Using Continuous Statistical Machine Learning to Enable Performance Prediction in Hybrid Functional/Cycle Accurate Instruction Set Simulators

Content Provider	Semantic Scholar
Author	Powell, Daniel
Copyright Year	2009
Abstract	Functional instruction set simulators perform instruction-accurate simulation of benchmarks at high instruction rates. Unlike their slower, but cycle-accurate, counterparts however, they are not capable of providing cycle counts due to the higher level of hardware abstraction. In this report we present a novel approach to performance prediction based on statistical machine learning utilising a hybrid functional and cycle-accurate simulator. We introduce the concept of continuous learning to simulation whereby new training data points are acquired on demand and used for on-the-fly updates of the prediction model. Furthermore, we show how statistical regression can be adapted to reduce the cost of these updates during a performance-critical simulation. For a stateof-the-art simulator modelling the ARC 750D embedded processor core we demonstrate that our approach is highly accurate, with average error < 2.5% whilst achieving a speed-up of ≈ 1.5 times that of the baseline cycle-accurate simulation. When extended to use a modern JIT-compiled simulation for the functional simulation, this would enable the full simulation of e.g. complex multimedia applications.
File Format	PDF HTM / HTML
Alternate Webpage(s)	http://homepages.inf.ed.ac.uk/s0347677/papers/Powell-UThesis.pdf
Language	English
Access Restriction	Open
Content Type	Text
Resource Type	Article