Presentation 2014-07-28
Auto Program Tuning for Improving Utilization of GPU Resources
Ryo TAKESHIMA, Tomoaki TSUMURA,
PDF Download Page PDF download Page Link
Abstract(in Japanese) (See Japanese page)
Abstract(in English) Utilizing a GPU to perform general purpose computation is called GPGPU. The high theoretical performance of GPU draws attention to GPGPU. CUBA supplies a platform for the developers of GPU applications. In CUBA programming model, massive threads are allocated to GPU's calculation units. Besides, CUBA has various kinds of memories on GPU. These memories have different features of access latency, capacity, and so on. Therefore, to produce high-performance GPU programs, developers should consider how to allocate the massive threads to cores and which memory should be used for storing data. Hence, developers should have deep understanding of the GPU architecture and CUBA APIs. To address this problem, we propose an auto tuning system for GPU programs, and explain an implementation of a preprocessor for the system, in this paper.
Keyword(in Japanese) (See Japanese page)
Keyword(in English) GPGPU / CUDA / tuning GPU programs / data transfer
Paper # CPSY2014-15
Date of Issue

Conference Information
Committee CPSY
Conference Date 2014/7/21(1days)
Place (in Japanese) (See Japanese page)
Place (in English)
Topics (in Japanese) (See Japanese page)
Topics (in English)
Chair
Vice Chair
Secretary
Assistant

Paper Information
Registration To Computer Systems (CPSY)
Language JPN
Title (in Japanese) (See Japanese page)
Sub Title (in Japanese) (See Japanese page)
Title (in English) Auto Program Tuning for Improving Utilization of GPU Resources
Sub Title (in English)
Keyword(1) GPGPU
Keyword(2) CUDA
Keyword(3) tuning GPU programs
Keyword(4) data transfer
1st Author's Name Ryo TAKESHIMA
1st Author's Affiliation Nagoya Institute of Technology()
2nd Author's Name Tomoaki TSUMURA
2nd Author's Affiliation Nagoya Institute of Technology
Date 2014-07-28
Paper # CPSY2014-15
Volume (vol) vol.114
Number (no) 155
Page pp.pp.-
#Pages 6
Date of Issue