VMD-L Mailing List

From: Benjamin Bouvier (benjamin.bouvier_at_ibcp.fr)
Date: Tue Oct 14 2008 - 11:34:44 CDT

On Tue, 14 Oct 2008 18:13:11 +0200, Axel Kohlmeyer
<akohlmey_at_cmm.chem.upenn.edu> wrote:

> On Tue, 14 Oct 2008, Benjamin Bouvier wrote:
>
> BB> Hi all,
>
> hi benjamin,
>
> BB> set startlist [ lsort [ glob *.coor ] ]
> BB> foreach start $startlist {
> BB> cd $coordir
> BB> set mnum [ mol load pdb $start ]
>
> please use "mol new" the "mol load" syntax is deprecated
> and will be removed at some point.

OK, good to know...

>
> BB> cd $trajdir
> BB> set traj [ join "[ lindex [ split $start . ] 0 ] dcd" . ]
> BB> mol addfile $traj first $startframe last $endframe step
> BB> $stepframe waitfor all molid mnum
> BB>
> BB> 2. Then comes some analysis (irrelevant to my problem)...
>
> i disagree. the vast majority of memory leak problems so far
> could be traced back to calls to atomselect without the
> proper deletions.

And you are right :-( ... which I found out minutes after posting my
initial message. A rogue undeleted atomselect inside a loop was the
culprit; the memory usage is now a flat curve.

Apologies for posting too fast!

> BB> 3. Before going on to the next trajectory, I try to free up memory:
> BB>
> BB> puts "Done with window $start..."
> BB> animate delete all
> BB> mol delete $mnum
> BB> }
> BB>
> BB> The whole thing seems to work, but memory usage keeps climbing and
> the
> BB> script runs out of memory after 4-5 trajectories... So it seems the
> "mol
> BB> delete" and "animate delete" commands do not release the memory, as I
>
> just "mol delete $mnum" should suffice.
>
> BB> thought they would.
> BB>
> BB> Any clues ?
>
> do you see the same problem _without_ the analysis?
>
> if yes, does it still happen with the latest alpha build?
>
> cheers,
> axel.
>
>
> BB>
> BB> BTW: This is VMD 1.8.6 on 64-bit Linux.
> BB>
> BB> Cheers,
> BB>
> BB> ----
> BB> Benjamin
> BB>
>

Cheers,

Benjamin.