* MPI examples *

** On Mac **

$ source .env

(be sure you have llvm & libomp installed, path in .env should match it)

** Build **

$ mkdir build && cd build

$ cmake ..

$ make

to build cuda samples configure with

$ cmake .. -DBUILD_CUDA=ON -DCMAKE_CUDA_ARCHITECTURES=<compute_cap>

where compute cap is

$ nvidia-smi --query-gpu=compute_cap --format=csv compute_cap

times 10 (e.g. 5.2 * 10 = 52)

e.g.

$ cmake .. -DBUILD_CUDA=ON -DCMAKE_CUDA_ARCHITECTURES=52

** Run **

$ mpirun -np <num_processes> <file name>

Example:

$ mpirun -np 4 hello

** Create sample **

create a function and then instantiate the sample as

SAMPLE(function_name, args)