[O-MPI users] Re: [Beowulf] Alternative to MPI ABI
joachim at ccrl-nece.de
Tue Mar 29 23:36:15 PST 2005
Greg Lindahl wrote:
> On Sat, Mar 26, 2005 at 06:47:41AM -0500, Jeff Squyres wrote:
>>I guess I don't understand your reluctance to accept a MorphMPI-like
> You have repeated your original MorphMPI attributes. I responded to
> them, and I don't see any sign that you've read my response. This is
> not the way discussions are usually held.
Jeff explained very well the problems he sees with your additional
F77-like-C-API (which, BTW, is largely realized in MPICH as all opaque
types there are integer handles): two interfaces means two mpi.h and two
libraries to maintain. I doubt that any implementor is keen on this.
Instead, a layer on top would be easier, and not unrealistic to
maintain: a serious MPI implementation will rarely (never!?) change its
mpi.h/mpif.h. We (for NEC's MPI implemenation across all platforms)
didn't do so for 5 years when we completed the full MPI-2 functionality.
And even before, we only added things, but didn't change existing
Joachim Worringen - NEC C&C research lab St.Augustin
fon +49-2241-9252.20 - fax .99 - http://www.ccrl-nece.de
More information about the Beowulf