PORTAL: a communication library for visualization of distributed, asynchronous data

We present a method for collecting and visualizing data generated by a parallel computational simulation during run time. Data distributed across multiple processes is sent across parallel communication lines to a workstation, which sorts and queues the data for visualization. We have implemented the method in a set of tools called PORTAL (parallel architecture data-transfer library). The tools comprise generic routines for sending data from a parallel program (callable from C or FORTRAN), and a run-time connection to the scientific visualization program AVS. The method is most valuable when used to examine large datasets that can be efficiently generated and do not need to be stored on disk.<
>