Skip to content

Problems with lpSolve/Memory ? R crashes

2 messages · TEMPL Matthias, Arne Henningsen

TEMPL Matthias
# 
Hello!

I have a curious problem, which I cannot solve.
With my code I solve thousands of small linear programs with the package lpSolve automatically.
But R crashes sometimes (~always, but always on different linear programs) in a strange way.
For illustration, I tried to prepare a simple example, which shows the nature of the problem.
The function aaa (see below) declares some constants (only in this special example) and in the end it solves the linear program.

aaa(1)
      [,1] [,2]
 [1,]    0    0
 [2,]    0    0
 [3,]    0    0
 [4,]    0    0
 [5,]   52    0
 [6,]    2    2
 [7,]    0    0
 [8,]    2    0
 [9,]    0    0
[10,]    0    0
[11,]    0    0
[12,]    0    0
[13,]   54    0

Works fine.
Now I make the *same* calculation, say 1000 times:
aaa(1000)
R (I have tried it with R2.0.1, 2.0.0, 1.9.1, 2.1.0dev) crashes completly - without warning and error message under Windows XP, Intel Pentium 3 with 256 MB RAM
Under Linux SuSe 8.2 R (2.0.1) it crashes again, but in this case I get the following message:
Calloc of 40004 bytes failed on line 114 of file lpkit.c
...
Calloc of 80008 bytes failed on line 113 of file lpkit.c
Error: cannot allocate vector of size 3 Kb

Now I?m completly lost. Solving the linear program one time makes no problem. Solving it twice in the same way makes no problem either. Running the same calculation, say 1000 times, causes a crash. Why should there be a problem with memory?

For any hint, I would be really happy. 
Thank you,
Matthias


### ---------- function aaa ------------

aaa <- function(amount=1){
f.obj <- rep(0,33)

w <- c(3,4,5,6,11,13,17,22,25,26,27,28,33)

m <- matrix(c(0,0,1,1,1,1,0,0,0,0,-1,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,

              0,0,0,0,0,0,0,0,0,0,0,0,1,0,0,0,1,0,0,0,0,-1,0,0,0,0,0,0,0,0,0,0,0,

              0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,-1,-1,-1,-1,0,0,0,0,1,

              0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,

              0,0,0,0,0,0,0,0,0,0,0,0,1,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,

              0,0,1,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,-1,0,0,0,0,0,0,0,0,

              0,0,0,1,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,-1,0,0,0,0,0,0,0,

              0,0,0,0,1,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,-1,0,0,0,0,0,0,

              0,0,0,0,0,1,0,0,0,0,0,0,0,0,0,0,1,0,0,0,0,0,0,0,0,0,0,-1,0,0,0,0,0,

              0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,

              0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,

              0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,

              0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,

              0,0,0,0,0,0,0,0,0,0,-1,0,0,0,0,0,0,0,0,0,0,-1,0,0,0,0,0,0,0,0,0,0,1

              ),ncol=33,byrow=TRUE)

f.dir <- c("=","=","=","=","=","=","=","=","=","=","=",
"=","=","=","=","=","=","=","=","=","=","=","=",
"=","=","=","=","=","=","=","=","=","=")

f.rhs <- c(-52,0,54,0,2,0,0,0,0,0,0,0,0,0)

lp.out <- matrix( ncol = 2, nrow = 13 )

for(ii in 1:amount){ # - simple iterate the same

for( i in 1:13 ){

	f.obj[ w[i] ] <- 1

	lp.out[ i, 1 ] <- lp("min", f.obj, m, f.dir, f.rhs)$objval

	lp.out[ i, 2 ] <- lp("max", f.obj, m, f.dir, f.rhs)$objval

	f.obj <- rep( 0, 33 )

}

}
lp.out

}
Arne Henningsen
# 
Unfortunately, I can't tell you how to solve this problem. However, I can 
confirm that this crash happens also on my machines (R 2.0.1 on Debian Sarge, 
AMD Athlon, 750 MHz, 256 MB RAM and R 2.0.0 on SuSE 9.0, Intel P4, 2.6 GHz, 
512 MB RAM):
R> aaa(1000)
alloc of 80000 bytes failed
alloc of 40000 bytes failed
Speicherzugriffsfehler

We had a model, which consists of 1000+ linear programs with each about 550 
rows and 1100 columns (http://www.unipr.it/arpa/dipseq/EAAE/PR/Parallel/ 
42b4_henningsen.pdf). On MS-Windows 2000/XP this model always crashed after 
some hundreds linear programs, but using Linux on the same machines it 
_never_ crashed. I really can't see the reason why our much larger programs 
worked, but these small programs let R crash. 
I hope that this information may help to solve this problem.

Arne
On Wednesday 30 March 2005 12:16, TEMPL Matthias wrote: