Efficient memory allocations on a many-core accelerator

Abstract

Memory management is one of the key challenges in the design of embedded systems where memory is a scarce resource. The problem scales disproportionally as new embedded systems incorporate many-core architectures where the cores have to struggle accessing an even more limited amount of resources. In this paper we present a way of creating custom memory allocators for many-core accelerators. We evaluated our approach in the P2012 platform, a many-core accelerator from ST. It is shown that a custom memory allocator created by our framework could save on average 62% of the total cycles spent on memory resource management when compared with the platform's current memory allocator without increasing the allocator's overhead.