Re: The naive reverse reality check (Was: Is Scryer Prologs failure measurable?)

The test case that crashes all these new
Prolog systems, is for example this one.
It still crashes Scryer Prolog:
$ ulimit -m 2000000
$ ulimit -v 2000000
$ target/release/scryer-prolog -v
v0.9.4-135-g7cfe8ee5
$ target/release/scryer-prolog
?- [user].
app([], X, X).
app([X|Y], Z, [X|T]) :- app(Y, Z, T).
garbage(0, [0]) :- !.
garbage(N, L) :- M is N-1, garbage(M, R), app(R, R, L).
foo :- garbage(12,_), foo.
?- foo.
memory allocation of 2147483648 bytes failed
Aborted
The above preferably runs indefinitely. For
example SWI-Prolog can run it indefinitely.
It also crashes other Prolog systems:
$ ~/go/bin/1pl
Top level for ichiban/prolog v1.2.1
This is for testing purposes only!
See https://github.com/ichiban/prolog for more details.
Type Ctrl-C or 'halt.' to exit.
?- ['bomb.p'].
true.
?- foo.
runtime: out of memory:
cannot allocate 4194304-byte block (802816000 in use)
Mild Shock schrieb:

 But I am nevertheless convinced Scryer Prolog
is dead. These good genes only appeared recently.
But I was comparing oranges and apples.
 I compared Dogelog Player which has garbage
collection to a Prolog system like Scryer Prolog
which doesn't have garbage collection.
 A Prolog system that doesn't have garbage collection
can run faster. The challenge is a runtime engine
that is fast AND has garbage collection.
 Mild Shock schrieb:

>
But Scryer Prolog must nevertheless have somewhere
some good genes. Even it posted about a Prolog compiler
written in Prolog itself. Lets make some reality check:
>
/* Ichiban Prolog */
real    0m39.635s
user    0m59.684s
sys     0m7.891s
>
/* Dogelog Player for Java */
?- time((between(1,6001,_), nrev, fail; true)).
% Zeit 588 ms, GC 0 ms, Lips 5113263, Uhr 11.08.2024 10:47
true.
>
/* Trealla Prolog */
?- time((between(1,6001,_), nrev, fail; true)).
% Time elapsed 0.549s, 3000503 Inferences, 5.468 MLips
    true.
>
/* Scryer Prolog */
?- time((between(1,6001,_), nrev, fail; true)).
    % CPU time: 0.302s, 3_024_526 inferences
    true.
>
/* SWI-Prolog */
?- time((between(1,6001,_), nrev, fail; true)).
% 2,994,499 inferences, 0.078 CPU in 0.089 seconds (88% CPU, 38329587 Lips)
true.
>
Dogelog Player has a Prolog compiler written in
Prolog itself. But the limiting factor is of course
always the runtime engine itself, that executes the
>
compiled code. You can inline and optimize whatever
you want, if the runtime engine, its architecture,
has some limitations performance wise, you won't
>
see aby speed. I don't know yet whether I will beat
SWI-Prolog in this test case ever in the near future?
Especially with some cheap effort?
>
Mild Shock schrieb:

>
Just look at GitHub issues and sort by "recent update".
I get for the last week the following figures:
>
- New tickets: 7 new tickets
- Closed tickets: 2 closed tickets
>
To get a turn around you the the 2nd number bigger
that the 1st number, and not the other way around.