DocumentCode
3142763
Title
Analysis of SPEC benchmark programs
Author
Kainaga, M. ; Yamada, K. ; Inayoshi, H.
Author_Institution
Syst. Dev. Lab., Hitachi Ltd., Kawasaki, Japan
fYear
1991
fDate
21-27 Nov 1991
Firstpage
208
Lastpage
215
Abstract
An analysis of the SPEC benchmark is performed from the standpoint of a CPU or compiler designer. Each program in SPEC is investigated in terms of its operation flow, friction by function. The execution time of each function, number of function calls, and the text sizes are studied. The higher locality in instruction fetches is confirmed and the effectiveness of a cache is studied. A normalized CPI is proposed to compare performances among different CPUs and compilers, and to study the possibility of performance enhancement. The possibility of obtaining a normalized CPI of less than 0.5 is discovered. Finally, the relation between Dhrystone and SPECint is studied, and a factor of 1.5 is found
Keywords
buffer storage; performance evaluation; program compilers; CPU; Dhrystone; SPEC benchmark programs; SPECint; cache; compiler; execution time; function calls; instruction fetches; Equations; Laboratories; Microprocessors; Operating systems; Optimizing compilers; Performance analysis; Performance evaluation; Program processors; System performance; Workstations;
fLanguage
English
Publisher
ieee
Conference_Titel
TRON Project Symposium, 1991. Proceedings., Eighth
Conference_Location
Tokyo
ISSN
1063-6749
Print_ISBN
0-8186-2475-2
Type
conf
DOI
10.1109/TRON.1991.213102
Filename
213102
Link To Document