Improved parallel QR method for large least squares problems involving Kronecker products (Q675993)

The authors develop a new algorithm for the efficient solution of a large least squares problem in which the coefficient matrix of the linear system is a Kronecker product of two matrices of smaller dimension. The solution algorithm is based on QR factorizations of the smaller dimensioned matrices. Nearly perfect load balancing is achieved by exploiting the commutativity property of the Kronecker product. Communication requirements are minimized by employing a binary exchange algorithm for matrix transposition. A parallel algorithm is presented, and timing results are shown from test runs on an Intel C860 computer.