Accelerating Astronomical Applications

FPGAs are processors that work very differently from the CPUs we are used to in ordinary computers. CPUs execute instructions (software) to perform a certain task; the hardware is fixed but the software offers a high degree of flexibility to perform all kinds of tasks.

With FPGAs, the hardware is still "malleable": in order to perform a certain task, all kinds of pieces of hardware on the FPGA are tied together in such a way that together they form a pipeline in which input data goes in and processed data comes out. The big advantage over ordinary CPUs is that they are much more energy-efficient; the disadvantage is that they are less flexible and more difficult to program.

New FPGA technologies are at the heart of this project. A high-level programming language (OpenCL), hardware support for floating-point numbers and close integration with CPUs will not only make it possible to program FPGAs much easier than before, it will also become possible to use them for much more complex applications than was previously possible. Intel wants to use this OpenCL/FPGA solution in data centres and for the IoT. In this project, we will develop a number of (radio-astronomic) applications for FPGAs.

The goals of this project are:

  • To master, evaluate and compare these technologies with other technologies (e.g. GPUs) with respect to performance, energy efficiency and programming effort;
  • To introduce these technologies in radio astronomy so that future projects can benefit from reduced programming effort and high energy efficiency.

Invited talks

  • John W. Romein, Accelerating Radio Astronomy (seminar), Vrije Universiteit, Amsterdam, October 11, 2017
  • Atze v.d. Ploeg, Bram Veenboer, and John W. Romein, Using FPGAs for HPC (poster), Netherlands eScience Symposium, Amsterdam, the Netherlands, October 18, 2017
  • John W. Romein, Bram Veenboer, and Atze v.d. Ploeg, Can FPGAs compete with GPUs?, GPU Technology Conference, San Jose, CA, March 26-29, 2018
  • John W. Romein, Bram Veenboer, and Atze v.d. Ploeg, Experiences with the Intel OpenCL/FPGA toolkit, BoF on Reconfigurable Computing, ISC'18, Frankfurt/Main, Germany, June 25-27, 2018
  • John W. Romein and Bram Veenboer, Can FPGAs compete with GPUs?, GPU Technology Conference, San Jose, CA, March 18-21, 2019 (https://developer.nvidia.com/gtc/2019/video/s9338)
  • John W. Romein, FPGA Programming in a High-Level Language, ASTRON TechnoLunch, Dwingeloo, the Netherlands, May 21, 2019
  • Johan Hidding, Merijn Verstraaten, Ben van Werkhoven, Bram Veenboer, Daniël v.d. Schuur, and John W. Romein, Using FPGAs for HPC (poster), Commit2Data, October 31, 2019
  • John W. Romein, High-Performance Computing in Radio Astronomy (seminar), Norwegian University of Life Sciences, Ȧs, Norway, January 21, 2020
  • John W. Romein, Exploring new GPU and FPGA Technologies for Radio Astronomy (colloquium) Radio Camera Initiative, online, September 11, 2020 (https://www.radiocamera.io/seminars/exploring)
  • John W. Romein, Technological Innovatons in the Processing of Radio-Astronomical Data, Dutch Hardware Acceleration Event, online, September 2020
  • John W. Romein, Bram Veenboer, Reinier v.d. Walle, Experiences with Programming FPGAs in a High-Level Programming Language (ASTRON TEASER talk), online, November 17, 2020
  • John W. Romein, Johan Hidding, Alessio Sclocco, Bram Veenboer, Merijn Verstraaten, Reinier v.d. Walle, Ben van Werkhoven, Experiences with the Intel OpenCL/FPGA Programming Environment, CASPER Workshop, online, May 19, 2021 (https://www.youtube.com/watch?v=fevhSBnDH8U @ 27'13'')
Project Team

Subcribe and stay informed about all our researchprojects and achievements

Publications

Radio-Astronomical Imaging: FPGAs vs GPUs, Euro-Par'19

15 December 2020

Radio-Astronomical Imaging on Accelerators

Radio Astronomy; Imaging; Algorithms; High-performance Computing; Graphics Processors; Field-Programmable Gate Arrays; Hardware/Software codesign; Performance optimization; Green computing

15 December 2020

News

Real-Time Data for Products to Move Data-Driven Real-Time Decision Making in Supply Chains and Logistics
The project, here referred to as Data2Move, was started in 2017. The goal of Data2Move is to develop and demonstrate real-time data-driven logistic methods and techniques, initially focusing on inventory and transportation. Over the past years, signi...
27 November 2023
Virtual Interiors as Interfaces for Big Historical Data Research
The Semantic Web needs interfaces for critical, reliable analyses of Big Data for research in humanities, cultural heritage, and creative industries. Utilizing data on the production and consumption of cultural goods, geodata, maps, and building plan...
21 November 2023
How can mathematics solve your data science challenge? Sign up for the Study Group
Mathematics provides a surprising amount of techniques to solve challenges in data science. Does your company have such challenges but no time to look into them? Or would you like to strengthen your contacts in the academic world? Would you like to m...
17 November 2023