Casa HPC/Parallelization meeting agenda/minutes

Thursday [03/03/2016], [ESO Pavo, A.2.01], [16:00 UT]

How to connect

• Dial in: +49 89 6834
• Video connection: Use this one (46101@134.171.42.27).
• Rooms:
  • CV: ER-209
  • Socorro:
  • ESO: Pavo

Attendees:

Agenda

1. Proposal for minor changes in flagdata when processing an MMS without MPI.
   • Now that simple_cluster is not used anymore, it will be faster if flagdata process an MMS as a monolithic MS instead of per SubMS if MPI is not available. Before, the task would decide to generate jobs for the SubMSs if the input was an MMS. Now, the decision should be made if MPI is enabled instead. The advantages are:
     • it is quicker to process an MMS as a monolithic MS. There is no overhead in creating jobs.
     • it is simpler to handle the writing of the HISTORY table (only MPI Client will write to HISTORY if running in parallel).
2. C++ and MPI integration in CASA.
   • Summary of Julian and Martin's tests.
     • We need to understand if with the current patch created by Julian, it is possible to implement Martin's code without any other implications in the Python side.
     • Is it possible to fulfill all imaging requirements with the current patch?
   • What to do next?

Minutes

1. Item

2. Item

3. Item

Action Item List

Useful Links

• Major blockers
• Current status
• CASA parallelization task list
• Long-term issue chart
• Bleeding tests area
• Documentation (see Reference Materials topic)