This is the mail archive of the
gcc-patches@gcc.gnu.org
mailing list for the GCC project.
Re: A new patch for floating point conditional mov on PPo
- To: john at feith dot com (John Wehle)
- Subject: Re: A new patch for floating point conditional mov on PPo
- From: Jeffrey A Law <law at cygnus dot com>
- Date: Fri, 19 Jun 1998 21:12:18 -0600
- cc: hjl at lucon dot org, egcs-patches at cygnus dot com
- Reply-To: law at cygnus dot com
In message <199806121748.NAA02655@jwlab.FEITH.COM>you write:
> > The current notice_update_cc0 works with my current patch. If
> > there is a change in notice_update_cc0, I think I may find a way
> > to accommodate it.
>
> That fact that your patch doesn't seem to break anything doesn't
> make it right.
Exactly! There is a fundamental difference between a patch "working"
and being "correct".
> I have illustrated a problem which will occur
> in the next few days due to notice_update_cc0 getting smarter.
Right. In fact, I think I installed that code last night!
> Changing the integer comparison patterns so that the assembler
> output causes cc0 to have completely different meanings (depending
> on the next INSN) without reflecting that in the RTL is dangerous.
> It is much safer to change tha meaning of cc0 when you output the
> actual floating point conditional move.
Agreed. In general, all side effects should be shown in RTL. We don't
always achieve it in practice, but that's a goal we strive for. The
more accurate the RTL, the less likely the compiler is going to screw
something up later because it didn't know about some obscure side
effect.
jeff