This is the mail archive of the
gcc-patches@gcc.gnu.org
mailing list for the GCC project.
Re: [PATCH] Remove unreachable nodes before IPA profile pass (PR ipa/87706).
- From: Jan Hubicka <hubicka at ucw dot cz>
- To: Richard Biener <richard dot guenther at gmail dot com>
- Cc: Martin Liška <mliska at suse dot cz>, GCC Patches <gcc-patches at gcc dot gnu dot org>
- Date: Thu, 8 Nov 2018 12:46:15 +0100
- Subject: Re: [PATCH] Remove unreachable nodes before IPA profile pass (PR ipa/87706).
- References: <8c883adb-9ec5-c9d5-f1cb-cc48a6ccd2ea@suse.cz> <20181108111948.uahgghjf3nteeiqw@kam.mff.cuni.cz> <0d9a363d-edb2-fce6-028b-7f15668bc6dc@suse.cz> <CAFiYyc2Q9D_zoNdSS4-RLwkC30XomZ76H=2QnF8m+q3nM=nUhA@mail.gmail.com>
> On Thu, Nov 8, 2018 at 12:39 PM Martin Liška <mliska@suse.cz> wrote:
> >
> > On 11/8/18 12:19 PM, Jan Hubicka wrote:
> > >> Hi.
> > >>
> > >> In order to fix the warnings mentioned in the PR, we need
> > >> to run remove_unreachable_nodes after early tree passes. That's
> > >> however possible only within a IPA pass. Thus I'm calling that
> > >> before the profile PASS.
> > >>
> > >> Patch survives regression tests on ppc64le-linux-gnu and majority
> > >> of warnings are gone in profiledbootstrap.
> > >>
> > >> Ready for trunk?
> > >
> > > I think we want to do that even with no -fprofile-generate because the
> > > unreachable code otherwise goes into all other IPA passes for no good
> > > reason. So perhaps adding it as todo after the early optimization
> > > metapass?
> >
> > That fails due to gcc_assert.
> > So one needs:
> >
> > diff --git a/gcc/passes.c b/gcc/passes.c
> > index d838d909941..be92a2f3be3 100644
> > --- a/gcc/passes.c
> > +++ b/gcc/passes.c
> > @@ -485,7 +485,7 @@ const pass_data pass_data_all_early_optimizations =
> > 0, /* properties_provided */
> > 0, /* properties_destroyed */
> > 0, /* todo_flags_start */
> > - 0, /* todo_flags_finish */
> > + TODO_remove_functions | TODO_rebuild_cgraph_edges, /* todo_flags_finish */
> > };
> >
> > class pass_all_early_optimizations : public gimple_opt_pass
> > @@ -1989,7 +1989,8 @@ execute_todo (unsigned int flags)
> > of IPA pass queue. */
> > if (flags & TODO_remove_functions)
> > {
> > - gcc_assert (!cfun);
> > + gcc_assert (!cfun
> > + || strcmp (current_pass->name, "early_optimizations") == 0);
> > symtab->remove_unreachable_nodes (dump_file);
> > }
> >
> >
> > Or do you prefer to a new pass_remove_functions pass that will be added after
> > pass_local_optimization_passes ?
>
> Can you make it todo_flags_start of pass_ipa_tree_profile instead?
It fails because all_early_optimizations are now gimple pass, so it
should be TODO after pass_local_optimization_passes?
Honza
>
> > Martin
> >
> > >
> > > Honza
> > >> Thanks,
> > >> Martin
> > >>
> > >> gcc/ChangeLog:
> > >>
> > >> 2018-11-08 Martin Liska <mliska@suse.cz>
> > >>
> > >> * tree-profile.c: Run TODO_remove_functions before "profile"
> > >> pass in order to remove dead functions that will trigger
> > >> -Wmissing-profile.
> > >> ---
> > >> gcc/tree-profile.c | 2 +-
> > >> 1 file changed, 1 insertion(+), 1 deletion(-)
> > >>
> > >>
> > >
> > >> diff --git a/gcc/tree-profile.c b/gcc/tree-profile.c
> > >> index d8f2a3b1ba4..c14ebc556a6 100644
> > >> --- a/gcc/tree-profile.c
> > >> +++ b/gcc/tree-profile.c
> > >> @@ -776,7 +776,7 @@ const pass_data pass_data_ipa_tree_profile =
> > >> 0, /* properties_required */
> > >> 0, /* properties_provided */
> > >> 0, /* properties_destroyed */
> > >> - 0, /* todo_flags_start */
> > >> + TODO_remove_functions, /* todo_flags_start */
> > >> TODO_dump_symtab, /* todo_flags_finish */
> > >> };
> > >>
> > >>
> > >
> >