https://hgpu.org/?p=6118
Parallel implematation of flow and matching algorithms