disabling simulations manually can give better results that disabling simulatiosn automatically
I noticed the following while working on #568 (closed).
Compare
% genaut --cycle-onehot=7..12 | autfilt -x simul=0 --parity -D --stats="%S -> %s (%r)"
49 -> 102 (0.00289605)
64 -> 133 (0.00459376)
81 -> 168 (0.00687837)
100 -> 207 (0.0115986)
121 -> 250 (0.0179563)
144 -> 297 (0.0296003)
to
% genaut --cycle-onehot=7..12 | autfilt --parity -D --stats="%S -> %s (%r)"
49 -> 13 (0.00428941)
64 -> 15 (0.0129772)
81 -> 17 (0.04455)
100 -> 19 (0.183336)
121 -> 4360 (0.660911)
144 -> 9481 (3.32305)
Notice how the cases 11 and 12 (last two lines of the output) are larger when simulation are enabled? This is related to the simul-max
parameter, that shut down simulations above 4096 states. E.g. if we crank this value up, we get a nice reduction there too:
% genaut --cycle-onehot=11..12 | autfilt -x simul-max=10000 --parity -D --stats="%S -> %s (%r)"
121 -> 21 (1.12528)
144 -> 23 (4.0981)
However if simulation are disabled because if the size, I would at least expect to obtain an output whose size is comparable to what I have in the first output, when simulation are explicitly disabled.