[R-sig-hpc] Possible memory leak for serialize?

[luke-tierney at uiowa.edu]

Clearly no memory leak as such: if you rm(out) and gc() then OS usage
drops down and repeating your example/rm(out) stays stable.

It does seem that 'out' is holding on to more memory than it should;
I'll try to have a look when I get a chance.

Best,

luke

On Wed, 5 Sep 2012, Hao Yu wrote:

> Hi.
>
> serialize may have memory leak issue. When I run the following codes
>
> x=split(1:40000, rep(1:8, 5000))
> fun=function(x)lapply(x, serialize, NULL)
> out =replicate(5000, fun(x))
>
> After using gc() severally times, R reports about 800MB memory usage
> (objects fun, out and x) but the real memory usage is 2.4GB (reported from
> OS). After deleting all objects, R reports about 12MB usage but real
> memory usage is reduced to 1.7GB and will stay there.
>
> Don't try with big size. I used up all my 16GB ram on my windows PC and
> caused other programs to crash.
>
> I can reproduce this behave on both Windows and Debian with R 2.15.1
> 64bit. If this memory leak does exist, it may have impact on jobs with
> lots of R objects exchanges among nodes, in particularly master node.
>
> Hao
>
> PS: Changing fun to fun2=function(x)lapply(1:8,
>    function(i)serialize(x[[i]],NULL)) has no effect.
>
>
>

-- 
Luke Tierney
Chair, Statistics and Actuarial Science
Ralph E. Wareham Professor of Mathematical Sciences
University of Iowa                  Phone:             319-335-3386
Department of Statistics and        Fax:               319-335-3017
    Actuarial Science
241 Schaeffer Hall                  email:   luke-tierney at uiowa.edu
Iowa City, IA 52242                 WWW:  http://www.stat.uiowa.edu