Re: optimization/7270: Redundant compares not eliminated on x86.

Re: optimization/7270: Redundant compares not eliminated on x86.

[giovannibajo]

Synopsis: Redundant compares not eliminated on x86.

State-Changed-From-To: open->feedback
State-Changed-By: bajo
State-Changed-When: Wed May 14 10:28:57 2003
State-Changed-Why:
    See Dara's question.

http://gcc.gnu.org/cgi-bin/gnatsweb.pl?cmd=view%20audit-trail&database=gcc&pr=7270

Re: optimization/7270: Redundant compares not eliminated on x86.

[Dara Hazeghi]

The following reply was made to PR optimization/7270; it has been noted by GNATS.

From: Dara Hazeghi <dhazeghi@yahoo.com>
To: gcc-gnats@gcc.gnu.org, anamax@earthlink.net
Cc:  
Subject: Re: optimization/7270: Redundant compares not eliminated on x86.
Date: Tue, 13 May 2003 11:52:35 -0700

 http://gcc.gnu.org/cgi-bin/gnatsweb.pl?cmd=view%20audit- 
 trail&database=gcc&pr=7270
 
 Hello,
 
 this report is against gcc 3.1 which is now a bit outdated. Would it be  
 possible for you to test this on gcc 3.3 prerelease or cvs, and see if  
 this problem still occurs? Also, can you provide a sample of the  
 assembly which is unsatisfactory, and specifically note the  
 unsatisfactory parts? Thanks,
 
 Dara
 

optimization/7270: Redundant compares not eliminated on x86.

[anamax]

>Number:         7270
>Category:       optimization
>Synopsis:       Redundant compares not eliminated on x86.
>Confidential:   no
>Severity:       serious
>Priority:       medium
>Responsible:    unassigned
>State:          open
>Class:          pessimizes-code
>Submitter-Id:   net
>Arrival-Date:   Wed Jul 10 22:46:00 PDT 2002
>Closed-Date:
>Last-Modified:
>Originator:     Andy Freeman
>Release:        gcc 3.1
>Organization:
>Environment:
Redhat 7.3
>Description:
A <= compare&branch followed by a == test on the same arguments uses a redundant compare for the == on x86.

Using built-in specs.
Configured with: ../gcc-3.1/configure 
Thread model: single
gcc version 3.1
>How-To-Repeat:
Look at the -O3 -S output for x86.  Notice the two cmpls on either side of the "ja .L4".  I believe that the second cmpl is unnecessary as the flags from the first should survive the "ja .L4".
>Fix:
Better value propagation.
>Release-Note:
>Audit-Trail:
>Unformatted:
----gnatsweb-attachment----
Content-Type: application/octet-stream; name="ii.cc"
Content-Transfer-Encoding: base64
Content-Disposition: attachment; filename="ii.cc"

dW5zaWduZWQgZm9vKHVuc2lnbmVkIGFbXSkgew0KICBmb3IoIHVuc2lnbmVkIGkgPSAwOyBpIDwg
MTA7ICsraSApIHsNCiAgICBpZiggaSA8PSBhW2ldICkgew0KICAgICAgaWYoIGkgPT0gYVtpXSAp
IHsNCglyZXR1cm4gOTsNCiAgICAgIH0gZWxzZSB7DQoJcmV0dXJuIDE1Ow0KICAgICAgfQ0KICAg
IH0NCiAgfQ0KfQ0K