Bill is probably referring to the changes made by IBM for the BG/P. Rajeev On Aug 14, 2012, at 12:55 PM, mpich2 wrote:

#1158: Fortran-Binding for MPI_Comm_get_attr(MPI_APPNUM) broken on bigendian/64 bits ------------------------------------+--------------------------------------- Reporter: nicolai.stange@… | Owner: gropp Type: bug | Status: reopened Priority: major | Milestone: mpich2-1.5 Component: mpich2 | Resolution: Keywords: | ------------------------------------+---------------------------------------

Comment(by gropp):

Well before r6346 - the current code confuses int and MPI_Fint and MPIR_Pint and MPIR_FAint, since the integer values are for the Fortran 77/9x interface respectively. The code is demonstrably incorrect if MPI_Fint is not the same as int, which is valid, we can handle, and some users have requested. A much earlier change removed the distinction between Fortran and C attributes in the code and introduce the problems that r6346 fixed. I'll look into designing and documenting a fix that will handle the various cases (there is an attribute test that Jeff Squyres wrote that checks for the various interlanguage cases).

-- Ticket URL: <http://trac.mcs.anl.gov/projects/mpich2/ticket/1158#comment:16> mpich2 <http://www.mcs.anl.gov/research/projects/mpich2> MPICH2 is a high-performance and widely portable implementation of MPI