Re: [patch, rfc] PR32172 - Update -funsafe-math-optimizations flag definition

[Brooks Moses <brooks dot moses at codesourcery dot com>]

Richard Guenther wrote:
> On 6/3/07, Dominique Dhumieres <dominiq@lps.ens.fr> wrote:
> > Thanks for the patch. However in my opinion reordering transformations and
> > noy obeying parentheses are two different issues: optimizing f+a-f as a is one
> > thing, optimizing (f+a)-f is another (bad) one.
>
> There is no difference between both cases for gcc.

Internally within the compiler, perhaps.  From the user's point of view, 
however, there is a big difference -- and, since the documentation is 
for the user, it's important to be clear that the "arbitrary reordering 
transformations" allowed by -funsafe-math-optimizations include 
disobeying parentheses.

- Brooks