Newer versions of CASA allow some amount of parallelization to speed up some parts of processing. If we could offer multi-core containers for these newer versions of CASA, users could see a significant speed-up in some of their work, which would be an attractive feature of ARCADE. This may require more effort to enable than available for even the beta prototype, so I've also noted this on our longer-term goals document. A few relevant notes for implementation:
CASA 5.4.0 is the first version where parallelization was implemented for imaging. As of the most current CASA version (5.6.2), parallelization for calibration remains possible but not recommended (untested). So, for now, we can expect that users would only be interested in parallelization for imaging. We would be looking to implement multi-core containers for CASA versions 5.4 and higher (only).
According to help documents that I have seen, no additional installation steps are required to implement parallelization in CASA. Instead, one simply needs to initiate casa with the command ‘mpicasa’ instead of 'casa', and when running tclean for imaging, include the setting ‘parallel=True’. That being said, I have tried this with the CASA5.4.0 version installed on ARCADE under demo.canfar.net and it crashes when I type 'mpicasa'. So we might need some time to dig in to installation issues more than advertised
Newer versions of CASA allow some amount of parallelization to speed up some parts of processing. If we could offer multi-core containers for these newer versions of CASA, users could see a significant speed-up in some of their work, which would be an attractive feature of ARCADE. This may require more effort to enable than available for even the beta prototype, so I've also noted this on our longer-term goals document. A few relevant notes for implementation: