From help-octave-request at bevo dot che dot wisc dot edu  Thu Oct 12 04:21:36 1995
Subject: Memory usage?
From: Eyal Doron <doron at goofy dot mpi-hd dot mpg dot de>
To: help-octave at bevo dot che dot wisc dot edu (Octave mailing list)
Date: Thu, 12 Oct 1995 10:20:29 +0100 (MET)

Hi guys,

I seem to have a memory leak problem of some kind. Can anybody explain?
I'm running the following script in the background:

a=0.4, delta=0.2, Nreturn=65:75

Rlist=1:0.01:1.5; klist=106+(0:200)/200*20;
klist=klist(35:length(klist));

Split=zeros(length(Nreturn)*length(Rlist)*length(klist),1); 
TunAmp=Split;
tic;
block=length(Nreturn)*length(Rlist);
for ik=1:length(klist)
  [split,tunamp]=splitstat(klist(ik),a,delta,Rlist,Nreturn);
  Split((ik-1)*block+(1:block))=split; 
  TunAmp((ik-1)*block+(1:block))=tunamp;
  save splittings3.mat a delta Nreturn Rlist klist Split TunAmp ik
  printf("Elapsed time is %g minutes\n",round(toc/60)), fflush(stdout);
end

Never mind what "splitstat" does, the main thing is that the main
variables "Split" and "TunAmp" are pre-allocated, before the loop is
invoked. Nevertheless, the "SIZE" reported by the "top" command 
increases as the loop progresses, until the program crashes. Currently
it is reporting:

goofy:annular> top -n
last pid: 32683;  load averages:  1.00,  1.02,  1.00    10:12:11
101 processes: 2 running, 15 sleeping, 81 idle, 3 stopped

Memory: Real: 30M/54M act/tot  Virtual: 179M/212M use/tot  Free: 3984K


  PID USERNAME PRI NICE   SIZE   RES STATE   TIME    CPU COMMAND
18687 doron     61   10119360K 20250K run  1187:44 94.90% octave.bin


and virtual memory is about to be exhausted. What is going on? Why is
the "SIZE" increasing if the matrices are pre-allocated?

Thanks,

     Eyal Doron