[Bug c/43756] New: Bit shifitng by a constant 32 isn't the same as bitshifting by a variable 32.

[Bug c/43756] New: Bit shifitng by a constant 32 isn't the same as bitshifting by a variable 32.

[mail dot alexhaase at gmail dot com]

I ran my demo code from bug id #43755 on a 32bit host for fun. All the casting
i could dream up for the value of ~0 << 32 resulted in a value of 0 when using
constants. However when using a variable to upshift by 32, I get the result ~0
(4294967295). I'm stumped.


-- 
           Summary: Bit shifitng by a constant 32 isn't the same as
                    bitshifting by a variable 32.
           Product: gcc
           Version: 4.4.1
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: c
        AssignedTo: unassigned at gcc dot gnu dot org
        ReportedBy: mail dot alexhaase at gmail dot com
 GCC build triplet: i486-linux-gnu
  GCC host triplet: i486-linux-gnu
GCC target triplet: i486-linux-gnu


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=43756

[Bug c/43756] Bit shifitng by a constant 32 isn't the same as bitshifting by a variable 32.

[mail dot alexhaase at gmail dot com]

------- Comment #1 from mail dot alexhaase at gmail dot com  2010-04-15 03:32 -------
Created an attachment (id=20385)
 --> (http://gcc.gnu.org/bugzilla/attachment.cgi?id=20385&action=view)
a demo of this bug


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=43756

[Bug c/43756] Bit shifitng by a constant 32 isn't the same as bitshifting by a variable 32.

[rguenth at gcc dot gnu dot org]

------- Comment #2 from rguenth at gcc dot gnu dot org  2010-04-15 09:27 -------
Shifting by an amount bigger than the width of the type invokes undefined
behavior (that is, target specific behavior in this case).  We usually
try to simulate target behavior here but the i?86 architecture is inconsistent
in itself:

/* Define if shifts truncate the shift count
   which implies one can omit a sign-extension or zero-extension
   of a shift count.  */
/* On i386, shifts do truncate the count.  But bit opcodes don't.  */

/* #define SHIFT_COUNT_TRUNCATED */


-- 

rguenth at gcc dot gnu dot org changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|UNCONFIRMED                 |RESOLVED
         Resolution|                            |INVALID
            Summary|Bit shifitng by a constant  |Bit shifitng by a constant
                   |32 isn't the same as        |32 isn't the same as
                   |bitshifting by a variable   |bitshifting by a variable
                   |32.                         |32.


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=43756