On-line Distributed Debugging on Scaleable Multicomputer Architectures

. Debugging parallel programs is one of the most tedious jobs in programming scalable multiprocessor architectures. Due to the distributed resources of these machines, programming is often architecture dependent. Most development tools still reflect this dependency even during the analysis phase of parallel programs. This paper presents the distributed debugger DETOP, which offers a global name space and hides architectural features like the mapping of processes. DETOP is part of the integrated tool environment TOPSYS implemented on iPSC hypercubes, networks of SPARCstations and partly on Transputer systems. 1 Introduction Some compute intensive applications, e.g. fluid dynamics and simulations, require computing performance far beyond the performance of today's supercomputers. Many computer scientists claim that the performance necessary for these "grand challenges" can only be reached with parallel distributed memory computers. Although some of these computers already offer hardwar..