The standard benchmark for Prolog is Lips or logical inferences per second. Happily
Shen Prolog under Scheme is doing between 2-3 MLips (2.5 x 10^6 Lips) on my 4.7GHz which is pretty good for version designed to port over umpteen languages and written at a high level.. Here is a computation running at 2.5 MLips.
(11-) (ftp.<-kb (ec))
run time: 1.0140064999999998 secs
2509580 inferences
true
However I find that SWI Prolog counts stuff like numeric comparision as an inference
?- time(5 > 7).
% 2 inferences, 0.000 CPU in 0.000 seconds (?% CPU, Infinite Lips)
false.
?- time(15 > 7).
% 1 inferences, 0.000 CPU in 0.000 seconds (?% CPU, Infinite Lips)
true.
Shen Prolog does not count invocations of when as inferences; ditto call and var? and the other walk on parts. So Lips comparisons are misleading.
On Einstein's Riddle I get SWI Prolog as about 2.5 times faster than Shen Prolog.
Mark