🚀 go-pugleaf

David Hawley got the following result when trying tak(24,16,8,X) on SICStus:

> SICstus Prolog(compiled)  - mem ovflw after > 1 min of real time

SICStus Prolog runs out of memory because a lot of choicepoints are
created and a lot of trailing is done.  If a cut is put into the first
clause of "tak" no choicepoints will be created we will not run out of
memory.  Presumably Van Roy's compiler has detected that this cut may be
safely inserted.

Running tak(24,16,8,X) using SICStus Prolog on a SPARCstation 1 having
compiled the program into emulated code then takes about 77 seconds.  We
also compiled the program to native code for SPARC (this version of
SICStus has not yet been released) and it took about 25 seconds.  As the
SPARCstation has about half the number of MIPS as Van Roy's machine, we
still have a factor of ten of execution performance improvements to get
down to his figure of just 1.2 seconds.

>So how did he do it? Inquiring minds (and inquisitive noses) want to know.

Yes, please tell us how it was done!

	Dan Sahlin
email: dan@sics.se