Uploaded image for project: 'SAFe Program'
  1. SAFe Program
  2. SP-1411

Enhance the GPU imaging/gain calibration pipeline to support visibility weighting schemes

Change Owns to Parent OfsSet start and due date...
    XporterXMLWordPrintable

Details

    • Enabler
    • Not Assigned
    • PI9
    • COM SDP SW
    • None
    • Data Processing
    • Hide

      The GPU imaging/gain calibration pipeline currently supports only natural weighting, resulting in high side lobes, and poor resolution. Including uniform weighting and the hybrid Briggs would improve the range of image tests that can be performed by end users.

      Show
      The GPU imaging/gain calibration pipeline currently supports only natural weighting, resulting in high side lobes, and poor resolution. Including uniform weighting and the hybrid Briggs would improve the range of image tests that can be performed by end users.
    • Hide

      TBD.

      Show
      TBD.
    • 3
    • 3
    • 2
    • Team_NZAPP
    • Sprint 5
    • Hide

      All tasks outlined for this work have been successfully completed.

      [NZAPP-285] Performed research and reading up on visibility weighting schemes. Useful sources included CASA imaging documentation, Dan Briggs PhD Thesis, and RASCIL source code.

      [NZAPP-286] Implemented visibility weight buffers within the imaging pipeline. At present these are stored with a configurable resolution of single/double precision. We would like revisit this in the future, to investigate the use of byte sized resolution for ingest weights (from correlator). Computation of adjusted weights (ie: uniform, robust) will remain single/double via compiler C define flag - https://gitlab.com/ska-telescope/sdp/sep_pipeline_imaging/-/blob/master/imaging/common.h#L91

      [NZAPP-289] Began parallelized implementation of Briggs/Robust visibility weighting. Robust weighting can be enabled via YAML configuration file by setting WEIGHTING_SCHEME to 2 (enum driven) and ROBUSTNESS to the desired range within [-2, +2] (emulates uniform, natural, or a mixture of the two).

      WEIGHTING_SCHEME: https://gitlab.com/ska-telescope/sdp/sep_pipeline_imaging/-/blob/master/imaging_configs/default.yaml#L22

      ROBUSTNESS: https://gitlab.com/ska-telescope/sdp/sep_pipeline_imaging/-/blob/master/imaging_configs/default.yaml#L26

      [NZAPP-304] Completed parallelized implementation of Briggs/Robust visibility weighting. Ironed out remaining bugs. Preliminary results compared to traditional natural/uniform weighting can be found on tickets [NZAPP-287] and [NZAPP-288]. The bulk of the source code for weighting functionality can be found here: https://gitlab.com/ska-telescope/sdp/sep_pipeline_imaging/-/blob/master/imaging/src/weighting.cu

      [NZAPP-287] Implemented and completed parallelized implementation of natural visibility weighting (default weights from correlator). Can be enabled using a WEIGHTING_SCHEME of 0, ROBUSTNESS has no effect on this scheme. See ticket for more information/graphs/images.

      [NZAPP-288] Implemented and completed parallelized implementation of uniform visibility weighting (adjusts weights from correlator by density of visibilities in the UV-plane). Can be enabled using a WEIGHTING_SCHEME of 1, ROBUSTNESS has no effect on this scheme. See ticket for more information/graphs/images.

      Show
      All tasks outlined for this work have been successfully completed. [NZAPP-285] Performed research and reading up on visibility weighting schemes. Useful sources included CASA imaging documentation, Dan Briggs PhD Thesis, and RASCIL source code. [NZAPP-286] Implemented visibility weight buffers within the imaging pipeline. At present these are stored with a configurable resolution of single/double precision. We would like revisit this in the future, to investigate the use of byte sized resolution for ingest weights (from correlator). Computation of adjusted weights (ie: uniform, robust) will remain single/double via compiler C define flag - https://gitlab.com/ska-telescope/sdp/sep_pipeline_imaging/-/blob/master/imaging/common.h#L91 [NZAPP-289] Began parallelized implementation of Briggs/Robust visibility weighting. Robust weighting can be enabled via YAML configuration file by setting WEIGHTING_SCHEME to 2 (enum driven) and ROBUSTNESS to the desired range within [-2, +2] (emulates uniform, natural, or a mixture of the two). WEIGHTING_SCHEME: https://gitlab.com/ska-telescope/sdp/sep_pipeline_imaging/-/blob/master/imaging_configs/default.yaml#L22 ROBUSTNESS: https://gitlab.com/ska-telescope/sdp/sep_pipeline_imaging/-/blob/master/imaging_configs/default.yaml#L26 [NZAPP-304] Completed parallelized implementation of Briggs/Robust visibility weighting. Ironed out remaining bugs. Preliminary results compared to traditional natural/uniform weighting can be found on tickets [NZAPP-287] and [NZAPP-288] . The bulk of the source code for weighting functionality can be found here: https://gitlab.com/ska-telescope/sdp/sep_pipeline_imaging/-/blob/master/imaging/src/weighting.cu [NZAPP-287] Implemented and completed parallelized implementation of natural visibility weighting (default weights from correlator). Can be enabled using a WEIGHTING_SCHEME of 0, ROBUSTNESS has no effect on this scheme. See ticket for more information/graphs/images. [NZAPP-288] Implemented and completed parallelized implementation of uniform visibility weighting (adjusts weights from correlator by density of visibilities in the UV-plane). Can be enabled using a WEIGHTING_SCHEME of 1, ROBUSTNESS has no effect on this scheme. See ticket for more information/graphs/images.
    • 14.2
    • Outcomes Reviewed
    • PI24 - UNCOVERED

    Description

      This feature will add CUDA-implemented visibility weighting schemes to the GPU imaging/gain calibration pipeline.

      Attachments

        Issue Links

          Structure

            Activity

              People

                b.nikolic Nikolic, Bojan
                A.Ensor Ensor, Andrew [X] (Inactive)
                Votes:
                0 Vote for this issue
                Watchers:
                4 Start watching this issue

                Feature Progress

                  Story Point Burn-up: (100.00%)

                  Feature Estimate: 3.0

                  IssuesStory Points
                  To Do00.0
                  In Progress   00.0
                  Complete614.0
                  Total614.0

                  Dates

                    Created:
                    Updated:
                    Resolved:

                    Structure Helper Panel