From Newsgroup: comp.lang.prolog

Hi,

The WebPL shunting experiment still resonates. Dogelog
Players assumption is implicit shunting through
unification variable bind preferences, derived

from how unification is called. So I rearranged
the implementation of member/2 a little bit, and
now I get, a less invasive version, doesn't modify L:

/* Dogelog Player 2.1.5 Preview */

/* Not Invasive */
?- T = T, L = [X,Y,Z], member(T, L), write(L), nl, fail; true.
[_0, _1, _2]
[_0, _1, _2]
[_0, _1, _2]
true.

/* Not Invasive */
?- L = [X,Y,Z], T = T, member(T, L), write(L), nl, fail; true.
[_3, _4, _5]
[_3, _4, _5]
[_3, _4, _5]
true.

We can compare with SWI-Prolog:

/* SWI-Prolog 10.1.1 */
/* Invasive */
?- T = T, L = [X,Y,Z], member(T, L), write(L), nl, fail; true. [_13280,_13294,_13300]
[_13288,_13280,_13300]
[_13288,_13294,_13280]
true.

/* Not Invasive */
?- L = [X,Y,Z], T = T, member(T, L), write(L), nl, fail; true. [_18762,_18768,_18774]
[_18762,_18768,_18774]
[_18762,_18768,_18774]
true.

And the result for Scryer Prolog:

/* Scryer Prolog 0.10 */
/* Invasive */
?- T = T, L = [X,Y,Z], member(T, L), write(L), nl, fail; true.
[_112,_131,_138]
[_123,_112,_138]
[_123,_131,_112]
true.

/* Invasive */
?- L = [X,Y,Z], T = T, member(T, L), write(L), nl, fail; true.
[_117,_125,_133]
[_117,_120,_133]
[_117,_125,_120]
true.

Bye

Mild Shock schrieb:

Hi,

Here some test results when testing
Desktop and not the Web. With Desktop
Prolog versions I find:

/* SWI-Prolog 9.3.28 */
% 7,506,637 inferences, 0.578 CPU in 0.567 seconds

/* Dogelog Player 1.3.6 for Java (16.08.2025) */
% Zeit 803 ms, GC 0 ms, Lips 9367988, Uhr 17.08.2025 18:03

/* Scryer Prolog 0.9.4-592 */
% CPU time: 0.838s, 7_517_613 inferences

/* Trealla Prolog 2.82.12 */
% Time elapsed 2.315s, 11263917 Inferences, 4.866 MLips

Bye

Mild Shock schrieb:

Hi,

The paper by Shalin and Carlson from 1991
did not yet ring a bell. But it suggest testing
something with primes and freeze. Lets do

primes as suggested but without freeze. SWI-Prolog
seems not to the OG of GC. Putting aside Shalin
and Carlson, its an typical example of a lot of

intermediate results, that can be discarded by
a garbage collection. Every candidate number that
is not a prime number can be remove from the

trail they get unreachable in the first clause
of search/3. Besides this obvious unreachability
task, I don't have statistics or don't see immediately

where large variable instantiation chains are supposed
to be created. At least not in my Prolog system, since
a result variable is passed without binding it to a

local variable, this "shunting" happens independent
of neck tests and the "shunting" there. The result variable
passing is extremly simple to implement and could

be what is effective here besides the reachability thingy.
At least the 1 ms GC time in Dogelog Player show that
the reachability thingy is the minor effort or optimization

to get nice performance:


/* WebPL GC */
(1846.1ms)

/* Dogelog Player 1.3.6 for JavaScript (16.08.2025) */
% Zeit 2992 ms, GC 1 ms, Lips 2514202, Uhr 17.08.2025 17:44

/* SWI-Prolog WASM */
(4204.2ms)

/* Trealla Prolog WASM */
(23568.9ms)

The test code was:

test :-
-a-a-a len(L, 1000),
-a-a-a primes(L, _).

primes([], 1).
primes([J|L], J) :-
-a-a-a primes(L, I),
-a-a-a K is I+1,
-a-a-a search(L, K, J).

search(L, I, J) :-
-a-a-a mem(X, L),
-a-a-a I mod X =:= 0, !,
-a-a-a K is I+1,
-a-a-a search(L, K, J).
search(_, I, I).

mem(X, [X|_]).
mem(X, [_|Y]) :-
-a-a-a mem(X, Y).

len([], 0) :- !.
len([_|L], N) :-
-a-a-a N > 0,
-a-a-a M is N-1,
-a-a-a len(L, M).

Bye

--- Synchronet 3.21b-Linux NewsLink 1.2