arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 1906.00874cs.MScs.DCcs.NAmath.NA
收藏

在 $\mathcal{H}-LU$ 分解中利用嵌套任务并行性

Exploiting nested task-parallelism in the $\mathcal{H}-LU$ factorization

Rocío Carratalá-Sáez, Sven Christophersen, José I. Aliaga, Vicenç Beltran, Steffen Börm, Enrique S. Quintana-Ortí

AI总结:

本文针对边界元法产生的分层矩阵LU分解并行化问题,提出利用OmpSs模型在执行时发现数据流并行性,通过将数据结构与依赖检测解耦,并引入OmpSs-2的弱操作数和依赖提前释放机制,有效加速了嵌套细粒度任务的并行执行。

摘要:

我们探讨了源自边界元法的分层矩阵($\mathcal{H}$-matrices)LU分解的并行化。我们的方法通过 OmpSs 编程模型和运行时来利用任务并行性,该运行时通过基于任务操作数内存地址的数据依赖性分析,在执行时发现操作固有的数据流并行性。这对于 $\mathcal{H}$-matrices 尤其具有挑战性,因为包含数据的结构在执行期间维度会发生变化。我们通过将数据结构与用于检测依赖性的结构解耦来解决这个问题。此外,我们利用 OmpSs-2 中最近引入的弱操作数和依赖性提前释放支持,来加速具有嵌套任务并行性和细粒度任务的并行代码执行。

英文摘要:

We address the parallelization of the LU factorization of hierarchical matrices ($\mathcal{H}$-matrices) arising from boundary element methods. Our approach exploits task-parallelism via the OmpSs programming model and runtime, which discovers the data-flow parallelism intrinsic to the operation at execution time, via the analysis of data dependencies based on the memory addresses of the tasks' operands. This is especially challenging for $\mathcal{H}$-matrices, as the structures containing the data vary in dimension during the execution. We tackle this issue by decoupling the data structure from that used to detect dependencies. Furthermore, we leverage the support for weak operands and early release of dependencies, recently introduced in OmpSs-2, to accelerate the execution of parallel codes with nested task-parallelism and fine-grain tasks.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑