r194405 - Improve the documentation of the optimization flags

[Hal Finkel]

----- Original Message -----
> On 11/11/2013 20:32, Arthur O'Dwyer wrote:
> > On Mon, Nov 11, 2013 at 11:04 AM, Sylvestre Ledru
> > <sylvestre at debian.org> wrote:
> >> Author: sylvestre
> >> Date: Mon Nov 11 13:04:47 2013
> >> New Revision: 194405
> > [...]
> >
> >> +=item B<-Ofast>
> >> +
> >> +Enables all the optimizations from B<-O3> along with other
> >> aggressive
> >> +optimizations that may violate strict compliance with language
> >> standards.
> > FWIW, this isn't a terribly helpful description. I assume it's
> > talking
> > (mostly?) about -ffast-math, but as it stands I don't think it
> > makes
> > sense to have an option whose only documentation is "This option
> > may
> > break your code in undocumented ways." ;)
> That is the original description (which does not make it better, I
> agree :)
> I am open to suggestion.

As I recall, this is currently -O3 -ffast-math -fstrict-aliasing, and we might as well say so.

 -Hal

> >> +=item B<-O4>
> >> +
> >> +Equivalent to B<-O3>.
> > You can roll this into the following item for "-O5 and higher". As
> > I
> > understand it, there's no longer anything unique about -O4. (And as
> > below: I don't think Clang intends to guarantee that -O4 will
> > *always*
> > be equivalent to -O3.)
> Well, I left it separated because -O4 used to do something (LTO) and
> it
> might be used again in the future (?).
> >= -O5 never had any meaning.
> Anyway, I saw Chandler's comment, I will fix that.
> 
> >
> >> +=item B<-O5> and upper
> >> +
> >> +Fall back on B<-O3>.
> > s/upper/higher/ (or /greater/, your choice).
> > I would prefer to see this documented as "Currently equivalent to
> > -O3", since (A) you've sensibly gotten away from the "fall back"
> > terminology in the code patch itself, and (B) I don't think Clang
> > intends to guarantee that these options will *always* be equivalent
> > to
> > -O3. It might even make sense to add an explicit caveat, such as
> > "These options may be assigned new semantics in future releases."
> >
> Right, thanks.
> 
> Sylvestre
> _______________________________________________
> cfe-commits mailing list
> cfe-commits at cs.uiuc.edu
> http://lists.cs.uiuc.edu/mailman/listinfo/cfe-commits
> 

-- 
Hal Finkel
Assistant Computational Scientist
Leadership Computing Facility
Argonne National Laboratory