Casa Parallelization meeting minutes

Monday December 17th Room 317, 3:30pm MDT

Attendees:

Targets/task

• Large channel count cube imaging
  • Brief discussion about how best to address MS tile shape.
  • Justo and Sandra are working on to be executed post calibration to do for 4.1
• Parallelization of MS-MFS major cycle
  • Urvashi is working on to do for 4.1 with more complete solution after 4.1
  • Will include A-projection added to MS-MFS
• Parallel pipeline (possibly out of scope for this group) (drop from here).
  • Discuss what if any details need to be addressed by this group
• Filler
  • Discussion of lazy filler potentially available by end of january.
  • Status and future plans discussion
• Remaining partition non-compliant tasks
  • Status update on tasks that still need work to be partition compliant
  • cvel, split, hanning-smooth being worked on. Targetted for 4.1
• Resource requirement heuristics by task (imaging first) targetted for 4.1
  • Discussion of how to proceed with making task resource (memory, number of cores etc) aware.
  • Kumar working on helper task
  • Looking at thread locking issues
• Other
  • Parallelize Cube imaging scatter gather (low level target for 4.1)
    • More consideration needed, possibly virtually combine planes
  • Expand threaded gridder case to >4 threads (not a 4.1 deliverable)
    • Sanjay looking at LOFAR approach

Deferred Items