[Bug tree-optimization/109535] New: internal compiler error: in finalize_new_accesses, at rtl-ssa/changes.cc:471

[Bug tree-optimization/109535] New: internal compiler error: in finalize_new_accesses, at rtl-ssa/changes.cc:471

[malat at debian dot org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=109535

            Bug ID: 109535
           Summary: internal compiler error: in finalize_new_accesses, at
                    rtl-ssa/changes.cc:471
           Product: gcc
           Version: 13.0
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: tree-optimization
          Assignee: unassigned at gcc dot gnu.org
          Reporter: malat at debian dot org
  Target Milestone: ---

I cannot compile highway on riscv/rv64gcv1p0 with gcc-13.

Compilation fais with:

% /usr/bin/g++-13 -freport-bug -DHWY_SHARED_DEFINE
-I"/home/malat/highway-1.0.4~git20230317.8681eb8" -g -O2
-ffile-prefix-map=/home/malat/highway-1.0.4~git20230317.8681eb8=.
-fstack-protector-strong -Wformat -Werror=format-security -DHWY_BROKEN_EMU128=0
-Wdate-time -D_FORTIFY_SOURCE=2 -fPIE -fvisibility=hidden
-fvisibility-inlines-hidden -Wno-builtin-macro-redefined
-D__DATE__=\"redacted\" -D__TIMESTAMP__=\"redacted\" -D__TIME__=\"redacted\"
-fmerge-all-constants -Wall -Wextra -Wconversion -Wsign-conversion -Wvla
-Wnon-virtual-dtor -fmath-errno -fno-exceptions -march=rv64gcv1p0 -Werror
-DHWY_IS_TEST=1 -DGTEST_HAS_PTHREAD=1 -MD -MT
CMakeFiles/transform_test.dir/hwy/contrib/algo/transform_test.cc.o -MF
CMakeFiles/transform_test.dir/hwy/contrib/algo/transform_test.cc.o.d -o
CMakeFiles/transform_test.dir/hwy/contrib/algo/transform_test.cc.o -c
'/home/malat/highway-1.0.4~git20230317.8681eb8/hwy/contrib/algo/transform_test.cc'
during RTL pass: vsetvl
/home/malat/highway-1.0.4~git20230317.8681eb8/hwy/contrib/algo/transform_test.cc:
In function 'void hwy::N_RVV::ForeachCountAndMisalign<Test>::operator()(T, D)
const [with T = unsigned char; D = hwy::N_RVV::Simd<unsigned char, 8192, 2>;
Test = hwy::N_RVV::TestGenerate]':
/home/malat/highway-1.0.4~git20230317.8681eb8/hwy/contrib/algo/transform_test.cc:133:3:
internal compiler error: in finalize_new_accesses, at rtl-ssa/changes.cc:471
  133 |   }
      |   ^
0xf4b18d rtl_ssa::function_info::finalize_new_accesses(rtl_ssa::insn_change&)
        ../../src/gcc/rtl-ssa/changes.cc:471
0xf4c0a7
rtl_ssa::function_info::change_insns(array_slice<rtl_ssa::insn_change*>)
        ../../src/gcc/rtl-ssa/changes.cc:659
0xb4f1e7 rtl_ssa::function_info::change_insn(rtl_ssa::insn_change&)
        ../../src/gcc/rtl-ssa/changes.cc:717
0xb4f1e7 change_insn
        ../../src/gcc/config/riscv/riscv-vsetvl.cc:1028
0xb4f1e7 pass_vsetvl::cleanup_insns() const
        ../../src/gcc/config/riscv/riscv-vsetvl.cc:3951
0xb59891 pass_vsetvl::lazy_vsetvl()
        ../../src/gcc/config/riscv/riscv-vsetvl.cc:4211
0xb59a7d pass_vsetvl::execute(function*)
        ../../src/gcc/config/riscv/riscv-vsetvl.cc:4241
0xb59a7d pass_vsetvl::execute(function*)
        ../../src/gcc/config/riscv/riscv-vsetvl.cc:4222
Please submit a full bug report, with preprocessed source.
Please include the complete backtrace with any bug report.
See <file:///usr/share/doc/gcc-13/README.Bugs> for instructions.

[Bug target/109535] internal compiler error: in finalize_new_accesses, at rtl-ssa/changes.cc:471

[malat at debian dot org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=109535

--- Comment #1 from Mathieu Malaterre <malat at debian dot org> ---
Created attachment 54874
  --> https://gcc.gnu.org/bugzilla/attachment.cgi?id=54874&action=edit
preprocessed source

% gcc-13 --version
gcc-13 (Debian 13-20230411-1) 13.0.1 20230411 (experimental) [master
r13-7137-gb8e32978e3d]

[Bug target/109535] internal compiler error: in finalize_new_accesses, at rtl-ssa/changes.cc:471

[rsandifo at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=109535

--- Comment #2 from rsandifo at gcc dot gnu.org <rsandifo at gcc dot gnu.org> ---
The assert in question fires if the pass creates an instruction
whose pattern uses a register or memory and if the pass doesn't
provide associated use information.  Let me know if it looks like
a bug in rtl-ssa rather than a bug in the vsetvl pass.

[Bug target/109535] internal compiler error: in finalize_new_accesses, at rtl-ssa/changes.cc:471

[juzhe.zhong at rivai dot ai]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=109535

--- Comment #3 from JuzheZhong <juzhe.zhong at rivai dot ai> ---
Thanks for reporting it. We will take a look at it.

[Bug target/109535] internal compiler error: in finalize_new_accesses, at rtl-ssa/changes.cc:471

[juzhe.zhong at rivai dot ai]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=109535

--- Comment #4 from JuzheZhong <juzhe.zhong at rivai dot ai> ---
(In reply to Mathieu Malaterre from comment #1)
> Created attachment 54874 [details]
> preprocessed source
> 
> % gcc-13 --version
> gcc-13 (Debian 13-20230411-1) 13.0.1 20230411 (experimental) [master
> r13-7137-gb8e32978e3d]

Hi, it seems that your attatchment is an excution file instead of 
a source code. It may not be easy for us to debug it.

Can you give me the source code? Thanks.

[Bug target/109535] internal compiler error: in finalize_new_accesses, at rtl-ssa/changes.cc:471

[kito at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=109535

--- Comment #5 from Kito Cheng <kito at gcc dot gnu.org> ---
Confirmed the the output is text file, it's just suffixed with .out

[Bug target/109535] internal compiler error: in finalize_new_accesses, at rtl-ssa/changes.cc:471

[pan2.li at intel dot com]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=109535

Li Pan <pan2.li at intel dot com> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |pan2.li at intel dot com

--- Comment #6 from Li Pan <pan2.li at intel dot com> ---
(In reply to rsandifo@gcc.gnu.org from comment #2)
> The assert in question fires if the pass creates an instruction
> whose pattern uses a register or memory and if the pass doesn't
> provide associated use information.  Let me know if it looks like
> a bug in rtl-ssa rather than a bug in the vsetvl pass.

Just sync with juzhe for the assertion failure. It tries to find the regno=8 in
the shared_uses=[0, 66, 67] in function_info::finalize_new_accesses. And then
it will hit the NOT_NULL assert.

While in the pass_vsetvl::cleanup_insns, it will clean up the AVL use similar
as below.

AVL regno 8
Before Uses Reg Nums => [0,8,66,67,]
After  Uses Reg Nums => [0,66,67,]

After vsetvl, the avl-related use, aka use Regno=8 will be removed, because the
instruction pattern in RVV will eliminate the dependencies of the operand.

Juzhe can help to correct me if any misleading or misunderstanding.

Thanks.

[Bug target/109535] internal compiler error: in finalize_new_accesses, at rtl-ssa/changes.cc:471

[kito at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=109535

Kito Cheng <kito at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
   Last reconfirmed|                            |2023-04-17
             Status|UNCONFIRMED                 |ASSIGNED
     Ever confirmed|0                           |1
           Assignee|unassigned at gcc dot gnu.org      |juzhe.zhong at rivai dot ai

[Bug target/109535] internal compiler error: in finalize_new_accesses, at rtl-ssa/changes.cc:471

[juzhe.zhong at rivai dot ai]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=109535

--- Comment #7 from JuzheZhong <juzhe.zhong at rivai dot ai> ---
(In reply to rsandifo@gcc.gnu.org from comment #2)
> The assert in question fires if the pass creates an instruction
> whose pattern uses a register or memory and if the pass doesn't
> provide associated use information.  Let me know if it looks like
> a bug in rtl-ssa rather than a bug in the vsetvl pass.

Hi, Richard.

As Li Pan said, the vsetvl pass is trying to remove regno = 8 in
old uses = [0,8,66,67,]

Here is the code:

access_array_builder uses_builder (attempt);
uses_builder.reserve (insn->num_uses () - 1);
for (use_info *use : insn->uses ())
if (use != find_access (insn->uses (), REGNO (avl)))
  uses_builder.quick_push (use);
use_array new_uses = use_array (uses_builder.finish ());
change.new_uses = new_uses;

You can see here, I add all uses except the use which has the same regno with
"avl". 

Am I using RTL_SSA in a wrong way ? Or I should remove RTL_SSA updating when
removing operands of uses ?

Looking forward your suggestions.
Thanks.

[Bug target/109535] internal compiler error: in finalize_new_accesses, at rtl-ssa/changes.cc:471

[rsandifo at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=109535

--- Comment #8 from rsandifo at gcc dot gnu.org <rsandifo at gcc dot gnu.org> ---
Can you quote a dump of the new insn pattern?  Or just:

  dump(change);

in function_info::finalize_new_accesses should do.

[Bug target/109535] internal compiler error: in finalize_new_accesses, at rtl-ssa/changes.cc:471

[juzhe.zhong at rivai dot ai]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=109535

--- Comment #9 from JuzheZhong <juzhe.zhong at rivai dot ai> ---
Confirm it is VSETVL pass bug in RISC-V port backend.
Fix patch here:
https://gcc.gnu.org/pipermail/gcc-patches/2023-April/615882.html
Hopefully, it will be merged soon.

[Bug target/109535] [13/14] internal compiler error: in finalize_new_accesses, at rtl-ssa/changes.cc:471

[kito at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=109535

Kito Cheng <kito at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
   Target Milestone|---                         |13.2
            Summary|internal compiler error: in |[13/14] internal compiler
                   |finalize_new_accesses, at   |error: in
                   |rtl-ssa/changes.cc:471      |finalize_new_accesses, at
                   |                            |rtl-ssa/changes.cc:471

[Bug target/109535] [13/14] internal compiler error: in finalize_new_accesses, at rtl-ssa/changes.cc:471

[cvs-commit at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=109535

--- Comment #10 from CVS Commits <cvs-commit at gcc dot gnu.org> ---
The master branch has been updated by Kito Cheng <kito@gcc.gnu.org>:

https://gcc.gnu.org/g:a2d12abedc89a9439fd6aadc38730fdadca0684f

commit r14-113-ga2d12abedc89a9439fd6aadc38730fdadca0684f
Author: Ju-Zhe Zhong <juzhe.zhong@rivai.ai>
Date:   Wed Apr 19 18:41:51 2023 +0800

    RISC-V: Fix wrong check of register occurrences [PR109535]

    count_occurrences will conly count same RTX (same code and same mode),
    but what we want to track is the occurrence of a register, a register
    might appeared in the insn with different mode or contain in SUBREG.

    Testcase coming from Kito.

    gcc/ChangeLog:

            PR target/109535
            * config/riscv/riscv-vsetvl.cc (count_regno_occurrences): New
function.
            (pass_vsetvl::cleanup_insns): Fix bug.

    gcc/testsuite/ChangeLog:

            PR target/109535
            * g++.target/riscv/rvv/base/pr109535.C: New test.
            * gcc.target/riscv/rvv/base/pr109535.c: New test.

    Signed-off-by: Ju-Zhe Zhong <juzhe.zhong@rivai.ai>
    Co-authored-by: kito-cheng <kito.cheng@sifive.com>

[Bug target/109535] [13/14] internal compiler error: in finalize_new_accesses, at rtl-ssa/changes.cc:471

[juzhe.zhong at rivai dot ai]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=109535

JuzheZhong <juzhe.zhong at rivai dot ai> changed:

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

--- Comment #11 from JuzheZhong <juzhe.zhong at rivai dot ai> ---
fixed

[Bug target/109535] [13/14] internal compiler error: in finalize_new_accesses, at rtl-ssa/changes.cc:471

[malat at debian dot org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=109535

--- Comment #12 from Mathieu Malaterre <malat at debian dot org> ---
@JuzheZhong 

Technically you are supposed to simply remove the keyword '14' from the title
and close when backported on 13...

[Bug target/109535] [13/14] internal compiler error: in finalize_new_accesses, at rtl-ssa/changes.cc:471

[juzhe.zhong at rivai dot ai]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=109535

JuzheZhong <juzhe.zhong at rivai dot ai> changed:

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

--- Comment #13 from JuzheZhong <juzhe.zhong at rivai dot ai> ---
reopen

[Bug target/109535] [13/14] internal compiler error: in finalize_new_accesses, at rtl-ssa/changes.cc:471

[juzhe.zhong at rivai dot ai]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=109535

--- Comment #14 from JuzheZhong <juzhe.zhong at rivai dot ai> ---
(In reply to Mathieu Malaterre from comment #12)
> @JuzheZhong 
> 
> Technically you are supposed to simply remove the keyword '14' from the
> title and close when backported on 13...

Thank you for correcting me. I reopen this bug and let wait for kito backport
to
GCC 13 to close this bug.

Thanks.

[Bug target/109535] [13 regression] internal compiler error: in finalize_new_accesses, at rtl-ssa/changes.cc:471

[cvs-commit at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=109535

--- Comment #15 from CVS Commits <cvs-commit at gcc dot gnu.org> ---
The releases/gcc-13 branch has been updated by Kito Cheng <kito@gcc.gnu.org>:

https://gcc.gnu.org/g:5821f378b58443e21c77c0f8c673067fc7655fcb

commit r13-7285-g5821f378b58443e21c77c0f8c673067fc7655fcb
Author: Ju-Zhe Zhong <juzhe.zhong@rivai.ai>
Date:   Wed Apr 19 18:41:51 2023 +0800

    RISC-V: Fix wrong check of register occurrences [PR109535]

    count_occurrences will conly count same RTX (same code and same mode),
    but what we want to track is the occurrence of a register, a register
    might appeared in the insn with different mode or contain in SUBREG.

    Testcase coming from Kito.

    gcc/ChangeLog:

            PR target/109535
            * config/riscv/riscv-vsetvl.cc (count_regno_occurrences): New
function.
            (pass_vsetvl::cleanup_insns): Fix bug.

    gcc/testsuite/ChangeLog:

            PR target/109535
            * g++.target/riscv/rvv/base/pr109535.C: New test.
            * gcc.target/riscv/rvv/base/pr109535.c: New test.

    Signed-off-by: Ju-Zhe Zhong <juzhe.zhong@rivai.ai>
    Co-authored-by: kito-cheng <kito.cheng@sifive.com>
    (cherry picked from commit a2d12abedc89a9439fd6aadc38730fdadca0684f)

[Bug target/109535] [13 regression] internal compiler error: in finalize_new_accesses, at rtl-ssa/changes.cc:471

[kito at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=109535

Kito Cheng <kito at gcc dot gnu.org> changed:

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

--- Comment #16 from Kito Cheng <kito at gcc dot gnu.org> ---
Fixed both on trunk and GCC 13 branch :)