This is the mail archive of the
mailing list for the GCC project.
Re: [patch] h8300.md: Add new peepholes.
- To: Kazu Hirata <kazu at hxi dot com>
- Subject: Re: [patch] h8300.md: Add new peepholes.
- From: "Zack Weinberg" <zackw at Stanford dot EDU>
- Date: Mon, 19 Feb 2001 22:14:12 -0800
- Cc: "'gcc-patches at gcc dot gnu dot org '" <gcc-patches at gcc dot gnu dot org>
- References: <879EA880A0FED411996B00B0D0B082EC0163B1@hxi_exch01.hxi.com>
On Mon, Feb 19, 2001 at 03:45:34PM -0500, Kazu Hirata wrote:
> Hi Zack,
> > > 2001-02-19 Kazu Hirata <firstname.lastname@example.org>
> > >
> > > * config/h8300/h8300.md: Add peepholes to optimize ior and
> > > xor.
> > Shouldn't you be using define_peephole2?
> Is it OK even if RTL generated by define_peephole2 is rather complex, like a
> parallel consisting of a zero extension and an ior? (I am guessing it is
> because if you emit assembly language with define_peephole, then no other
> pass can modify it anyway.)
I believe it is okay, as long as the RTL is recognizable insns. Worst
case, no later pass will be able to improve it. If you use text
peepholes, then nothing gets a chance to improve it. Also we'd like
to get rid of the text peephole machinery eventually.