This is the mail archive of the gcc-patches@gcc.gnu.org mailing list for the GCC project.

Index Nav:	[Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav:	[Date Prev] [Date Next]	[Thread Prev] [Thread Next]
Other format:	[Raw text]

Re: [PATCH] Fix regression for missed combine opportunity

From: Andrew Pinski <pinskia at physics dot uc dot edu>
To: Pat Haugen <pthaugen at us dot ibm dot com>
Cc: gcc-patches at gcc dot gnu dot org
Date: Fri, 15 Oct 2004 15:43:10 -0400
Subject: Re: [PATCH] Fix regression for missed combine opportunity
References: <OF905F19C9.DC92F0DB-ON86256F2E.0069DC6F-86256F2E.006ACC05@us.ibm.com>

On Oct 15, 2004, at 3:26 PM, Pat Haugen wrote:

A regression was introduced when changes were made to canonicalize XOR/MINUS signbit to PLUS signbit. Code to combine an XOR feeding a ZERO_EXTRACT was not recognizing the opportunity because the XOR was canonicalized into a PLUS. An example of what this fixes: if ((u & 16) == 0) ... Old: xori 3,3,16 andi. 0,3,16 beq 0,.L2 New: andi. 0,3,16 bne 0,.L2 Bootstrapped/regtested on powerpc64-unknown-linux-gnu with no new regressions.


Just a note that this is PR 18002 (I noticed it while looking into
bug reports).  This also effects 3.4.x.

Thanks,
Andrew Pinski

References:
- [PATCH] Fix regression for missed combine opportunity
  - From: Pat Haugen

Index Nav:	[Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav:	[Date Prev] [Date Next]	[Thread Prev] [Thread Next]