In SWI-Prolog, there is no difference larger than 1%:

?- set_prolog_flag(double_quotes, codes).
true.

?- length(L,1000), maplist(=("abcabcabc"),L),
time((between(1,1000,_), flatten(L,_,[]), fail; true)), fail.
% 20,002,000 inferences, 1.000 CPU in 1.000 seconds (100% CPU, 20002000 Lips)
false.

?- set_prolog_flag(double_quotes, chars).
true.

?- length(L,1000), maplist(=("abcabcabc"),L),
time((between(1,1000,_), flatten(L,_,[]), fail; true)), fail.
% 20,002,000 inferences, 1.000 CPU in 0.999 seconds (100% CPU, 20002000 Lips)
false.

Although I didn't consult Scryer Prolog GitHub, I have
a suspicion why chars are slower than codes in Scryer
Prolog. But I wonder what the experts say.

Mostowski Collapse schrieb am Sonntag, 17. April 2022 um 22:39:14 UTC+2:
> Somehow flatten/2 can be used for atom_list_concat/3,
> at least this would work, when atoms itself are lists.
> But do chars perform better than codes?
>
> I did a little test for Scryer Prolog:
>
> $ target/release/scryer-prolog --version
> "v0.9.0-101-gb4486818"
> $ target/release/scryer-prolog
> ...
> ?- length(L,1000), maplist(=([97, 98, 99, 97, 98, 99, 97, 98, 99]),L),
> time((between(1,1000,_), flatten(L,_,[]), fail; true)), fail.
> % CPU time: 4.730s
> false.
>
> ?- length(L,1000), maplist(=("abcabcabc"),L),
> time((between(1,1000,_), flatten(L,_,[]), fail; true)), fail.
> % CPU time: 4.956s
> false.
>
> Using:
>
> flatten([]) --> !, [].
> flatten([X|Y]) --> !, flatten(X), flatten(Y).
> flatten(X) --> [X].
>
> What is the explanation of the ca. 5% slow-down?
> Mostowski Collapse schrieb am Sonntag, 17. April 2022 um 14:10:27 UTC+2:
> > Here is a challenge. From SWI-Prolog I am used to:
> >
> > ?- atomic_list_concat([a,b,c],',',A).
> > A = 'a,b,c'.
> >
> > ?- use_module(library(dialect/ifprolog)).
> > true.
> >
> > ?- atom_split(A, ',', [a,b,c]).
> > A = 'a,b,c'.
> >
> > Something similar available for Scryer Prolog? These predicates
> > offer the opportunity for some O(n) implementation, compared
> > to a naive Prolog bootstrapping with atom_concat/3 which is O(n^2).
> >
> > This holds for the modes atom_split(-, +, +) and atomic_list_concat(+, +, -).
> > The other mode I am not exactly sure whether there is really an opportunity.
> > The predicate mode atom_split(-, +, +) also brings back to Prolog
> >
> > the join() idiom, widely discussed in JavaScript and Python places.
> > Mostowski Collapse schrieb am Sonntag, 17. April 2022 um 02:30:23 UTC+2:
> > > Oh my dear, Scryer Prolog lost the battle with its
> > > char lists = strings idea and his preference for chars APIs.
> > > I don't have a single chars API in Dogelog Player, was
> > >
> > > fully betting on codes only, makes life much easier.
> > > Now this here. BTW: A library charsio does not anymore
> > > exist in SICStus Prolog 4.7.1:
> > >
> > > | :- use_module(library(charsio)).
> > > ! Existence error in argument 1 of use_module/1
> > > ! file library(charsio) does not exist
> > > ! goal: use_module(user:library(charsio))
> > >
> > > Oh the irony, I did find a library codesio:
> > >
> > > | ?- use_module(library(codesio)).
> > > % loading d:/software/logic/sics/library/codesio.po...
> > > % module codesio imported into user
> > > % loading d:/software/logic/sics/library/types.po...
> > > % module types imported into codesio
> > > % loaded d:/software/logic/sics/library/types.po in module types, 0 msec 6432 bytes
> > > % loading foreign resource d:/software/logic/sics/library/x86_64-win32-nt-4/codesio.dll in module codesio
> > > % loaded d:/software/logic/sics/library/codesio.po in module codesio, 0 msec 64768 bytes
> > > yes
> > > | ?- read_from_codes("X+Y.", T).
> > > T = _A+_B ?
> > >
> > > LoL