12.07.2015 Views

Mapping and Scheduling of Parallel Tasks for ... - ComplexHPC.org

Mapping and Scheduling of Parallel Tasks for ... - ComplexHPC.org

Mapping and Scheduling of Parallel Tasks for ... - ComplexHPC.org

SHOW MORE
SHOW LESS

Create successful ePaper yourself

Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.

The Mixed <strong>Parallel</strong> Task (M-Task) Programming ModelProgramming Support M-Task Programming<strong>Mapping</strong> on Multicore Plat<strong>for</strong>msSummaryThe M-Task Programming ModelThe Programming Model<strong>Scheduling</strong>Programming ModelApplication consists <strong>of</strong> mixed parallel tasks (M-<strong>Tasks</strong>).M-<strong>Tasks</strong> can be:basic (dataparallel implementations), e.g. dataparallel matrix multiplication,orcomplex (consist <strong>of</strong> other M-<strong>Tasks</strong>), e.g. matrix multiplication by Strassen.Basic M-<strong>Tasks</strong> can have multiple implementations, e.g. due to differentdata distributions.M-<strong>Tasks</strong> can be executed in a taskparallel manner when no precedenceconstraints are violated.M-<strong>Tasks</strong> can run on an arbitrary number <strong>of</strong> processors.M-<strong>Tasks</strong> may have input <strong>and</strong> output parameters.Communication between M-<strong>Tasks</strong> through these parameters. Dependencies between M-<strong>Tasks</strong>.Data re-distribution operations are required ifData distribution changes from producing M-Task to consuming M-Task orProducing <strong>and</strong> consuming M-Task are executed on different process<strong>org</strong>roups.Gudula Rünger<strong>Mapping</strong> <strong>and</strong> <strong>Scheduling</strong> <strong>of</strong> <strong>Parallel</strong> <strong>Tasks</strong> <strong>for</strong> Multicore Systems5

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!