@minhqdao

I'll have a look into it

Hi, I wanted to share a thought and ask for opinions. I'm using the npy reader and for my application I can run the process on multiple processes with MPI. I need all processes to retrieve the same data from my list of .npy at the very beginning. So for testing I added the shared option to the open statement as I'm using intel compilers, it works perfect.

So, given that each compiler has different extensions for shared IO, I was wondering if it would be worth adding this features as an optional argument to load_npy which would then be translated to the appropriate option compiler-wise. Any thoughts?

Thanks