Dear lammps-user:
I have built 3 executables with both the GPU and USER-CUDA packages installed by using different precisions,single,mixed,double.But when I run the script of gpu, the single one is OK,but the mixed and the double ones will get errors,
Cuda driver error 701 in call at file ‘geryon/nvd_kernel.h’ in line 234.
application called MPI_Abort(MPI_COMM_WORLD,-1) - process 0
I have no idea how to solve this problem.
I will very much appreciate any input and help!
Thanks,
Chen
I don't know - I've done this successfully. Maybe Mike
has an idea.
Steve
Hi Chen,
Cuda driver error 701 corresponds to "launch exceeded resources".
Could you provide more details: What is the output of nvc_get_devices?
Which Makefile.foo you are using to compile libgpu.a? Did you do a
"make -f Makefile.foo clean" before compiling libgpu.a with other
precision options? What is the version of lammps you are using? Which
input script generated the crash with mixed and double?
Regards,
-Trung