Hierarchical Scheduling of DAG Structured Computations on Manycore Processors with Dynamic Thread Grouping

Abstract

Abstract. Many computational solutions can be expressed as directed acyclic graphs (DAGs) with weighted nodes. In parallel computing, scheduling such DAGs onto manycore processors remains a fundamental challenge, since synchronization across dozens of threads and preserving precedence constraints can dramatically degrade the performance. In order to improve scheduling performance on manycore processors, we propose a hierarchical scheduling method with dynamic thread grouping, which schedules DAG structured computations at three different levels. At the top level, a supermanager separates threads into groups, each consisting of a manager thread and several worker threads. The supermanager dynamically merges and partitions the groups to adapt the scheduler to the input task dependency graphs. Through group merging and partitioning, the proposed scheduler can dynamically adjust to become a centralized scheduler, a distributed scheduler or somewhere in between, depending on the input graph. At the group level, managers collaboratively schedule tasks for their workers. At the within-group level, workers perform self-scheduling within their respective groups and execute tasks. We evaluate the proposed scheduler on the Sun Ultra-SPARC T2 (Niagara 2) platform that supports up to 64 hardware threads. With respect to various input task dependency graphs, the proposed scheduler exhibits superior performance when compared with other various baseline methods, including typical centralized and distributed schedulers.

Memory aware load balance strategy on a parallel branch‐and‐bound application

“…In this case, recall that main memory is being shared. Nonetheless, this model does not consider memory size, and at the end, too many threads can be allocated to share the same cache, and as a consequence, the amount of cache miss might be increased. The importance of accurately representing the communication costs depending on the memory hierarchy regarding the evaluation carried out by Chai et al .…”

“…In this case, recall that main memory is being shared. Nonetheless, this model does not consider memory size, and at the end, too many threads can be allocated to share the same cache, and as a consequence the amount of cache miss might be increase [34,39]. The importance of accurately representing the communication costs depending on the memory hierarchy regarding the evaluation carried out by [34] on various applications, suggested that intra and inter-processor communication is as important as inter-machine communication, and data locality techniques that avoid memory contention must be designed to improve application performance.…”

Elastic Places

“…Xia et al [47] propose Dynamic Thread Grouping (DTG), one of the few schemes to support both automated grouping and automatic hardware partitioning. To achieve this, Xie et al rely on a DAG described as a vector of tasks with dependencies, which is then partitioned at runtime.…”