search

omlins / ParallelStencil.jl

Package for writing high-level code for parallel high-performance stencil computations that can be deployed on both GPUs and CPUs
BSD 3-Clause "New" or "Revised" License
311 stars 31 forks source link

Example for init_global_grid_usage #89

Closed LaurentPlagne closed 1 year ago

LaurentPlagne commented 1 year ago

how do you to launch an example (e.g. diffusion3D_multigpucpu_novis.jl) on multiple processes ?

luraess commented 1 year ago

That's rather an ImplicitGlobalGrid related thing. Depending on your server install you can simply launch diffusion3D_multigpucpu_novis.jl using e.g.

mpirun -n 8 julia --project ./diffusion3D_multigpucpu_novis.jl

Check before hand your MPI install is working and you may enable GPU-aware MPI upon setting IGG_CUDAAWARE_MPI=1.

LaurentPlagne commented 1 year ago

Thank you very much !

I read the README.mdin both ParallelStencil.jl and ImplicitGlobalGrid.jl and I could not find the command you proposed. It could be interesting to add a section on how to launch the examples (you type this, you should obtain this).

LaurentPlagne commented 1 year ago

A quick note in case someone fall on the same problem. If your local MPI installation does not match the MPI.jl's one then the command mpirun -n 8 julia --project ./diffusion3D_multigpucpu_novis.jl launches 8 times the same monoprocess execution with a warning. Fixing the MPI mismatch fix the pb.