When I run the nvprof command like this
nvprof --concurrent-kernels on --profile-api-trace all --profile-from-start on
–system-profiling off --unified-memory-profiling per-process-device
–track-memory-allocations off --device-buffer-size 8 --device-cdp-buffer-size 8
–openacc-profiling on --devices 0 --aggregate-mode on
–metrics sysmem_write_utilization -o run11
mpirun -np 4 /opt/lammps-22Aug18/src/lmp_ubuntu -sf gpu -in in.lj
The execution time is the same as when I run program normally without nvprof. Also the output file is about 100KB and nothing is shown when I open that with the visual profiler.
Is something missing?