I found the error, at least vor MPI_type_vector:
the stride is the one between elements, not blocks, therefore
call MPI_TYPE_VECTOR(NK, NI, NI*NJ, MPI_REAL8, y_slice_vec, err)
cheers
Hi there, I hope someone can help me here. This is the problem:
I need to communicate between processes a slice of a 3D array (i,j,k). The slice is a plane with j=constant. The memory is not contiguos on the slice since j changes faster then k.
the dimensions of the array are (NI,NJ,NK)
My...
This site uses cookies to help personalise content, tailor your experience and to keep you logged in if you register.
By continuing to use this site, you are consenting to our use of cookies.