18.5 Specialized Solvers

Function File: bicgstab (A, b)

Function File: bicgstab (A, b, tol, maxit, M1, M2, x0)

This procedure attempts to solve a system of linear equations A*x = b for x. The A must be square, symmetric and positive definite real matrix N*N. The b must be a one column vector with a length of N. The tol specifies the tolerance of the method, the default value is 1e-6. The maxit specifies the maximum number of iterations, the default value is min(20,N). The M1 specifies a preconditioner, can also be a function handler which returns M\X. The M2 combined with M1 defines preconditioner as preconditioner=M1*M2. The x0 is the initial guess, the default value is zeros(N,1).

The value x is a computed result of this procedure. The value flag can be 0 when we reach tolerance in maxit iterations, 1 when we don't reach tolerance in maxit iterations and 3 when the procedure stagnates. The value relres is a relative residual - norm(b-A*x)/norm(b). The value iter is an iteration number in which x was computed. The value resvec is a vector of relres for each iteration.

Function File: cgs (A, b)

Function File: cgs (A, b, tol, maxit, M1, M2, x0)

This procedure attempts to solve a system of linear equations A*x = b for x. The A must be square, symmetric and positive definite real matrix N*N. The b must be a one column vector with a length of N. The tol specifies the tolerance of the method, default value is 1e-6. The maxit specifies the maximum number of iteration, default value is MIN(20,N). The M1 specifies a preconditioner, can also be a function handler which returns M\X. The M2 combined with M1 defines preconditioner as preconditioner=M1*M2. The x0 is initial guess, default value is zeros(N,1).