Today, VLSI systems for computationally demanding applications are being built as Systems-on-Chip (SoCs) with a distributed memory sub-system which is shared by a large number of processing elements. The memory sub-system is a potential performance bottle-neck in the system. In this paper, we consider such a distributed memory sub-system and study the impact of address space distribution on system performance. For a given application on such a system, we introduce the notion of address assignment quality. We show that this assignment quality metric is strongly correlated with memory sub-system throughput over large regions of the design space. We show this using open loop performance modeling of the memory sub-system, and justify this using a queueing and a Markov chain analysis. Further, we develop a detailed memory sub-system model for a multi-processor simulation system built on the Augmint [14] framework. Using two (highly parallel) applications (matrix multiplication and bubble s...
G. Hazari, Madhav P. Desai, H. Kasture