On Apr 15, 2016, at 8:49 AM, Kawashima, Takahiro <t-kawash...@jp.fujitsu.com> wrote: > >> I just checked MPICH 3.2, and they *do* include MPI_SIZEOF interfaces for >> CHARACTER and LOGICAL, but they are missing many of the other MPI_SIZEOF >> interfaces that we have in OMPI. Meaning: OMPI and MPICH already diverge >> wildly on MPI_SIZEOF. :-\ > > And OMPI 1.6 also had MPI_SIZEOF interfaces for CHARACTER and LOGICAL. :-) > > > https://github.com/open-mpi/ompi-release/blob/v1.6/ompi/mpi/f90/scripts/mpi_sizeof.f90.sh#L27
Fair point. Doing a little spelunking through git history, I can't find a reason cited as to why https://github.com/open-mpi/ompi/commit/d7eaca83fac0d9783d40cac17e71c2b090437a8c removed the support for CHARACTER and LOGICAL in the mpif.h and use mpi interfaces. However, it does look like when we added the use mpi_f08 interface (https://github.com/open-mpi/ompi/commit/253444c6d0cb105ce6d4e9fb81da42dbb8904d18), we *only* added interfaces for the numeric types -- we did not add interfaces for CHARACTER and LOGICAL (although no reason was cited in the commit logs as to why). Here's what I'm guessing happened: - When Craig and I were writing the mpi_f08 stuff, we saw the same definition for MPI_SIZEOF that I cited earlier: only numeric types - So we only wrote interfaces for the numeric types for mpi_f08 - When the MPI_SIZEOF stuff was later totally revamped, we kept the "only numeric types" philosophy -- Jeff Squyres jsquy...@cisco.com For corporate legal information go to: http://www.cisco.com/web/about/doing_business/legal/cri/
