Title :
Analysis of parallel algorithms using continuous job profiles
Author :
Weems, Mason L. ; Ullah, Nasr ; Gonzalez, Mario J., Jr.
Author_Institution :
Dept. of Electr. & Comput. Eng., Texas Univ., Austin, TX, USA
Abstract :
The increased use of parallelism in computer systems has created the need for models which can be used to evaluate the performance of jobs executing on multiple processor systems. The paper proposes a geometric model, referred to as the continuous job profile, which reflects the allocation of resources needed to process a job. Continuous job profiles are geometric models of a job which can be used to determine how a job´s structure limits its potential speedup when executed on a multiple processor system. Job classes are analyzed and shown to provide additional information which is specific to certain classes of problems
Keywords :
multiprocessing systems; parallel algorithms; parallel processing; performance evaluation; continuous job profile; continuous job profiles; geometric models; job classes; multiple processor system; multiple processor systems; parallel algorithms; potential speedup; resource allocation; Algorithm design and analysis; Computer architecture; Concurrent computing; Information analysis; Parallel algorithms; Parallel processing; Power engineering computing; Resource management; Shape; Solid modeling;
Conference_Titel :
Parallel and Distributed Processing, 1990. Proceedings of the Second IEEE Symposium on
Conference_Location :
Dallas, TX
Print_ISBN :
0-8186-2087-0
DOI :
10.1109/SPDP.1990.143553