STRESS tensor

[carlo antonio pignedoli]

Dear all,
is there a way to have the stress tensor in
a GEO_OPT run for a bulk
calculation whithin QS DFT?

Ciao

Carlo

[Rad]

I just ran a geo_opt calculation using a version of cp2k that I
compiled on January 24th and included "STRESS TENSOR"
keyword under force evaulation section. The output didn't have any
stress tenor printed. Here is my input:

&GLOBAL
BLACS_GRID SQUARE
PROGRAM_NAME CP2K
PROJECT_NAME CD-NVE
# TRACE T
RUN_TYPE GEO_OPT
PREFERRED_FFT_LIBRARY FFTW
&END GLOBAL
&MOTION
&GEO_OPT
OPTIMIZER CG
MAX_ITER 200
&END GEO_OPT
&END MOTION
&FORCE_EVAL
STRESS_TENSOR ANALYTICAL
METHOD Quickstep
&DFT
BASIS_SET_FILE_NAME BASIS_MOLOPT
&MGRID
NGRIDS 9
CUTOFF 350
&END MGRID
&QS
MAP_CONSISTENT T
&END QS

Teo - what am I missing in my input?

Thanks
Rad


On Feb 2, 8:02 am, "carlo antonio pignedoli" <c.pigned...@gmail.com>
wrote:

On Feb 2, 8:02 am, "carlo antonio pignedoli" <c.pigned...@gmail.com>
wrote:

[Teodoro Laino]

Ciao Carlo, Rad,

The printing of STRESS tensor was not enabled in the GEO_OPT module.
It is now.. just update and recompile..

As Rad said you've to specify the "STRESS_TENSOR" keyword and you've to
switch on the print_key

&MOTION

&PRINT
&STRESS
&END
&END
&END

Cheers,
Teo

[carlo antonio pignedoli]

Ciao Teo, thanks a lot!

Carlo

[Rad]

Teo,

It is still not printing the pressure/stress tensor. I pulled out a
copy of cp2k from the cvs respository (not the nightly build) and
compiled it for the run. When will the nightly build with the updated
version be available?
Is there a way I can quickly look at the code (for specifice code
changes) to make sure that I am working with the correct version that
prints the tensor in an optimization run?

Thanks
Rad

> >> Carlo- Hide quoted text -
>
> - Show quoted text -

[Teodoro Laino]

Hi Rad,

On 4 Feb 2008, at 14:47, Rad wrote:

It is still not printing the pressure/stress tensor. I pulled out a

copy of cp2k from the cvs respository (not the nightly build) and

compiled it for the run. When will the nightly build with the updated

version be available?

I have really no idea rad.. It should be done every night but that is out of my control.

Unfortunately it depends on many things and due to the relatively instability of berlios

this could be delayed .. 

Is there a way I can quickly look at the code (for specifice code

changes) to make sure that I am working with the correct version that

prints the tensor in an optimization run?

If I'm not wrong (pls correct me in case) you should see when running cp2k the LAST CVS entry

at the very beginning of the output file.. you can check with the cvs homepage to see if that file corresponds

really to the one you're interested..

Anyway I'm able to print the stress tensor with the yesterday version.. Please check about the keywords

you've to specify (see the previous message)..

Otherwise just post the input file and I will have a look.

Ciao,

Teo

[Axel]

On Feb 4, 8:47 am, Rad <rad.b...@arl.army.mil> wrote:
> Teo,
>
> It is still not printing the pressure/stress tensor. I pulled out a

rad,
please check the files that are created. stress is by default
written to file and not to the regular output.

> copy of cp2k from the cvs respository (not the nightly build) and
> compiled it for the run. When will the nightly build with the updated
> version be available?

why don't you create your own nightly builds? or on-demand builds?

just create a small script and have it executed through cron.
here is what i run (manually) to update and compile on my desktop.
executables for production are then copied over from the exe/
directory (after testing them ok).

#!/bin/sh

PATH=${HOME}/bin:${PATH}
export PATH

cvs update -d
pushd makefiles
make ARCH=Linux-x86-64-intel popt
make ARCH=Linux-x86-64-intel sopt
make ARCH=Linux-x86-64-intel ssgl
make ARCH=Linux-x86-64-g95 sopt
make ARCH=Linux-x86-64-g95 ssgl
make ARCH=Linux-x86-64-g95 sdbg
make ARCH=Linux-x86-64-intel leap2fist
make ARCH=Linux-x86-64-intel fes
popd

test exe/Linux-x86-64-intel/cp2k.popt -nt ChangeLog && perl ./tools/
cvs2cl.pl -b

> Is there a way I can quickly look at the code (for specifice code
> changes) to make sure that I am working with the correct version that

a very useful tool to follow the changes in a cvs archive is 'cvsps'.
it creates "changesets" from cvs commits and you can easily generate
diffs for those changesets. very useful to see what people are doing.

cheers,
axel.

[Rad]

I am sorry Teo, you are right. The stress was printed in separate
file. I am still getting used to look for it.


Axel - we can't do cvs from our environment, can't do cron jobs
either. There are so many restrictions. Nightly build download from
the web is the easiest option. When i am working from home then I can
do cvs update
from an university machine and move the stuff back to our hpc
mcahines. At workplace we are restricted to only few options.

Thanks
Rad

[Axel]

On Feb 4, 12:29 pm, Rad <rad.b...@arl.army.mil> wrote:
> I am sorry Teo, you are right. The stress was printed in separate
> file. I am still getting used to look for it.
>
> Axel - we can't do cvs from our environment, can't do cron jobs
> either. There are so many restrictions. Nightly build download from
> the web is the easiest option. When i am working from home then I can
> do cvs update
> from an university machine and move the stuff back to our hpc
> mcahines. At workplace we are restricted to only few options.

you just have to be a little more creative!
if you can download from a web sever you can have your own nightly
builds on another machine where you have access and then upload the
executable to some webserver. of course this can even be done from
home or elsewhere. it just has to be set up to run
(semi-)automatically.

a.