Problem when using MKL

[ingstra]

Hello! I'm trying to use the steadystate solver. When not using MKL, my program runs (albeit a bit slow). When using MKL, I get the following error message. I have tried to figure it out myself, to no avail.

Traceback (most recent call last):
File "cubicH_quantumPump_loop.py", line 54, in
rho = steadystate(H, c_ops)
File "/home/ingrid/anaconda3/lib/python3.6/site-packages/qutip/steadystate.py", line 234, in steadystate
return _steadystate_direct_sparse(A, ss_args)
File "/home/ingrid/anaconda3/lib/python3.6/site-packages/qutip/steadystate.py", line 400, in _steadystate_direct_sparse
v = mkl_spsolve(L, b, perm = in_perm, verbose = ss_args['verbose'])
File "/home/ingrid/anaconda3/lib/python3.6/site-packages/qutip/_mkl/spsolve.py", line 384, in mkl_spsolve
x = lu.solve(b, verbose=verbose)
File "/home/ingrid/anaconda3/lib/python3.6/site-packages/qutip/_mkl/spsolve.py", line 167, in solve
raise Exception(pardiso_error_msgs[str(error)])
KeyError: '[-4]'

[nonhermitian]

You may want to try the development version. However, it is saying that the MKL solver is failing with a zero pivot. At present, it is not possible to modify the underlying parameters for the solver, but it is something I want to do. The MKL solver is quite fast but sometimes runs into trouble.

[pranavm1502]

Hi! I am facing the same problem with steadystate solver. I am also unaware of how to "not use" MKL for the solver. Is there some settings which can be changed to turn it off?

[nonhermitian]

You can set

qutip.settings.has_mkl = False

[nonhermitian]

This should be addressed in Pull #801. The only other reported instance of error when using the MKL solver was resolved when the new kwarg weighted_matching=False is used in the steady state solver.