What is it about?
In our study, we looked at how a specific type of computer chip called an FPGA can help with processing complex network diagrams (graphs) more efficiently. FPGAs are great for tasks that require handling many operations at once, and they use less energy compared to other types of chips, which is especially useful in scientific research.
Featured Image
Photo by GuerrillaBuzz on Unsplash
Why is it important?
We focused on a common graph-processing task called Breadth-First Search, using a special set of tools provided by Intel, known as oneAPI, that helps make programming these chips easier. We aimed to make this task run faster and more efficiently on the FPGA.
Perspectives
Initially, we made some standard improvements recommended by Intel and used a tool that helps manage data better to speed up processing. This first attempt made the chip perform better in some cases but not as well as other methods that use different programming approaches or tools. So, we then tried a more detailed approach by organizing the data in a way that takes up less space and managing how data flows through the chip more tightly. This second method proved to be much more effective, making our FPGA perform as well as or better than other approaches in most of the tests we conducted. Overall, our research shows that with the right techniques, FPGAs can be a powerful tool for handling complex graph-processing tasks quickly and efficiently, which could be beneficial in various scientific and technological applications.
Kaan Olgu
University of Bristol
Read the Original
This page is a summary of: Optimisation and Evaluation of Breadth First Search with oneAPI/SYCL on Intel FPGAs: from Describing Algorithms to Describing Architectures, April 2024, ACM (Association for Computing Machinery),
DOI: 10.1145/3648115.3648134.
You can read the full text:
Resources
Contributors
The following have contributed to this page
