A parallel direct method is discussed for solving linear equations with a general sparse matrix. The algorithm works on a distributed multiprocessor (MIMD) with a square mesh communication network. At each step of the \(LU\) decomposition, a parallel search for a set of \(m\) compatible pivot elements is performed. The compatibility of the \(m\) pivots enables the simultaneous elimination of \(m\) pivot rows and \(m\) pivot columns in a rank \(m\) update of the reduced matrix. Tests on a Parsytec 400 processors transputer network show a speedup of up to the factor 107.