BFS and Pagerank using SYCL for a GPU

 
Project Contacts: Gurbinder Gill (gill@cs.utexas.edu)

Project Description:
GPUs have become popular platform for improving the performance of graph analytical systems. CUDA is the popular language for developing applications for GPUs. However, this language is restricted to NVIDIA GPUs. Recently, a new language called SYCL [1] has been introduced that supports different kinds of GPUs as well as other accelerators. There are many graph analytics systems [2, 3] for GPUs that are implemented using CUDA. In this project, you will implement two graph analytical applications using SYCL for a single GPU: breadth first search (bfs) and pagerank (pr). You can look at existing C++ and CUDA implementations for these applications in the Galois repository [4]. The goal should be to match the performance of these CUDA applications on NVIDIA GPUs.

Hardware:

• Bridges cluster P100 nodes: 1 machines each with 2 NVIDIA P100 GPUs.
• Internal Machine located in UT called Tuxedo: 1 machine with 4 K80 GPUs and 2 GTX 1080 GPUs.

Suggested timeline:

• November 6: A clear description of your planned project.
• November 13: An understanding of the SYCL language and CUDA implementations of breadth first search and pagerank.
• December 6: An implementation of breadth first search and pagerank for a GPU using SYCL.
• December 6: A project report, written like an ACM conference paper, that summarizes your work.

Papers:

[1] https://www.khronos.org/sycl/

[2] Sreepathi Pai, Keshav Pingali: A compiler for throughput optimization of graph algorithms on GPUs. OOPSLA 2016: 1-19

[3] Roshan Dathathri, Gurbinder Gill, Loc Hoang, Hoang-Vu Dang, Alex Brooks, Nikoli Dryden, Marc Snir, Keshav Pingali, Gluon: a communication-optimizing substrate for distributed heterogeneous graph analytics. PLDI 2018: 752-768

[4] https://github.com/IntelligentSoftwareSystems/Galois