Denoising Fixpoint Operator in Prolog (Was: BigNums: JavaScript is killing it [Node.js v24.0.1])

Prolog and logic programming has traditionally
payed a lot of attention to fix point theory. For
example one can defin the least herbrand model
of a definite program.
Anybody already experimented in pure Prolog code
with stable diffusion, aka a fixpoint operator that
is a denoising autoencoder and do the thing as a
kernel operator over a grid?
P.S.: Interesting fixpoint scenarios and
maybe even new hardware:
GENERATIVE PSEUDO-INVERSE MEMORY
https://openreview.net/pdf?id=Harn4_EZBw
Mild Shock schrieb:

Hi,
 Just noticed that Node.js v24.0.1 has mostlikely
faster BigNums. Here is some test:
 /* Dogelog Player 1.3.3 for JavaScript */
?- time(_ is 7^(8^9)).
% Zeit 3813 ms, GC 0 ms, Lips 5, Uhr 11.05.2025 11:31
true.
 Not bad , current bun.js cannot do it. And it is
almost like this here based on GMP:
 /* SWI-Prolog 9.3.21 */
?- time(_ is 7^(8^9)).
% 0 inferences, 1.594 CPU in 1.677 seconds
true
 Bye