Uploaded image for project: 'SAFe Program'
  1. SAFe Program
  2. SP-3804

Develop the first version of the PST Flow Through mode in the digital signal processing component

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

Details

    • Feature
    • Should have
    • PI21
    • COM PST SW
    • None
    • Data Processing
    • Hide

      See Why? in description

      Show
      See  Why?  in description
    • Hide

      See Success Metrics in description

      Show
      See  Success Metrics  in description
    • 5
    • 5
    • 0
    • Team_PST
    • Sprint 5
    • Hide

      All of the above functionality has been released as ska-pst version 0.11.0 via REL-1245.

      Show
      Reviewed the Design for AA1 Architecture of PST for supporting multiple applications and libDSPSR integration. Defined the Flow Through mode Monitoring Points to be implemented. Added Flow Through mode to the Configure Scan schema in ska-telmodel. Added Flow Through mode to the PST Beam TANGO device Rationalised the PST monitoring and control architecture , combining the 4 sub-component TANGO devices into single Beam device Developed the skeleton for the C++ Signal Processing Pipeline Monitoring and Control functionality developed, from the protobuf message defintions to the application manager that implements the PST state model . Library and Application architecture fleshed out Developed BDD tests that exercise the happy-path scan life cycle of Flow Through mode: XTP-22309 PST can be configured to perform a scan. XTP-22311 PST can perform a scan XTP-22322 PST can stop a scan XTP-22313 PST can return to the Idle State Added CUDA / GPU support to PST Configuration to use the Gitlab GPU runners (V100 and A100 @ STFC) Migrated PST containers to use Nvidia development and runtime base images Add chart configurations to reserve GPU resources of k8s nodes Added CUDA support to the ska-pst-docker image that enables offline signal processing Developed proof of concept for integrating libdspsr with PST’s DSP package. All of the above functionality has been released as ska-pst version 0.11.0 via REL-1245 .
    • 21.6
    • Stories Completed, Solution Intent Updated, Outcomes Reviewed, Demonstrated, Satisfies Acceptance Criteria, Accepted by FO
    • PI22 - UNCOVERED

    • Low G4

    Description

      https://confluence.skatelescope.org/display/SE/Develop+the+first+iteration+of+the+PST+%27Flow-through%27+mode+in+the+digital+signal+processing+component 

      Who?

      • SKA software integrators; Commissioning scientists

      Why?

      • In preparation for testing the Flow Through, dynamic spectrum and pulsar timing modes of PST, support for GPU processing needs to be implemented. Early implementation of the GPU infrastructure will enable testing of GPU code with Gitlab runners in a Continuous Integration environment, which will retire the risk of a correctly functioning DSP component in PST, ahead of deployment at AA1 for commissioning tests on astronomical objects.
      • Typical tied-array beam observations at the full temporal resolution and over the full frequency band generate large amounts of data. The PST Flow Through mode is a useful implementation that allows a commissioning scientist to record only a configurable portion of the band and with reduced bit-depth per sample. The ability to record a compressed version of the data is critical for reducing the turn-around time between obtaining and processing the data of an observation, and shortens the testing-learning-development cycles during telescope commissioning. Early commissioning tests of long-track pulsar observations, drift-scans of bright sources, as well as investigations of low-level RFI in the tied-array beam will benefit from such an implementation.

      What?

      • Many of the real-time processing functions of PST will run on GPU in the digital processing component (DSP). Before implementing those functions, the supporting infrastructure for GPU processing needs to be built into DSP. This Goal will deliver the building blocks for GPU processing in PST, which is essential in enabling future implementation of (a) the PST Flow Through mode, (b) dynamic spectrum recording mode, (c) pulsar timing mode and (d) VLBI processing.

      The PST Flow Through mode enables the compression and archiving of a tied-array beam observation of complex voltages by re-sampling the data to a configurable bit-depth. It also enables the operator to select only a portion of the frequency band for archiving. The Flow Through mode is developed on GPU and is part of the Digital Signal Processing component of PST. This mode is important for commissioning and verification tests with PST in early Array Assemblies, when in an observation of an astronomical object only a small data set is needed for the verification of telescope functionality and tied-array beamforming performance.

      Success Metrics

      • Review the design and existing code for the Flow Through mode and agree on the implementation details (platform considerations, etc.).
      • Update the telescope model to support Flow Through
      • Define the BDD tests for the Flow-through mode
      • 'Flow through' control functionality (C++ API) is integrated into the DSP application
      • PST can be configured via PST BEAM for a 'Flow through' mode observation (TANGO device)
      • The Flow through mode code can be executed on GPU runners in Gitlab CI

      Attachments

        Issue Links

          Structure

            Activity

              People

                A.Noutsos Noutsos, Aristeidis
                A.Noutsos Noutsos, Aristeidis
                Votes:
                0 Vote for this issue
                Watchers:
                2 Start watching this issue

                Feature Progress

                  Story Point Burn-up: (100.00%)

                  Feature Estimate: 5.0

                  IssuesStory Points
                  To Do00.0
                  In Progress   00.0
                  Complete2241.5
                  Total2241.5

                  Dates

                    Created:
                    Updated:
                    Resolved:

                    Structure Helper Panel