scientific article; zbMATH DE number 15287
aggregationautomated problem mappingdirected acyclic graphEncore Multimaxmappingparallel architecturespartitioningperformanceschedulingshared memory architecturesvery sparse triangular system
Graphs and linear algebra (matrices, eigenvalues, etc.) (05C50) Direct numerical methods for linear systems and matrix inversion (65F05) Computational methods for sparse matrices (65F50) Parallel numerical computation (65Y05) Numerical algorithms for specific classes of architectures (65Y10) Complexity and performance of numerical algorithms (65Y20) Graph theory (including graph drawing) in computer science (68R10)
[For the entire collection see Zbl 0694.00025.] The author investigates methods appropriate for the aggregation, mapping and scheduling of relatively fine grained computations specified by a directed acyclic graph. A framework for partitioning very sparse triangular systems of linear equations that is designed to produce favourable performance results in a wide variety of parallel architectures is presented. Simple expressions are givenen that describe how to schedule computational work with varying degree of granularity. The Encore Multimax is used as a hardware simulator to investigate the performance effects of using partitioning techniques in shared memory architectures with varying relative synchronization costs.
- Parallel mapping of sparse computations
- scientific article; zbMATH DE number 4055009 (Why is no real title available?)
- Rescheduling support for mapping dynamic scientific computation onto distributed memory multiprocessors
- scientific article; zbMATH DE number 934541 (Why is no real title available?)
- A layered method for process and code mapping
This page was built for publication:
Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q3977609)