Driven by several real-life case studies and in-lab developments,
synthetic memory reference generation has a
long tradition in computer science research. The goal is
that of reproducing the running of an arbitrary program,
whose generated traces can later be used for simulations
and experiments. In this paper we investigate this research
context and provide principles and algorithms of a
Markov-Model-based framework for supporting real-time
generation of synthetic memory references effectively and
efficiently. Specifically, our approach is based on a novel
Machine Learning algorithm we called Hierarchical Hidden/
non Hidden Markov Model (HHnHMM). Experimental
results conclude this paper.
