This is the mail archive of the
gcc@gcc.gnu.org
mailing list for the GCC project.
Re: Experimenting with tree inliner parameter settings
Op zo 04-05-2003, om 22:57 schreef Jakub Jelinek:
> On Sun, May 04, 2003 at 10:44:45PM +0200, Richard Guenther wrote:
> > [Summary / analysis of inline params snipped]
> >
> > The SPEC results show, that at least defaults for C and Fortran do not
> > necessarily apply to C++ code (as the testcases from the PRs show).
> >
> > So may I suggest again splitting the defaults per language? Especially for
> > templated C++ code with exceptions the instruction counting can be _very_
> > wrong.
>
> I strongly second this.
> Furthermore, I still think that e.g. in C explicit inline should be given
> bigger limits than -O3 inlining (and nested inline functions even bigger).
There was a long discussion about giving the "inline" keyword more
weight, and I did not want to start over so I just choose them to be the
same.
Giving more weight to nested inline functions seems like a good idea,
and it's probably not very hard to do. Just see if TREE_CONTEXT != NULL
in tree-inline. That would probably also give more weight to inline
functions inside class declaration in C++. Certainly something we
should try.
Greetz
Steven