On Jul 20, 2013, at 04:30, Vincent Habchi wrote: > I am beginning to patch some ports to add a special variant called ‘perf’ > that picks clang-3.3 and dragonegg-3.3-gfortran-4.8 (if necessary) instead of > standard compilers then replace standard flags by ‘-O3 -march=native’. The > idea is to provide a way for users that seek performance to optimize the > executable code by selecting the most recent compilers and aggressive > optimization flags (the above combination enables loop vectorizing and AVX > instructions if possible). Obviously, this is only relevant to some math or > science ports; possible candidates include: metis, cgal, boost (?), geos, > atlas, arpack… > > Since this option results in non-portable code (because of the -march=native > flag) it will always be disabled by default, so the buildbot can carry on > cranking out widely compatible code. > > Thoughts?
Would the idea be to make this "perf" variant replace some of the other variants in the atlas port? That would be good. _______________________________________________ macports-dev mailing list [email protected] https://lists.macosforge.org/mailman/listinfo/macports-dev
