[Bug tree-optimization/111648] New: Wrong code at -O2/3 on x86_64-linux-gnu since r14-3243-ga7dba4a1c05

[Bug tree-optimization/111648] New: Wrong code at -O2/3 on x86_64-linux-gnu since r14-3243-ga7dba4a1c05

[shaohua.li at inf dot ethz.ch]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=111648

            Bug ID: 111648
           Summary: Wrong code at -O2/3 on x86_64-linux-gnu since
                    r14-3243-ga7dba4a1c05
           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: shaohua.li at inf dot ethz.ch
                CC: prathamesh3492 at gcc dot gnu.org
  Target Milestone: ---

gcc at -O2/3 produced the wrong code

Bisected to r14-3243-ga7dba4a1c05

Compiler explorer: https://godbolt.org/z/1bvWTrEK1

$ cat a.c
int printf(const char *, ...);
int a;
int *b = &a;
static int **c = &b;
static int d;
short e;
char f;
int main() {
  f = -21;
  for (; f < -5; f++) {
    e = f ^ 3;
    d = *b;
    **c = e;
  }
  printf("%d\n", d);
}
$
$ gcc -O0 a.c && ./a.out
-6
$ gcc -O2 a.c && ./a.out
2
$

[Bug tree-optimization/111648] Wrong code at -O2/3 on x86_64-linux-gnu since r14-3243-ga7dba4a1c05

[prathamesh3492 at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=111648

--- Comment #1 from prathamesh3492 at gcc dot gnu.org ---
Hi,
Sorry for the breakage, will take a look.

Thanks,
Prathamesh

[Bug tree-optimization/111648] [14 Regression] Wrong code at -O2/3 on x86_64-linux-gnu since r14-3243-ga7dba4a1c05

[pinskia at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=111648

Andrew Pinski <pinskia at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
   Target Milestone|---                         |14.0
             Status|UNCONFIRMED                 |ASSIGNED
     Ever confirmed|0                           |1
   Last reconfirmed|                            |2023-10-01
            Summary|Wrong code at -O2/3 on      |[14 Regression] Wrong code
                   |x86_64-linux-gnu since      |at -O2/3 on
                   |r14-3243-ga7dba4a1c05       |x86_64-linux-gnu since
                   |                            |r14-3243-ga7dba4a1c05

--- Comment #2 from Andrew Pinski <pinskia at gcc dot gnu.org> ---
.

[Bug tree-optimization/111648] [14 Regression] Wrong code at -O2/3 on x86_64-linux-gnu since r14-3243-ga7dba4a1c05

[prathamesh3492 at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=111648

--- Comment #3 from prathamesh3492 at gcc dot gnu.org ---
Created attachment 56037
  --> https://gcc.gnu.org/bugzilla/attachment.cgi?id=56037&action=edit
Untested fix

The issue is that when a1 is a multiple of vector length, we end up creating
following encoding in result: { base_elem, arg[0], arg[1], ... } where arg is
chosen input vector, which is incorrect.

For above case, vectorizer pass creates VEC_PERM_EXPR<arg0, arg, sel> where:
arg0: { -16, -9, -10, -11 } 
arg1: { -12, -5, -6, -7 } 
sel = { 3, 4, 5, 6 }

arg0, arg1 and sel are encoded with npatterns = 1 and nelts_per_pattern = 3.
Since a1 = 4 and arg_len = 4, it ended up creating the result with
following encoding:
res = { arg0[3], arg1[0], arg1[1] } // npatterns = 1, nelts_per_pattern = 3
    = { -11, -12, -5 }

So for res[4], it used S = (-5) - (-12) = 7
And hence computed it as -5 + 7 = 2.
instead of arg1[2], ie, -6.
which is the difference we see in output at -O0 vs -O2.

The patch tweaks the constratints in valid_mask_for_fold_vec_perm_cst_p to punt
if a1 is a multiple of vector length, so a1 ... ae only selects from stepped
part of the input vector, which seems to fix this issue.
I will run a proper bootstrap+test and post it upstream.

[Bug tree-optimization/111648] [14 Regression] Wrong code at -O2/3 on x86_64-linux-gnu since r14-3243-ga7dba4a1c05

[prathamesh3492 at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=111648

--- Comment #4 from prathamesh3492 at gcc dot gnu.org ---
(In reply to prathamesh3492 from comment #3)
> Created attachment 56037 [details]
> Untested fix
> 
> The issue is that when a1 is a multiple of vector length, we end up creating
> following encoding in result: { base_elem, arg[0], arg[1], ... } where arg
> is chosen input vector, which is incorrect.
> 
> For above case, vectorizer pass creates VEC_PERM_EXPR<arg0, arg, sel> where:
> arg0: { -16, -9, -10, -11 } 
> arg1: { -12, -5, -6, -7 } 
> sel = { 3, 4, 5, 6 }
> 
> arg0, arg1 and sel are encoded with npatterns = 1 and nelts_per_pattern = 3.
> Since a1 = 4 and arg_len = 4, it ended up creating the result with
> following encoding:
> res = { arg0[3], arg1[0], arg1[1] } // npatterns = 1, nelts_per_pattern = 3
>     = { -11, -12, -5 }
> 
> So for res[4], it used S = (-5) - (-12) = 7
Typo: I meant res[3], not res[4]. Sorry.
> And hence computed it as -5 + 7 = 2.
> instead of arg1[2], ie, -6.
> which is the difference we see in output at -O0 vs -O2.
> 
> The patch tweaks the constratints in valid_mask_for_fold_vec_perm_cst_p to
> punt if a1 is a multiple of vector length, so a1 ... ae only selects from
> stepped part of the input vector, which seems to fix this issue.
> I will run a proper bootstrap+test and post it upstream.

[Bug tree-optimization/111648] [14 Regression] Wrong code at -O2/3 on x86_64-linux-gnu since r14-3243-ga7dba4a1c05

[rguenth at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=111648

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
           Priority|P3                          |P1

[Bug tree-optimization/111648] [14 Regression] Wrong code at -O2/3 on x86_64-linux-gnu since r14-3243-ga7dba4a1c05

[cvs-commit at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=111648

--- Comment #5 from CVS Commits <cvs-commit at gcc dot gnu.org> ---
The master branch has been updated by Prathamesh Kulkarni
<prathamesh3492@gcc.gnu.org>:

https://gcc.gnu.org/g:3ec8ecb8e92faec889bc6f7aeac9ff59e82b4f7f

commit r14-4726-g3ec8ecb8e92faec889bc6f7aeac9ff59e82b4f7f
Author: Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
Date:   Thu Oct 19 00:29:38 2023 +0530

    PR111648: Fix wrong code-gen due to incorrect VEC_PERM_EXPR folding.

    gcc/ChangeLog:
            PR tree-optimization/111648
            * fold-const.cc (valid_mask_for_fold_vec_perm_cst_p): If a1
            chooses base element from arg, ensure that it's a natural stepped
            sequence.
            (build_vec_cst_rand): New param natural_stepped and use it to
            construct a naturally stepped sequence.
            (test_nunits_min_2): Add new unit tests Case 6 and Case 7.

[Bug tree-optimization/111648] [14 Regression] Wrong code at -O2/3 on x86_64-linux-gnu since r14-3243-ga7dba4a1c05

[prathamesh3492 at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=111648

prathamesh3492 at gcc dot gnu.org changed:

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

--- Comment #6 from prathamesh3492 at gcc dot gnu.org ---
Fixed.