ACM Transactions on Mathematical Software · 2009 · 58 citations · 22 references
ComputingComputational EngineeringComputer ScienceFault ToleranceParallel AlgorithmsParallel ComputingConcurrent ProgrammingData ParallelismDistributed Data ProcessingFault Tolerance In Parallel SystemsParallel Data StructuresScalable Algorithm DesignCheckpoint MechanismsCheckpointingCheckpointing SupportCheckpointsData Dependency ManagementDeterministic Global OptimizationDeterministic OptimizationFault Tolerance Mechanisms
VTDIRECT95 is a Fortran 95 implementation of D. R. Jones' deterministic global optimization algorithm called DIRECT , which is widely used in multidisciplinary engineering design, biological science, and physical science applications. The package includes both a serial code and a data-distributed massively parallel code for different problem scales and optimization (exploration vs. exploitation) goals. Dynamic data structures are used to organize local data, handle unpredictable memory requirements, reduce the memory usage, and share the data across multiple processors. The parallel code employs a multilevel functional and data parallelism to boost concurrency and mitigate the data dependency, thus improving the load balancing and scalability. In addition, checkpointing features are integrated into both versions to provide fault tolerance and hot restarts. Important algorithm modifications and design considerations are discussed regarding data structures, parallel schemes, error handling, and portability. Using several benchmark functions and real-world applications, the software is evaluated on different systems in terms of optimization effectiveness, data structure efficiency, parallel performance, and checkpointing overhead. The package organization and usage are also described in detail.
22
Layne T. Watson, Maria Sosonkina, Robert Melville et al. · ACM Transactions on Mathematical Software · 1997 · 167 citations · Full text