Re: Will adding RAM help "No more memory available" error?
- To: mathgroup at smc.vnet.net
- Subject: [mg90167] Re: Will adding RAM help "No more memory available" error?
- From: RiffenistRiffRaff at gmail.com
- Date: Tue, 1 Jul 2008 06:59:46 -0400 (EDT)
- References: <g3agpd$pka$1@smc.vnet.net>
Hey there,
Here is some information that might be useful. Depending on the code
you are running it is possible to obtain the message ~
No more memory available.
Mathematica kernel has shut down.
Try quitting other applications and then retry.
if you have not correctly handled the local variables in Modules[ ]
called from the main program.
I experienced a similar error message working on a large data set, and
after some trouble shooting I discovered the following. I am by no
means an expert at Mathematica, nonetheless some users might find this
helpful. If I have made errors in my discussion please let me know.
I have included some small "artificial" code that simply explains what
can go wrong if modules are not written within context (pun intended ~
see below) of Mathematica. You may also use the code snippets to do
your own experiments.
The MAIN program is simple. It tracks memory usage and performs a
While[ ] loop creating a large list (see I below). This list is
called x and is constructed with x = Append[ x, Range[1, 10^6] ]. At
the end of the program, before exiting, the variable x is cleared and
one last call to MemoryInUse[ ] is made.
Within the MAIN loop a call is made to the function MSTEPPER (see IIa
below) which returns its own large list. In fact MSTEPPER is just the
main program written as a module, with two local variables { m0, y0 }.
The list it builds is returned to the calling program as y0. Both y0
and m0 appear upfront in a ClearAll[ ] statement.
If you run MAIN you find that at the start there is around 2 meg of
overhead. Just prior to the Exit[ ] of MAIN there is 63 meg of memory
used. It would appear that all variables have been cleared and so
MemoryInUse[ ] should be the overhead of around 2 meg.
At least part of the answer, is that although y0 is a local variable,
memory is not reclaimed. There is a memory leak. To get a clearer
picture try uncommenting the two lines in the loop of the MAIN
program. The MemoryInUse[ ] might be expected to stay low since we
are continually clearing x. It does not.
So what have I done wrong? Look at MSTEPPER of IIb. Here I build a
package using BeginPackage and EndPackage. I create a subexpression
using Begin and End. This creates y0 as a private variable with
context limited to the module MSTEPPER. Now when you run main you will
see that the memory used for y0 is reclaimed.
Hope this helps. It plugged the leak I created, which caused the out
of memory error.
Pete
I. MAIN PROGRAM
RUTE = "C:/Program Files/Wolfram Research/Mathematica/5.2/
myMathematica/";
SetDirectory[RUTE];
<<"mstepper.m";
mem = { };
mem = Append[ mem, MemoryInUse[ ] ];
byte = { };
meg = Power[ 2.0, 20 ];
x = { };
n = 1;
While[ n =98 4,
{
(* x = { ]; *)
x = Append[x, mstepper[ ] ];
byte = Append[ byte, ByteCount[ x ] ];
mem = Append[ mem, MemoryInUse[ ] ];
(* ClearAll[ x ]; *)
};
n++
];
Dimensions[ x ]
mem/meg
byte/meg
ClearAll[ x ];
MemoryInUse[ ]/meg
Exit[ ];
IIa. MODULE MSTEPPER WITHOUT CONTEXT
mstepper[ ] := Module[ { m0, y0 },
ClearAll[ m0, y0 ];
y0 = { };
m0 = 1;
While[ m0 =98 4,
{
y0 = Append[ y0, Range[ 1, 10^6 ] ];
};
m0++
];
Return[ y0 ];
];
IIb. MODULE MSTEPPER WITH CONTEXT
BeginPackage["mstepper`"];
mstepper::"usage" = "Will this work?";
Begin["`Private`"];
mstepper[ ] := Module[ { m0, y0 },
ClearAll[ m0, y0];
y0 = { };
m0 = 1;
While[ m0 =98 4,
{
y0 = Append[ y0, Range[ 1, 10^6 ] ];
};
m0++
];
Return[ y0 ];
];