Subject: Re: What is expensive about consing
From: rpw3@rpw3.org (Rob Warnock)
Date: Fri, 12 Sep 2008 02:02:41 -0500
Newsgroups: comp.lang.lisp
Message-ID: <36OdnTIpq8aMiVfVnZ2dnUVZ_oLinZ2d@speakeasy.net>
Scott Burson  <FSet.SLB@gmail.com> wrote:
+---------------
| rpw3@rpw3.org (Rob Warnock) wrote:
| > With this change, (REDUCE #'+ NUMBER-LIST :MAX-ARITY T) could be
| > implemented *VERY* efficiently, and could even be coded by the
| > compiler to use whatever SIMD or vector (e.g., SSE2/3) instructions
| > were available on the platform.
| 
| Why can't (REDUCE #'+ NUMBER-LIST) be compiled just as efficiently now?
+---------------

Hmmm... You're right, it could. And, in fact, could also be for
any function whose signature is lexically apparent to the compiler.

But :MAX-ARGS might still be useful if the REDUCE call were inside
a separately-compiled routine that accepted a function argument...


-Rob

-----
Rob Warnock			<rpw3@rpw3.org>
627 26th Avenue			<URL:http://rpw3.org/>
San Mateo, CA 94403		(650)572-2607