Pragma Directed Shared Memory Centric Optimizations on GPUs

GPUs become a ubiquitous choice as coprocessors since they have excellent ability in concurrent processing. In GPU architecture, shared memory plays a very important role in system performance as it can largely improve bandwidth utilization and accelerate memory operations. However, even for affine...

Full description

Saved in:
Bibliographic Details
Published inJournal of computer science and technology Vol. 31; no. 2; pp. 235 - 252
Main Author Jing Li CCF, Lei Liu Yuan Wu Xiang-Hua Liu Yi Gao Xiao-Bing Feng Cheng-YongWu
Format Journal Article
LanguageEnglish
Published New York Springer US 01.03.2016
Springer Nature B.V
State Key Laboratory of Computer Architecture, Institute of Computing Technology, Chinese Academy of Sciences Beijing 100190, China%Beijing Samsung Telecom Research and Development Center, Beijing 100028, China
Subjects
Analysis
Architecture
Arrays
Artificial Intelligence
Bandwidths
Central processing units
Compilers
Computer programming
Computer Science
Computer simulation
Concurrent processing
Conveying
CPUs
Data management
Data Structures and Information Theory
Design
Design optimization
GPU
Graphics processing units
Information Systems Applications (incl.Internet)
Mathematical models
Optimization
Optimization techniques
Parallel processing
Parameters
Performance enhancement
Programmers
R&D
Regular Paper
Research & development
Resource utilization
Science
Software Engineering
Studies
Theory of Computation
共享内存
内存优化
多面体模型
带宽利用率
并行处理能力
程序员
资源利用率
data centric
pragma directed
GPU
shared memory
Online AccessGet full text

Cover

More Information
Summary:GPUs become a ubiquitous choice as coprocessors since they have excellent ability in concurrent processing. In GPU architecture, shared memory plays a very important role in system performance as it can largely improve bandwidth utilization and accelerate memory operations. However, even for affine GPU applications that contain regular access patterns, optimizing for shared memory is not an easy work. It often requires programmer expertise and nontrivial parameter selection. Improper shared memory usage might even underutilize GPU resource： Even using state-of-the-art high level programming models （e.g., OpenACC and OpenHMPP）, it is still hard to utilize shared memory since they lack inherent support in describing shared memory optimization and selecting suitable parameters, let alone maintaining high resource utilization. Targeting higher productivity for affine applications, we propose a data centric way to shared memory optimization on GPU. We design a pragma extension on OpenACC so as to convey data management hints of programmers to compiler. Meanwhile, we devise a compiler framework to automatically select optimal parameters for shared arrays, using the polyhedral model. We further propose optimization techniques to expose higher memory and instruction level parallelism. The experimental results show that our shared memory centric approaches effectively improve the performance of five typical GPU applications across four widely used platforms by 3.7x on average, and do not burden programmers with lots of pragmas.
Bibliography:GPUs become a ubiquitous choice as coprocessors since they have excellent ability in concurrent processing. In GPU architecture, shared memory plays a very important role in system performance as it can largely improve bandwidth utilization and accelerate memory operations. However, even for affine GPU applications that contain regular access patterns, optimizing for shared memory is not an easy work. It often requires programmer expertise and nontrivial parameter selection. Improper shared memory usage might even underutilize GPU resource： Even using state-of-the-art high level programming models （e.g., OpenACC and OpenHMPP）, it is still hard to utilize shared memory since they lack inherent support in describing shared memory optimization and selecting suitable parameters, let alone maintaining high resource utilization. Targeting higher productivity for affine applications, we propose a data centric way to shared memory optimization on GPU. We design a pragma extension on OpenACC so as to convey data management hints of programmers to compiler. Meanwhile, we devise a compiler framework to automatically select optimal parameters for shared arrays, using the polyhedral model. We further propose optimization techniques to expose higher memory and instruction level parallelism. The experimental results show that our shared memory centric approaches effectively improve the performance of five typical GPU applications across four widely used platforms by 3.7x on average, and do not burden programmers with lots of pragmas.
11-2296/TP
GPU, shared memory, pragma directed, data centric
ObjectType-Article-1
SourceType-Scholarly Journals-1
ObjectType-Feature-2
content type line 14
content type line 23
ISSN:1000-9000
1860-4749
DOI:10.1007/s11390-016-1624-8