A variety of control and tuning options can be injected into the direct solvers in /multifrontal:

• Number of OpenMP threads used internally, see .set_nthreads()
• Maximum blocksize of supernodes, see .set_blocksize()
• Size below which leaves of the AssemblyTree are aggregated into dense blocks, see .set_globsize()
• Callback routines for progress monitoring, see .set_progress()
• .. more options are forthcoming, see source-level documentation of multifrontal::Options for further details

Generally speaking, the use of non-default Options does not have any permanent effects upon the underlying solver. Instead, Options are interpreted on a "one-call-at-a-time" basis. For example, the #threads used during numeric factorization could be different from the #threads used later during backsolution, if you passed in different Options to the two calls.