Dear lammps users,
I tried building gpu acceleration for the RTX 3080. On the official website, the architecture for Ampere is sm_80. However, when I try to build a gpu package using traditional make such as make lib-gpu args="-b -m mpi -a sm_80 -p double", the error message says “invalid choice sm_80”. I tried lammps stable version of 29 Oct 2020. Can anyone help me to figure this problem out?
With my best regards
Dear lammps users,
I tried building gpu acceleration for the RTX 3080. On the official website, the architecture for Ampere is sm_80. However, when I try to build a gpu package using traditional make such as make lib-gpu args="-b -m mpi -a sm_80 -p double", the error message says “invalid choice sm_80”. I tried lammps stable version of 29 Oct 2020. Can anyone help me to figure this problem out?
You need a newer version of the LAMMPS source for that step to work with a new architecture. Please try the latest patch version 14 May 2021, for example.
axel.