fix 0 all gpu force/neigh 1 1 -1.0

I just tested the latest code (Apr29,2011) and still got the same result.

I have some more recent files from Mike Brown which I haven't
yet released. Maybe that is what Axel is talking about. Will
do so next week.

Steve

I have some more recent files from Mike Brown which I haven't
yet released. Maybe that is what Axel is talking about. Will
do so next week.

it appears to be fixed in the current version already.
i just updated my lammps checkout and the gpu selection works fine.

jan-michael, did you recompile you gpu library after updating the sources?

cheers,
     axel.

Axel,

Yes i did

Jan Michael

Axel,

Yes i did

strange. it _should_ work.

i've double and triple checked that i am not by
accident using any files that have a workaround
implemented.

either i have overlooked something or you,.
or there is something else going on, but
i cannot imagine what.

at this point the only suggestion that i can make
is to compile with debug info and use a debugger
to check which gpu id is computed to be used for
real when the gpu device is opened.

axel.

Axel,

If i did a run with "fix 0 all gpu force/neigh 1 1 -1.0" i get the output:

"Initializing GPU and compiling on process 0...Done.
Initializing GPU 1 on core 0...Done."

which looks like it is using Device 1. But if I look at nvidia-smi
using "nvidie-smi -d" command i get the following output:

ok. you are right. i get the same behavior.
i must have overlooked this in your earlier mail.
my bad. thanks for your persistence.

the patch that i was talking about is thus still missing.
the change is trivial. check out:

http://git.icms.temple.edu/git/?p=lammps-icms.git;a=commitdiff;h=e03cd3d8854a3c9e2466aa321d4120792aee4549

that should fix your problem. please give it a try and let us know.
i was under the impression that it was already applied, but since
i didn't have the time yet, to synchronize LAMMPS-ICMS with the
recent LAMMPS development, i didn't know for sure.

again. my apologies.

axel.

[...]

Ok, thanks for the link. I will test it soon. Thanks again for your help.