VEGAS Pulsar Mode Integration - Planning Page

Actions:

  • John Ford: Set up a telecon with Dave McMahon
  • All: Check for missing tasks; complete effort and duration estimates

Introduction

This page is intended for gathering plans related to the proposed VEGAS Pulsar Mode Integration.

The main activity is to add pulsar modes to VEGAS. A more detailed description of what is required is available here: VEGAS Pulsar Mode Integration - Initial Specifications.

Here are some hardware notes from Randy

Here in some information from Joe on software tasks and associated software effort estimates.

Context of this work

The majority of this work can be considered "VEGAS Continued Maintenance and Enhancement" (e.g. cleaning up the HPC code for better maintainability), or "GUPPI Continued Maintenance and Enhancement" (transferring the pulsar modes to more modern, maintainable hardware). Therefore we assume that this will be an internally-managed, GB Ops project. We acknowledge the need to keep track of effort expended, and top-level milestones (which might be reported up as part of the POP process).

Top Level Strategy

We propose that the two major developments of "a generic 1600MHz sampling approach with digital downconversion", and "HPC Clean-up / Merge" should proceed first. Each of these will have a prototyping phase.These can proceed independently from each other; existing interfaces will be maintained unchanged. At the same time, the VEGAS Manager can be upgraded to support coherent and incoherent pulsar modes. These three developments will be tested using internal/unit tests, then will be integrated together for extensive system-level tests.

I/O (lustre) and CPU Performance Issues should also be addressed early in the project.

Once this major development has taken place, we will add ancillary modes (such as FAST4K and FAST8K); improved status displays and so on.

1. Major tasks - Phase I (can be done in parallel)

1.1. Develop 1600 MHz sampling with tunable Digital Downconverters approach

  • Seems to be the most flexible approach, which also allows easier future upgrades (e.g. to Roach III).
  • Allows us to dispense with Analog Filter Rack
  • 800 / 400 / 200 / 100 MHz bandwidth modes delivered simply by adjusting filter coefficients.
  • Will be based initially on a VEGAS L1 mode.

  • Subtasks:
    1. Prototype Approach (Who: ? Effort: fte-weeks Elapsed Time: fte weeks)
    2. Compare required effort against "rebuild for 400 MHz and decimate data" approach (Who: ? Effort: fte-weeks Elapsed Time: fte weeks)
    3. Complete implementation of basic Coherent mode
    4. Complete implementation of basic Incoherent mode
    5. Add remaining Coherent modes
    6. Add remaining Incoherent modes

1.2. HPC Cleanup / Merge

  • We want this to be done first so that we do not need extensive retesting of without then with this.
  • We considered three approaches:
    1. Quick and dirty
    2. Clean implementation based on hashpipe
    3. Above plus actively try to support other existing developments.
  • We decided to go with approach 2. If this is done well, then others will take our work and continue to extend it.
  • We noted Dave McMahon as a key player; we should talk to him as soon as possible: Action John Ford to arrange.
  • May wish to discuss further at the CASPER Workshop in Berkeley in June.

  • Subtasks:
    1. List of tasks from Joe's document.
    2. Need to add task - liaise / coordinate with others who are working with the HPC code
    3. Need to add task "documentation" (may be performed by Berkeley)

1.3. Address I/O Performance Issues

  • Wolfgang has set up different options for lustre striping

  • Subtasks:
    1. Subtask One
    2. Subtask Two
    3. etc

1.4. Address CPU performance Issues

  • Subtasks:
    1. Subtask One
    2. Subtask Two
    3. etc

1.5. Develop VEGAS Manager / Config_Tool to support additional pulsar modes.

  • Initially, continue to use the shared memory FITS keyword=value interface
  • This would be released with the merged HPC code
  • Could be tested with existing pulsar / HPC code, since interfaces will be the same.

  • Subtasks:
    1. VEGAS Manager:
      1. Develop MR
      2. Implement a CoherentBackend Class
      3. Implement an IncoherentBackend Class
      4. Add Parameters Unique to Pulsar Modes
      5. Add Samplers for Pulsar BOF Signal Level
      6. Update VEGAS Coordinator
    2. Config_Tool
      1. Develop MR
      2. Specify how to run GUPPI and VEGAS in parallel
      3. Implement ability to configure for VEGAS Pulsar Modes
      4. Implement ability to configure for VEGAS and GUPPI in parallel
    3. IF Manager
      1. Determine if any modifications are required
    4. Anything else?

2. Major Tasks - Phase II (Initial Commissioning)

  • We need to spot-check all of the VEGAS modes to confirm the changes have not broken anything, and perform a thorough check of all of the pulsar modes to ensure they are working correctly.

2.1 (Re-) Commissioning Spectral Line Modes

  • subtasks:
    1. Spectral Line Modes
      1. n tests x m hours per test x 29 modes = q hours of testing
      2. one maintenance day and one astronomy session per week = 29 weeks of elapsed time.

2.2 Commissioning Pulsar Modes

  • subtasks:
    1. Pulsar Modes
      1. do similar calculations to estimate effort and elapsed times.

3. Major Tasks - Phase III (Post Commissioning Development)

3.1 Detailed VEGAS / GUPPI Intercomparisons

  • subtasks
    1. VEGAS / GUPPI Intercomparisons
      1. Start comparing VEGAS and GUPPI data and timing results.

3.2. Develop and release 100 MHz Bandwidth FAST4K and FAST8K modes

  • subtasks
    1. Develop firmware
    2. Extend IncoherentBackend class to support these modes
    3. Extend Config_Tool to support these modes
    4. Commission new modes
    5. Anything else?

3.3. Revise Status Memory Interface to HPC

Update the shared memory status/con guration keyword interface with an interface to the value-keyword 'service'. (The service being hosted in the HPC program.)
  • subtasks
    1. Task One
    2. Task Two
    3. Etc

3.4. Updates to Real-Time-Displays

Notes on VEGAS pulsar real-time displays by Paul Demorest

  • subtasks
    1. Task One
    2. Task Two
    3. Etc

3.5. Develop and release 400, 200 and 100 MHz Incoherent and Coherent Modes

  • subtasks
    1. Task One
    2. Task Two
    3. Etc

3.6. Register naming conventions

We have decided not to re-write existing code. However, any new development will be compatible with VEGAS Spectral Line Modes.

Other areas for which we have not yet made effort estimates

Proposal Submission Tool

User Documentation

CLEO application

What Else?

Open Questions

  • Where do we store VEGAS Pulsar data?
  • What if anything do we do about VEGAS Pulsar data archiving

Topic attachments
I Attachment Action Size DateSorted ascending Who Comment
VegasPulsarIntegrationV4.pdfpdf VegasPulsarIntegrationV4.pdf manage 157 K 2014-03-11 - 15:10 RichardPrestage VEGAS Pulsar Mode Integration - Initial Specifications
Implementing_Pulsar_Modes_in_VEGAS.pdfpdf Implementing_Pulsar_Modes_in_VEGAS.pdf manage 63 K 2014-03-11 - 15:11 RichardPrestage Randy's hardware notes
tasks.pdfpdf tasks.pdf manage 30 K 2014-03-11 - 15:11 RichardPrestage Joe's software tasks
vp_estimates.pdfpdf vp_estimates.pdf manage 33 K 2014-03-11 - 15:12 RichardPrestage Joe's software effort estimates
This topic: CICADA > WebHome > VegasPulsarModeIntegration
Topic revision: 2014-03-12, RichardPrestage
This site is powered by FoswikiCopyright © by the contributing authors. All material on this collaboration platform is the property of the contributing authors.
Ideas, requests, problems regarding NRAO Public Wiki? Send feedback