Method of computing on graphics processors
Abstract
An overview of a simple and efficient method to abort execution of a running kernel on graphics processing unit when it is required for fast release of its resources is presented. Required conditions of the method and kernel feedback approaches needed for a kernel restart on another processing unit are described in details.
About the Authors
I. N. Hubchyk
Белорусский государственный университет информатики и радиоэлектроники, Республика Беларусь
Belarus
N. N. Ivanov
Белорусский государственный университет информатики и радиоэлектроники, Республика Беларусь
Belarus
References
1. Dynamic Checkpointing Policy in Heterogeneous Real-Time Standby Systems / G. Levitin [et al.] // IEEE Transactions on Computers. 2017. Vol. 66, iss. 8. P. 1449-1456.
2. Sousa Dynamic Load Balancing for Real-Time Video Encoding on Heterogeneous CPU+GPU Systems / S. Momcilovic [et al.] // IEEE Transactions on Multimedia. 2014. Vol. 16, iss. P. 108-121.
3. Fu C., Wang Z., Zhai Y. A CPU-GPU Data Transfer Optimization Approach Based on Code Migration and Merging // 16th International Symposium on Distributed Computing and Applications to Business, Engineering and Science (DCABES). 13-16 October 2017.
4. Cheng J., Grossman M., McKercher T. Professional CUDA C Programming. John Wiley & Sons, Inc. 2014. 499 p.
5. Sanders Dzh., Kendrot E. Tekhnologiya CUDA v primerakh. Vvedenie v programmirovanie graficheskikh protsessorov. M.: DMK Press, 2013. 232 s.
6. Shevtsov M. OpenCL™ and OpenGL* Interoperability Tutorial [Electronic resource]. URL: https://software.intel.com/en-us/articles/opencl-and-opengl-interoperability-tutorial (Date of access: 07.05.2018).
For citations:
Hubchyk I.N.,
Ivanov N.N.
Method of computing on graphics processors. Doklady BGUIR. 2018;(4):94-99.
(In Russ.)
Views: 414