The paper by I. S. Duff and J. K. Reid extends the frontal method for solving linear systems of equations to handle more than one front simultaneously, enabling the development of a code for general symmetric systems. The authors discuss the organization and implementation of a multifrontal code that uses the minimum-degree ordering to solve indefinite systems stably. They illustrate the performance of their code on both the IBM 3033 and the CRAY-1, highlighting the advantages of vector processing for large problems. The paper covers the overall strategy, minimum-degree ordering, management of the tree search, generating a tree from a given pivotal sequence, numerical factorization, and numerical solution. The authors aim to develop a general-purpose sparse solver for symmetric positive-definite systems, solve indefinite symmetric systems stably with minimal overhead, and ensure good vectorization performance. The results show that their code is competitive with other routines and performs well on both scalar and vector machines.The paper by I. S. Duff and J. K. Reid extends the frontal method for solving linear systems of equations to handle more than one front simultaneously, enabling the development of a code for general symmetric systems. The authors discuss the organization and implementation of a multifrontal code that uses the minimum-degree ordering to solve indefinite systems stably. They illustrate the performance of their code on both the IBM 3033 and the CRAY-1, highlighting the advantages of vector processing for large problems. The paper covers the overall strategy, minimum-degree ordering, management of the tree search, generating a tree from a given pivotal sequence, numerical factorization, and numerical solution. The authors aim to develop a general-purpose sparse solver for symmetric positive-definite systems, solve indefinite symmetric systems stably with minimal overhead, and ensure good vectorization performance. The results show that their code is competitive with other routines and performs well on both scalar and vector machines.