At Zenotech we are keen to collaborate with both academic and commercial partners to discover new applications for our technology and optimise processes. Therefore we were excited to be invited to work on a project with the Oxford e-Research Centre, which draws on its multidisciplinary expertise to harness innovative tech, striving to drive research in new directions.

We chose to approach the problem of peak finding within the data recorded from a radio telescope. This is part of the process in determining if there is an astronomical signal present in the data recorded. Last month James Sharpe, our Lead on Big Data and Security, gave a presentation on the positive outcome of the project, which resulted in a 3.5 times increase in the speed of data-processing.

The University of Oxford’s e-Research Centre is a multidisciplinary applied research department, developing and applying innovative computational and information technology in both academic research and industrial applications.

The e-Research Centre had the following aims from the collaboration:

• To assess the quality of work that can be delivered by the collaboration
• To assess the ability to get to grips with complex signal processing problems and associated algorithms, their software engineering skills and their ability to deliver highly optimised code.
• The collaboration’s ability to work as part of a wider team that might include commercial companies and academics.
• The collaboration’s ability to overcome differences in working methodologies and also differences in technical language used.

We completed implementation of the peak finding algorithm on an NVidia TitanX GPU, and this work was integrated into the open source AstroAccelerate library. This is a library of many-core accelerated real-time data processing modules, supporting multiple hardware architectures and providing automatic optimisation for the given system configuration.

The AstroAccelerate library is being used in the Oxford e-Research Centre project, “Pulsar Acceleration on the GPU for the Square Kilometre Array” (Sofia Dimoudi, Wes Armour, Karel Adamek) which is utilsiing the world’s largest telescope, the Square Kilometre array with a collecting area of 1 km2. The project has far-reaching scientific objectives, including discovering more about the first objects in the universe, galaxy evolution and Dark Energy.

The results

The project achieved a 3.5x speed up compared to the Nvidia NPP library, reaching a throughput of 138GB/s. The algorithm will be applied in the future via the AstroAccelerate library into the Square Kilometre array project where the library will be deployed to process radio signals in real-time.

If you want to find out more about the project, check out our presentation.