[Bug c/28473] New: with -O, casting result of round(x) to uint64_t produces wrong values for x > INT_MAX

[Bug c/28473] New: with -O, casting result of round(x) to uint64_t produces wrong values for x > INT_MAX

[maxp at alum dot mit dot edu]

When compiling with -O or greater optimization, and if x > INT_MAX, then code
like this:
     uint64_t y = (uint64_t)round(x);
assigns the wrong value to y (the top 32 bits are all 1s).  But this code
assigns the right value to z:
     double dz = round(x);
     uint64_t z = dz;

It almost seems as if gcc -O in some cases compiles using a built-in
declaration of round() that returns a 32-bit int.

(1) GCC VERSION

gavia% gcc -v
Using built-in specs.
Target: i386-redhat-linux
Configured with: ../configure --prefix=/usr --mandir=/usr/share/man
--infodir=/usr/share/info --enable-shared --enable-threads=posix
--enable-checking=release --with-system-zlib --enable-__cxa_atexit
--disable-libunwind-exceptions --enable-libgcj-multifile
--enable-languages=c,c++,objc,obj-c++,java,fortran,ada --enable-java-awt=gtk
--disable-dssi --with-java-home=/usr/lib/jvm/java-1.4.2-gcj-1.4.2.0/jre
--with-cpu=generic --host=i386-redhat-linux
Thread model: posix
gcc version 4.1.1 20060525 (Red Hat 4.1.1-1)

This is on Fedora Core 5.  Same problem occurs with gcc 4.0.2 on FC4.

(2) COMMAND LINE, COMPILER OUTPUT, AND PROGRAM OUTPUT

gavia% gcc -Wall -std=c99 -O -save-temps -o x-opt x.c -lm
gavia% ./x-opt
x: 2147483648
y: 18446744071562067968
z: 2147483648

(3) C SOURCE

#include <stdio.h>
#include <math.h>
#include <inttypes.h>
int main(int c, char **v)
{
     uint64_t x = 2147483648ULL; /* INT_MAX+1 */
     printf("x: %llu\n", x);

     uint64_t y = (uint64_t)round(x);
     printf("y: %llu\n", y);

     double dz = round(x);
     uint64_t z = dz;
     printf("z: %llu\n", z);

     return 0;
}

(4) PREPROCESSED FILE

Included as attachment.


-- 
           Summary: with -O, casting result of round(x) to uint64_t produces
                    wrong values for x > INT_MAX
           Product: gcc
           Version: 4.1.1
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: c
        AssignedTo: unassigned at gcc dot gnu dot org
        ReportedBy: maxp at alum dot mit dot edu
 GCC build triplet: i386-redhat-linux
  GCC host triplet: i386-redhat-linux
GCC target triplet: i386-redhat-linux


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=28473

[Bug c/28473] with -O, casting result of round(x) to uint64_t produces wrong values for x > INT_MAX

[maxp at alum dot mit dot edu]

------- Comment #1 from maxp at alum dot mit dot edu  2006-07-24 20:54 -------
Created an attachment (id=11930)
 --> (http://gcc.gnu.org/bugzilla/attachment.cgi?id=11930&action=view)
preprocessor output


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=28473

Re: [Bug c/28473] New: with -O, casting result of round(x) to uint64_t produces wrong values for x > INT_MAX

[Andrew Pinski]

> 
> When compiling with -O or greater optimization, and if x > INT_MAX, then code
> like this:
>      uint64_t y = (uint64_t)round(x);
> assigns the wrong value to y (the top 32 bits are all 1s).  But this code
> assigns the right value to z:
>      double dz = round(x);
>      uint64_t z = dz;
> 
> It almost seems as if gcc -O in some cases compiles using a built-in
> declaration of round() that returns a 32-bit int.

Yes this is wrong code, GCC is generating for some reason lround instead of
llround.


Thanks,
Andrew Pinski

[Bug c/28473] with -O, casting result of round(x) to uint64_t produces wrong values for x > INT_MAX

[pinskia at physics dot uc dot edu]

------- Comment #2 from pinskia at physics dot uc dot edu  2006-07-24 22:01 -------
Subject: Re:   New: with -O, casting result of round(x) to uint64_t produces
wrong values for x > INT_MAX

> 
> When compiling with -O or greater optimization, and if x > INT_MAX, then code
> like this:
>      uint64_t y = (uint64_t)round(x);
> assigns the wrong value to y (the top 32 bits are all 1s).  But this code
> assigns the right value to z:
>      double dz = round(x);
>      uint64_t z = dz;
> 
> It almost seems as if gcc -O in some cases compiles using a built-in
> declaration of round() that returns a 32-bit int.

Yes this is wrong code, GCC is generating for some reason lround instead of
llround.


Thanks,
Andrew Pinski


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=28473

[Bug c/28473] with -O, casting result of round(x) to uint64_t produces wrong values for x > INT_MAX

[pinskia at gcc dot gnu dot org]

------- Comment #3 from pinskia at gcc dot gnu dot org  2006-07-25 02:12 -------
Hmm, this works correctly on powerpc-darwin, where we get no round function at
all.


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=28473

[Bug middle-end/28473] [4.0/4.1/4.2 Regression] with -O, casting result of round(x) to uint64_t produces wrong values for x > INT_MAX

[rguenth at gcc dot gnu dot org]

------- Comment #4 from rguenth at gcc dot gnu dot org  2006-07-25 08:46 -------
Confirmed.  Testcase:

typedef unsigned uint64_t __attribute__((mode(DI)));
uint64_t foo(void)
{
     uint64_t x = 2147483648ULL; /* INT_MAX+1 */
     uint64_t y = (uint64_t)__builtin_round(x);
     return y;
}

where we fold it to

uint64_t y = (long long unsigned int) __builtin_lround ((double) x);


-- 

rguenth at gcc dot gnu dot org changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |rguenth at gcc dot gnu dot
                   |                            |org
             Status|UNCONFIRMED                 |NEW
          Component|c                           |middle-end
     Ever Confirmed|0                           |1
           Keywords|                            |wrong-code
      Known to work|                            |3.4.6
   Last reconfirmed|0000-00-00 00:00:00         |2006-07-25 08:46:28
               date|                            |
            Summary|with -O, casting result of  |[4.0/4.1/4.2 Regression]
                   |round(x) to uint64_t        |with -O, casting result of
                   |produces wrong values for x |round(x) to uint64_t
                   |> INT_MAX                   |produces wrong values for x
                   |                            |> INT_MAX
   Target Milestone|---                         |4.0.4


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=28473

[Bug middle-end/28473] [4.0/4.1/4.2 Regression] with -O, casting result of round(x) to uint64_t produces wrong values for x > INT_MAX

[rguenth at gcc dot gnu dot org]

------- Comment #5 from rguenth at gcc dot gnu dot org  2006-07-25 08:58 -------
I have a patch.


-- 

rguenth at gcc dot gnu dot org changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
         AssignedTo|unassigned at gcc dot gnu   |rguenth at gcc dot gnu dot
                   |dot org                     |org
             Status|NEW                         |ASSIGNED
   Last reconfirmed|2006-07-25 08:46:28         |2006-07-25 08:58:05
               date|                            |


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=28473

[Bug middle-end/28473] [4.0/4.1/4.2 Regression] with -O, casting result of round(x) to uint64_t produces wrong values for x > INT_MAX

[roger at eyesopen dot com]

------- Comment #6 from roger at eyesopen dot com  2006-07-25 20:02 -------
Grr.  I've just noticed richi has just assigned this patch to himself.
I also have a patch that been bootstrapped and has nearly finished
regression testing, that I was just about to post/commit.  richi what
does your fix look like?

Mine contains several copies of:

          if (!TARGET_C99_FUNCTIONS)
            break;
!         if (outprec < TYPE_PRECISION (long_integer_type_node)
!             || (outprec == TYPE_PRECISION (long_integer_type_node)
!                 && !TYPE_UNSIGNED (type)))
            fn = mathfn_built_in (s_intype, BUILT_IN_LCEIL);
+         else if (outprec == TYPE_PRECISION (long_long_integer_type_node)
+                  && !TYPE_UNSIGNED (type))
+           fn = mathfn_built_in (s_intype, BUILT_IN_LLCEIL);
          break;

[Serves me right for not assigning this when pinkia asked me to investigate.
 I knew there was a good reason I don't normally bother with recent PRs].


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=28473

[Bug middle-end/28473] [4.0/4.1/4.2 Regression] with -O, casting result of round(x) to uint64_t produces wrong values for x > INT_MAX

[roger at eyesopen dot com]

------- Comment #7 from roger at eyesopen dot com  2006-07-25 20:08 -------
Ahh, I've just found the Richard's patch submission posting at
http://gcc.gnu.org/ml/gcc-patches/2006-07/msg01065.html
I agree with Andrew Pinski, I think my changes are the better fix.

We also need to investigate wether (unsigned int)round(x) is better
implemented as (unsigned int)llround(x).  For the time being, my
patch doesn't perform this transformation, and using lround is unsafe.


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=28473

[Bug middle-end/28473] [4.0/4.1/4.2 Regression] with -O, casting result of round(x) to uint64_t produces wrong values for x > INT_MAX

[sayle at gcc dot gnu dot org]

------- Comment #8 from sayle at gcc dot gnu dot org  2006-07-25 23:22 -------
Subject: Bug 28473

Author: sayle
Date: Tue Jul 25 23:21:56 2006
New Revision: 115742

URL: http://gcc.gnu.org/viewcvs?root=gcc&view=rev&rev=115742
Log:

        PR middle-end/28473
        * convert.c (convert_to_integer): When transforming (T)foo(x) into
        bar(x) check that bar's result type can represent all the values of T.

        * gcc.dg/fold-convround-1.c: New test case.


Added:
    trunk/gcc/testsuite/gcc.dg/fold-convround-1.c
Modified:
    trunk/gcc/ChangeLog
    trunk/gcc/convert.c
    trunk/gcc/testsuite/ChangeLog


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=28473

[Bug middle-end/28473] [4.0/4.1 Regression] with -O, casting result of round(x) to uint64_t produces wrong values for x > INT_MAX

[pinskia at gcc dot gnu dot org]

------- Comment #9 from pinskia at gcc dot gnu dot org  2006-07-26 00:24 -------
Fixed on the mainline at least.


-- 

pinskia at gcc dot gnu dot org changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
      Known to work|3.4.6                       |3.4.6 4.2.0
            Summary|[4.0/4.1/4.2 Regression]    |[4.0/4.1 Regression] with -
                   |with -O, casting result of  |O, casting result of
                   |round(x) to uint64_t        |round(x) to uint64_t
                   |produces wrong values for x |produces wrong values for x
                   |> INT_MAX                   |> INT_MAX


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=28473

[Bug middle-end/28473] [4.0/4.1 Regression] with -O, casting result of round(x) to uint64_t produces wrong values for x > INT_MAX

[sayle at gcc dot gnu dot org]

------- Comment #10 from sayle at gcc dot gnu dot org  2006-07-29 19:07 -------
Subject: Bug 28473

Author: sayle
Date: Sat Jul 29 19:07:40 2006
New Revision: 115812

URL: http://gcc.gnu.org/viewcvs?root=gcc&view=rev&rev=115812
Log:

        PR middle-end/28473
        Backport from mainline.
        * convert.c (convert_to_integer): When transforming (T)foo(x) into
        bar(x) check that bar's result type can represent all the values of T.
        * builtins.c (fold_fixed_mathfn): When long and long long are the
        same size, canonicalize llceil*, llfloor*, llround* and llrint*
        functions to their lceil*, lfloor*, lround* and lrint* forms.

        * gcc.dg/fold-convround-1.c: New test case.
        * gcc.dg/builtins-55.c: New test case.


Added:
    branches/gcc-4_1-branch/gcc/testsuite/gcc.dg/builtins-55.c
    branches/gcc-4_1-branch/gcc/testsuite/gcc.dg/fold-convround-1.c
Modified:
    branches/gcc-4_1-branch/gcc/ChangeLog
    branches/gcc-4_1-branch/gcc/builtins.c
    branches/gcc-4_1-branch/gcc/convert.c
    branches/gcc-4_1-branch/gcc/testsuite/ChangeLog


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=28473

[Bug middle-end/28473] [4.0 Regression] with -O, casting result of round(x) to uint64_t produces wrong values for x > INT_MAX

[sayle at gcc dot gnu dot org]

------- Comment #11 from sayle at gcc dot gnu dot org  2006-07-30 13:22 -------
Subject: Bug 28473

Author: sayle
Date: Sun Jul 30 13:21:59 2006
New Revision: 115821

URL: http://gcc.gnu.org/viewcvs?root=gcc&view=rev&rev=115821
Log:

        PR middle-end/28473
        Backport from mainline.
        * convert.c (convert_to_integer): When transforming (T)foo(x) into
        bar(x) check that bar's result type can represent all the values of T.
        * builtins.c (fold_fixed_mathfn): When long and long long are the
        same size, canonicalize llceil*, llfloor*, llround* and llrint*
        functions to their lceil*, lfloor*, lround* and lrint* forms.

        * gcc.dg/fold-convround-1.c: New test case.
        * gcc.dg/builtins-55.c: New test case.


Added:
    branches/gcc-4_0-branch/gcc/testsuite/gcc.dg/builtins-55.c
    branches/gcc-4_0-branch/gcc/testsuite/gcc.dg/fold-convround-1.c
Modified:
    branches/gcc-4_0-branch/gcc/ChangeLog
    branches/gcc-4_0-branch/gcc/builtins.c
    branches/gcc-4_0-branch/gcc/convert.c
    branches/gcc-4_0-branch/gcc/testsuite/ChangeLog


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=28473

[Bug middle-end/28473] [4.0 Regression] with -O, casting result of round(x) to uint64_t produces wrong values for x > INT_MAX

[pinskia at gcc dot gnu dot org]

------- Comment #12 from pinskia at gcc dot gnu dot org  2006-07-30 13:31 -------
Fixed.


-- 

pinskia at gcc dot gnu dot org changed:

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


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=28473