Implementation of node debugger in hypercube computer system

Abstract

Mistakes(bugs) are found in a computer program for many reasons. They are classified as syntactic errors, logical errors, algorithmic errors. The program in the parallel computer has more increased complexity and this increased complexity makes debugging a parallel program more difficult than debugging a sequential program. Using a debugger, a programmer can easily locate, analyze, correct suspected errors. Therefore we implement the debugger for KAICUBE II hypercube computer. Our debugger consists of Host debugger and Node debugger. This thesis describes the implementation of Node debugger for KAICUBE II. The requirements for debugging a parallel program are examining and changing the distributed data, finding lost messages and other message related bugs, and controlling multiple independent processes. Our Node debugger satisfies these requirements of a parallel debugger. Compared with other commercial hypercube debugger, our Node debugger is as good as others on the whole. And our Node debugger can suspend a individual debugged process or continue a individual debugged process. This point is superior to others.