Path: utzoo!utgpu!water!watmath!clyde!att!osu-cis!tut.cis.ohio-state.edu!cwjcc!hal!nic.MR.NET!shamash!nis!ems!srcsip!orion!rogers
From: rogers@orion.SRC.Honeywell.COM (Brynn Rogers)
Newsgroups: comp.lang.lisp
Subject: Followup to The BEST way to sum
Message-ID: <10848@srcsip.UUCP>
Date: 26 Oct 88 15:03:33 GMT
Sender: news@src.honeywell.COM
Reply-To: rogers@orion.UUCP (Brynn Rogers)
Organization: Honeywell Systems & Research Center, Camden, MN
Lines: 29
References:


 Thanks to everyone who responded. I posted this problem not
 because of any great need for speed, but to learn the best way
 to program.

(setq nn '(1 2 3 4 5 6 7 8 9 10 11 12 13 14 15))
(everything was wrapped in (dotimes (x 20000) (testalgorithm)) and compiled.
results with TIME function:
algorithm          time
---------          ----
(apply #'+ nn)     4.18 
(dolist ...)       4.22
recursive sum      8.35
(eval (cons '+ nn) 8.40
(reduce #'+ nn)   16.75

The dolist and the apply came in first with virtually the same time
The recursive sum function surprised me (but not some who responded)
 by being so fast.
I don't know why reduce was slow.

I am going to use the apply function because it is the most elegant.
I realize that how good the compiler is the major determineing factor here.

Thanks for responding!
                      Brynn Rogers rogers@src.honeywell.com
P.S. we have clocked GC COMMON LISP on our 386/20 to be about
1/4 the speed of a Symbolics 3640 machine (at 1/40 the price approx.)
(this was a Very rough comparison. Someday I'll clock it against the SUNs)