Academic Journals Database
Disseminating quality controlled scientific knowledge

Detection of Weak Spots in Benchmarks Memory Space by using PCA and CA

Author(s): Abdul Kareem PARCHUR | Fazal NOORBASHA | Ram Asaray SINGH

Journal: Leonardo Electronic Journal of Practices and Technologies
ISSN 1583-1078

Volume: 9;
Issue: 16;
Start page: 43;
Date: 2010;
VIEW PDF   PDF DOWNLOAD PDF   Download PDF Original page

Keywords: SPEC CPU INT 2006 | Principal Component Analysis (PCA) | Cluster Analysis (CA) | Performance

This paper describes the weak spots in SPEC CPU INT 2006 Benchmarks memory space by using Principal Component Analysis and Cluster Analysis. We used recently published SPEC CPU INT 2006 Benchmark scores of AMD Opteron 2000+ and AMD Opteron 8000+ series processors. The four most significant PCs, which are retained for 72.6% of the variance, PC2, PC3, and PC4 covers 26.5%, 2.9%, 0.91% and 0.019% variance respectively. The dendrogram is useful to identify the similarities and dissimilarities between the benchmarks in workload space. These results and analysis can be used by performance engineers, scientists and developers to better understand the benchmark behavior in workload space and to design a Benchmark Suite that covers the complete workload space.
Download IEC Standards