Impact study of data locality on task-based applications through the Heteroprio scheduler

Bramas, Bérenger

doi:10.7717/peerj-cs.190

“…We provide a detailed example of an execution with Heteroprio in the “Heteroprio execution example”. In 2019, an enhancement has been brought to Heteroprio to take into account the data locality ( Bramas, 2019 ). The original version treats all workers of the same type exactly equally, which completely discards memory management and can lead to massive and sometimes avoidable data movement.…”

Section: Introductionmentioning

confidence: 99%

Automated prioritizing heuristics for parallel task graph scheduling in heterogeneous computing

Flint

¹

,

Paillat

²

,

Bramas

³

2022

PeerJ Computer Science

Self Cite

1

0

View full text Add to dashboard Cite

High-performance computing (HPC) relies increasingly on heterogeneous hardware and especially on the combination of central and graphical processing units. The task-based method has demonstrated promising potential for parallelizing applications on such computing nodes. With this approach, the scheduling strategy becomes a critical layer that describes where and when the ready-tasks should be executed among the processing units. In this study, we describe a heuristic-based approach that assigns priorities to each task type. We rely on a fitness score for each task/worker combination for generating priorities and use these for configuring the Heteroprio scheduler automatically within the StarPU runtime system. We evaluate our method’s theoretical performance on emulated executions and its real-case performance on multiple different HPC applications. We show that our approach is usually equivalent or faster than expert-defined priorities.

show abstract

“…This is a difficult problem, especially when using heterogeneous computing nodes as it cannot be solved optimally in general. Much research is continuously conducted by the HPC and the scheduling communities to provide better generic schedulers (Bramas, 2019a). The granularity issue is related to the size of the tasks.…”

Section: Introductionmentioning

confidence: 99%

Improving parallel executions by increasing task granularity in task-based runtime systems using acyclic DAG clustering

Bramas

¹

,

Ketterlin

²

2020

PeerJ Computer Science

Self Cite

View full text Add to dashboard Cite

The task-based approach is a parallelization paradigm in which an algorithm is transformed into a direct acyclic graph of tasks: the vertices are computational elements extracted from the original algorithm and the edges are dependencies between those. During the execution, the management of the dependencies adds an overhead that can become significant when the computational cost of the tasks is low. A possibility to reduce the makespan is to aggregate the tasks to make them heavier, while having fewer of them, with the objective of mitigating the importance of the overhead. In this paper, we study an existing clustering/partitioning strategy to speed up the parallel execution of a task-based application. We provide two additional heuristics to this algorithm and perform an in-depth study on a large graph set. In addition, we propose a new model to estimate the execution duration and use it to choose the proper granularity. We show that this strategy allows speeding up a real numerical application by a factor of 7 on a multi-core system.

show abstract

“…This is a difficult problem, especially when using heterogeneous computing nodes as it cannot be solved optimally in general. Much research is continuously conducted by the HPC and the scheduling communities to provide better generic schedulers [16]. The granularity issue is related to the size of the tasks.…”

Section: Introductionmentioning

confidence: 99%

Peer Review #2 of "Improving parallel executions by increasing task granularity in task-based runtime systems using acyclic DAG clustering (v0.1)"

2020

0

View full text Add to dashboard Cite

The task-based approach is a parallelization paradigm in which an algorithm is transformed into a direct acyclic graph of tasks: the vertices are computational elements extracted from the original algorithm and the edges are dependencies between those. During the execution, the management of the dependencies adds an overhead that can become significant when the computational cost of the tasks is low. A possibility to reduce the makespan is to aggregate the tasks to make them heavier, while having fewer of them, with the objective of mitigating the importance of the overhead. In this paper, we study an existing clustering/partitioning strategy to speed up the parallel execution of a task-based application. We provide two additional heuristics to this algorithm and perform an in-depth study on a large graph set. In addition, we propose a new model to estimate the execution duration and use it to choose the proper granularity. We show that this strategy allows speeding up a real numerical application by a factor of 7 on a multi-core system.

show abstract

Impact study of data locality on task-based applications through the Heteroprio scheduler

Cited by 8 publications

References 25 publications

Automated prioritizing heuristics for parallel task graph scheduling in heterogeneous computing

Automated prioritizing heuristics for parallel task graph scheduling in heterogeneous computing

Improving parallel executions by increasing task granularity in task-based runtime systems using acyclic DAG clustering

Peer Review #2 of "Improving parallel executions by increasing task granularity in task-based runtime systems using acyclic DAG clustering (v0.1)"

Contact Info

Product

Resources

About