Remote execution and debugging of computationally intensive programs in an inhomogeneous, distributed computing environment

An ideal debugging system should provide the programmer with an interface to his parallel/distributed application that is consistent with that level of abstraction used in creating the application, while insulating him from differences in architecture, compiler, and operating system technology characterizing inhomogeneous computing environments. The nature of this abstraction problem is explored for various debugging systems, using a modeling method that interprets compilation and debugging as reciprocal mappings between adjoining levels of abstraction for the abstract data state of a program. Also, a new program (RXDM) is described that extends the capabilities of the existing debugger in the $\IR\sp{\rm n}$ programming environment to remote domains. Finally, a completely new system for execution and debugging is proposed, that will facilitate high-level debugging of parallel/distributed, optimized code in inhomogeneous environments. This new system uses a graduated hierarchy of abstraction layers, and a new intermediate language that provides primitives for communicating debugging information