GAP

The Berkeley Graph Algorithm Platform (GAP) Project spans the entire stack, and it aims to accelerate graph algorithms through software optimization and hardware acceleration.

Graph algorithms are becoming increasingly important, with applications covering a wide range of scales. Warehouse-scale computers use graph algorithms to reason about vast amounts of data for analytics and recommendation systems. Mobile clients use graph algorithms as important components of recognition and machine-learning applications.

Some of the outcomes from the GAP project so far are:

Direction-Optimizing Breadth-First Search - fastest BFS for low-diameter graphs
GAP Benchmark Suite - standardize evaluations with specification + reference code
GBSP - graph DSL embedded in Python that achieves native performance thanks to SEJITS