion led by Walter:
  • 12:00 PM Hardware/Software discussion led by Adam:
    • Cluster performance/tuning, hardware/software perils, OS/kernel/library gotchas, Mk5 pitfalls, configuration control
  • 12:30 PM Lunch at the M Mountain grill (walk up from AOC)
  • 1:45 PM Tour of the VLBA correlator
  • 2:15 PM Monitoring/control discussion led by Walter/Chris:
    • new performance/cpu/network info (DIFX_DIAGNOSTIC message?), standard logging, aggregation/summary?
    • real time monitoring of visibilities, phase cals (inc. demo by Chris)
    • Possibility of control of some parameters via difxmessage e.g. clocks; accountability
  • 3:00 PM Discussion of human interaction led by Walter:
    • Source name allowed characters, length etc
    • Case sensitivity
    • Any rationalizing of parameter names in eg vex2difx
  • 3:30 PM Coffee break
  • 3:45 PM Discussion/investigation of bug tracking on Trac led by Cormac
  • 4:00 PM Discussion of testing led by John Morgan:
    • Unit testing on svn
    • Revision testing for e.g. one reasonably large geodetic correlation all the way through both AIPS and fourfit, automated summaries
  • 4:30 PM Discussing useful things not currently on SVN or not currently general, led by Adam:
    • runmpifxcorr scripts
    • splitting setup into compile and run, better generalization
    • running versioned correlations, storing old correlations
  • 4:45 PM Discussion on changing band references and LSB/USB conventions, led by Walter:
    • Potential to change all LSB to USB at unpack stage
    • The current perils with LSB
    • Potential for shifted FFTs, totally avoiding aliasing
    • Changing reference frequencies to band center
  • 5:00 PM Discussion on the problems of Mark5B, led by Walter
  • 5:15 PM Review by hand to-do list, add new requests (Adam will update webpage later)
  • This item should be moved up as necessary to ensure it will be done as the last item of the day, even if some of the other discussion points are moved to following days
  • 5:30 PM End
  • Dinner at Socorro Springs, 7:00 PM (there will be a sign up sheet for reservations)

Day 3:

  • Hack-a-thon, with initial discussion covering:
    • Getting pcals into FITS
    • Potential for refactoring in difx2fits, difx2mark4
    • Allowing correctly formatted clock models and EOPs in vex file
    • Overlapped FFTs - possibilities and priority
  • Wednesday lunch in Auditorium (reconfiguration of tables needed 12:00-1:00)
    • All workshop attendees should feel free to attend and introduce themselves
  • Hack-a-thon, with initial discussion covering:
    • Auditing/refactoring of amplitude scaling
    • Cleaning up SVN
    • Improving the runtime predictor
  • Tour of electronics lab
  • Hack-a-thon
  • BBQ at Walter's, 6:30 PM (tentative)

Day 4:

  • Doc-a-thon, with focus on:
    • Getting better description of mpifxcorr internals, including buffer lengths and layouts, onto the wiki and easier to find
    • Getting Walter's DIFX User's Guide onto svn (versioned) and the wiki (latex2html?)
    • Man pages for the various DiFX programs?
    • Doxygen output for mpifxcorr (first complete all methods, then put somewhere useful too)
    • Plus a list made at the end of the session of the remaining documentation to be done
  • Hack-a-thon, with initial discussion covering:
    • Dealing with IPP7
    • Completing VDIF support
    • Should we support K5
    • Should we seriously look at only reading input files at manager and transmitting to other nodes

Day 5: (Friday Nov. 19)

  • Hack-a-thon
  • Colloquium in Auditorium (reconfiguration of tables needed 10:30 to 12:15)
    • Narrowband Imaging Observations of Lyman-Continuum Radiation at z~3 (Alice Shapley)
  • Hack-a-thon
  • Discussion of next meeting
difx/socorro2010/agenda.txt · Last modified: 2010/11/16 13:27 by walterbrisken