> > Still, using non-standard hacks like bin32 does not sound like a good idea. > > I think the actual issue is there is no common approach for this on various > platforms. >
On platforms I've seen, there are 2 sets of libraries but only 64 bit executables provided. This is what we had for OFED 1.0, OFED 1.1, and I don't see the reason to change that - adding more executables to install on production systems will double QA work. > > > Maybe an option to *only* make 32 bit userspace might make sense though. > > Something like --disable-32bit, --disable-64bit. > > This would solve your problem, would it not? > > Does that mean if I don't specify one of them, I'll get 32- and 64bit > execs? > If yes, that's fine. No, by default we build 2 sets of libraries, and only 64 bit execs. But for your development purposes (I think you mentioned testing user/kernel context switch) I think we could have --disable-32bit flag to configure to get only 32 bit userspace. -- MST _______________________________________________ openib-general mailing list openib-general@openib.org http://openib.org/mailman/listinfo/openib-general To unsubscribe, please visit http://openib.org/mailman/listinfo/openib-general
