On Tue, Apr 24, 2012 at 03:02:02PM -0500, Bruce Dubbs wrote: > > http://gcc.gnu.org/onlinedocs/gcc/Optimize-Options.html > > -O3 turns on > -finline-functions > -funswitch-loops > -fpredictive-commoning > -fgcse-after-reload > -ftree-vectorize > -fipa-cp-clone > > I know that inline functions can cause problems at times. Personally, I go > with > the upstream default optimizations. I don't think any improvements would be > worth the risk. If I were developing compute intensive sw, I'd pay > attention, > but for a general purpose library, the devs know best. > > Note that I *have* done compute intensive sw and sometimes it does make the > difference of an 18 hour run instead of a 24 hour run. However, I don't > think > the standard libraries are the place for this type of optimization. > > That said, have fun... > > -- Bruce
Thanks. So, why do we do it on 32-bit x86 ? My memory says that Greg proposed it, something along the lines of "it's what upstream recommend", but I could be mistaken. 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
