Linpack solves dense systems of linear equations using Gaussian elimination and back substitution. Lincube is the iPSC
version. The matrix factor algorithm has a natural broadcast communication pattern. The hypercube code stores the matrix by columns with a wrap mapping (so that node zero gets columns 1, p t 1, 2p t 1, etc (where p is the number of processors)). At each stage in the factor algorithm, the node with the pivot column finds the pivot, computes the multipliers, and sends the pivot column to all the other nodes (using the system broadcast which is implemented as a fanout tree). The largest message is length n (the order of the matrix) but the messa details on the implementation o Be size decreases during the algorithm. For more Lincube see [5]. The performance results are summarized in Figure 1.