[Bug tree-optimization/114190] New: wrong code with -O2 -fno-dce -fharden-compares -mvpclmulqdq --param=max-rtl-if-conversion-unpredictable-cost=136

[Bug tree-optimization/114190] New: wrong code with -O2 -fno-dce -fharden-compares -mvpclmulqdq --param=max-rtl-if-conversion-unpredictable-cost=136

[zsojka at seznam dot cz]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=114190

            Bug ID: 114190
           Summary: wrong code with -O2 -fno-dce -fharden-compares
                    -mvpclmulqdq
                    --param=max-rtl-if-conversion-unpredictable-cost=136
           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
  Target Milestone: ---
              Host: x86_64-pc-linux-gnu
            Target: x86_64-pc-linux-gnu

Created attachment 57585
  --> https://gcc.gnu.org/bugzilla/attachment.cgi?id=57585&action=edit
reduced testcase

Output:
$ x86_64-pc-linux-gnu-gcc -O2 -fno-dce -fharden-compares -mvpclmulqdq
--param=max-rtl-if-conversion-unpredictable-cost=136 testcase.c -Wno-psabi
$ ./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-r14-9248-20240301110451-gd3d0fcb6527-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-r14-9248-20240301110451-gd3d0fcb6527-checking-yes-rtl-df-extra-nobootstrap-amd64
Thread model: posix
Supported LTO compression algorithms: zlib zstd
gcc version 14.0.1 20240301 (experimental) (GCC)

[Bug target/114190] wrong code with -O2 -fno-dce -fharden-compares -mvpclmulqdq --param=max-rtl-if-conversion-unpredictable-cost=136

[pinskia at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=114190

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


is missing with `-fno-dce`

[Bug target/114190] wrong code with -O2 -fno-dce -fharden-compares -mvpclmulqdq --param=max-rtl-if-conversion-unpredictable-cost=136

[pinskia at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=114190

--- Comment #2 from Andrew Pinski <pinskia at gcc dot gnu.org> ---
So after reload, it looks ok:
(insn 22 21 380 2 (set (reg:CCZ 17 flags)
        (compare:CCZ (reg:SI 4 si [orig:111 _21+4 ] [111])
            (const_int 0 [0]))) "/app/example.cpp":8:8 discrim 1 7
{*cmpsi_ccno_1}
     (nil))
(insn 380 22 381 2 (parallel [
            (set (reg:DI 1 dx [353])
                (plus:DI (reg/f:DI 7 sp)
                    (const_int 212 [0xd4])))
            (clobber (reg:CC 17 flags))
        ]) "/app/example.cpp":8:65 discrim 1 272 {*adddi_1}
     (expr_list:REG_EQUIV (plus:DI (reg/f:DI 16 argp)
            (const_int 76 [0x4c]))
        (nil)))
(insn 381 380 466 2 (set (reg:CCZ 17 flags)
        (compare:CCZ (reg:SI 4 si [orig:111 _21+4 ] [111])
            (const_int 0 [0]))) "/app/example.cpp":8:65 discrim 1 7
{*cmpsi_ccno_1}
     (nil))

Note without -fno-dce, `insn 22` is not in the IR and that is the only
difference so far. THis is ok at this point.

[Bug rtl-optimization/114190] wrong code with -O2 -fno-dce -fharden-compares -mvpclmulqdq --param=max-rtl-if-conversion-unpredictable-cost=136

[pinskia at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=114190

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|UNCONFIRMED                 |NEW
     Ever confirmed|0                           |1
   Last reconfirmed|                            |2024-03-01
          Component|target                      |rtl-optimization

--- Comment #3 from Andrew Pinski <pinskia at gcc dot gnu.org> ---
cmpelim produces:
```
(insn 22 21 486 2 (set (reg:CCZ 17 flags)
        (compare:CCZ (reg:SI 4 si [orig:111 _21+4 ] [111])
            (const_int 0 [0]))) "/app/example.cpp":8:8 discrim 1 7
{*cmpsi_ccno_1}
     (expr_list:REG_UNUSED (reg:CCZ 17 flags)
        (nil)))
(insn 486 22 466 2 (set (reg:DI 1 dx [353])
        (plus:DI (reg/f:DI 7 sp)
            (const_int 212 [0xd4]))) "/app/example.cpp":8:65 discrim 1 254
{*leadi}
     (nil))
(insn 466 486 383 2 (set (reg:DI 4 si [354])
        (const:DI (plus:DI (symbol_ref:DI ("u") [flags 0x2]  <var_decl
0x7fe775410c60 u>)
                (const_int 1 [0x1])))) "/app/example.cpp":8:65 discrim 1 84
{*movdi_internal}
     (expr_list:REG_EQUIV (const:DI (plus:DI (symbol_ref:DI ("u") [flags 0x2] 
<var_decl 0x7fe775410c60 u>)
                (const_int 1 [0x1])))
        (nil)))
(insn 383 466 384 2 (set (reg:DI 1 dx [348])
        (if_then_else:DI (eq (reg:CCZ 17 flags)
                (const_int 0 [0]))
            (reg:DI 1 dx [353])
            (reg:DI 4 si [354]))) "/app/example.cpp":8:65 discrim 1 1451
{*movdicc_noc}
     (expr_list:REG_DEAD (reg:CCZ 17 flags)
        (expr_list:REG_DEAD (reg:DI 4 si [354])
            (nil))))
```

Which is fine (but note the REG_UNUSED note which was not updated).

From pro_and_epilogue's dump:
deleting insn with uid = 22.

Which is totally bogus; I think someone didn't redo REG_UNUSED notes again.

[Bug rtl-optimization/114190] [14 regression] Wrong code with -O2 -fno-dce -fharden-compares -mvpclmulqdq --param=max-rtl-if-conversion-unpredictable-cost=136

[rguenth at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=114190

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
   Target Milestone|---                         |14.0

[Bug rtl-optimization/114190] [14 regression] Wrong code with -O2 -fno-dce -fharden-compares -mvpclmulqdq --param=max-rtl-if-conversion-unpredictable-cost=136

[rguenth at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=114190

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

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

[Bug rtl-optimization/114190] [14 regression] Wrong code with -O2 -fno-dce -fharden-compares -mvpclmulqdq --param=max-rtl-if-conversion-unpredictable-cost=136

[jakub at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=114190

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |jakub at gcc dot gnu.org

--- Comment #4 from Jakub Jelinek <jakub at gcc dot gnu.org> ---
Started with r14-5355-g3cd3a09b3f91a1d023cb180763d40598d6bb274b

[Bug rtl-optimization/114190] [14 regression] Wrong code with -O2 -fno-dce -fharden-compares -mvpclmulqdq --param=max-rtl-if-conversion-unpredictable-cost=136

[jakub at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=114190

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

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

--- Comment #5 from Jakub Jelinek <jakub at gcc dot gnu.org> ---
Created attachment 57611
  --> https://gcc.gnu.org/bugzilla/attachment.cgi?id=57611&action=edit
gcc14-pr114190.patch

Untested fix.  After removing the REG_DEAD/REG_UNUSED notes the pass calls
df_analyze, but as problems are normally only removed during df_finish_pass,
the note problem is still requested and the df_analyze recreates the notes
again.
We could just move the df_analyze before the manual note removal, but I think
it is a waste of time to compute the notes again there.

[Bug rtl-optimization/114190] [14 regression] Wrong code with -O2 -fno-dce -fharden-compares -mvpclmulqdq --param=max-rtl-if-conversion-unpredictable-cost=136

[cvs-commit at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=114190

--- Comment #6 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:1157d5de35b41eabe5ee51d532224864173c37bd

commit r14-9329-g1157d5de35b41eabe5ee51d532224864173c37bd
Author: Jakub Jelinek <jakub@redhat.com>
Date:   Wed Mar 6 09:35:37 2024 +0100

    i386: Fix up the vzeroupper REG_DEAD/REG_UNUSED note workaround [PR114190]

    When writing the rest_of_handle_insert_vzeroupper workaround to manually
    remove all the REG_DEAD/REG_UNUSED notes from the IL, I've missed that
    there is a df_analyze () call right after it and that the problems added
    earlier in the pass, like df_note_add_problem () done during mode
switching,
    doesn't affect just the next df_analyze () call right after it, but all
    other df_analyze () calls until the end of the current pass where
    df_finish_pass removes the optional problems.

    So, as can be seen on the following patch, the workaround doesn't actually
    work there, because while rest_of_handle_insert_vzeroupper carefully
removes
    all REG_DEAD/REG_UNUSED notes, the df_analyze () call at the end of the
    function immediately adds them in again (so, I must say I have no idea
    why the workaround worked on the earlier testcases).

    Now, I could move the df_analyze () call just before the
REG_DEAD/REG_UNUSED
    note removal loop, but I think the following patch is better, because
    the df_analyze () call doesn't have to recompute the problem when we don't
    care about it and will actively strip all traces of it away.

    2024-03-06  Jakub Jelinek  <jakub@redhat.com>

            PR rtl-optimization/114190
            * config/i386/i386-features.cc (rest_of_handle_insert_vzeroupper):
            Call df_remove_problem for df_note before calling df_analyze.

            * gcc.target/i386/avx-pr114190.c: New test.

[Bug rtl-optimization/114190] [14 regression] Wrong code with -O2 -fno-dce -fharden-compares -mvpclmulqdq --param=max-rtl-if-conversion-unpredictable-cost=136

[jakub at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=114190

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

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

--- Comment #7 from Jakub Jelinek <jakub at gcc dot gnu.org> ---
Fixed.