On Wed, Apr 06, 2011 at 05:41:48PM -0600, Matthew Burgess wrote: > On Wed, 6 Apr 2011 19:29:48 -0400, Jeremy Huntwork > <jhuntw...@lightcubesolutions.com> wrote: > > On Apr 6, 2011, at 6:14 PM, Ken Moffat <k...@linuxfromscratch.org> wrote: > > > >> From my viewpoint (minimalist, extra packages which aren't > >> mandatory need to be justified), the lack of applications (that I > >> know of, maybe there is now something) which detect and use > >> Cloog-whichever was the problem. > > > > The number of applications directly using these libs isn't now, nor likely > > ever to be, the argument for inclusion. It will be rare to find more > > packages that want or need these. > > > > The main driving force for wanting these is the compilation optimizations > > and performance gains that they bring to the toolchain. > > But, my current understanding (from a quick grep of the GCC manual) is that > those > optimizations have to be explicitly requested. The following appear to be > dependent on > CLooG/PPL: > > -ftree-loop-linear > -floop-interchange > -floop-strip-mine > -floop-block > -fgraphite-identity > -floop-parallelize-all > > None of those flags are enabled by any of the more commonly used -O, -O2, -O3 > switches, > so they'd have to be explicitly requested in a package's Makefiles. > > Regards, > > Matt. > Thanks Matt, that matches my memory of what I found when I last looked at this.
ĸen -- das eine Mal als Tragödie, das andere Mal als Farce -- http://linuxfromscratch.org/mailman/listinfo/lfs-dev FAQ: http://www.linuxfromscratch.org/faq/ Unsubscribe: See the above information page
