Quartus Fitter Seed Setting

The seed value only initializes a random number generator that is used by the fitter algorithm to reshuffle the placement every once in a while.

Thus, setting the seed value may or may not improve fitter results - it completely depends on the netlist, constraints, and of course the state of the random number generator. You should only use the seed variable to try to shave those last few nanoseconds off your critical path - and every time you change something in your design you may get either better or worse results with a given seed value.

Best regards,

Ben

Ben's reply is correct -- the fitter seed randomly perturbs the fitting algorithms in order to generate a different result. Since no placement or routing algorithm can guarantee an optimal result (these are NP-complete problems), trying again from a different starting point can produce a slightly better or worse fit. Hence by running multiple seeds and taking the best result you can get a better answer. The average design improves by about 5% if you run several seeds rather than just taking the result of the first (default) seed.

Running multiple seeds, and trying different optimization options in Quartus (there are lots), and taking the best result is automated by the Design Space Explorer (DSE) script that ships with Quartus. See

Regards,

Vaughn Betz Altera