We define several generic cluster-computing algorithms, each of them encapsulating a general purpose control structure. By extending these generic algorithms with sequential code, we derive specific cluster-computing algorithms for diverse problems. We use a specification and prototyping language to test the validity of the derived parallel algorithms before finally converting them into efficient C code that runs in a cluster-computing environment, such as PVM. Our performance experiments demonstrate that highly generic cluster-computing algorithms can produce efficient cluster-computing applications.
展开▼