This is the mail archive of the
gcc-patches@gcc.gnu.org
mailing list for the GCC project.
Re: [RFA] new code size estimation for inlining
- From: Jan Hubicka <jh at suse dot cz>
- To: Richard Henderson <rth at redhat dot com>, Jan Hubicka <jh at suse dot cz>,pfeifer at dbai dot tuwien dot ac dot at, gcc-patches at gcc dot gnu dot org
- Date: Tue, 8 Jul 2003 23:34:13 +0200
- Subject: Re: [RFA] new code size estimation for inlining
- References: <20030630205631.GD10852@kam.mff.cuni.cz> <20030708191915.GC32690@redhat.com>
> On Mon, Jun 30, 2003 at 10:56:31PM +0200, Jan Hubicka wrote:
> > + walk_tree_without_duplicates (&DECL_SAVED_TREE (decl), c_estimate_num_insns_1, &num);
>
> Wrap. Why "without_duplicates"? Seems to me if the same CALL_EXPR
> node shows up twice, ew get two calls, and so need to count it twice.
I was told that CALL_EXPRs are never shared. C++ frontend does not
share it in one case and it causes tree inliner to go crazy, so i think
it is bug. I sent a fix for that in other email.
I am not sure what is the best behaviour here. I expected that things
that gets shared would likely get CSEed too, but I am not quite sure it
is the case. I can simply remove the _without_duplcates and we see what
works better.
OK?
Honza
>
> The rest looks ok.
>
>
> r~