You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
That seems like it could be a good idea. Something like immediate_receive(count)?
There's also the immediate_receive_into_*() set of functions, which would allow you to do an MPI_IRecv with a count. These take a mutable reference to some buffer of a fixed size and then return a request. To ensure that this buffer isn't dropped while MPI is still using it, you have to wrap your code in a closure that takes a Scope object and ensure that you allocate any buffers outside the closure. Before exiting the closure, all requests must be waited on, or else it will panic.
examples/immediate.rs is a good example that demonstrates this.
For the current version of immediate_receive(), I think you might be able to work around this if you derived Equivalence on a struct containing a buffer. Something like this:
Hi all,
First off, thanks for maintaining this package!
I'm trying to understand why
immediate_receive()
is hardcoded with a count of 1 - I think this makes it impossible to receive a vector of, say,u8
.a) is there a reason for the restriction?
b) is there another way to do an
MPI_IRecv
with a count?If the answer to a) is "There's no particular reason", then I might just try my hand at a PR (I'll probably need some help).
Cheers!
The text was updated successfully, but these errors were encountered: