problem with hybrid sw and lj with GPU get

Sorry for my mistake to give these information
1. the version I used often is 15May2015 , but i also try the 3May2016 version
    and also have these problem
2. (a) Linux version 2.6.32-573.12.1.el6.x86_64 ([email protected])
         (gcc version 4.4.7 20120313 (Red Hat 4.4.7-16) (GCC) )
    (b) CPU : Intel(R) Core(TM) i7-4770 CPU @ 3.40GHz ( 8 core)
    (c) GPU : Tesla C2075
                   Nvidia-smi 352.68
    (d) the package is gpu on pair force ( sw/gpu and lj/gpu)
    (e) at first, I use the double precision, and also single give me the same problem
thank for your help

Yi-Xian

Sorry for my mistake to give these information
1. the version I used often is 15May2015 , but i also try the 3May2016 version
    and also have these problem
2. (a) Linux version 2.6.32-573.12.1.el6.x86_64 ([email protected])
         (gcc version 4.4.7 20120313 (Red Hat 4.4.7-16) (GCC) )
    (b) CPU : Intel(R) Core(TM) i7-4770 CPU @ 3.40GHz ( 8 core)
    (c) GPU : Tesla C2075
                   Nvidia-smi 352.68
    (d) the package is gpu on pair force ( sw/gpu and lj/gpu)
    (e) at first, I use the double precision, and also single give me the same problem
thank for your help

thanks for providing this information and i can confirm your issue
with the latest development sources on my desktop.
unfortunately, i don't see an obvious fix. this will need some careful
debugging by an expert in programming the GPU package and some time.

please stay tuned,

axel.

Hi Yi-Xian,

thanks for reporting the issue and sending the input deck. Please apply the attached patch to the LAMMPS directory (at the same level with src/, lib/, etc.):

patch -p1 < hybrid-3body.diff

Then, go to lib/gpu:

make -f Makefile.your_machine veryclean

make -f Makefile.your_machine -j4

Go to src/:

make package-update
make your_machine

Let me know if this patch resolves the issue you got. If so, I will submit a pull request on github.

Best,
-Trung

hybrid-3body.diff (69.1 KB)

hi trung,

i've tested your patch against the input deck available at

https://github.com/lammps/lammps/issues/69

and can now complete that simulation without a problem. i'll be
including the changes (and i've removed some dead code that the
compiler warned about in lal_sw.cu) in lammps-icms now.

axel.

Great! Thanks Axel!

-Trung