Author :
Gracia Perez, D. ; Berry, H. ; Temam, Olivier
Abstract :
This sampling technique, which is hardware-independent and almost entirely transparent to the user, employs a budget-based approach that jointly considers warm-up and sampling costs, presents them as a single parameter to the user, and distributes simulated instructions between warm-up and sampling based on region partitioning and clustering information. Although it focuses on practicality, the technique delivers nearly state-of-the-art speed and accuracy. In this article, we explain how even the most successful of the existing sampling techniques fall short on practicality, especially when it comes to warm-up. We propose a more practical alternative: a sampling technique that focuses specifically on transparency, while also delivering nearly state-of-the-art speed and accuracy. The technique offers a hardware-independent, integrated approach to warm-up and sampling that requires no modification of the functional simulator and relies solely on the performance simulator for warm-up
Keywords :
computer architecture; sampling methods; architecture simulation; budget-based approach; clustering information; functional simulator; performance simulator; region partitioning; sampling costs; sampling method; Clustering methods; Computational modeling; Computer simulation; Costs; Error analysis; Sampling methods; Surges; Timing; evaluation; measurement; modeling; simulation;