PROJECT TITLE:

Trusted Performance Analysis on Systems With a Shared Memory - 2015

ABSTRACT:

With the increasing complexity of each information structures and pc architectures, the performance of applications wants fine tuning so as to achieve the expected runtime execution time. Performance tuning is traditionally based on the analysis of performance information. The analysis results might not be accurate, relying on the quality of the information and also the applied analysis approaches. Therefore, application developers might ask: Will we have a tendency to trust the analysis results? This paper introduces our analysis work in performance optimization of the memory system, with a target the cache locality of a shared memory and the memory locality of a distributed shared memory. The standard of the information analysis is guaranteed by using each real performance data acquired at the runtime while the application is running and well-established data analysis algorithms in the sphere of Bioinformatics and Data Mining. We verified the quality of the proposed approaches by optimizing a group of benchmark applications. The experimental results show a vital performance gain.