From mboxrd@z Thu Jan  1 00:00:00 1970
Return-Path: <gcc-prs-return-25731-listarch-gcc-prs=gcc.gnu.org@gcc.gnu.org>
Received: (qmail 20371 invoked by alias); 5 Dec 2002 01:36:02 -0000
Mailing-List: contact gcc-prs-help@gcc.gnu.org; run by ezmlm
Precedence: bulk
List-Archive: <http://gcc.gnu.org/ml/gcc-prs/>
List-Post: <mailto:gcc-prs@gcc.gnu.org>
List-Help: <mailto:gcc-prs-help@gcc.gnu.org>
Sender: gcc-prs-owner@gcc.gnu.org
Received: (qmail 20338 invoked by uid 71); 5 Dec 2002 01:36:00 -0000
Resent-Date: 5 Dec 2002 01:36:00 -0000
Resent-Message-ID: <20021205013600.20336.qmail@sources.redhat.com>
Resent-From: gcc-gnats@gcc.gnu.org (GNATS Filer)
Resent-Cc: gcc-prs@gcc.gnu.org, gcc-bugs@gcc.gnu.org
Resent-Reply-To: gcc-gnats@gcc.gnu.org, squelart@hotmail.com
Received: (qmail 18800 invoked by uid 61); 5 Dec 2002 01:32:32 -0000
Message-Id: <20021205013232.18799.qmail@sources.redhat.com>
Date: Wed, 04 Dec 2002 17:36:00 -0000
From: squelart@hotmail.com
Reply-To: squelart@hotmail.com
To: gcc-gnats@gcc.gnu.org
X-Send-Pr-Version: gnatsweb-2.9.3 (1.1.1.1.2.31)
Subject: optimization/8815: C switch statement produces unnecessary code in some cases
X-SW-Source: 2002-12/txt/msg00260.txt.bz2
List-Id: <gcc-prs.sourceware.org>


>Number:         8815
>Category:       optimization
>Synopsis:       C switch statement produces unnecessary code in some cases
>Confidential:   no
>Severity:       non-critical
>Priority:       low
>Responsible:    unassigned
>State:          open
>Class:          change-request
>Submitter-Id:   net
>Arrival-Date:   Wed Dec 04 17:36:00 PST 2002
>Closed-Date:
>Last-Modified:
>Originator:     squelart@hotmail.com
>Release:        gcc-3.2.1
>Organization:
>Environment:
i686-cygwin, arm-linux
>Description:
I've got some switch statements like this one:
switch(value & 0xFF) { /* all 256 cases... */ }.
The assembly code generated on both intel and arm (with -O2 or -O3) is something like:
and r,$255
cmp r,$255
if> jump after_switch
else jump (table+value*4)
Since the value has been ANDed by 0xFF, comparing it to 0xFF is unnecessary and could be optimized out.
>How-To-Repeat:
The simplest code seems to be:
int main(int argc, char** argv)
{
  switch(argc&7)
  {
  case 0: return 0;
  case 1: return 1;
  case 2: return 2;
  case 3: return 3;
  case 4: return 4;
  case 5: return 5;
  case 6: return 6;
  case 7: return 7;
  }
}

gcc -S -O2 test.c
>Fix:
Remove the comparison and jump, when the switch value has been ANDed by a constant, and all cases&constant are handled.
Also, with -Wall, the code above produces a "warning: control reaches end of non-void function", which is not true since all cases return (but maybe that's another problem, though it seems related).
[Newbie disclaimer: I know _nothink_ about gcc sources, I'm just trying to help by pointing out a possible optimization, I hope it's not too difficult to implement, but my life doesn't depend on it, so please don't answer "DIY"!]
>Release-Note:
>Audit-Trail:
>Unformatted: