Back to Search Start Over

Upper Bounds on Number of Steals in Rooted Trees

Authors :
Massachusetts Institute of Technology. Computer Science and Artificial Intelligence Laboratory
Massachusetts Institute of Technology. Department of Electrical Engineering and Computer Science
Leiserson, Charles E.
Schardl, Tao Benjamin
Suksompong, Warut
Massachusetts Institute of Technology. Computer Science and Artificial Intelligence Laboratory
Massachusetts Institute of Technology. Department of Electrical Engineering and Computer Science
Leiserson, Charles E.
Schardl, Tao Benjamin
Suksompong, Warut
Source :
Springer US
Publication Year :
2016

Abstract

Inspired by applications in parallel computing, we analyze the setting of work stealing in multithreaded computations. We obtain tight upper bounds on the number of steals when the computation can be modeled by rooted trees. In particular, we show that if the computation with n processors starts with one processor having a complete k-ary tree of height h (and the remaining n−1 processors having nothing), the maximum possible number of steals is ∑ni=1(k−1)i(hi).

Details

Database :
OAIster
Journal :
Springer US
Notes :
application/pdf, English
Publication Type :
Electronic Resource
Accession number :
edsoai.on1018412092
Document Type :
Electronic Resource