We chose to evaluate three applications from the benchmark suite. They are Breadth-First Search, Needleman-Wunsch, and Streamcluster. These were selected as they cover a broad application space and are diverse in their execution behaviours.
Breadth-First Search is a graph traversal algorithm that is widely used in many disciplines and application domains, and is also part of the graph500 set of benchmarks. Needleman-Wunsch is a bioinformatics workload and follows a dynamic programming algorithm: it is a non-linear global optimisation method for DNA sequence alignments and works on a 2D matrix. Streamcluster is a dense linear algebra algorithm used in data mining applications.
These benchmarks will be used with different implementations and systems to evaluate how different systems perform under the diverse workloads represented by the suite. Adept will optimise and streamline the benchmark codes and investigate how to improve the efficiency of these types of workload.