[O-MPI users] [Beowulf] MPI ABI

Greg Lindahl lindahl at pathscale.com
Mon Oct 10 12:16:28 PDT 2005


On Mon, Oct 10, 2005 at 02:43:15PM -0400, Patrick Geoffray wrote:

> Furthermore, there is no standard binary interface in Fortran, so you 
> cannot safely use a library compiled with one compiler and an 
> application compiled with another.

Our InfiniPath MPICH-based libraries have this solved on i386 and
AMD64: we have a single MPI library that works with PathScale, g77,
PGI, and Intel's Fortran compilers. So there's your existence
proof. There were no issues other than the ones I've brought up: name
mangling, LOGICAL, command line.

> In retrospect, the choice of the MPI forum to not put constraints on the 
> implementation was a good one. Now that MPI is pervasive, it's 
> legitimate to care about an ABI, but it would certainly have been a 
> weight for the adoption of the standard back in the days.

I whole-heartedly agree with this -- I hope no one thinks that the
notion of an MPI ABI implies that the MPI forum did something wrong.

-- greg




More information about the Beowulf mailing list