I am using the 15May2020 Patch for the LAMMPS & KOKKOS packages.
The error message I am recieving is the following:
ERROR: Cannot yet use pair hybrid with Kokkos (src/pair_hybrid.cpp:425)
Last command: pair_coeff * * reax/c/kk ffield.reax.SiOCH Si NULL NULL
Is this due to an incorrect version of LAMMPS I am using ?
[Re - Sent to the mailing list, sorry for sending it a 2nd time with the correct receivers]
no, this specific error due to an incorrect use of LAMMPS.
if you are not using suffixes via command line flags of the suffix command, but apply them explicitly in your input, then you must use pair style hybrid/kk instead of hybrid to get the KOKKOS version and proper KOKKOS support for substyles.
As Axel said, the easiest is to use “-sf kk” on the command line, which enables the Kokkos suffix for all supported styles. See https://lammps.sandia.gov/doc/Speed_kokkos.html for more details. Let us know if you have other issues.
In fact my full command line is the following one:
mpirun -np 8 $PROGNAME -pk kokkos newton on neigh half comm no -k on g 2 -sf kk -in in.si.sputtering
I have removed the /kk of my script and put the -sf kk flag, but I still receive the error message saying “ERROR: Cannot yet use pair hybrid with Kokkos (src/pair_hybrid.cpp:425)”.
On the other side, I have removed the -sf kk option and but /kk for all accepted commands (so with hybrid/kk reax/c/kk morse/kk, as well as my fixes and qeq/reax/kk) but the same error message appears (explicitly saying that hybrid cannot be used with KOKKOS). I am still using LAMMPS version of the 15May2020.
Is it due to a wrong compilation of my KOKKOS package ? Since you’re saying that hybrid has KOKKOS support since a few years, I don’t understand where is my mistake.
Thank you for your answers,
“Moore, Stan” —02/07/2020 22:03:38—As Axel said, the easiest is to use “-sf kk” on the command line, which enables the Kokkos suffix fo
indeed, there appears to be an oversight in how the test is applied.
as a temporary workaround, please delete the test in the source (lines 424 and 425 in src/pair_hybrid.cpp) and recompile.
that will allow you to run this setup under any condition (even bad ones), so please be careful.
we will work to have a better check in the next patch release.
Alright I will do the recompilation and come back if there is further problems. Thank you for your answers.
“Axel Kohlmeyer” —03/07/2020 15:10:58—indeed, there appears to be an oversight in how the test is applied. as a temporary workaround, plea