Understanding co-run performance on CPU-GPU integrated processors： observations, insights, directions

Recent years have witnessed a processor development trend that integrates central processing unit (CPU) and graphic processing unit (GPU) into a single chip. The integration helps to save some host-device data copying that a discrete GPU usually requires, but also introduces deep resource sharing an...

Full description

Saved in:

Bibliographic Details
Published in:	Frontiers of Computer Science Vol. 11; no. 1; pp. 130 - 146
Main Authors:	ZHU, Qi, WU, Bo, SHEN, Xipeng, SHEN, Kai, SHEN, Li, WANG, Zhiying
Format:	Journal Article
Language:	English
Published:	Beijing Higher Education Press 01-02-2017 Springer Nature B.V
Subjects:	Central processing units co-run degradation Computer Science Copying CPUs fused processor GPGPU Graphics processing units Microprocessors Operating systems performance analysis Performance degradation Power management Processors program transformation Research Article 中央处理器图形处理单元定时机制性能影响操作系统数据复制经营业绩集成 fused processor program transformation co-run degradation performance analysis GPGPU
Online Access:	Get full text
Tags:	Add Tag No Tags, Be the first to tag this record!

Description
Summary:	Recent years have witnessed a processor development trend that integrates central processing unit (CPU) and graphic processing unit (GPU) into a single chip. The integration helps to save some host-device data copying that a discrete GPU usually requires, but also introduces deep resource sharing and possible interference between CPU and GPU. This work investigates the performance implications of independently co-running CPU and GPU programs on these platforms. First, we perform a comprehensive measurement that covers a wide variety of factors, including processor architectures, operating systems, benchmarks, timing mechanisms, inputs, and power management schemes. These measurements reveal a number of surprising observations.We analyze these observations and produce a list of novel insights, including the important roles of operating system (OS) context switching and power management in determining the program performance, and the subtle effect of CPU-GPU data copying. Finally, we confirm those insights through case studies, and point out some promising directions to mitigate anomalous performance degradation on integrated heterogeneous processors.
Bibliography:	performance analysis, GPGPU, co-run degrada-tion, fused processor, program transformation 11-5731/TP Recent years have witnessed a processor develop- ment trend that integrates central processing unit （CPU） and graphic processing unit （GPU） into a single chip. The inte- gration helps to save some host-device data copying that a discrete GPU usually requires, but also introduces deep re- source sharing and possible interference between CPU and GPU. This work investigates the performance implications of independently co-running CPU and GPU programs on these platforms. First, we perform a comprehensive measurement that covers a wide variety of factors, including processor ar- chitectures, operating systems, benchmarks, timing mecha- nisms, inputs, and power management schemes. These mea- surements reveal a number of surprising observations. We an- alyze these observations and produce a list of novel insights, including the important roles of operating system （OS） con- text switching and power management in determining the program performance, and the subtle effect of CPU-GPU data copying. Finally, we confirm those insights through case studies, and point out some promising directions to mitigate anomalous performance degradation on integrated heteroge- neous processors. co-run degradation Document received on :2015-11-05 fused processor Document accepted on :2016-02-29 program transformation performance analysis GPGPU
ISSN:	2095-2228 2095-2236
DOI:	10.1007/s11704-016-5468-8