Algorithm-based fault-tolerant programming in scientific computation on multiprocessors

EEcient parallel algorithms proposed to solve many fundamental problems in scientiic computation are sensitive to processor failures. Because of its low costs, algorithm-based fault tolerance i s a n i n t e r esting concept for introducing fault tolerance into existing multi-processors. To facilitate fault{tolerant programming in scientiic computation, we have modiied and developed further an existing parallel run{time environment. In this paper the aspect of tuning known error processing techniques to the algorithm{based approach is primarily examined. Design issues for implementation and execution time overhead of a fault{tolerant application in our run{time environment are s t u d i e d. In contrast to many other environments for parallel fault{ tolerant programming, which use the master/slave programming model, our environment enables one to add fault tolerance to existing parallel applications in sci-entiic computation.