Just to clarify: we primarily use c99 features in our plugins as a means of directly specifying which functions are being implemented, and which are not. In c89, this can only be done by maintaining positional alignment - c99 allows us to do this using the function names. Thus, the c99 method is much more maintainable, which was a very large factor in our decision. It wasn’t just cosmetic.
We have not found anyone raising an issue with this decision in the four years since it was implemented. This includes installations running legacy Fortran and C codes, all of which have updated their OMPI installations. So I don’t think it is quite the dire situation you are implying. > On Aug 29, 2016, at 8:32 AM, C Bergström <cbergst...@pathscale.com> wrote: > > On Mon, Aug 29, 2016 at 11:22 PM, Jeff Squyres (jsquyres) > <jsquy...@cisco.com <mailto:jsquy...@cisco.com>> wrote: >> On Aug 29, 2016, at 11:06 AM, C Bergström <cbergst...@pathscale.com> wrote: >>> >>> If the patches are performance impacting I would never burden >>> upstream, but I do hope that regardless you'll consider them. Based on >>> the patch for 1.x it seems cosmetic. I'll take the most honest and >>> unbiased look at the patches against 2.x and master to see if I feel >>> guilty for asking for review. >> >> We've used a lot more C99 in master/v2.x (i.e., since we forked for v1.7). >> It would be a much, much harder sell to remove all the C99 from there. >> >> Also, if SLES 10 is EOL, that also somewhat detracts from the desire to add >> a bunch of engineering work to support a 27-year-old version of C. >> >> As it is, I am surprised that your patches are so small for v1.10 -- that >> can't possibly remove all the C99 stuff from the entire code base. Are you >> are only selectively removing *some* of the C99 from the parts of Open MPI >> that you are compiling that make it work on your compiler? If so, that's a >> bit more of an oddball case: i.e., you're not proposing strict C89 adherence >> across the entire code base. > > I'm not intentionally doing anything oddball. I am however testing > with clang and our compiler. If upstream clang is doing anything > oddball (which I'd be a bit surprised) then we probably follow along. > > The features you're using in c99 appear to be cosmetic candy and > non-performance impacting. 27-year-old standards and older are quite > frequently still in use for HPC. *cough* Fortran *cough*... > ----------- > Based on the latest response - it seems that we'll just fork OMPI and > maintain those patches on top. I'll advise our customers not to use > OMPI and document why. > > Thanks again > _______________________________________________ > devel mailing list > devel@lists.open-mpi.org <mailto:devel@lists.open-mpi.org> > https://rfd.newmexicoconsortium.org/mailman/listinfo/devel > <https://rfd.newmexicoconsortium.org/mailman/listinfo/devel>
_______________________________________________ devel mailing list devel@lists.open-mpi.org https://rfd.newmexicoconsortium.org/mailman/listinfo/devel
