Please do not reply directly to this email. All additional comments should be made in the comments box of this bug report. Summary: Review Request: ladspa-caps-plugins - The C* Audio Plugin Suite https://bugzilla.redhat.com/show_bug.cgi?id=292371 ------- Additional Comments From j.w.r.degoede@xxxxxx 2007-09-21 02:26 EST ------- (In reply to comment #4) > (In reply to comment #3) > > Erm, explain? Passing RPM_OPT_FLAGS is always needed to enable things like > > FORTIFY_SOURCE and stack-smashing protection, which are not only usefull for > > security reasons, but also for catching normal bugs. Normally RPM_OPT_FLAGS > > should be the only optimalisation affecting flags, but when upstream uses > > -ffast-math, that may be added. > > Ah, ok, sorry. > Still, '-funroll-loops' is in upstream but missing in my build logs (unless _I_ > am missing something yet again Above kept for context of question below. > - maybe that compiler option no longer makes sense?) I believe it doesn't, some upstreams tend to use all kind of funky optimalization options, and usually those are a bad idea. For one gcc as in Fedora gets extensively tested with $RPM_OPT_FLAGS, setting other options might trigger unknown bugs. In this specific case -funroll-loops may have helped when upstream last did some benchmarking, but on modern CPU's smaller code is usually faster as modern CPU's are bandwidth limited and smaller code uses less bandwitdh (and cache, and instruction trace cache, and brench predictor slots and ...) -funroll-loops causes code to get bigger and thus usually slower. gcc with -O2 will already unroll some loops, but only when this is beneficially (for example a small loop with 2 iterations might actually have less code when unrolled). I think -O2 includes -funroll-loops, but not -funroll-all-loops, but I'm not sure. Either way usually its a good idea to not out smart the gcc developers and just use -O2. -ffast-math is an exception here as -ffast-math is never enabled by default as it breaks certain aspects of the C-standard causing (very small) precision loss in floating point operations, but since upstream is using it I guess the slightly lower lever of precision is acceptable. -- Configure bugmail: https://bugzilla.redhat.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug, or are watching someone who is. _______________________________________________ Fedora-package-review mailing list Fedora-package-review@xxxxxxxxxx http://www.redhat.com/mailman/listinfo/fedora-package-review