lammps-daily can run with reax type force fields?

This is what the builder (Anton Gladky) of lammps-daily ppa has to say about above question. I forwarded to mailing list as it may helpful for other user of lammps-daily ppa.
Best wishes,
Jeffrey

Hi,

packages for ubuntu are building with the following command [1]:

cd src; (MAKE) yes-all; (MAKE) no-lib; $(MAKE) openmpi

As I understand, all modules should be compiled during this build.
If it is not, please, let me know, what should be done to enable “reax”.

[1] http://bazaar.launchpad.net/~gladky-anton/lammps/lammps-debian/view/head:/rules#L15

Best regards,

Anton

PS Feel free to repost this email to Axel or Lammps-mailing list, if you want.
Do not forget to CC me, I am not subscribed there.

reax requires an external library built before making LAMMPS, and when
you use "make no-lib" you have disabled reax.

Why not use reax/c?

Ray

jeffrey,

first off, let me reiterate that the reax pair style *should* not be
used unless there is a very, *very* good reason (and i currently
cannot think of one). the fortran based version provides no additional
functionality, runs slower, occasionally needs adjustments to array
dimensions and needlessly blows up stack space requirements.

thus particularly prebuild binary packages *should* not include the
reax package simply to enforce the issue, since users of prebuilt
executable packages are even less likely to read the documentation and
cannot easily adjust the fortran library to their needs. reax/c has
been thoroughly tested to yield consistent results with reax.

This is what the builder (Anton Gladky) of lammps-daily ppa has to say about
above question. I forwarded to mailing list as it may helpful for other user
of lammps-daily ppa.
Best wishes,
Jeffrey

Hi,

packages for ubuntu are building with the following command [1]:

cd src; \(MAKE\) yes\-all; (MAKE) no-lib; $(MAKE) openmpi

As I understand, all modules should be compiled during this build.

that is not quite correct. building an as complete as (technically and
legally) possible LAMMPS executable requires a bit more fiddling and
tweaking.
you can see what is done with the corresponding RPM builds here:
http://git.icms.temple.edu/git/?p=lammps-icms.git;a=blob;f=tools/rpm/lammps.spec;hb=HEAD

If it is not, please, let me know, what should be done to enable "reax".

as explained above, you *should* not enable reax in your precompiled
packages. it is a "bad idea(tm)". so far nobody has yet been able to
offer a convincing reason to not use reax/c instead.

that being said, it is possible to add more packages to the build that
*do* make sense, and i am happy to provide advice on how to do that
off-list, if the link to the build instructions of the spec file for
LAMMPS-ICMS is not sufficient.

axel.

axel.

Axel and Ray
Thank you for the explanations. My simulation works well with reax/c. However, Van Duin and co-workers are actively update their reaxFF force fields for specific problems which may be helpful to simulate certain systems (no specific example in my mind though) without having bogus results as they tested them for the system of interest.
Best wishes,
Jeffrey

jeffrey,

first off, let me reiterate that the reax pair style should not be
used unless there is a very, very good reason (and i currently
cannot think of one). the fortran based version provides no additional
functionality, runs slower, occasionally needs adjustments to array
dimensions and needlessly blows up stack space requirements.

thus particularly prebuild binary packages should not include the
reax package simply to enforce the issue, since users of prebuilt
executable packages are even less likely to read the documentation and
cannot easily adjust the fortran library to their needs. reax/c has
been thoroughly tested to yield consistent results with reax.

This is what the builder (Anton Gladky) of lammps-daily ppa has to say about
above question. I forwarded to mailing list as it may helpful for other user
of lammps-daily ppa.
Best wishes,
Jeffrey

Hi,

packages for ubuntu are building with the following command [1]:

cd src; (MAKE) yes-all; (MAKE) no-lib; $(MAKE) openmpi

As I understand, all modules should be compiled during this build.

that is not quite correct. building an as complete as (technically and
legally) possible LAMMPS executable requires a bit more fiddling and
tweaking.
you can see what is done with the corresponding RPM builds here:
http://git.icms.temple.edu/git/?p=lammps-icms.git;a=blob;f=tools/rpm/lammps.spec;hb=HEAD

If it is not, please, let me know, what should be done to enable “reax”.

as explained above, you should not enable reax in your precompiled
packages. it is a “bad idea™”. so far nobody has yet been able to
offer a convincing reason to not use reax/c instead.

that being said, it is possible to add more packages to the build that
do make sense, and i am happy to provide advice on how to do that
off-list, if the link to the build instructions of the spec file for
LAMMPS-ICMS is not sufficient.

axel.

axel.

Axel and Ray
Thank you for the explanations. My simulation works well with reax/c.
However, Van Duin and co-workers are actively update their reaxFF force
fields for specific problems which may be helpful to simulate certain
systems (no specific example in my mind though) without having bogus results
as they tested them for the system of interest.

what does that mean? what bogus results and where? please don't speak
in riddles.

thanks,
     axel.