> I believe the concern is that we not define what the MPI_MAX/MIN of a
> complex variable be. We removed this in the 1.1 spec (June 12, 1995,
> p. 116, line 23). The statement at the meeting (maybe the one I said)
> may have been too general for what I was really concerned about.
Right. Since, in the abstract, complex numbers cannot be ordered, it
is erroneous to define max/min of them. Thus, we just have to specify
how complex numbers are handled in reductions. My previous reference
to page 115 is the one you indicate here, I believe (for some reason,
the copy of the June 12 doc I have does not seem to be paginated
correctly).
Are we saying the same thing?
Regards,
Andrew Lumsdaine