s-news
[Top] [All Lists]

[S] Memory management in 5.1

To: s-news@wubios.wustl.edu
Subject: [S] Memory management in 5.1
From: Paul Murtaugh <murtaugh@stat.orst.edu>
Date: Mon, 26 Jul 1999 10:01:15 -0700 (PDT)
Cc: murtaugh@stat.orst.edu
Sender: owner-s-news@wubios.wustl.edu

Hi,

I have been using S+ 5.1 on a Sun Ultra 10 (333MHz, 128MB memory),
and have found that the speed and memory management in this version
are substantially worse than they were in Version 3.4, at least for
the kinds of things I do.    
 
For example, I ran an identical simulation in both versions on my
machine, at different times, with no competing processes:
 
   v5.1:  8 iterations completed in 12.4 h (93 min per iteration)
          Using 12.8% CPU and 81.8% MEM after 8 iterations
 
   v3.4:  15 iterations completed in 4.4 h (18 minutes per iteration)
          Using 99.8% CPU and 19.2% MEM after 15 iterations
 
So, these simulations ran more than five times faster in 3.4 than in
5.1, and it appears that the 5.1 process was headed towards paging
death after just eight iterations.
 
The Mathsoft blurb posted here last month boasted that S+ 5 has the
"latest S engine from Lucent Technologies with an improved
object-oriented programming model", resulting in "improved memory
management".  Version 5.1 *is* improved compared to 5.0, which would
get bogged down with even very simple simulations, but, for my 
purposes, it's a big step backwards from 3.4.  In fact, we are back 
to using 3.4 as our "default" version.
 
Are other people having the same experience?  Are there plans to
improve Version 5?  Does anyone care about S+ in UNIX any more?!
 
-Paul Murtaugh
 Statistics, Oregon State Univ.
-----------------------------------------------------------------------
This message was distributed by s-news@wubios.wustl.edu.  To unsubscribe
send e-mail to s-news-request@wubios.wustl.edu with the BODY of the
message:  unsubscribe s-news

<Prev in Thread] Current Thread [Next in Thread>