No, GMIN affects simulator behavior; it will not try to calculate a value below GMIN. RSHUNT is essentially an additional device, a resistor, between every node in the circuit and ground.

@Tim Edwards So, GMIN would not solve singular matrix issue ?

@promach, @Tim Edwards is right, gmin is the 'minimum conductance allowed by the program', it is used to calculate off resistance of some ideal switches and other components, but this minimum conductance is not added to floating nodes, this is done (on any nodes, not only floating ones) with .options RSHUNT=1e11

@Tim Edwards @Stefan Schippers So, what should I do now to remove singular matrix issue since

.option rshunt=1e11

does not help ?

@promach the only effective solution is adding in the schematic a 1e11 resistor to GND on all nodes ngspice reports as problematic. at least until the RSHUNT option gets fixed on ''large pdks'

@Stefan Schippers the ngspice fix is not working for me in my environment. Could you reproduce the same issue https://sourceforge.net/p/ngspice/discussion/133842/thread/e27767b336/#9d50/9ea0 in your environment ?