[Bug c/16790] New: Integer down cast ignored in larger expression

[Bug c/16790] New: Integer down cast ignored in larger expression

[andyb at mathworks dot com]

GCC sometimes ignores integer down casts in larger expressions
starting around version 3.2.2
In GCC 2.95.4, these same downcasts were honored.

The following code snippet is from file int_ops.c

  y_final_1 = (int16_T)(  (int16_T)(u1 << 1)   * mul_const >> nshift_right);
  y_middle  =             (int16_T)(u1 << 1);
  y_final_2 = (int16_T)(       y_middle        * mul_const >> nshift_right);

A programmer would expect y_final_1 and y_final_2 to have the same values
under all conditions.   This is not true with the GCC version 3.2.2 through
to the current.  

y_final_1 = -16384   0xFFFFC000
y_final_2 =  16384   0x00004000

Same sequence of integer operations produced DIFFERENT result.

Compiler log shown below.  I don't see any place on this bugzilla
page to enter the .i file that the bug writing guidelines demand???
I'm hoping the "Commit" button will ask me to upload it or something!
Cheers,
Andy B

bartletta-linux:> /hub/Linux/glibc-2.2.5/i686/apps/gcc-3.4.0/bin/gcc -v
-save-temps -g3 -O0 -o int_ops ../int_ops.c ; int_ops
Reading specs from
/hub/Linux/glibc-2.2.5/i686/apps/gcc-3.4.0/lib/gcc/i686-pc-linux-gnu/3.4.0/specs
Configured with: /home/greg/src/gcc-3.4.0/configure
--prefix=/hub/Linux/glibc-2.2.5/i686/apps/gcc-3.4.0
Thread model: posix
gcc version 3.4.0
/hub/Linux/glibc-2.2.5/i686/apps/gcc-3.4.0/libexec/gcc/i686-pc-linux-gnu/3.4.0/cc1
-E -quiet -v -dD ../int_ops.c -mtune=pentiumpro -fworking-directory -O0 -o int_ops.i
ignoring nonexistent directory
"/hub/Linux/glibc-2.2.5/i686/apps/gcc-3.4.0/lib/gcc/i686-pc-linux-gnu/3.4.0/../../../../i686-pc-linux-gnu/include"
#include "..." search starts here:
#include <...> search starts here:
 /usr/local/include
 /hub/Linux/glibc-2.2.5/i686/apps/gcc-3.4.0/include
 /hub/Linux/glibc-2.2.5/i686/apps/gcc-3.4.0/lib/gcc/i686-pc-linux-gnu/3.4.0/include
 /usr/include
End of search list.
 /hub/Linux/glibc-2.2.5/i686/apps/gcc-3.4.0/libexec/gcc/i686-pc-linux-gnu/3.4.0/cc1
-fpreprocessed int_ops.i -quiet -dumpbase int_ops.c -mtune=pentiumpro -auxbase
int_ops -g3 -O0 -version -o int_ops.s
GNU C version 3.4.0 (i686-pc-linux-gnu)
        compiled by GNU C version 3.4.0.
GGC heuristics: --param ggc-min-expand=98 --param ggc-min-heapsize=128673
 as -V -Qy -o int_ops.o int_ops.s
GNU assembler version 2.12 (i686-pc-linux-gnu) using BFD version 2.12
 /hub/Linux/glibc-2.2.5/i686/apps/gcc-3.4.0/libexec/gcc/i686-pc-linux-gnu/3.4.0/collect2
--eh-frame-hdr -m elf_i386 -dynamic-linker /lib/ld-linux.so.2 -o int_ops
/usr/lib/crt1.o /usr/lib/crti.o
/hub/Linux/glibc-2.2.5/i686/apps/gcc-3.4.0/lib/gcc/i686-pc-linux-gnu/3.4.0/crtbegin.o
-L/hub/Linux/glibc-2.2.5/i686/apps/gcc-3.4.0/lib/gcc/i686-pc-linux-gnu/3.4.0
-L/hub/Linux/glibc-2.2.5/i686/apps/gcc-3.4.0/lib/gcc/i686-pc-linux-gnu/3.4.0/../../..
int_ops.o -lgcc -lgcc_eh -lc -lgcc -lgcc_eh
/hub/Linux/glibc-2.2.5/i686/apps/gcc-3.4.0/lib/gcc/i686-pc-linux-gnu/3.4.0/crtend.o
/usr/lib/crtn.o

-- 
           Summary: Integer down cast ignored in larger expression
           Product: gcc
           Version: 3.4.0
            Status: UNCONFIRMED
          Severity: normal
          Priority: P2
         Component: c
        AssignedTo: unassigned at gcc dot gnu dot org
        ReportedBy: andyb at mathworks dot com
                CC: gcc-bugs at gcc dot gnu dot org


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=16790

[Bug c/16790] Integer down cast ignored in larger expression

[andyb at mathworks dot com]

------- Additional Comments From andyb at mathworks dot com  2004-07-27 22:42 -------
Created an attachment (id=6842)
 --> (http://gcc.gnu.org/bugzilla/attachment.cgi?id=6842&action=view)
preprocessor output .i file


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=16790

[Bug c/16790] Integer down cast ignored in larger expression

[pinskia at gcc dot gnu dot org]

------- Additional Comments From pinskia at gcc dot gnu dot org  2004-07-27 23:03 -------
Note 0x8000 is out of range for 16bit ints.

-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=16790

[Bug c/16790] [3.3/3.4/3.5 regression] Integer down cast ignored in larger expression

[bangerth at dealii dot org]

------- Additional Comments From bangerth at dealii dot org  2004-07-27 23:12 -------
Confirmed. A regression in 3.3/3.4/mainline against 2.95. Here's  
a reduced testcase: 
---------------- 
void abort (); 
 
static void compute(unsigned int u1) 
{ 
  unsigned char same_result; 
 
  signed short y_final_1; 
  signed short y_middle; 
  signed short y_final_2; 
 
  y_final_1 = (signed short)( (signed short)(u1 << 1) * 3 >> 1); 
  y_middle  =                 (signed short)(u1 << 1); 
  y_final_2 = (signed short)( y_middle * 3 >> 1); 
 
  if (y_final_1 != y_final_2) 
    abort (); 
} 
 
main() 
{ 
  compute(0x4000U); 
} 
----------------------------- 
 
Note that (0x4000<<1) is just the corner case when casted to signed int, 
since it has only the sign bit set, none of the other bits are set. We 
seem to get this wrong somehow. 
 
W. 

-- 
           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|UNCONFIRMED                 |NEW
     Ever Confirmed|                            |1
      Known to fail|                            |3.2.3 3.3.3 3.4.0 3.5.0
      Known to work|                            |2.95
   Last reconfirmed|0000-00-00 00:00:00         |2004-07-27 23:12:21
               date|                            |
            Summary|Integer down cast ignored in|[3.3/3.4/3.5 regression]
                   |larger expression           |Integer down cast ignored in
                   |                            |larger expression
   Target Milestone|---                         |3.3.5


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=16790

[Bug middle-end/16790] [3.3/3.4/3.5 regression] Integer down cast ignored in larger expression

[pinskia at gcc dot gnu dot org]

------- Additional Comments From pinskia at gcc dot gnu dot org  2004-07-27 23:39 -------
To say what happens is this (u1 * 2 is the same as u1 << 1:
(((int) (signed short)(u1 * 2)) * 3)
is being changed by fold (or in the C front-end) into
u1 * 6
most likely because there is a STRIP_NOP or (int)(short)(int_variable) is being changed into int_variable.

-- 
           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |gdr at gcc dot gnu dot org
          Component|c                           |middle-end
           Keywords|                            |wrong-code
   Target Milestone|3.3.5                       |3.4.2


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=16790

[Bug middle-end/16790] [3.3/3.4/3.5 regression] Integer down cast ignored in larger expression

[cvs-commit at gcc dot gnu dot org]

------- Additional Comments From cvs-commit at gcc dot gnu dot org  2004-08-03 21:24 -------
Subject: Bug 16790

CVSROOT:	/cvs/gcc
Module name:	gcc
Changes by:	sayle@gcc.gnu.org	2004-08-03 21:24:32

Modified files:
	gcc            : ChangeLog fold-const.c 
	gcc/testsuite  : ChangeLog 
Added files:
	gcc/testsuite/gcc.c-torture/execute: pr16790-1.c 

Log message:
	PR middle-end/16790
	* fold-const.c (extract_muldiv_1) <NOP_EXPR>: Disallow local
	truncations, not just global truncations.
	
	* gcc.c-torture/execute/pr16790-1.c: New test case.

Patches:
http://gcc.gnu.org/cgi-bin/cvsweb.cgi/gcc/gcc/ChangeLog.diff?cvsroot=gcc&r1=2.4785&r2=2.4786
http://gcc.gnu.org/cgi-bin/cvsweb.cgi/gcc/gcc/fold-const.c.diff?cvsroot=gcc&r1=1.432&r2=1.433
http://gcc.gnu.org/cgi-bin/cvsweb.cgi/gcc/gcc/testsuite/ChangeLog.diff?cvsroot=gcc&r1=1.4090&r2=1.4091
http://gcc.gnu.org/cgi-bin/cvsweb.cgi/gcc/gcc/testsuite/gcc.c-torture/execute/pr16790-1.c.diff?cvsroot=gcc&r1=NONE&r2=1.1



-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=16790

[Bug middle-end/16790] [3.3/3.4 regression] Integer down cast ignored in larger expression

[cvs-commit at gcc dot gnu dot org]

------- Additional Comments From cvs-commit at gcc dot gnu dot org  2004-08-07 21:56 -------
Subject: Bug 16790

CVSROOT:	/cvs/gcc
Module name:	gcc
Branch: 	gcc-3_4-branch
Changes by:	sayle@gcc.gnu.org	2004-08-07 21:56:55

Modified files:
	gcc            : ChangeLog fold-const.c 
	gcc/testsuite  : ChangeLog 
Added files:
	gcc/testsuite/gcc.c-torture/execute: pr16790-1.c 

Log message:
	PR middle-end/16790
	Backport from mainline.
	* fold-const.c (expand_muldiv_1) <NOP_EXPR>: Disallow local
	truncations, not just global truncations.
	
	* gcc.c-torture/execute/pr16790-1.c: New test case.

Patches:
http://gcc.gnu.org/cgi-bin/cvsweb.cgi/gcc/gcc/ChangeLog.diff?cvsroot=gcc&only_with_tag=gcc-3_4-branch&r1=2.2326.2.575&r2=2.2326.2.576
http://gcc.gnu.org/cgi-bin/cvsweb.cgi/gcc/gcc/fold-const.c.diff?cvsroot=gcc&only_with_tag=gcc-3_4-branch&r1=1.322.2.12&r2=1.322.2.13
http://gcc.gnu.org/cgi-bin/cvsweb.cgi/gcc/gcc/testsuite/ChangeLog.diff?cvsroot=gcc&only_with_tag=gcc-3_4-branch&r1=1.3389.2.242&r2=1.3389.2.243
http://gcc.gnu.org/cgi-bin/cvsweb.cgi/gcc/gcc/testsuite/gcc.c-torture/execute/pr16790-1.c.diff?cvsroot=gcc&only_with_tag=gcc-3_4-branch&r1=NONE&r2=1.1.4.1



-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=16790

[Bug middle-end/16790] [3.3 regression] Integer down cast ignored in larger expression

[pinskia at gcc dot gnu dot org]

------- Additional Comments From pinskia at gcc dot gnu dot org  2004-08-07 22:03 -------
Fixed also in 3.4.2.

-- 
           What    |Removed                     |Added
----------------------------------------------------------------------------
      Known to work|2.95 3.5.0                  |2.95 3.5.0 3.4.2
            Summary|[3.3/3.4 regression] Integer|[3.3 regression] Integer
                   |down cast ignored in larger |down cast ignored in larger
                   |expression                  |expression
   Target Milestone|3.4.2                       |3.3.5


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=16790

[Bug middle-end/16790] [3.3 regression] Integer down cast ignored in larger expression

[cvs-commit at gcc dot gnu dot org]

------- Additional Comments From cvs-commit at gcc dot gnu dot org  2004-08-08 18:55 -------
Subject: Bug 16790

CVSROOT:	/cvs/gcc
Module name:	gcc
Branch: 	gcc-3_3-branch
Changes by:	sayle@gcc.gnu.org	2004-08-08 18:55:35

Modified files:
	gcc            : ChangeLog fold-const.c 
	gcc/testsuite  : ChangeLog 
Added files:
	gcc/testsuite/gcc.c-torture/execute: pr16790-1.c 

Log message:
	2004-08-08  Roger Sayle  <roger@eyesopen.com>
	
	PR middle-end/16790
	Backport from mainline.
	* fold-const.c (expand_muldiv_1) <NOP_EXPR>: Disallow local
	truncations, not just global truncations.
	
	* gcc.c-torture/execute/pr16790-1.c: New test case.

Patches:
http://gcc.gnu.org/cgi-bin/cvsweb.cgi/gcc/gcc/ChangeLog.diff?cvsroot=gcc&only_with_tag=gcc-3_3-branch&r1=1.16114.2.1009&r2=1.16114.2.1010
http://gcc.gnu.org/cgi-bin/cvsweb.cgi/gcc/gcc/fold-const.c.diff?cvsroot=gcc&only_with_tag=gcc-3_3-branch&r1=1.227.2.7&r2=1.227.2.8
http://gcc.gnu.org/cgi-bin/cvsweb.cgi/gcc/gcc/testsuite/ChangeLog.diff?cvsroot=gcc&only_with_tag=gcc-3_3-branch&r1=1.2261.2.382&r2=1.2261.2.383
http://gcc.gnu.org/cgi-bin/cvsweb.cgi/gcc/gcc/testsuite/gcc.c-torture/execute/pr16790-1.c.diff?cvsroot=gcc&only_with_tag=gcc-3_3-branch&r1=NONE&r2=1.1.6.1



-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=16790

[Bug middle-end/16790] [3.3 regression] Integer down cast ignored in larger expression

[pinskia at gcc dot gnu dot org]

------- Additional Comments From pinskia at gcc dot gnu dot org  2004-08-08 18:57 -------
Fixed.

-- 
           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|NEW                         |RESOLVED
         Resolution|                            |FIXED


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=16790