High productivity multi-device exploitation with the Heterogeneous Programming Library

详细信息    查看全文

• 作者：Moisé ; s Viñ ; as ^{moises.vinas@udc.es}Author Vitae ; Basilio B. Fraguela ; ^{basilio.fraguela@udc.es}Author Vitae ; Diego Andrade ^{diego.andrade@udc.es}Author Vitae ; Ramó ; n Doallo ^{doallo@udc.es}Author Vitae
• 关键词：Programmability ; Heterogeneity ; Parallelism ; Portability ; Libraries ; Load balancing ; OpenCL
• 刊名：Journal of Parallel and Distributed Computing
• 出版年：2017
• 出版时间：March 2017
• 年：2017
• 卷：101
• 期：Complete
• 页码：51-68
• 全文大小：1840 K
• 卷排序：101

文摘

Three approaches to develop multi-device heterogeneous applications are proposed. Easy, efficient and coherent subarray usage for kernels and movements is implemented. Simple argument annotations allow to easily split kernels and arrays among devices. Accurate automatic workload balancing is provided by means of a friendly API. The results are very promising both in terms of performance and programmability.