[Pw_forum] insufficient virtual memory for a Zn48O48 system!!!!!!!!

[Axel Kohlmeyer]

On Mon, 28 Feb 2005 16:47:00 +0800  "Shujun Hu" wrote:

[...]

> Dear Drs,
> I try to do the calculation for a 96-atom system, Zn48O48 (before doping, for the
> purpose of comparing). There are totle 864 electrons, 12 for Zn and 6 for O
> respectively. And the number of bands is set to be 450. After running the memory.x

dear shujun hu,

wow, this is quite a large system.

> system gave such information:
> **********************************************************
>      Program memory    v.2.1.2  starts ...
>      Today is 28Feb2005 at 15:49:49
>      Estimated Max memory (k-point code): 14690.79Mb
>      nonscalable memory =  123.65Mb
>         scalable memory = 3654.75Mb
>      nonscalable wspace =  148.32Mb
>         scalable wspace =10764.07Mb   (diag:10764.07Mb, mix:  287.04Mb)
>      Estimates based on 16-byte complex,  8-byte reals,  4-byte integers
> **********************************************************
> As my pc taking only 512M physical momory, I selected "conjugate-gradient

whoa. with that kind of machine, you'll have to wait _ages_
until you'll see a result. also the machine will grind
your harddrive easily to death.


> Noting that there is insufficient virtual memory (swap partition is working, but
> only 2G for Redhat linux), I set up an 5G active swapfile to add to the virtual
> memory. But same information appeared as above.
>     Maybe such system is too large. I want to know why it cannt work as there is
> enough virtual memory. Should i add the physical memory?

you may have enough virtual memory in theory, but you have to remember
that you have only a 32bit cpu, and that will restrict your maximum
available address space. under the best circumstances (the proper kernel
and a compiler with a 'clean' runtime), there should be a limit of
3GB minus some change for your maximum address space. most of the time,
you'll be limited to 2GB or less. what you need is a 64bit machine 
(or even better a 64bit SMP machine) with at least 8GB of memory.
with anything less, your machine will come almost to a halt, as
swapspace is so much slower than physical memory.

for testing, i'd recommend you cut the system size in half in 
each direction so that you have a unitcell of 12 atoms, take 
a denser k-point mesh and use that to pre-optimize your structure
and get a feeling for what works (and what doesn't) and how long
the 8x larger system will take to compute.

regards,
	axel kohlmeyer.
> 
>     Best wishes
> 
>                               Shujun Hu
> ____________________________
> hushujun at mail.sdu.edu.cn
> 
> 
> 
> 
> 
> 
> 
> _______________________________________________
> Pw_forum mailing list
> Pw_forum at pwscf.org
> http://www.democritos.it/mailman/listinfo/pw_forum
> 

--

=======================================================================
Axel Kohlmeyer      e-mail:  axel.kohlmeyer at theochem.ruhr-uni-bochum.de
Lehrstuhl fuer Theoretische Chemie          Phone: ++49 (0)234/32-26673
Ruhr-Universitaet Bochum - NC 03/53         Fax:   ++49 (0)234/32-14045
D-44780 Bochum                   http://www.theochem.ruhr-uni-bochum.de
=======================================================================
If you make something idiot-proof, the universe creates a better idiot.