Performance Evaluation of SNPs Machine-Learning Workload on Intelreg; Pentiumreg; 4 Hyper-Threading Architectures

摘要

This paper analyzes a Pentium 4 hyper-threading processor and a Pentium 4 hyper-threading processor on 90nm technology with a machine learning workload parallelized with OpenMP~* and Intel compiler. The focus is to understand SNPs performance and the underlying reasons behind that performance. The particular attention is paid to micro-architecture metrics and comparison to examine and evaluate, where appropriate, how those two types of processors perform relative to expectation on SNP machine learning workloads. Results include parallel speedup, micro-architecture metrics comparison.