TY - JOUR

T1 - Multicore performance prediction with MPET

T2 - Using scalability characteristics for statistical cross-architecture prediction

AU - Arndt, Oliver Jakob

AU - Lüders, Matthias

AU - Riggers, Christoph

AU - Blume, Holger

PY - 2020/9

Y1 - 2020/9

N2 - Multicore processors serve as target platforms in a broad variety of applications ranging from high-performance computing to embedded mobile computing and automotive applications. But, the required parallel programming opens up a huge design space of parallelization strategies each with potential bottlenecks. Therefore, an early estimation of an application’s performance is a desirable development tool. However, out-of-order execution, superscalar instruction pipelines, as well as communication costs and (shared-) cache effects essentially influence the performance of parallel programs. While offering low modeling effort and good simulation speed, current approximate analytic models provide moderate prediction results so far. Virtual prototyping requires a time-consuming simulation, but produces better accuracy. Furthermore, even existing statistical methods often require detailed knowledge of the hardware for characterization. In this work, we present a concept called Multicore Performance Evaluation Tool (MPET) and its evaluation for a statistical approach for performance prediction based on abstract runtime parameters, which describe an application’s scalability behavior and can be extracted from profiles without user input. These scalability parameters not only include information on the interference of software demands and hardware capabilities, but indicate bottlenecks as well. Depending on the database setup, we achieve a competitive accuracy of 20% mean prediction error (11% median), which we also demonstrate in a case study.

AB - Multicore processors serve as target platforms in a broad variety of applications ranging from high-performance computing to embedded mobile computing and automotive applications. But, the required parallel programming opens up a huge design space of parallelization strategies each with potential bottlenecks. Therefore, an early estimation of an application’s performance is a desirable development tool. However, out-of-order execution, superscalar instruction pipelines, as well as communication costs and (shared-) cache effects essentially influence the performance of parallel programs. While offering low modeling effort and good simulation speed, current approximate analytic models provide moderate prediction results so far. Virtual prototyping requires a time-consuming simulation, but produces better accuracy. Furthermore, even existing statistical methods often require detailed knowledge of the hardware for characterization. In this work, we present a concept called Multicore Performance Evaluation Tool (MPET) and its evaluation for a statistical approach for performance prediction based on abstract runtime parameters, which describe an application’s scalability behavior and can be extracted from profiles without user input. These scalability parameters not only include information on the interference of software demands and hardware capabilities, but indicate bottlenecks as well. Depending on the database setup, we achieve a competitive accuracy of 20% mean prediction error (11% median), which we also demonstrate in a case study.

KW - Multicore Software Migration

KW - Parallelization

KW - Performance Prediction

KW - Scalability

UR - http://www.scopus.com/inward/record.url?scp=85087405083&partnerID=8YFLogxK

U2 - 10.1007/s11265-020-01563-w

DO - 10.1007/s11265-020-01563-w

M3 - Article

AN - SCOPUS:85087405083

VL - 92

SP - 981

EP - 998

JO - Journal of Signal Processing Systems

JF - Journal of Signal Processing Systems

SN - 1939-8018

IS - 9

ER -