A Scalable Parallel Computation of Sparse Approximate Inverse Preconditioners