Casa Parallelization meeting minutes

Monday May 16th Room 317, 3;30pm MDT

Attendees:

Targets/tasks for May 16th - May 23rd

• Examine async thread failure problem Jim
• Work toward implementing streaming framework in applycal Jim
• Refine and distribute cluster scheduler interface James
• Spectral serial improvements (avail mem, don't grid uncleaned channels) Kumar
• Schedule Parallel discussion/meeting Brian

Discussion/Agenda

• LOFAR visitors
  • Examining W and A projection suitability for LOFAR
  • Ger here to discuss
    • table mods for caching storage manager
    • Help with async issues.
    • Whether to change auxiliary tables from standard to AIPSio by default
    • Build / compiler issues (talk with Scott Rankin)
• async status Jim
  • serializing/deserializing objects: decouple them such that destruction is safe
  • Ger may have a solution
  • Build against Ger's checkin's planned for last week had no effect
• Imaging issues
  • Any outstanding problems regarding continuum multi-ms imaging
  • Time to revisit spectral imaging
    • Serial improvements
    • Discussion about interaction with split and spectral windows
• Parallel gridder status Sanjay
  • Should work with various gridder algorithms
    • error in distributing grids to cores, results in missing data fixed
  • Inner loop vs outer loop performance
  • C++ vs Fortran performance, possibly a compiler issue
    • May just be a reflection of GCC optimization flags, test against gcc 4.6.0
  • Need code snippet to show to some external folks and possibly give to nVidia for GPU purposes (memory impact)
• Examine streaming framework design Jim
  • Continue to work on in parallel with async split
  • Apply new framework to applycal
    • Main hotspot is ordering of two iteration loops, gaincal and spectral windows triggers
    • Week or two for initial tests
• Complete parallel user interface frame work Jeff/Honglin
• Cluster scheduler scripting with CV James
  • Clean up and make available James
  • Focusing on inhouse script for now
• Test async performance against parallel continuum case James
  • Need to test vs various hardware configs
  • Need 300+GB continuum case with complex field
    • Bryan will look into observing (spectral line and extended emission) IRC216
• General discussion on what's next

Older Issues

• Filling I/O issues update
  • Michel is optimistic of having a streamed I/O approach and use sub-ms infrastructure
  • Waiting on Francois to change data access methods
• Plot-ms suffers similar problem to split but needs lower level solution Ger
• Timing scripts
  • Large spectral (Bryan talked with Claire, Miriam available)
    • Less interesting in absence of parallel cube imaging
• Time domain partitioning with multiple MSes per imaging process Kumar
  • Cube case regression status
  • Trying to identify 10% small reads, possibly metadata column (either main or sub table)