I don't understand the temporary buffers. From what I can see, they don't belong to the application because the application itself has no need or knowledge of them. But if that is the case, how are these buffers ever allocated? Is the buffer defined at compile time or is it only allocated when an MPI function is called? I assume the buffer only exists if you look at the internals of the MPI functions?
ferozenaina
I would think the buffers are implemented inside MPI internally. MPI acts as a message passing layer and would allocate these buffers dynamically in memory.
tcm
Yes, they are inside the message-passing software layer (e.g., MPI).
I don't understand the temporary buffers. From what I can see, they don't belong to the application because the application itself has no need or knowledge of them. But if that is the case, how are these buffers ever allocated? Is the buffer defined at compile time or is it only allocated when an MPI function is called? I assume the buffer only exists if you look at the internals of the MPI functions?
I would think the buffers are implemented inside MPI internally. MPI acts as a message passing layer and would allocate these buffers dynamically in memory.
Yes, they are inside the message-passing software layer (e.g., MPI).