Procedure Mapping Using Static Call Graph Estimation

Procedure Mapping Using Static Call Graph Estimation (1997)

Cached

Download Links

by Amir Hooshang Hashemi , David R. Kaeli , Brad Calder

Citations:

8 - 2 self

Versions

Version History

Metadata Version 1

Datum	Value	Source
TITLE	Procedure Mapping Using Static Call Graph Estimation	SVM HeaderParse 0.1
AUTHOR NAME	Amir Hooshang	SVM HeaderParse 0.1
AUTHOR AFFIL	Dept. of Electrical and Computer Engineering Dept. of Computer Science and Engineering; Northeastern University University of California, San Diego	SVM HeaderParse 0.2
AUTHOR ADDR	Boston, MA La Jolla, CA	SVM HeaderParse 0.1
AUTHOR NAME	Hashemi David	SVM HeaderParse 0.1
AUTHOR AFFIL	Dept. of Electrical and Computer Engineering Dept. of Computer Science and Engineering; Northeastern University University of California, San Diego	SVM HeaderParse 0.2
AUTHOR ADDR	Boston, MA La Jolla, CA	SVM HeaderParse 0.1
AUTHOR NAME	R. Kaeli	SVM HeaderParse 0.1
AUTHOR AFFIL	Dept. of Electrical and Computer Engineering Dept. of Computer Science and Engineering; Northeastern University University of California, San Diego	SVM HeaderParse 0.2
AUTHOR ADDR	Boston, MA La Jolla, CA	SVM HeaderParse 0.1
AUTHOR NAME	Brad Calder	SVM HeaderParse 0.1
AUTHOR AFFIL	Dept. of Electrical and Computer Engineering Dept. of Computer Science and Engineering; Northeastern University University of California, San Diego	SVM HeaderParse 0.2
AUTHOR ADDR	Boston, MA La Jolla, CA	SVM HeaderParse 0.1
ABSTRACT	As the gap between memory and processor performance continues to grow, it becomes increasingly important to exploit cache memory effectively. One technique used by compiler and linkers to improve the performance of the cache is code reordering. Code reordering optimizations rearrange a program so that sections of the program with temporal locality will be place next to each other in the final program layout. A number of software approaches to code reordering have been proposed. Their goal is to reduce the number of cache line conflicts. Most of these schemes use profile data in order to reposition the code in the address space. In this paper we present a link-time procedure mapping algorithm which uses a call graph constructed without the use of profile data. We will refer to this scheme as static call graph estimation. In this approach we use program-based heuristics to statically estimate the behavior of the call graph. Then once the estimated weighted call graph is formed, we can employ various procedure remapping algorithms. Our results show that we were able to reduce instruction cache miss rates by 20 % on average when using our estimated static call graph with modern procedure reordering algorithms.	SVM HeaderParse 0.1
CITATIONS	14 found	ParsCit 1.0