Re: more non-blocking collective discussion

Eric Salo (salo@mrjones.engr.sgi.com)
Fri, 7 Feb 1997 13:23:29 -0800

> This is nice as long as there is no try_connect to go along with it.
> Otherwise, if a try_connect and a try_accept are hoping to synchronize,
> it may take a while. :-) Same scenario with a blocking connect and
> a collective try_accept that can only succeed if all acceptors are
> "there" as well.

Yes! This is exactly the kind of ugliness we get into by trying to reinvent the
wheel, IMHO. Unless we are able to provide every feature that is currently
available in the BSD socket interface, someone will be unhappy. And slapping an
MPI wrapper around every possible variation will be extremely messy.

Since a tidal wave of support has yet to appear for the SAPHIR_JOIN idea, I'd
like to propose the following alternative: How about if we only define
blocking, collective versions of ACCEPT and CONNECT to handle the simple/common
cases, but then also add SAPHIR_JOIN to provide all of the fancier
functionality for those who need it? Seems like a solid, middle-ground
solution.

-- 
Eric Salo         Silicon Graphics Inc.             "Do you know what the
(415)933-2998     2011 N. Shoreline Blvd, 8U-802     last Xon said, just
salo@sgi.com      Mountain View, CA   94043-1389     before he died?"