[Bug tree-optimization/114425] New: wrong code with _BitInt() __builtin_add_overflow_p() at -O2

[Bug tree-optimization/114425] New: wrong code with _BitInt() __builtin_add_overflow_p() at -O2

[zsojka at seznam dot cz]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=114425

            Bug ID: 114425
           Summary: wrong code with _BitInt() __builtin_add_overflow_p()
                    at -O2
           Product: gcc
           Version: 14.0
            Status: UNCONFIRMED
          Keywords: wrong-code
          Severity: normal
          Priority: P3
         Component: tree-optimization
          Assignee: unassigned at gcc dot gnu.org
          Reporter: zsojka at seznam dot cz
                CC: jakub at gcc dot gnu.org
  Target Milestone: ---
              Host: x86_64-pc-linux-gnu
            Target: x86_64-pc-linux-gnu

Created attachment 57764
  --> https://gcc.gnu.org/bugzilla/attachment.cgi?id=57764&action=edit
reduced testcase

Output:
$ x86_64-pc-linux-gnu-gcc -O2 testcase.c
$ ./a.out 
Aborted

$ x86_64-pc-linux-gnu-gcc -v
Using built-in specs.
COLLECT_GCC=/repo/gcc-trunk/binary-latest-amd64/bin/x86_64-pc-linux-gnu-gcc
COLLECT_LTO_WRAPPER=/repo/gcc-trunk/binary-trunk-r14-9595-20240321132344-ga2fe34e0b99-checking-yes-rtl-df-extra-nobootstrap-amd64/bin/../libexec/gcc/x86_64-pc-linux-gnu/14.0.1/lto-wrapper
Target: x86_64-pc-linux-gnu
Configured with: /repo/gcc-trunk//configure --enable-languages=c,c++
--enable-valgrind-annotations --disable-nls --enable-checking=yes,rtl,df,extra
--disable-bootstrap --with-cloog --with-ppl --with-isl
--build=x86_64-pc-linux-gnu --host=x86_64-pc-linux-gnu
--target=x86_64-pc-linux-gnu --with-ld=/usr/bin/x86_64-pc-linux-gnu-ld
--with-as=/usr/bin/x86_64-pc-linux-gnu-as --enable-libsanitizer
--disable-libstdcxx-pch
--prefix=/repo/gcc-trunk//binary-trunk-r14-9595-20240321132344-ga2fe34e0b99-checking-yes-rtl-df-extra-nobootstrap-amd64
Thread model: posix
Supported LTO compression algorithms: zlib zstd
gcc version 14.0.1 20240321 (experimental) (GCC)

[Bug tree-optimization/114425] wrong code with _BitInt() __builtin_add_overflow_p() at -O2

[zsojka at seznam dot cz]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=114425

--- Comment #1 from Zdenek Sojka <zsojka at seznam dot cz> ---
Created attachment 57771
  --> https://gcc.gnu.org/bugzilla/attachment.cgi?id=57771&action=edit
another testcase, using __builtin_mul_overflow_p

Output for second testcase:
$ x86_64-pc-linux-gnu-gcc -O2 testcase2.c
$ ./a.out 
Aborted

[Bug tree-optimization/114425] wrong code with _BitInt() __builtin_add_overflow_p() and __builtin_mul_overflow_p() at -O2

[jakub at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=114425

--- Comment #2 from Jakub Jelinek <jakub at gcc dot gnu.org> ---
Strangely it is dependent on the printf loop, without that it works fine.

Slightly adjusted testcase:

#if __BITINT_MAXWIDTH__ >= 2000
_BitInt(8) a;
_BitInt(300) b;
_BitInt(2000) c;

unsigned
foo (_BitInt(2000) d)
{
  int o = __builtin_add_overflow_p (d, 0, b);
  _BitInt(2000) m = c * a;
  unsigned u = m;
  return u + o;
}

__attribute__((noipa)) void
bar (int x)
{
  (void) x;
}
#endif

int
main ()
{
#if __BITINT_MAXWIDTH__ >= 2000
  unsigned x = foo
(0xfa7ac16f2613255eeb217e871c1f02221e26ce11f82d6a33206ec0ad5d4414722019933c0e2wb);
  for (unsigned i = 0; i < sizeof (x); i++)
    bar (((volatile unsigned char *) &x)[i]);
  if (x != 1)
    __builtin_abort ();
#endif
}

[Bug tree-optimization/114425] wrong code with _BitInt() __builtin_add_overflow_p() and __builtin_mul_overflow_p() at -O2

[zsojka at seznam dot cz]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=114425

--- Comment #3 from Zdenek Sojka <zsojka at seznam dot cz> ---
Created attachment 57772
  --> https://gcc.gnu.org/bugzilla/attachment.cgi?id=57772&action=edit
unreduced testcase ; output differs between -O0 and -O2

[Bug tree-optimization/114425] wrong code with _BitInt() __builtin_add_overflow_p() and __builtin_mul_overflow_p() at -O2

[zsojka at seznam dot cz]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=114425

--- Comment #4 from Zdenek Sojka <zsojka at seznam dot cz> ---
Created attachment 57773
  --> https://gcc.gnu.org/bugzilla/attachment.cgi?id=57773&action=edit
another unreduced testcase ; output differs between -O0 and -O2

[Bug tree-optimization/114425] wrong code with _BitInt() __builtin_add_overflow_p() and __builtin_mul_overflow_p() at -O2

[jakub at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=114425

Jakub Jelinek <jakub at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
     Ever confirmed|0                           |1
   Last reconfirmed|                            |2024-03-22
           Assignee|unassigned at gcc dot gnu.org      |jakub at gcc dot gnu.org
             Status|UNCONFIRMED                 |ASSIGNED

--- Comment #5 from Jakub Jelinek <jakub at gcc dot gnu.org> ---
Created attachment 57779
  --> https://gcc.gnu.org/bugzilla/attachment.cgi?id=57779&action=edit
gcc14-pr114425.patch

Untested fix.

[Bug tree-optimization/114425] wrong code with _BitInt() __builtin_add_overflow_p() and __builtin_mul_overflow_p() at -O2

[jakub at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=114425

--- Comment #6 from Jakub Jelinek <jakub at gcc dot gnu.org> ---
Though, guess it would help if evrp avoided undesirable propagation here:
It is changing:
   <bb 2> :
   # DEBUG BEGIN_STMT
   _8 = .ADD_OVERFLOW (d_7(D), 0);
   _1 = IMAGPART_EXPR <_8>;
   _2 = (_Bool) _1;
   # DEBUG o => (int) _2
   # DEBUG BEGIN_STMT
   a.0_3 = a;
   _4 = (_BitInt(2000)) a.0_3;
   c.1_5 = c;
   m_11 = _4 * c.1_5;
   # DEBUG m => m_11
   # DEBUG BEGIN_STMT
   u_12 = (unsigned int) m_11;
   # DEBUG u => u_12
   # DEBUG BEGIN_STMT
-  o.2_6 = (unsigned int) _2;
+  o.2_6 = (unsigned int) _1;
   _13 = o.2_6 + u_12;
   return _13;

 }
which is surely possible because IMAGPART_EXPR of .{ADD,SUB,MUL}_OVERFLOW has
[0, 1]
range, but for such large/huge _BitInt IMAGPART_EXPR it extends the lifetime of
an expensive large variable rather than just being a boolean.

[Bug tree-optimization/114425] wrong code with _BitInt() __builtin_add_overflow_p() and __builtin_mul_overflow_p() at -O2

[cvs-commit at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=114425

--- Comment #7 from GCC Commits <cvs-commit at gcc dot gnu.org> ---
The master branch has been updated by Jakub Jelinek <jakub@gcc.gnu.org>:

https://gcc.gnu.org/g:f92cf8cbbe199bda70d0dd7893e8c8836777e2d0

commit r14-9640-gf92cf8cbbe199bda70d0dd7893e8c8836777e2d0
Author: Jakub Jelinek <jakub@redhat.com>
Date:   Sat Mar 23 11:19:09 2024 +0100

    bitint: Handle complex types in build_bitint_stmt_ssa_conflicts [PR114425]

    The task of the build_bitint_stmt_ssa_conflicts hook for
    tree-ssa-coalesce.cc next to special casing the
    multiplication/division/modulo is to ignore statements with
    large/huge _BitInt lhs which isn't in names bitmap and on the
    other side pretend all uses of the stmt are used in a later stmt
    (single user of that SSA_NAME or perhaps single user of lhs of
    the single user etc.) where the lowering will actually emit the
    code.

    Unfortunately the function wasn't handling COMPLEX_TYPE of the large/huge
    BITINT_TYPE, while the FE doesn't really support such types, they are
    used under the hood for __builtin_{add,sub,mul}_overflow{,_p}, they are
    also present or absent from the names bitmap and should be treated the
same.

    Without this patch, the operands of .ADD_OVERFLOW were incorrectly
pretended
    to be used right in that call statement rather than on the cast stmt from
    IMAGPART_EXPR of .ADD_OVERFLOW return value to some integral type.

    2024-03-23  Jakub Jelinek  <jakub@redhat.com>

            PR tree-optimization/114425
            * gimple-lower-bitint.cc (build_bitint_stmt_ssa_conflicts): Handle
            _Complex large/huge _BitInt types like the large/huge _BitInt
types.

            * gcc.dg/torture/bitint-67.c: New test.

[Bug tree-optimization/114425] wrong code with _BitInt() __builtin_add_overflow_p() and __builtin_mul_overflow_p() at -O2

[jakub at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=114425

Jakub Jelinek <jakub at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|ASSIGNED                    |RESOLVED
   Target Milestone|---                         |14.0
         Resolution|---                         |FIXED

--- Comment #8 from Jakub Jelinek <jakub at gcc dot gnu.org> ---
Fixed.