[Bug tree-optimization/107269] New: wrong code at -O1 and above with "-fno-tree-ccp" on x86_64-linux-gnu

[Bug tree-optimization/107269] New: wrong code at -O1 and above with "-fno-tree-ccp" on x86_64-linux-gnu

[zhendong.su at inf dot ethz.ch]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=107269

            Bug ID: 107269
           Summary: wrong code at -O1 and above with "-fno-tree-ccp" on
                    x86_64-linux-gnu
           Product: gcc
           Version: unknown
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: tree-optimization
          Assignee: unassigned at gcc dot gnu.org
          Reporter: zhendong.su at inf dot ethz.ch
  Target Milestone: ---

It appears to be a recent regression.

Compiler Explorer: https://godbolt.org/z/WdcsavxTh


[593] % gcctk -v
Using built-in specs.
COLLECT_GCC=gcctk
COLLECT_LTO_WRAPPER=/local/suz-local/software/local/gcc-trunk/libexec/gcc/x86_64-pc-linux-gnu/13.0.0/lto-wrapper
Target: x86_64-pc-linux-gnu
Configured with: ../gcc-trunk/configure --disable-bootstrap
--enable-checking=yes --prefix=/local/suz-local/software/local/gcc-trunk
--enable-sanitizers --enable-languages=c,c++ --disable-werror --enable-multilib
--with-system-zlib
Thread model: posix
Supported LTO compression algorithms: zlib
gcc version 13.0.0 20221014 (experimental) [master r13-3307-g8efc38347a7] (GCC) 
[594] % 
[594] % gcctk -O1 small.c; ./a.out
[595] % 
[595] % gcctk -O1 -fno-tree-ccp small.c
[596] % ./a.out
Floating point exception
[597] % 
[597] % cat small.c
int a, b, f;
char c = 3;
unsigned d = 1;
short e;
int main() {
  int h = 2357999721;
  char i = a = 3;
  for (; a >= 0; a--) {
    int *j = &h;
    char k = i || c;
    short l = ~h;
    b = ~(d % h);
    f = ~i / b;
    if (h >= -1936967575) {
      int m = h / d / 110000000 ^ (f & c);
      short n = c;
      if (h <= -1936967575)
        n = e;
      h = m;
      c = ~(n + f);
      d = -1;
    }
    i = ~l - k;
  }
  return 0;
}

[Bug tree-optimization/107269] wrong code at -O1 and above with "-fno-tree-ccp" on x86_64-linux-gnu

[zhendong.su at inf dot ethz.ch]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=107269

--- Comment #1 from Zhendong Su <zhendong.su at inf dot ethz.ch> ---
A related instance:

[537] % gcctk -v
Using built-in specs.
COLLECT_GCC=gcctk
COLLECT_LTO_WRAPPER=/local/suz-local/software/local/gcc-trunk/libexec/gcc/x86_64-pc-linux-gnu/13.0.0/lto-wrapper
Target: x86_64-pc-linux-gnu
Configured with: ../gcc-trunk/configure --disable-bootstrap
--enable-checking=yes --prefix=/local/suz-local/software/local/gcc-trunk
--enable-sanitizers --enable-languages=c,c++ --disable-werror --enable-multilib
--with-system-zlib
Thread model: posix
Supported LTO compression algorithms: zlib
gcc version 13.0.0 20221014 (experimental) [master r13-3307-g8efc38347a7] (GCC) 
[538] % 
[538] % gcctk -O1 small.c; ./a.out
[539] % 
[539] % gcctk -O1 -fno-tree-ccp small.c
[540] % timeout -s 9 5 ./a.out
Killed
[541] % 
[541] % cat small.c
int a, c, d;
unsigned b;
int main() {
  int f = 1;
 L1:
  c = f;
 L2:
  b & f;
  short i = -f;
  if (!a) {
    d = f;
    f = ~(b % f);
    if (d > -1)
      goto L2;
  }
  f = i;
  if (a)
    goto L1;
  int *j = &f;
  return 0;
}

[Bug tree-optimization/107269] wrong code at -O1 and above with "-fno-tree-ccp" on x86_64-linux-gnu

[zhendong.su at inf dot ethz.ch]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=107269

--- Comment #2 from Zhendong Su <zhendong.su at inf dot ethz.ch> ---
Another related instance:

[642] % gcctk -v
Using built-in specs.
COLLECT_GCC=gcctk
COLLECT_LTO_WRAPPER=/local/suz-local/software/local/gcc-trunk/libexec/gcc/x86_64-pc-linux-gnu/13.0.0/lto-wrapper
Target: x86_64-pc-linux-gnu
Configured with: ../gcc-trunk/configure --disable-bootstrap
--enable-checking=yes --prefix=/local/suz-local/software/local/gcc-trunk
--enable-sanitizers --enable-languages=c,c++ --disable-werror --enable-multilib
--with-system-zlib
Thread model: posix
Supported LTO compression algorithms: zlib
gcc version 13.0.0 20221015 (experimental) [master r13-3311-gbaeec7cc83b] (GCC)
[643] %
[643] % gcctk -O1 small.c; ./a.out
[644] %
[644] % gcctk -O1 -fno-tree-ccp small.c
[645] % ./a.out
Aborted
[646] %
[646] % cat small.c
char a;
unsigned b;
int d = 39381789, e, f, g = 1;
long h, i;
int main() {
  int j = 38;
  long k = 1;
  for (; e < 8; e++)
    if (k)
      while (1) {
        int l = j * h;
        char m = j;
        if (g)
          m = a;
        f = j;
        b = d;
        j = d = b * f;
        if (d < 665)
          break;
        if (f > d)
          k = m;
        long n = ~(b - i);
        i = ~h;
        if (b <= 46295405)
          break;
        h = n;
        int *c = &j;
        if (i > 10843)
          __builtin_abort();
      }
  return 0;
}

[Bug tree-optimization/107269] wrong code at -O1 and above with "-fno-tree-ccp" on x86_64-linux-gnu

[crazylht at gmail dot com]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=107269

Hongtao.liu <crazylht at gmail dot com> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |crazylht at gmail dot com

--- Comment #3 from Hongtao.liu <crazylht at gmail dot com> ---
Looks like the same issue as PR107172 since below change can also pass all the
testcase in the PR.

diff --git a/gcc/config/i386/i386.md b/gcc/config/i386/i386.md
index 8e847520491..be815341af5 100644
--- a/gcc/config/i386/i386.md
+++ b/gcc/config/i386/i386.md
@@ -21335,7 +21335,7 @@ (define_insn "*x86_mov<mode>cc_0_m1_neg"
 (define_expand "x86_mov<mode>cc_0_m1_neg"
   [(parallel
     [(set (match_operand:SWI48 0 "register_operand")
-         (neg:SWI48 (ltu:SWI48 (reg:CCC FLAGS_REG) (const_int 0))))
+         (neg:SWI48 (ltu:SWI48 (reg:CC FLAGS_REG) (const_int 0))))
      (clobber (reg:CC FLAGS_REG))])])

 (define_split

[Bug tree-optimization/107269] wrong code at -O1 and above with "-fno-tree-ccp" on x86_64-linux-gnu

[rguenth at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=107269

Richard Biener <rguenth at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
         Resolution|---                         |DUPLICATE
             Status|UNCONFIRMED                 |RESOLVED

--- Comment #4 from Richard Biener <rguenth at gcc dot gnu.org> ---
dup then

*** This bug has been marked as a duplicate of bug 107172 ***

[Bug tree-optimization/107269] wrong code at -O1 and above with "-fno-tree-ccp" on x86_64-linux-gnu

[hjl.tools at gmail dot com]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=107269

H.J. Lu <hjl.tools at gmail dot com> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
            Version|unknown                     |13.0

--- Comment #5 from H.J. Lu <hjl.tools at gmail dot com> ---


*** This bug has been marked as a duplicate of bug 107273 ***

[Bug tree-optimization/107269] wrong code at -O1 and above with "-fno-tree-ccp" on x86_64-linux-gnu

[crazylht at gmail dot com]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=107269

--- Comment #6 from Hongtao.liu <crazylht at gmail dot com> ---
(In reply to Hongtao.liu from comment #3)
> Looks like the same issue as PR107172 since below change can also pass all
> the
> testcase in the PR.
> 
> diff --git a/gcc/config/i386/i386.md b/gcc/config/i386/i386.md
> index 8e847520491..be815341af5 100644
> --- a/gcc/config/i386/i386.md
> +++ b/gcc/config/i386/i386.md
> @@ -21335,7 +21335,7 @@ (define_insn "*x86_mov<mode>cc_0_m1_neg"
>  (define_expand "x86_mov<mode>cc_0_m1_neg"
>    [(parallel
>      [(set (match_operand:SWI48 0 "register_operand")
> -         (neg:SWI48 (ltu:SWI48 (reg:CCC FLAGS_REG) (const_int 0))))
> +         (neg:SWI48 (ltu:SWI48 (reg:CC FLAGS_REG) (const_int 0))))
>       (clobber (reg:CC FLAGS_REG))])])
> 
>  (define_split

Sorry, my patch against an older gcc before the guilty commit, and I compare
the result between it and latest GCC, that's why it show the patch pass the
testcase.