s-news
[Top] [All Lists]

Diagnosing memory problems

To: s-news@wubios.wustl.edu
Subject: Diagnosing memory problems
From: Glenn Treacy <gatreacy@maths.tcd.ie>
Date: Wed, 21 May 2008 16:14:37 +0100 (BST)
In-reply-to: <Pine.BSF.4.58L0.0805211518170.7966@bell.maths.tcd.ie>
References: <Pine.BSF.4.58L0.0805211518170.7966@bell.maths.tcd.ie>
 I have a large statistical problem that runs every day on S+6.2 on XP
 with 500MB RAM . It was recently altered to run on a very slightly larger
 dataset and now it fails at the very end (at the function wrap up I
 think) . The log file output from the Splus batch job ends with "Problem
 in isObject(data,"vector"): Unable to obtain requested dynamic memory".
 It is a long time since I've had such  problems and I'm at a loss where to
 start.

 I would really aprreciate the views of S+ users on the following
 questions.

 Does S+7 and S+8 use memory in a more efficient manner and could my
 problem be solved by upgrading the software on the PC in question?

 Adding extra RAM seems an obvious and lazy option. In the past I was
 disappointed with the difference in performance after RAM upgrades. What
 is the maximum amount of RAM that S+ can address under Windows?

 The functions allocated(), storage(), memory.size(), mem.tally.report(),
 object.size() and Patrick Burn's interlude function all offer ways to try
 to pin down where memory is being used. Does anyone have any comments on
 the pros and cons of these particular functions?

 Finally does anyone know of the existence of a profiling function that
 helps to diagnose these memory problems?

 Thanks in advance,
 Glenn


<Prev in Thread] Current Thread [Next in Thread>
  • Diagnosing memory problems, Glenn Treacy <=