Thesis Topic Details

Topic ID:
1054
Title:
Compiler Techniques for Managing Stream Register Files in Stream Processors
Supervisor:
Jingling Xue
Research Area:
Programming Languages
Associated Staff
Assessor:
Manuel Chakravarty
Topic Details
Status:
Active
Type:
R & D
Programs:
CS CE SE
Group Suitable:
Yes
Industrial:
No
Pre-requisites:
You should be enthusiastic about and interested in programming language implementations. You are expected to have some good understanding about comput
Description:
The stream architecture represents a promising alternative to cache-based architectures in achieving high performance in signal processing, multimedia and graphics. In stream
programming, a program must be expressed as a sequence of data streams flowing through computation kernels so that parallelism can be exploited by multiple ALUs in a SIMD fashion and locality enhanced by a software-managed
on-chip memory known as the Stream Register File (SRF). Efficient management of the SRF is crucial for good performance.


This project aims at developing novel algorithms and techniques for automatic allocation of the SRF to overlap kernel execution and memory operations, maximising performance, and to optimise use of scarce on-chip SRF memory, significantly reducing extranl memory bandwidth.

Comments:
You will work together with a PhD student working in this research area.
Past Student Reports
 
No Reports Available. Contact the supervisor for more information.

Check out all available reports in the CSE Thesis Report Library.

NOTE: only current CSE students can login to view and select reports to download.