Delay in the WAM and the TOAM: redoing a ten year old experiment

A 10 year old experiment related to freeze/2 and its implementation in B-Prolog and SICStus Prolog is redone. B-Prolog uses the TOAM abstract machine and its implementation of delayed goals is by means of suspension frames on the execution stack. The traditional WAM approach partly established by SICStus Prolog implementors is to put suspension terms on the heap. Earlier experiments comparing the two approaches/systems indicated a clear edge for the B-Prolog approach. We first redo these experiments and show to what extent the old conclusions still hold after 10 years. Subsequently, we use hProlog as an environment for showing to what extent the traditional WAM approach to freeze/2 can be made competitive to the B-Prolog approach. We do the same for the event mechanism of B-Prolog. Delay in the WAM and the TOAM: redoing a ten year old experiment Bart Demoen and Phuong-Lan Nguyen ∗ Dept. of Computer Science, K.U.Leuven, Belgium [email protected] † Inst. de Mathématiques Appliquées, UCO, Angers, France [email protected] Abstract. A 10 year old experiment related to freeze/2 and its implementation in B-Prolog and SICStus Prolog is redone. B-Prolog uses the TOAM abstract machine and its implementation of delayed goals is by means of suspension frames on the execution stack. The traditional WAM approach partly established by SICStus Prolog implementors is to put suspension terms on the heap. Earlier experiments comparing the two approaches/systems indicated a clear edge for the B-Prolog approach. We first redo these experiments and show to what extent the old conclusions still hold after 10 years. Subsequently, we use hProlog as an environment for showing to what extent the traditional WAM approach to freeze/2 can be made competitive to the B-Prolog approach. We do the same for the event mechanism of B-Prolog. A 10 year old experiment related to freeze/2 and its implementation in B-Prolog and SICStus Prolog is redone. B-Prolog uses the TOAM abstract machine and its implementation of delayed goals is by means of suspension frames on the execution stack. The traditional WAM approach partly established by SICStus Prolog implementors is to put suspension terms on the heap. Earlier experiments comparing the two approaches/systems indicated a clear edge for the B-Prolog approach. We first redo these experiments and show to what extent the old conclusions still hold after 10 years. Subsequently, we use hProlog as an environment for showing to what extent the traditional WAM approach to freeze/2 can be made competitive to the B-Prolog approach. We do the same for the event mechanism of B-Prolog.