Re: RFC: #pragma optimization_level

[Geoffrey Keating <geoffk at geoffk dot org>]

Dale Johannesen <dalej@apple.com> writes:

> On Apr 1, 2005, at 1:38 PM, Geoffrey Keating wrote:
> > If you do get a solution for preprocessor macros, you can remove the
> > restriction that the -Ox settings must match for PCH files, another
> > much-requested feature.
> 
> This is outside the scope of what I'm planning to do.

No, I mean that "if you can do this for -Ox, then all you need to do
for PCH is document that it works".

Likewise, as you verify that each optimization pass controlled by a -f
flag works in this environment, you can add it to the list of -f flags
that are PCH-safe.

> >   Another example: what happens if strict-aliasing is
> > switched off half-way through the compilation?  And then back on
> > again?  Look at the assert in mems_in_disjoint_alias_sets_p.
> 
> Another good point.  That optimization may be one that has to be
> disallowed, for the first cut.  Unfortunately, controlling it may very
> well be something people will want to do.  (Limiting the pragma
> to -On may keep them from asking for it :)

I mentioned strict-aliasing because it is part of -O2 but not -O1.

It sounds like what you'll really end up with is a *different* set of
optimisation levels than the command-line -O flags produce.  It seems
like that might be confusing for users.