This benchmark counts the number of triangles in an undirected graph.
The algorithm is called extend-reduce. A initial worklist contains initial embedding (single edge). It computes the common neighbors of the two endpoints of each edge using set intersection, and accumulates the count to the global counter (total number) of triangles.
We present timing on NVIDIA V100 GPU (32G memory)
Please read the following paper for detailed performance evaluation:
Xuhao Chen, Roshan Dathathri, Gurbinder Gill, Keshav Pingali,
Pangolin: An Efficient and Flexible Graph Mining System on CPU and GPU,
PVLDB, 13(8): 1190-1205, 2020