Heterogeneous MPI application optimization with ITAC

Autor: Vadim Karpusenko
Rok vydání: 2015
Předmět:
DOI: 10.1016/b978-0-12-802118-7.00025-x
Popis: Utilization of Intel Xeon Phi coprocessor within the heterogeneous cluster environment with Intel Xeon processors or other compute devices requires corresponding workload balancing, to prevent idling of the coprocessors. This chapter demonstrates static and dynamic load balancing techniques, which can be used to address this issue. Static (manual) load balancing redistributes workload proportional to the compute capabilities of the compute units participating in the heterogeneous cluster. Dynamic load balancing, implemented with Boss-Workers workload distribution model, allows simple scaling of the Monte Carlo calculation of Asian options payoffs, demonstrated in this chapter.
Databáze: OpenAIRE