[Date Index] [Thread Index] [Author Index]

Re: Mathemaica 5.2 are the 64 Bit an illusion !!!....?

• To: mathgroup at smc.vnet.net
• Subject: [mg60683] Re: Mathemaica 5.2 are the 64 Bit an illusion !!!....?
• From: Dave Seaman <dseaman at no.such.host>
• Date: Sat, 24 Sep 2005 02:55:26 -0400 (EDT)
• Organization: Information Technology at Purdue
• References: <dgr2cl$85f$1@smc.vnet.net> <dgtivg$266$1@smc.vnet.net> <dh0el3$q65$1@smc.vnet.net>
• Sender: owner-wri-mathgroup at wolfram.com

On Fri, 23 Sep 2005 08:33:39 +0000 (UTC), LumisROB wrote:
> On Thu, 22 Sep 2005 06:29:04 +0000 (UTC), Dave Seaman
><dseaman at no.such.host> wrote:

>>
>>Incorrect.  First of all, your arithmetic is wrong, since 10^4 * 10^4 * 8 is
>>only 800 MB, not 4 GB.  Secondly, I have created arrays bigger than 9 GB using
>>Mathematica 5.2 on a Mac G5.

> ....and 800 MB * 4 = 3.2 GB  (Attention, I have expressly said 4
> matrixes ).
>  What type of matrixes? I hope is not those all of 0 as you had
> already told me in another newsgroup (sci.math.symbolic). Test please
> using Table and an expression of the single element an a little more
> complicated (or that from me suitable): it would interest me to know
> the result in this case.  
> However, for the precision and correctness I have not verified on Mac
> but only on WinXP and Linux.  

As I have already explained, I used "0.", not "0".  The difference is
that 0. takes the same amount of space as a Random[] value, while 0 takes
only half as much.

I don't see the point of generating 4 matrices for the test instead of a
single matrix that is 4 times as large.  Perhaps it makes a difference in
the paging performance, but it shouldn't make a difference in whether the
calculation is able to complete or not.

Since my previous posting I have upgraded to Mathematica 5.2 on several
64-bit systems (AIX, SunOS, and Linux).  I couldn't get the computation
to finish on any of those.  The IBM SP version at least ran quite a bit
longer than Mathematica 5.1 did on the same platform, but so far only the
Mac OS X version has performed the test successfully.


-- 
Dave Seaman
Judge Yohn's mistakes revealed in Mumia Abu-Jamal ruling.
<http://www.commoncouragepress.com/index.cfm?action=book&bookid=228>