Improving Performance of GPGPU Considering Reliability Requirements

Please enable javascript in your browser.

Motallebi, Maryam | 2018

564 Viewed

Type of Document: M.Sc. Thesis
Language: Farsi
Document No: 50433 (19)
University: Sharif University of Technology
Department: Computer Engineering
Advisor(s): Hesabi, Shahin
Abstract:
In recent years, GPUs are becoming ideal candidates for processing a variety of high performance applications. By relying on thousands of concurrent threads in applications and the computational power of large numbers of computing units, GPGPUs have provided high efficiency and throughput. To achieve the potential computational power of GPGPUs in broader types of applications, we need to apply some modifications. By understanding the features and properties of applications, we can execute them in a more proper way on GPUs. Therefore, considering applications’ behavior, we define 5 different categories for them. Every category has special definitions, and we change the configuration of GPU according to the category which the application belongs to. Using this method, we can improve performance by 39% on average. In some cases we observed 98% improvement. On the other hand, there are some applications that need some kind of fault tolerance. In this way, in our proposed method, an application can use DMR of its own choice in order to improve reliability
Keywords:
Performance ; Power Consumption ; Fault Tolerance ; Cache Memory ; Frequencies ; General Purpose Graphic Processing Units (GPGPU)

No TOC