native implementation of variant/2 can be fast (Was: When do two Prolog terms marry?)

Ok, I tried a native implementation of variant/2.
I did it in Dogelog Player for Java. Thats a
Prolog system that doesn't have unification for
cyclic terms. So the situation is a little simpler.
Comparing on the same machine and with the data
we have already gathered it is shockingly fast!
/* Dogelog Player 1.2.4, JDK 22 */
?- time(test).
% Zeit 211 ms, GC 0 ms, Lips 7567459, Uhr 15.10.2024 01:30
true.
?- time(test2).
% Zeit 413 ms, GC 0 ms, Lips 10703217, Uhr 15.10.2024 01:30
true.

Here some results:
 - SWI-Prolog 9.3.11:
 ?- time(test).
% 2,126,498 inferences, 0.734 CPU in 0.752 seconds
(98% CPU, 2895657 Lips)
true.
 ?- time(test2).
% 2,159,795 inferences, 0.234 CPU in 0.236 seconds
(99% CPU, 9215125 Lips)
true.
 - Trealla Prolog 2.57.16:
 ?- time(test).
% Time elapsed 3.128s, 14827949 Inferences, 4.741 MLips
    true.
 ?- time(test2).
% Time elapsed 1.079s, 7775516 Inferences, 7.206 MLips
    true.
 - Scryer Prolog :
 ?- time(test3).
    % CPU time: 5.653s, 5_192_831 inferences
    true.
 ?- time(test2).
    % CPU time: 1.544s, 6_116_163 inferences
    true.
 Note: test3 is like test, only it uses builtins:variant/2.