GraphBLAST is a GPU implementation of GraphBLAS , an open standard for building blocks of graph algorithms. It gives data scientists without GPU programming experience the power to implement graph algorithms on the GPU. We are the leading graph framework in the following metrics: Performance : The first high-performance GPU implementation of GraphBLAS Composable : A library with building blocks for expressing most graph algorithms Concise : Single-source shortest path (SSSP) on GPU can be expressed in a mere 25 lines of code Innovative : Combines state-of-the-art graph optimizations from Gunrock with the direction-optimization heuristic of Ligra github地址： https://github.com/gunrock/graphblast