Xref: utzoo comp.lang.prolog:2824 comp.lang.lisp:3314 comp.lang.c:29611
Path: utzoo!attcan!uunet!samsung!usc!snorkelwacker!bu.edu!xylogics!transfer!crackers!m2c!umvlsi!dime!cs.umass.edu!pop
From: pop@cs.umass.edu
Newsgroups: comp.lang.prolog,comp.lang.lisp,comp.lang.c
Subject: Timings for  tak
Keywords: tak benchmark
Message-ID: <15643@dime.cs.umass.edu>
Date: 15 Jun 90 14:05:17 GMT
References: <13207 of comp.lang.c>
Sender: news@dime.cs.umass.edu
Reply-To: pop@cs.umass.edu (Robin Popplestone)
Followup-To: comp.lang.prolog
Organization: University of Massachusetts, Amherst
Lines: 16

[copy to pop-forum%uk.co.hp.lb@nsfnet-relay.ac.uk]

I obtained the following results for  tak(18,12,6) on a SUN/3 running POPLOG.
Memoisation is not in this case a big win over C, as it is for tak(24,16,8).
I do not have access to the compiler for the POP-11 system dialect, which
does considerably more optimisation. Roughly speaking, generating
fixed-integer code rather than code for general numbers (Common Lisp spec)
gives a factor of 2 speedup, while any other optimisations available to
the user of the incremental compiler gain about 10% extra.                     

POP-11(POPLOG,optimised,incremental)  SUN3/60        0.83
POP-11(POPLOG,memoised,generic)       SUN3/60        0.67
C     (cc -c tak.c)                   SUN3/60        0.37


Robin Popplestone.