Re: dynamic counter-proposal
William Gropp (gropp@mcs.anl.gov)
Sat, 11 May 1996 02:16:14 -0500
| > I consider it a major departure because it changes MPI_Spawn, the most
| > fundamental function in the chapter, in a fundamental way.
|
| This has been mentioned before, but I'll just ask it again here because I am
| hopeful that it might bring us all quickly into agreement:
|
| Would adding an INDEP/NOINDEP flag to the new spawn call make it more
| acceptable?
|
Possibly, but it would depend on how the flag interacted with the
detach/attach calls. For it not to constrain the implementation, there would
probably need to be additional constraints connecting the two calls, which
would introduce more opportunities for the user (and the documentation) to
make an error. And it will look like a wierd separation of function.
Bill