State of the art computational tools are in especially high demand in the field of Neuroscience. However, bottleneck exists in terms of how much data can be transferred between hard disk and memory for computation . As research increasingly relies on processing huge volumes of data, this issue demands attention. One strategy which addresses this is memory compression of data in memory (DRAM). Algorithms that are effective do a good job of decompressing exactly as much of the data as are needed for the calculations, so that they still minimize the memory footprint of the program without significant speed drops.
This proposal exists to address this bottleneck within the Neuromapp program created by members of the Blue Brain Project team. The project proposed is split into research and implementation of in-memory compression along the lines of compression library selection, interface algorithm development, and block data structure design. The resulting compression mini-app is intended to relieve this bottleneck, and provide accelerated calculation capacity for the suite of associated mini-apps that come with Neuromapp.