From mboxrd@z Thu Jan  1 00:00:00 1970
X-Spam-Checker-Version: SpamAssassin 3.4.4 (2020-01-24) on polar.synack.me
X-Spam-Level: 
X-Spam-Status: No, score=-1.9 required=5.0 tests=BAYES_00 autolearn=ham
	autolearn_force=no version=3.4.4
X-Google-Language: ENGLISH,ASCII-7-bit
X-Google-Thread: 103376,8c8550b9f2cf7d40
X-Google-Attributes: gid103376,public
X-Google-ArrivalTime: 2003-06-10 12:23:20 PST
Path: 
 archiver1.google.com!news1.google.com!newsfeed.stanford.edu!news-spur1.maxwell.syr.edu!news.maxwell.syr.edu!newsfeed.icl.net!newsfeed.fjserv.net!proxad.net!news.tele.dk!news.tele.dk!small.news.tele.dk!uninett.no!newsfeed1.e.nsc.no!nsc.no!nextra.com!news4.e.nsc.no.POSTED!53ab2750!not-for-mail
Newsgroups: comp.lang.ada
Subject: Re: Is ther any sense in *= and matrices?
References: <bebbba07.0306081051.5f3ac24b@posting.google.com>
 <AmNEa.67177$d51.133670@sccrnsc01>
 <bebbba07.0306082024.7cebb5df@posting.google.com>
 <bc15fg$ccu@library2.airnews.net> <7visre2rwv.fsf@vlinux.voxelvision.no>
 <i6qcnalrSvKcnnujXTWcpw@gbronline.com>
From: Ole Kristensen <ole@linux.local>
Message-ID: <m3of153fdd.fsf@linux.local>
Organization: Linux Private Site
User-Agent: Gnus/5.09 (Gnus v5.9.0) Emacs/21.2
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
NNTP-Posting-Host: 148.122.14.228
X-Complaints-To: news-abuse@telenor.net
NNTP-Posting-Date: Tue, 10 Jun 2003 21:23:20 MEST
X-Trace: news4.ulv.nextra.no 1055273000 148.122.14.228
Date: 10 Jun 2003 21:24:30 +0200
Xref: archiver1.google.com comp.lang.ada:38928
Date: 2003-06-10T21:24:30+02:00
List-Id: <comp.lang.ada>

Wesley Groleau <wesgroleau@despammed.com> writes:

> > Yes, on modern hardware, having a pointer for each row of the matrix
> > usually leads to slower code. The reason is that the indexing
> > operation by multilply/add is actually faster than fetching a pointer
> > from an array and then adding, thereby incurring an extra memory fetch
> > and cache pollution. The above holds for large matrices, I haven't
> 
> So Java's handling of arrays is inefficient
> for matrices, even if bypassing the JVM by
> compiling to native code?

Can't speak for Java, but when I tried a 128**3 matrix in Ada and C++,
the plain indexing scheme beat the clever pointer array by a
significant amount. This was on a 1.8GHz Pentium 4. I cannot say
anything definite about other systems, but in general, it pays to
think a lot about memory access patterns on fast CPU's.