Software abort and multiprocessor debugging

Halting all processes distributed machines, simultaneously, is difficult due to the communication delay. This problem can be solved or alleviated by existing halting algorithms, which however cannot avoid a minimal change when executing timing sensitive programs. It is important, especially in detecting time-dependent errors, that the debugger be able to stop a set of processes as fast as possible in order to preserve the relative states of the programmer inspection. This paper suggests a methodology and its hardware implementation, Conductor, as a laboratory model. This methodology enables the user to use a local debugging tool without modifications. So, the user can debug distributed processes in a multiprocessor debugging environment in a similar way to a single processor environment. By using Conductor, the user can set up breakpoints in distributed processes, and stop almost simultaneously at an instruction at an appropriate point where the user wants to see the status of the processes. It preserves relative consistency among processes and is important in inspecting time-dependent errors.<>