Estimating the Impact of Architectural and Software Design Choices on Dynamic Allocation of Heterogeneous Memories

Reducing energy consumption is a key challenge to the realization of the Internet of Things. While emerging memory technologies may offer power reduction, they come with major drawbacks such as high latency or limited endurance. As a result, system designers tend to juxtapose several memory technologies on the same chip. This paper studies the interactions between dynamic memory allocation and architectural choices regarding this heterogeneity. We provide cycle accurate simulations of embedded platforms with various memory technologies and we show that different dynamic allocation strategies have a major impact on performance. We demonstrate that interesting performance gains can be achieved even for a low fraction of heap objects in fast memory, but only with a clever data placement strategy between memory banks.