Re: definition of collective communication on intercommunicators
Tony Skjellum (tony@Aurora.CS.MsState.Edu)
Sun, 3 Dec 1995 22:25:13 -0600
It is clear that our interface looks half-duplex, and Marc's is
full-duplex. One should also consider whether more protocol is
needed for the full-duplex interface or not. I agree with Nathan
that the simplification is quite nice, but that it is not as trivial
as Marc made it out to be. Therefore, we should recast the whole
section to reflect on these options.
I found the receipt of this message both timely and helpful in
view of the upcoming talk at Supercomputing '95.
-Tony