Path: utzoo!utgpu!news-server.csri.toronto.edu!rpi!usc!sdd.hp.com!think.com!barmar
From: barmar@think.com (Barry Margolin)
Newsgroups: comp.lang.lisp
Subject: Re: LUCID bitvector operations??
Message-ID: <1991Jun27.064742.23624@Think.COM>
Date: 27 Jun 91 06:47:42 GMT
References: <3070@cantor.ACA.MCC.COM>
Sender: news@Think.COM
Reply-To: barmar@think.com
Distribution: usa
Organization: Thinking Machines Corporation, Cambridge MA, USA
Lines: 19

In article <3070@cantor.ACA.MCC.COM> dnewman@radar.UUCP (David Vincent Newman) writes:
>	This  is the crux of my problem.  I can't decide
>conclusively whether or not the speed of the bitvector logical
>operations is responsible for the poor performance of the "new"
>algorithim on the Sun4.  

I suspect your initial hypothesis, that the bitvector operations are the
bottleneck, was correct.

One reason you may not have seen the bottleneck when you ran a separate
test of bitvector operations is that the problem may relate to interactions
with memory management.  In the real algorithm, enough other stuff could be
taking place that EGCs are copying these bitvectors around alot, while
there might not be many EGCs in the simplified test.
-- 
Barry Margolin, Thinking Machines Corp.

barmar@think.com
{uunet,harvard}!think!barmar