I would like your opinion in the following problem. I' m trying to solve a complex, unsymmetric matrix (588x588) and as a test for the result accuracy I used ZGESV, PARDISO and DSS. ZGESV seems to give the best result according to the expected solution. DSS in the begining gives slightly different numbers and then converges to the results of ZGESV while PARDISO gives the first numbers same as ZGESV and then diverges from the ZGESV solution.
Please clarify what you mean by "solve a matrix". Are you solving Ax = b for x, where A is the matrix that you write about? Or are you finding singular values? Similarly, what do you mean by "result"? How was the "expected solution" obtained?
Let us first establish that there is something wrong before we ask what caused it.