Path: utzoo!utgpu!jarvis.csri.toronto.edu!mailrus!tut.cis.ohio-state.edu!ucbvax!decwrl!labrea!mcnc!rti!xyzzy!bert!meissner
From: meissner@bert.dg.com (Michael Meissner)
Newsgroups: gnu.gcc.bug
Subject: Re: Enhancements to strength-reduction code.
Message-ID: <4561@xyzzy.UUCP>
Date: 27 Mar 89 16:19:17 GMT
References: <8903162359.AA05408@bayes.arc.nasa.gov>
Sender: usenet@xyzzy.UUCP
Reply-To: meissner@xyzzy.UUCP (Michael Meissner)
Distribution: gnu
Organization: Data General (Languages @ Research Triangle Park, NC.)
Lines: 21

I just tried this out with GCC 1.34, and found that it exposes a
latent bug in loop optimization caused by 1.34.  I tried to recompile
the 88k compiler with it, and when I got to global-alloc.c, it aborted
with an insn not found.  The insn involved looked like:

	(set cc0 (minus (match_operand:SI 0 "register_operand" "r")
			(const_int 0)))

This was how you set the condition code under 1.33 and before.  However,
under 1.34 a new RTL operator was added 'compare', so the RTX should
look like:

	(set cc0 (compare (match_operand:SI 0 "register_operand" "r")
			  (const_int 0)))

I will look at this on my own, but this quick description might set off
bells for somebody more familar with that section of code.

Michael Meissner, Data General.
Uucp:		...!mcnc!rti!xyzzy!meissner		If compiles were much
Internet:	meissner@dg-rtp.DG.COM			faster, when would we
Old Internet:	meissner%dg-rtp.DG.COM@relay.cs.net	have time for netnews?