The Effect of Reordering Multi-Dimensional Array Data on CPU Cache Utilization

Memory, as a shared resource, has always been a high latency and bandwidth limited bottleneck of the execution pipeline in multi-core systems. This project analyzes data reordering in multi-dimensional arrays for a more ecient memory allocation method to improve cache utilization and reduce memory access bandwidth. While single-threaded run-time improvements are limited, we demonstrate up to 30% improved run-time and energy consumption in multi-threaded applications when the processing cores are competing for cache space and memory bandwidth.