The International Conference for High Performance Computing, Networking, Storage, and Analysis
The International Conference for High Performance Computing, Networking, Storage, and Analysis
Workshop:PyHPC 2020: 9th Workshop on Python for High-Performance and Scientific Computing
Authors: Ali Kazmi (Georgia Institute of Technology)
Abstract: Graph algorithms are a central part of many applications, and their speed is critical in real world use. Linear Algebraic formulations have been shown to provide speedups, and the GraphBLAS standard provides their building blocks.
GraphBLAS does not cover everything, though, and it can be difficult to formulate a problem in linear algebra. In such cases, we must implement graph algorithm code by hand. To investigate Python’s performance, I implemented Floyd Warshall’s algorithm in Python in two ways: Numba and Cython. I will compare these to more efficient C++/CUDA code, which was developed over several months by Vijay Thakkar, a graduate student in our lab.