[molpro-user] Memory Problems

[Andy May]

Sam,

The memory you specify for Molpro is per-core and is the size of the stack
that most of the large arrays are allocated to, but there will be some
additional static memory too.

2000M is approximately 16Gb of memory, so for 32 cores this is 512Gb which
exceeds the maximum available on a single node on your system. See:

http://www.molpro.net/pipermail/molpro-user/2010-April/003723.html

for more details about specifying memory.

Of course you can always tell the scheduler to run less processes per node
so that each process can have more memory.

Best wishes,

Andy

On 16 May 2016 at 23:47, MD Simulation <mdsimulationgroup at gmail.com> wrote:

> Hello Molpro Users,
>
> I'm trying to run an UCCSD(T) on a C4H8O radical-cation compound with
> aug-cc-pVTZ.  This has 414 basis functions.
>
> Each compute node has 32 cores and 126GB usable memory.  I have 16 nodes
> for a total of 512 cpus.
>
> I believe I am having memory problems because the calculation terminates
> saying something along the lines of "add X words for a total of Y words".
> (I don't have the full error right now, at home, but I will tomorrow).
>
> While using all 512 cpus, I've tried setting the memory from 250Mwords to
> 2000Mwords but I can't seem to get the calculation to run.
>
> Are there any suggestions in how to handle the memory?
>
> Thanks,
> Sam
>
>
>
>
>
> _______________________________________________
> Molpro-user mailing list
> Molpro-user at molpro.net
> http://www.molpro.net/mailman/listinfo/molpro-user
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://www.molpro.net/pipermail/molpro-user/attachments/20160517/8e36fe84/attachment.html>