[Skip to Content]

Project area/S

  • Fast Transients and Pulsars
  • The Local Universe
  • Data Intensive Astronomy

Project Details

In astronomy, amazing things happen in the blink of an eye. Gamma-Ray Bursts, pulsars, and Fast Radio Bursts were all discovered by equipping sensitive telescopes with powerful high-resolution signal processors. The Murchison Widefield Array (MWA) telescope has a wide field of view, giving it tre-mendous capability for real-time searches. However, search-ing the MWA’s wide field of view is computationally prohibitive. Real-time searches for the MWA and other widefield tele-scopes will require a combination of new data processing hardware and signal processing algorithms optimised for speed and sensitivity.
In this internship, the student will work with Dr. Danny Price to bring Pawsey’s latest-generation hardware to bear on the problem. Specifically, the student will investigate the use of tensor cores — processing units available on NVIDIA graphics cards — for radio astronomy application. Tensor cores excel at matrix multiplication and are well suited radio astronomy data. The student will learn how tensor cores work, and how to tar-get their use in GPU codes. The student will then examine existing code from the MeerKAT telescope, which performs pair-wise correlation of signals from many antennas. The stu-dent will run this code on the Topaz supercomputer, and adapt them for the MWA. The primary output will be tensor core utilities that can be used in the BLINK PaCER program.

Student Attributes

Academic Background

Computer science, engineering or astrophysics background.

Computing Skills
Python, bash scripting, Linux environment. Experience with GPU programming preferred, C/C++ also desirable.

Training Requirement
The student should participate in the standard PAWSEY training usually provided to summer students as it always has very relevant set of topics covered.

Project Timeline

  • Week 1 Inductions and project introduction
  • Week 2 Initial Presentation
  • Week 3 Getting familiar with the data to be used in the project
  • Week 4 Learning about tensor cores and GPU programming
  • Week 5 Installation and testing of the tensor core correlator pipe-line.
  • Week 6 Running correlator on MWA/EDA data
  • Week 7  Running correlator and analysis of MWA/EDA data
  • Week 8 Making images from MWA correlated visibilities (using existing tools)
  • Week 9 Final Presentation
  • Week 10 Final Report


Dr Marcin Sokolowski

Research Fellow

Read More