[PATCH] regrename: Skip renaming if instruction is noop move.

Jojo R rjiejie@linux.alibaba.com
Wed Nov 17 02:20:41 GMT 2021


— Jojo
在 2021年11月16日 +0800 PM8:12,Richard Biener <richard.guenther@gmail.com>,写道:
> On Tue, Nov 16, 2021 at 12:45 PM Jojo R via Gcc-patches
> <gcc-patches@gcc.gnu.org> wrote:
> >
> > Skip renaming if instruction is noop move, and it will
> > been removed for performance.
>
> Is there any (target specific) testcase you can add? Such commits are
> problematic
> when later bisected to since the intent isn't clear.

I made a issue in bugzilla, please check it, thanks.

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=103296
>
> > gcc/
> > * regrename.c (find_rename_reg): Return satisfied regno
> > if instruction is noop move.
> > ---
> > gcc/regrename.c | 3 +++
> > 1 file changed, 3 insertions(+)
> >
> > diff --git a/gcc/regrename.c b/gcc/regrename.c
> > index b8a9ca36f22..cb605f5176b 100644
> > --- a/gcc/regrename.c
> > +++ b/gcc/regrename.c
> > @@ -394,6 +394,9 @@ find_rename_reg (du_head_p this_head, enum reg_class super_class,
> > this_head, *unavailable))
> > return this_head->tied_chain->regno;
> >
> > + if (noop_move_p (this_head->first->insn))
> > + return best_new_reg;
> > +
> > /* If PREFERRED_CLASS is not NO_REGS, we iterate in the first pass
> > over registers that belong to PREFERRED_CLASS and try to find the
> > best register within the class. If that failed, we iterate in
> > --
> > 2.24.3 (Apple Git-128)


More information about the Gcc-patches mailing list