Efficient tracing for on-the-fly space-time displays in a debugger for message passing programs

R. Hood, G. Matthews
Proceedings First IEEE/ACM International Symposium on Cluster Computing and the Grid  
In this work we describe the implementation of a An Example Usage Scenario In order to illustrate how trace-driven debugging appears to a user, consider the following scenario. The user is experiencing problems with an MPI program that iteratively solves for the vector X in the matrix equation A * X = B. The program seems to hang indefinitely, with no warning messages or errors produced by MPI. The program breaks down into two main phases: a workload distribution phase where the rows of the
more » ... the rows of the matrix A are split up among the processes, and a computation phase where values for the next iteration of the X vector are found at each process and then collected together by a master process (process 0), which then distributes X for the next iteration.
doi:10.1109/ccgrid.2001.923219 dblp:conf/ccgrid/HoodM01 fatcat:tqcdm7amgzcubmwevyhrh65cne