Ntroduction

this paper we address how the communication and computational characteristics of a given computer architecture affect the performance of the numerical factorization phase of the multifrontal method. The computational graphs associated with such algorithms typically resemble an unbalanced tree with hundreds or thousands of independent paths, and the granularity of each task increasing towards the root. Thus an efficient implementation onto distributed memory architectures requires the partitioning (or parallelization) of large nodes near the tree root, and th