As a guest user you are not logged in or recognized by your IP address. You have
access to the Front Matter, Abstracts, Author Index, Subject Index and the full
text of Open Access publications.
Modern high performance platforms increasingly exhibit a hybrid system architecture with two types of execution resources: multicore CPUs and additional GPUs. To obtain the maximum performance on these CPU+GPU architectures, it is important to structure parallel applications in such a way that all execution resources can be exploited efficiently. This includes the decomposition of the application into subtasks and the selection of appropriate execution resources for these subtasks.
This article considers the execution of three benchmarks (LU-MZ, SP-MZ, BT-MZ) from the Multi-Zone NAS Parallel Benchmark suite on hybrid CPU+GPU architectures. In particular, it proposes two hybrid execution schemes for these benchmarks that are based on a decomposition into standard sequential tasks and into coarse-grained parallel tasks, respectively. Each of these schemes includes a static scheduling algorithm that assigns the tasks to the available execution resources. Experimental results show that the proposed hybrid CPU+GPU execution schemes achieve a considerably higher performance compared to pure GPU and different multithreaded CPU implementations on two platforms.
This website uses cookies
We use cookies to provide you with the best possible experience. They also allow us to analyze user behavior in order to constantly improve the website for you. Info about the privacy policy of IOS Press.
This website uses cookies
We use cookies to provide you with the best possible experience. They also allow us to analyze user behavior in order to constantly improve the website for you. Info about the privacy policy of IOS Press.