RE: optimization/10362: Optimization with constant shifts and com pares fails on MIPS

[Wolfgang Bangerth <bangerth@ices.utexas.edu>]

The following reply was made to PR optimization/10362; it has been noted by GNATS.

From: Wolfgang Bangerth <bangerth@ices.utexas.edu>
To: Berger David-MGI2063 <dberger@motorola.com>
Cc: "'gcc-bugs@gcc.gnu.org'" <gcc-bugs@gcc.gnu.org>,
   "'gcc-gnats@gcc.gnu.org'" <gcc-gnats@gcc.gnu.org>
Subject: RE: optimization/10362: Optimization with constant shifts and com
 pares fails on MIPS
Date: Wed, 9 Apr 2003 18:07:54 -0500 (CDT)

 > volatile unsigned long i;
 > 
 > void test(void)
 > {
 >   i = (((i & 0xf0) >> 4) >= 14); /* This one generates incorrect code with -O2 */
 >   i = ((i & 0xf0) >= (14 << 4)); /* This is what the optimizer is trying to do */
 > }
 
 I think what the compiler actually does, is to transform it to 
   i = ((i & 0xf0) > (14<<4-1));
 However, since the lowest 4 bits of the left hand side of the comparison 
 are provably zero, this is equivalent to
   i = ((i & 0xf0) > (13<<4));
 
 Do you have a value for "i", where you actually get a wrong comparison?
 
 W.
 
 -------------------------------------------------------------------------
 Wolfgang Bangerth              email:            bangerth@ices.utexas.edu
                                www: http://www.ices.utexas.edu/~bangerth/