Hi Gabriel, I do not fully understand the mechanisms for all processors, and being that is the case, perhaps you can add a lot of comments to configure.ac. I believe that the getconf command uses information from glibc/sysdeps/x86/cacheinfo.c, which uses information from the CPUID instruction of the processor, and tabulated information based on same; for the intel x86 case. I am under the impression that there is no ARM support, as there is none under glibc/sysdeps, and based upon https://bugzilla.redhat.com/show_bug.cgi?id=1190638. (Of course, I am referring to native compilation. And I don't have an ARM here to test.)
The bottom line is that, in my opinion, it would not hurt to add some hints and clues using comments in configure.ac so that somebody has a reasonably easy chance to figure out where the settings are originating from, for all interesting cases. And remember that the reader should not have to depend on searching through git commit comments for that purpose. Thanks, Gabriel. Burt On Fri, Feb 10, 2017 at 5:52 AM, Gabriel Ganne <gabriel.ga...@enea.com> wrote: > Hi, > > > I am currently working on a patch to auto-detect the cache line size with > a m4 macro in autoconf (https://gerrit.fd.io/r/#/c/4576/) > > Part of it updates cache.h tests so that it raises an error if the cache > line length is undefined. > > > I am stuck at the libpneum compilation: it's a python module, so it has > its compilation flags defined in setup.py. > > This means that the vpp flags are not propagated to the libpneum, and I > believe this is a bug. > > I think the python api and the libpneum will get big changes soon anyway > ... > > > How do you think I should handle this ? > > > Best regards, > > > Gabriel Ganne > > _______________________________________________ > vpp-dev mailing list > vpp-dev@lists.fd.io > https://lists.fd.io/mailman/listinfo/vpp-dev >
_______________________________________________ vpp-dev mailing list vpp-dev@lists.fd.io https://lists.fd.io/mailman/listinfo/vpp-dev
