Critical resources management and scheduling under StarPU

Arthur Chevalier 1, 2
1 STORM - STatic Optimizations, Runtime Methods
LaBRI - Laboratoire Bordelais de Recherche en Informatique, Inria Bordeaux - Sud-Ouest
2 HiePACS - High-End Parallel Algorithms for Challenging Numerical Simulations
LaBRI - Laboratoire Bordelais de Recherche en Informatique, Inria Bordeaux - Sud-Ouest
Abstract : The advanced users of StarPU have the ability to regulate the task submission to limit/control the consumption of critical resources (memory, network communication). This technique can penalize performance (parallelism deficit) and is not adapted to the concurrent management of several types of resources. The use of hierarchical tasks allow to refocus this management at the level of the scheduling engine: high-level tasks are used to express parallelism and resources requests. Once this graph is submitted, StarPU has the clairvoyance necessary to optimize the parallel execution of the application while respecting the constraints expressed.
Keywords : StarPU
Complete list of metadatas

Cited literature [8 references]  Display  Hide  Download

https://hal.inria.fr/hal-01718280
Contributor : Nathalie Furmento <>
Submitted on : Tuesday, February 27, 2018 - 11:42:56 AM
Last modification on : Tuesday, April 2, 2019 - 1:45:43 AM
Long-term archiving on : Monday, May 28, 2018 - 4:26:21 PM

File

Memoire.pdf
Files produced by the author(s)

Identifiers

  • HAL Id : hal-01718280, version 1

Citation

Arthur Chevalier. Critical resources management and scheduling under StarPU. Distributed, Parallel, and Cluster Computing [cs.DC]. 2017. ⟨hal-01718280⟩

Share

Metrics

Record views

181

Files downloads

168