Efficient GPU resource-sharing between multiple kernels has recently been a critical factor on overall performance. While previous works mainly focused on how to allocate resources to two kernels, there has been limited amount of work on determining which workloads to concurrently execute among multiple workloads. Therefore, we first demonstrate on a real GPU system how the selection of concurrent workloads can have significant impact on overall performance. We then propose GPU Navigator - a lookup-table-based dynamic multi-kernel scheduler that maximizes overall performance through online profiling. Our evaluation shows that GPU Navigator outperforms a greedy policy by 29.3% on average.
|Title of host publication||2020 57th ACM/IEEE Design Automation Conference, DAC 2020|
|Publisher||Institute of Electrical and Electronics Engineers Inc.|
|Publication status||Published - 2020 Jul|
|Event||57th ACM/IEEE Design Automation Conference, DAC 2020 - Virtual, San Francisco, United States|
Duration: 2020 Jul 20 → 2020 Jul 24
|Name||Proceedings - Design Automation Conference|
|Conference||57th ACM/IEEE Design Automation Conference, DAC 2020|
|City||Virtual, San Francisco|
|Period||20/7/20 → 20/7/24|
Bibliographical noteFunding Information:
We would like to thank to Jason Jong Kyu Park for his valuable feedback. This work was supported in part by the National Research Foundation of Korea (NRF) grant funded by the Korea government (MSIP)(No.2018R1D1A1B07050609, No.2015M3C4A7065647, No.2020R1A2B5B01001687), ICT R&D program of MSIP/IITP (No.2017-0-00142), and the R&D program of MOTIE/KEIT (No.10077609). Yongjun Park is the corresponding author.
© 2020 IEEE.
All Science Journal Classification (ASJC) codes
- Computer Science Applications
- Control and Systems Engineering
- Electrical and Electronic Engineering
- Modelling and Simulation