This is the mail archive of the gcc-patches@gcc.gnu.org mailing list for the GCC project.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]
Other format: [Raw text]

Re: Deprecate -funit-at-a-time


> Richard Guenther wrote:
> 
> >>I am not sure how to allow the message.  Also I noticed that
> >>-fno-unit-at-a-time was always documented as being dropped in future
> >>release and kernel+glibc already switched to other options, so perhaps
> >>we can just go ahead and drop it now.  It would simplify maintenance of
> >>cgraph code for sure.
> >
> >That would be of course nice, as this is the point of the whole excercise.
> >I vote +1 for doing that.  Mark, what do you think?
> 
> I think that we could drop it now, given that we've already documented 
> it as obsolete.  I think that we should implement "drop" by continuing 
> to accept the option, so that existing Makefiles and such work.  In 
> fact, I think we should probably treat it as equivalent to 
> -fno-toplevel-reorder.  From the user's point of view, that's the 
> visible effect of -fno-unit-at-a-time.

It is not exactly true, since, for instance, kernel people was more
concerned about inlining functions called once causing their stack frame
usage to grow than about reordering for a while.
However I will be happy to change the spec line I have to imply
-fno-toplevel-reorder -fno-section-anchors and output message (or not
output it and just silently accept it as alias).

Concerning the testcases, those are really directed to no-unit-at-a-time
problems.  Shall I just remove them or modify to use
-fno-toplevel-reorder?

Honza
> 
> -- 
> Mark Mitchell
> CodeSourcery
> mark@codesourcery.com
> (650) 331-3385 x713


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]