Keywords: sparse linear solver, python, scipy, testcase, matrix, GMRES, poisson, plot
Wikipedia GMRES says
fast convergence occurs when the eigenvalues of A are clustered away from the origin
But what if they're not ?
(There's a great divide between solvers for Ax = b
for A
symmetric, or not, or non-normal; see
how-does-gmres-method-iteration-behave-for-this-non-diagonalizable-matrix
.)