The following tak/4 program uses lemmas to avoid heap explosion in the case of
of a particularly AND intensive program with 4 recursive calls, a problem particularly severe in the case of the continuation passing binarization that BinProlog uses to simplify the WAM. To encode the 2 first arguments in a unique integer some bit-shifting is needed as it can be seen in tak_encode/3. To avoid such problems, hashing on arbitrary terms like Quintus Prolog's term_hash/2 looks a very useful addition to BinProlog.
tak(X,Y,Z,A) :- X =< Y, !, Z = A.
tak(X,Y,Z,A) :-
X1 is X - 1,
Y1 is Y - 1,
Z1 is Z - 1,
ltak(X1,Y,Z,A1),
ltak(Y1,Z,X,A2),
ltak(Z1,X,Y,A3),
ltak(A1,A2,A3,A).
ltak(X,Y,Z,A):-
tak_encode(X,Y,XY),
tak_lemma(XY,Z,tak(X,Y,Z,A),A).
tak_encode(Y,Z,Key):-Key is Y<<16 \/ Z.
tak_decode(Key,Y,Z):-Y is Key>>16, Z is Key <<17>>17 .
%optimized lemma <P,I,G> --> O (instantiated executing G)
tak_lemma(P,I,_,O):-val(P,I,X),!,X=O.
tak_lemma(P,I,G,O):-G,!,def(P,I,O).
go:- statistics(runtime,_),
tak(24,16,8,X),
statistics(runtime,[_,T]),statistics,
write([time=T,tak=X]), nl.
We hope that we showed the practicality of BinProlog's blackboard for basic work on data structures and problem solving.
BinProlog's blackboard primitives make a clear separation between the copying and the naming intent overloaded in Prolog's assert and retract.
Our blackboard primitives give most of the time simpler and more efficient solutions to current programming problems than assert and retract while being closer to a logical semantics and more cooperative to partial evaluation.