TY - GEN
T1 - JAWS
T2 - 20th ACM SIGPLAN Symposium on Principles and Practice of Parallel Programming, PPoPP 2015
AU - Piao, Xianglan
AU - Kim, Channoh
AU - Oh, Younghwan
AU - Li, Huiying
AU - Kim, Jincheon
AU - Kim, Hanjun
AU - Lee, Jae W.
PY - 2015/1/24
Y1 - 2015/1/24
N2 - This paper introduces JAWS, a JavaScript framework for adaptive work sharing between CPU and GPU for data-parallel workloads. Unlike conventional heterogeneous parallel programming environments for JavaScript, which use only one compute device when executing a single kernel, JAWS accelerates kernel execution by exploiting both devices to realize full performance potential of heterogeneous multicores. JAWS employs an efficient work partitioning algorithm that finds an optimal work distribution between the two devices without requiring offline profiling. The JAWS runtime provides shared arrays for multiple parallel contexts, hence eliminating extra copy overhead for input and output data. Our preliminary evaluation with both CPU-friendly and GPU-friendly benchmarks demonstrates that JAWS provides good load balancing and efficient data communication between parallel contexts, to significantly outperform best single-device execution.
AB - This paper introduces JAWS, a JavaScript framework for adaptive work sharing between CPU and GPU for data-parallel workloads. Unlike conventional heterogeneous parallel programming environments for JavaScript, which use only one compute device when executing a single kernel, JAWS accelerates kernel execution by exploiting both devices to realize full performance potential of heterogeneous multicores. JAWS employs an efficient work partitioning algorithm that finds an optimal work distribution between the two devices without requiring offline profiling. The JAWS runtime provides shared arrays for multiple parallel contexts, hence eliminating extra copy overhead for input and output data. Our preliminary evaluation with both CPU-friendly and GPU-friendly benchmarks demonstrates that JAWS provides good load balancing and efficient data communication between parallel contexts, to significantly outperform best single-device execution.
UR - http://www.scopus.com/inward/record.url?scp=84939156090&partnerID=8YFLogxK
UR - http://www.scopus.com/inward/citedby.url?scp=84939156090&partnerID=8YFLogxK
U2 - 10.1145/2688500.2688525
DO - 10.1145/2688500.2688525
M3 - Conference contribution
AN - SCOPUS:84939156090
T3 - Proceedings of the ACM SIGPLAN Symposium on Principles and Practice of Parallel Programming, PPOPP
SP - 251
EP - 252
BT - 20th ACM SIGPLAN Symposium on Principles and Practice of Parallel Programming, PPoPP 2015 - Proceedings
PB - Association for Computing Machinery
Y2 - 7 February 2015 through 11 February 2015
ER -