Use TriangleEdgeScore instead which is parallelized and has a similar performance even in the sequential case.
