Understanding how an application’s data moves through a system can support debugging and performance optimization, and may give insight into the design of future systems. We have developed an on-line, automated technique called value influence tracking that provides a data-centric perspective on how a value contributes to later computation, including how that value's influence is transferred from node to node within a distributed memory computing system. We have implemented a proof of concept value influence tracking tool that supports multithreaded and MPI-based applications.
- Philip C. Roth, Oak Ridge National Laboratory
DOE Office of Science's Office of Advanced Scientific Computing Research, SUPER Institute project within the Scientific Discovery through Advanced Computing (SciDAC) program
- P.C. Roth and J.S. Meredith, “Value Influence Analysis for Message Passing Applications,” Proceedings of the 28th ACM International Conference on Supercomputing (ICS’14), Munich, Germany, 2014, pp. 145-154.
- P.C. Roth, “Tracking a Value’s Influence on Later Computation,” Proceedings of the 6th Workshop on Productivity and Performance (PROPER 2013), Aachen, Germany, Aug. 2013.