This is the mail archive of the gcc@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]

Re: A (minor) doc problem?


> On Aug 15, 2001, Jack Lloyd <lloyd@acm.jhu.edu> wrote:
>
> > In the GCC 3.0 info pages, the option -finline-limits is documented as
> > controlling how large of a function is inlined for functions explicetely
> > marked as being inline. However, it seems like this actually controls the
> > inlining of all functions
>
> I doubt it.  Do you have a code snippet and a command-line that would
> demonstrate the problem?  Make sure you're not using -O3, because this
> implies -finline.

I was actually using -O3. But playing around with -finline-limits, it
seemed to affect the code regardless (ie, -O3 -finline-limit=0, the code
runs slower; but nominally, since inline-limit only affects things marked
inline, that shouldn't happen, since most of the code is not marked inline
[unless these two things are interfering with each other somehow, I
guess]).

I'll play around with it some more and see what it's actually doing.

> Oh, I recall this problem.  It affected a subset of sparc32 boxes, and
> is fixed for 3.0.1.

OK, great.

> > some code was looking for strcmp and friends and couldn't find them,
>
> I don't recall this one.

I that case, I would guess that it probably came about because I was using
"make -k", something tried to include the generated limits file which
didn't exist, and everything got messed up.

Jack


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