[Bug rtl-optimization/113044] New: [14 Regression] wrong code with vector shift at -O1

[Bug rtl-optimization/113044] New: [14 Regression] wrong code with vector shift at -O1

[zsojka at seznam dot cz]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=113044

            Bug ID: 113044
           Summary: [14 Regression] wrong code with vector shift at -O1
           Product: gcc
           Version: 14.0
            Status: UNCONFIRMED
          Keywords: wrong-code
          Severity: normal
          Priority: P3
         Component: rtl-optimization
          Assignee: unassigned at gcc dot gnu.org
          Reporter: zsojka at seznam dot cz
  Target Milestone: ---
              Host: x86_64-pc-linux-gnu
            Target: x86_64-pc-linux-gnu

Created attachment 56891
  --> https://gcc.gnu.org/bugzilla/attachment.cgi?id=56891&action=edit
reduced testcase

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

In the assembly:
...
foo:
...
        mov     al, dl  # tmp110, v
        shr     al, cl  # tmp110, _1
        movzx   ecx, ch # tmp117, _1
        mov     ah, dl  # tmp110, v
        shr     ah, cl  # tmp110, tmp117
...

'dl' contains the low byte of the vector; dh should be used for the second
shift:

$ diff -u a-testcase.s.BAD a-testcase.s
--- a-testcase.s.BAD    2023-12-16 10:52:17.108156425 +0100
+++ a-testcase.s        2023-12-16 10:52:22.871458132 +0100
@@ -23,7 +23,7 @@
        mov     al, dl  # tmp110, v
        shr     al, cl  # tmp110, _1
        movzx   ecx, ch # tmp117, _1
-       mov     ah, dl  # tmp110, v
+       mov     ah, dh  # tmp110, v
        shr     ah, cl  # tmp110, tmp117
 # testcase.c:7:   volatile char d = c;
        mov     BYTE PTR [rsp+31], dil  # d, c

fixes the assembly


$ 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-r14-6619-20231216150950-g39f9c426f58-checking-yes-rtl-df-extra-nobootstrap-amd64/bin/../libexec/gcc/x86_64-pc-linux-gnu/14.0.0/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 --disable-libstdcxx-pch
--prefix=/repo/gcc-trunk//binary-trunk-r14-6619-20231216150950-g39f9c426f58-checking-yes-rtl-df-extra-nobootstrap-amd64
Thread model: posix
Supported LTO compression algorithms: zlib zstd
gcc version 14.0.0 20231216 (experimental) (GCC)

[Bug target/113044] [14 Regression] wrong code with vector shift at -O1

[pinskia at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=113044

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
          Component|rtl-optimization            |target
   Target Milestone|---                         |14.0

[Bug target/113044] [14 Regression] wrong code with vector shift at -O1

[pinskia at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=113044

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
     Ever confirmed|0                           |1
   Last reconfirmed|                            |2023-12-16
             Status|UNCONFIRMED                 |NEW
           Keywords|                            |needs-bisection

--- Comment #1 from Andrew Pinski <pinskia at gcc dot gnu.org> ---
Confirmed.

[Bug target/113044] [14 Regression] wrong code with vector shift at -O1 since r14-5254

[jakub at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=113044

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |jakub at gcc dot gnu.org,
                   |                            |uros at gcc dot gnu.org
           Keywords|needs-bisection             |
            Summary|[14 Regression] wrong code  |[14 Regression] wrong code
                   |with vector shift at -O1    |with vector shift at -O1
                   |                            |since r14-5254

--- Comment #2 from Jakub Jelinek <jakub at gcc dot gnu.org> ---
Started with r14-5254-gdced5ae64703507a7159972316a1dde48e5f7470

[Bug target/113044] [14 Regression] wrong code with vector shift at -O1 since r14-5254

[ubizjak at gmail dot com]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=113044

Uroš Bizjak <ubizjak at gmail dot com> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
           Assignee|unassigned at gcc dot gnu.org      |ubizjak at gmail dot com
             Status|NEW                         |ASSIGNED

--- Comment #3 from Uroš Bizjak <ubizjak at gmail dot com> ---
Created attachment 56917
  --> https://gcc.gnu.org/bugzilla/attachment.cgi?id=56917&action=edit
Proposed patch

Patch in testing.

[Bug target/113044] [14 Regression] wrong code with vector shift at -O1 since r14-5254

[cvs-commit at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=113044

--- Comment #4 from GCC Commits <cvs-commit at gcc dot gnu.org> ---
The master branch has been updated by Uros Bizjak <uros@gcc.gnu.org>:

https://gcc.gnu.org/g:2766b83759a02572b7b303aae3d4b54a351f8f96

commit r14-6787-g2766b83759a02572b7b303aae3d4b54a351f8f96
Author: Uros Bizjak <ubizjak@gmail.com>
Date:   Thu Dec 21 13:50:26 2023 +0100

    i386: Fix shifts with high register input operand [PR113044]

    The move to the output operand should use high register input operand.

            PR target/113044

    gcc/ChangeLog:

            * config/i386/i386.md (*ashlqi_ext<mode>_1): Move from the
            high register of the input operand.
            (*<insn>qi_ext<mode>_1): Ditto.

    gcc/testsuite/ChangeLog:

            * gcc.target/i386/pr113044.c: New test.

[Bug target/113044] [14 Regression] wrong code with vector shift at -O1 since r14-5254

[ubizjak at gmail dot com]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=113044

Uroš Bizjak <ubizjak at gmail dot com> changed:

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

--- Comment #5 from Uroš Bizjak <ubizjak at gmail dot com> ---
Fixed.