Groups | Search | Server Info | Keyboard shortcuts | Login | Register [http] [https] [nntp] [nntps]

Groups > comp.lang.prolog > #14864

Trealla drastically beats Scryer [Jaffar's Algorithm] (Re: Who will win the new Prolog race? SWI or Scryer?

Show all headers | View raw

Hi,

Now I have the feeling that Trealla implements
the extreme of Jaffar's Unification, to the extend
that it keeps and trails functor linking? Was

studying the source of Trealla . Have also the
impression that it uses a very large native stack.
Is this trailing what gave it the name Trealla,

despite the geographic beach location with such a name ?
BTW: I have new benchmark results (*) = Dogelog:

Test | Java (*) | JavaScript (*) | Python (*) | Scryer | Trealla
-- | -- | -- | -- | -- | --
(=)/2 | 184 | 514 | 1750 | 210 | 53
unify_with_occurs_check/2 | 213 | 601 | 2344 | N/A | 53
(==)/2 | 95 | 227 | 1265 | 250 | 65
compare/3 | 86 | 221 | 1328 | 250 | 65
copy_term/2 | 409 | 589 | 2579 | 108 | N/A
term_variables/2 | 168 | 318 | 1469 | 166 | 90
ground/1 | 40 | 120 | 516 | 70 | 121
acyclic_term/1 | 129 | 290 | 1281 | 386 | N/A
numbervars/3 | 170 | 327 | 1516 | 164 | 100

Pitty `copy_term/2` and `acyclic_term/1` do not
yet work with a hydra example.

I am using a hydra variant, which puts less
pressure on the memory, but does benchmark
the built-ins in a loop, so that I see some

numbers. My problem was that Trealla was
below the 1 ms mesurememt:
```
hydra(0, _) :- !.
hydra(N, h(X, X)) :- N > 0, N0 is N-1, hydra(N0, X).

hydra(0, A, A) :- !.
hydra(N, h(X, X), A) :- N > 0, N0 is N-1, hydra(N0, X, A).

/* Binary Built-ins */
bench :-
    hydra(65536, X), hydra(65536, Y, Y),
    time((between(1,64,_), X = Y, fail; true)),
    time((between(1,64,_), unify_with_occurs_check(X, Y), fail; true)),
    time((between(1,64,_), X == Y, fail; true)),
    time((between(1,64,_), compare(_, X, Y), fail; true)), fail; true.

/* Unary Built-ins */
bench2 :-
    hydra(65536, X), hydra(65536, Y, Y),
    time((between(1,64,_), copy_term(X-Y,_), fail; true)),
    time((between(1,64,_), term_variables(X-Y,_), fail; true)),
    time((between(1,64,_), ground(X-Y), fail; true)),
    time((between(1,64,_), acyclic_term(X-Y), fail; true)),
    fail; true.

/* Misc Built-ins */
bench3 :-
    hydra(65536, X), hydra(65536, Y, Y),
    time((between(1,64,_), numbervars(X-Y,0,_), fail; true)),
    fail; true.
```
Bye

Mild Shock schrieb:
> Hi,
> 
> Who will win the new Prolog race?
> SWI or Scryer? Is it enough that SWI-Prolog
> has teamed up with Ciao Prolog?
> 
> What about resources that really can
> code a Prolog systems. SWI-Prolog is
> now doomed to have professional moron
> 
> Julio Di Egidio on board. Not sure whether
> he will ever become a heavy committer to
> SWI-Prolog source. But who knows, we should
> 
> also give Julio Di Egidio a chance. My
> gut feeling at the moment:
> 
> - SWI-Prolog:
>    Pull Requests 10 Open / 560 Closed
> 
> - Scryer Prolog:
>    Pull Requests 40 Open / 751 Closed
> 
> In the past I would measure active committers.
> But the new "liveness" indicator of a Prolog
> system could be indeed Pull Requests.
> 
> So who will win?
> 
> P.S.: Data sources:
> 
> https://github.com/mthom/scryer-prolog/pulls
> 
> https://github.com/SWI-Prolog/swipl-devel/pulls
> 

Back to comp.lang.prolog | Previous | Next — Previous in thread | Next in thread | Find similar | Unroll thread

Thread

Who will win the new Prolog race? SWI or Scryer? Mild Shock <janburse@fastmail.fm> - 2025-09-28 08:57 +0200
  The only problem is, Rust is shiiiit (Was: Who will win the new Prolog race? SWI or Scryer?) Mild Shock <janburse@fastmail.fm> - 2025-09-28 09:01 +0200
    Recent hybrid: Python beats Rust (Was: The only problem is, Rust is shiiiit) Mild Shock <janburse@fastmail.fm> - 2025-09-28 09:13 +0200
  Trealla drastically beats Scryer [Jaffar's Algorithm] (Re: Who will win the new Prolog race? SWI or Scryer? Mild Shock <janburse@fastmail.fm> - 2025-09-29 12:19 +0200
    Boris the Loris and the American Fuzzy Lop (Was: Trealla drastically beats Scryer [Jaffar's Algorithm]) Mild Shock <janburse@fastmail.fm> - 2025-09-29 12:54 +0200
      Was Boris living under a rock, how many Fuzzy Testing tools? (Was: Boris the Loris and the American Fuzzy Lop) Mild Shock <janburse@fastmail.fm> - 2025-09-29 12:59 +0200
        Not a Nazi-Retard, maybe a payed Troll on Putins payroll (Was: Was Boris living under a rock, how many Fuzzy Testing tools?) Mild Shock <janburse@fastmail.fm> - 2025-09-29 13:07 +0200
      Boris / EricGT: Why dont you tap into TAP? (Was: Boris the Loris and the American Fuzzy Lop) Mild Shock <janburse@fastmail.fm> - 2025-11-18 23:00 +0100
        Closing the abstract interpretation GAP (Re: Boris / EricGT: Why dont you tap into TAP?) Mild Shock <janburse@fastmail.fm> - 2025-11-18 23:12 +0100
    Fast, but not correct implementation (Was: Trealla drastically beats Scryer [Jaffar's Algorithm]) Mild Shock <janburse@fastmail.fm> - 2025-09-30 21:39 +0200
      Boris the Loris saw it all comming [Weak Prolog Community] (Was: Fast, but not correct implementation) Mild Shock <janburse@fastmail.fm> - 2025-09-30 21:53 +0200
  War between Scryer Prolog and Ciao Prolog [9:0 Triska vs Hermenegildo] (Was: Who will win the new Prolog race? SWI or Scryer?) Mild Shock <janburse@fastmail.fm> - 2025-11-19 12:49 +0100
    Useless as the Thirty Years War / Generalized Facts (Was: War between Scryer Prolog and Ciao Prolog) Mild Shock <janburse@fastmail.fm> - 2025-11-19 13:11 +0100
  The Guerilla tactics of SWI are amazing: Shell India? (Was: Who will win the new Prolog race? SWI or Scryer?) Mild Shock <janburse@fastmail.fm> - 2025-11-24 15:22 +0100
    The SWI7 exotism (dicts, etc..) has no value? (Was: The Guerilla tactics of SWI are amazing: Shell India?) Mild Shock <janburse@fastmail.fm> - 2025-11-24 15:41 +0100
      My new TinyCLP is much faster [100% Prolog] (Was: The SWI7 exotism (dicts, etc..) has no value?) Mild Shock <janburse@fastmail.fm> - 2025-11-24 21:27 +0100
  Railgun CLP versus Scryer Prolog [Euler #30 Variant] (Was: Who will win the new Prolog race?) Mild Shock <janburse@fastmail.fm> - 2026-01-30 02:10 +0100

csiph-web