This is the mail archive of the
gcc-patches@gcc.gnu.org
mailing list for the GCC project.
Re: [cond-optab] update
- From: Paolo Bonzini <bonzini at gnu dot org>
- To: DJ Delorie <dj at redhat dot com>, GCC Patches <gcc-patches at gcc dot gnu dot org>
- Date: Thu, 19 Mar 2009 22:58:23 +0100
- Subject: Re: [cond-optab] update
- References: <gpu2vp$cah$1@ger.gmane.org> <200903191848.n2JImQrN002998@greed.delorie.com>
DJ Delorie wrote:
>> Besides obvious register allocation differences
>
> m32c is very sensitive to register allocation issues.
Quote entirely: "Right now I aim at 100% identical assembly, maybe I'll
have to relax that" -- and then register allocation differences is the
first obvious thing to relax. But I still aim at 100% identical assembly.
>> you basically duplicate the cmp patterns into cbranch and
>
> m32c already has a cbranch, though. It gets split after reload.
Yes, in fact m32c was one of the simplest.
> Also, m32c needs a separate compare RTL insn in the end
> because it's not always emitted.
Yes, I have no plans to eliminate RTL compare insns; that would be
stupid. Only compare optabs, and the associated "stash away" operands
tricks.
>> config/m32c 4 files changed, 50 insertions(+), 150 deletions(-)
>
> Please don't commit this one until you've built all of libiberty,
> newlib, and libgcc with the changed compiler (-mcpu=m16c and
> -mcpu=m32c). Usually one of those fails to build if you're not
> careful about register allocation.
I'm going to ask for approval to all maintainers, of course. I have no
authority to commit anything on my own. Still, thanks for suggesting
the multilibs to try.
Paolo