Scalable Programming models for Battle Command App

Technology trends from the last decade indicate that future processing platforms will become increasingly parallel, driven by the development of multi-core chips and multi-chip processing systems. Such architectures have the potential for large speedups for many applications, but parallel programming techniques are required to exploit the parallel computational resources. However, in most industries, rewriting software to fit this new programming model is impractical. In an effort to conserve the applicability and improve the performance of legacy software, then it would be desirable to develop a technology that facilitates a distributed runtime environment for groups of concurrently running but cooperative codes that share data efficiently. Traditionally, this is done with coarse-grained synchronous I/O, where cooperating codes access files as a shared data medium. This is often inefficient, however, because a process may block on the availability of a complete file when in fact the data it needs is some subset of the file that is ready. Maxentric proposes to develop the Corevoyant Runtime system for facilitating data sharing between concurrently running software with no knowledge a priori of the sharing interface. This will enable legacy codes to run concurrently (e.g. on different cores or processors), significantly improving overall system performance.