Using synthetic perturbations and statistical screening to assay shared-memory programs

Synthetic-perturbation screening (SPS) is a diagnostic technique employing artificial code placed within segments of an MIMD program. These insertions simulate code changes in suspected program bottlenecks. Screening techniques based upon statistical experimental design then flag those program segme...

Full description

Saved in:
Bibliographic Details
Published inInformation processing letters Vol. 54; no. 3; pp. 147 - 153
Main Authors Snelick, Robert, JáJá, Joseph, Kacker, Raghu, Lyon, Gordon
Format Journal Article
LanguageEnglish
Published Amsterdam Elsevier B.V 12.05.1995
Elsevier Science
Elsevier Sequoia S.A
Subjects
Applied sciences
Computer programming
Computer science; control theory; systems
Computer systems and distributed systems. User interface
Distributed computing
Distributed processing
Exact sciences and technology
Methods
Parallel processing
Performance evaluation
Programming theory
Software
Studies
Systems analysis
Theoretical computing
Tuning
Performance evaluation
Parallel processing
Distributed computing
Tuning
Statistical method
Screening
Experimental design
Program performance
Synchronization
Shared memory
Online AccessGet full text

Cover

More Information
Summary:Synthetic-perturbation screening (SPS) is a diagnostic technique employing artificial code placed within segments of an MIMD program. These insertions simulate code changes in suspected program bottlenecks. Screening techniques based upon statistical experimental design then flag those program segments that are most sensitive to perturbation (delay). A subset of program segments so flagged can be candidates for improvement. The results are sensitivity analyses of specimen programs in terms of their questionable sections of code. This provides a portable, scalable and generic basis for assaying MIMD programs; the approach is quite powerful. SPS is adapted to evaluate some specific sources of poor performance on shared-memory programming systems: 1. insufficient parallelism, 2. load imbalance, 3. resource contention, 4. overhead introduced by synchronizations, and 5. remote memory accesses.
ISSN:0020-0190
1872-6119
DOI:10.1016/0020-0190(95)00012-2