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

Re: Odd memory behavior, and uncontrollable memory growth

• To: mathgroup at smc.vnet.net
• Subject: [mg40053] Re: [mg40019] Odd memory behavior, and uncontrollable memory growth
• From: Kirk Reinholtz <kirk.reinholtz at jpl.nasa.gov>
• Date: Mon, 17 Mar 2003 03:35:33 -0500 (EST)
• References: <200303160721.CAA02875@smc.vnet.net>
• Sender: owner-wri-mathgroup at wolfram.com

That looks like the classic Out[] syndrome: Each expression has its
result appended to Out[].  The (...;1) saves less than the
(...tmpFunk[10000]) because the "%" of the former is an int, the "%" of
the latter is a big array (8bytes/double * 10,000 doubles, or something
like that).

nafod40 wrote:
> 
> Hola,
> 
> We are writing code where semi-complicated functions get run thousands
> of times in packages. when we do this, we are getting memory use growth
> that eventually drives the system to virtual memory, and severe speed
> penalties.
> 
> I have been trying to gain an understanding of this process, and came up
> with the following odd example that makes no sense to me. Can someone
> explain?
> 
> a=MemoryInUse[];
> (q=tempFunk[10000];1);
> d=MemoryInUse[];
> d-a
> 
> The memory bumps up by 1080 bytes per iteration. It  climbs, but slowly.
> I figure that is justified by the In[] and Out[] storage. Now one small
> change...
> 
> a=MemoryInUse[];
> (q=tempFunk[10000];);
> d=MemoryInUse[];
> d-a
> 
> Now the memory jumps by 80,000 bytes per iteration! All I did was remove
> the "1" from within the parentheses. What's up with that? I tried this...
> 
> a=MemoryInUse[];
> (q=tempFunk[10000];);
> q=1;
> d=MemoryInUse[];
> d-a
> 
> To force the clearing of 'q', still it jumps by leaps and bounds. Any
> ideas why?

• References:
  • Odd memory behavior, and uncontrollable memory growth
    • From: nafod40 <may106@SPAMAWAY.psu.edu>