POAS: a framework for exploiting accelerator level parallelism in heterogeneous environments

Martínez Sánchez, Pablo Antonio; Bernabé García, Gregorio; García Carrasco, José Manuel

Publication:
POAS: a framework for exploiting accelerator level parallelism in heterogeneous environments

dc.contributor.author	Martínez Sánchez, Pablo Antonio
dc.contributor.author	Bernabé García, Gregorio
dc.contributor.author	García Carrasco, José Manuel
dc.contributor.department	Ingeniería y Tecnología de Computadores
dc.date.accessioned	2024-04-08T10:52:06Z
dc.date.available	2024-04-08T10:52:06Z
dc.date.issued	2024-03-25
dc.description	© The Author(s) 2024. This manuscript version is made available under the CC-BY 4.0 license http://creativecommons.org/licenses/by/4.0/ This document is the Published Manuscript version of a Published Work that appeared in final form in The Journal of Supercomputing. To access the final edited and published work see https://doi.org/10.1007/s11227-024-06008-w
dc.description.abstract	In the era of heterogeneous computing, a new paradigm called accelerator level parallelism (ALP) has emerged. In ALP, accelerators are used concurrently to provide unprecedented levels of performance and energy efficiency. To reach that there are many problems to be solved, one of the most challenging being co-execution. In this paper, we present a new scheduling framework called POAS, a general method for providing co-execution to applications. Our proposal consists of four steps: predict, optimize, adapt and schedule. With POAS, an unseen application can be executed concurrently in ALP with little effort. We evaluate POAS on a heterogeneous environment consisting of CPUs, GPUs (CUDA cores), and XPUs (Tensor cores) on two different fields, namely linear algebra (matrix multiplication benchmark) and deep learning (convolution benchmark). Our experiments prove that POAS provides excellent performance and completes the tasks within a time very close to the optimal time for the hardware and applications used, with a negligible execution time overhead. Moreover, the POAS predictor performed exceptionally well, achieving very low RMSE values for both use cases. Therefore, POAS can be a valuable tool for fully exploiting ALP and improving overall performance over offloading in heterogeneous settings.	es
dc.embargo.terms	2025-03-25
dc.format	application/pdf	es
dc.format.extent	28	es
dc.identifier.citation	The Journal of Supercomputing, 2024
dc.identifier.doi	https://doi.org/10.1007/s11227-024-06008-w
dc.identifier.issn	Print: 0920-8542
dc.identifier.issn	Electrónic: 1573-0484
dc.identifier.uri	http://hdl.handle.net/10201/140581
dc.language	eng	es
dc.publisher	Springer
dc.relation	Open Access funding provided thanks to the CRUE-CSIC agreement with Springer Nature. Grant No. (TED2021-129221B-I00) funded by MCIN/AEI/10.13039/501100011033 and by the “European Union NextGenerationEU/PRTR,” and Grant No. (PID2022-136315OB-I00) funded by MCIN/AEI/10.13039/501100011033/ and by “ERDF A way of making Europe,” EU.	es
dc.rights	info:eu-repo/semantics/openAccess	es
dc.rights	Atribución 4.0 Internacional	*
dc.rights.uri	http://creativecommons.org/licenses/by/4.0/	*
dc.subject	High performance computing	es
dc.subject	Heterogeneous computing	es
dc.subject	Accelerator level parallelism	es
dc.subject	Scheduling	es
dc.subject	Co execution	es
dc.title	POAS: a framework for exploiting accelerator level parallelism in heterogeneous environments	es
dc.type	info:eu-repo/semantics/article	es
dspace.entity.type	Publication	es