[Bug c++/95229] New: ice in mark_jump_label_1

[Bug c++/95229] New: ice in mark_jump_label_1

[dcb314 at hotmail dot com]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=95229

            Bug ID: 95229
           Summary: ice in mark_jump_label_1
           Product: gcc
           Version: unknown
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: c++
          Assignee: unassigned at gcc dot gnu.org
          Reporter: dcb314 at hotmail dot com
  Target Milestone: ---

For this C++ code:

struct a {
  unsigned long b;
  unsigned long c;
};
class {
public:
  a d;
} e;
struct f {
  unsigned g;
  unsigned h;
  int i();
};
int f::i() {
  e.d.b += g;
  e.d.c += h;
}

compiled with recent gcc trunk and compiler flag -O3 -march=native,
I get

$ /home/dcb/gcc/results/bin/gcc -c -O3 -march=native bug609.cc
bug609.cc: In member function ‘int f::i()’:
bug609.cc:17:1: warning: no return statement in function returning non-void
[-Wreturn-type]
   17 | }
      | ^
during RTL pass: expand
bug609.cc:17:1: internal compiler error: Segmentation fault
0xf4bbb9 crash_signal(int)
        ../../trunk.git/gcc/toplev.c:328
0xd5878a mark_jump_label_1(rtx_def*, rtx_insn*, bool, bool)
        ../../trunk.git/gcc/jump.c:1087
0xd588b0 mark_jump_label_1(rtx_def*, rtx_insn*, bool, bool)
        ../../trunk.git/gcc/jump.c:1211
0xd588b0 mark_jump_label_1(rtx_def*, rtx_insn*, bool, bool)
        ../../trunk.git/gcc/jump.c:1211

The bug first seems to occur on today's gcc. Yesterdays seems fine:

/home/dcb/gcc/results.20200519/bin/gcc
/home/dcb/gcc/results.20200520/bin/gcc
during RTL pass: expand
...

Native is x86_64, AMD FX(tm)-8350, Piledriver, bdver2.

[Bug middle-end/95229] ice in mark_jump_label_1

[pinskia at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=95229

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
          Component|c++                         |middle-end

--- Comment #1 from Andrew Pinski <pinskia at gcc dot gnu.org> ---
>Native is x86_64, AMD FX(tm)-8350, Piledriver, bdver2.

Can you add -v and provide the full output as -march=native is not really
helpfull.

[Bug middle-end/95229] ice in mark_jump_label_1

[dcb314 at hotmail dot com]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=95229

--- Comment #2 from David Binderman <dcb314 at hotmail dot com> ---
Using built-in specs.
COLLECT_GCC=/home/dcb/gcc/results/bin/gcc
Target: x86_64-pc-linux-gnu
Configured with: ../trunk.git/configure --prefix=/home/dcb/gcc/results.20200520
--disable-bootstrap --disable-multilib --disable-werror
--enable-checking=df,extra,fold,rtl,yes --enable-languages=c,c++,fortran
Thread model: posix
Supported LTO compression algorithms: zlib zstd
gcc version 11.0.0 20200520 (experimental) (GCC) 
COLLECT_GCC_OPTIONS='-v' '-c' '-O3' '-w' '-march=native'
 /home/dcb/gcc/results.20200520/libexec/gcc/x86_64-pc-linux-gnu/11.0.0/cc1plus
-quiet -v -D_GNU_SOURCE /home/dcb/gcc/foundBugs/bug609.cc -march=bdver2 -mmmx
-mno-3dnow -msse -msse2 -msse3 -mssse3 -msse4a -mcx16 -msahf -mno-movbe
 -maes -mno-sha -mpclmul -mpopcnt -mabm -mno-lwp -mfma -mfma4 -mxop -mbmi
-mno-sgx -mno-bmi2 -mno-pconfig -mno-wbnoinvd -mtbm -mavx -mno-avx2 -msse4.2
-msse4.1 -mlzcnt -mno-rtm -mno-hle -mno-rdrnd -mf16c -mno-fsgsbase -mno-rdsee
d -mprfchw -mno-adx -mfxsr -mxsave -mno-xsaveopt -mno-avx512f -mno-avx512er
-mno-avx512cd -mno-avx512pf -mno-prefetchwt1 -mno-clflushopt -mno-xsavec
-mno-xsaves -mno-avx512dq -mno-avx512bw -mno-avx512vl -mno-avx512ifma -mno-avx5
12vbmi -mno-avx5124fmaps -mno-avx5124vnniw -mno-clwb -mno-mwaitx -mno-clzero
-mno-pku -mno-rdpid -mno-gfni -mno-shstk -mno-avx512vbmi2 -mno-avx512vnni
-mno-vaes -mno-vpclmulqdq -mno-avx512bitalg -mno-movdiri -mno-movdir64b -mno-
waitpkg -mno-cldemote -mno-ptwrite -mno-avx512bf16 -mno-enqcmd
-mno-avx512vp2intersect -mno-serialize -mno-tsxldtrk --param l1-cache-size=16
--param l1-cache-line-size=64 --param l2-cache-size=2048 -mtune=bdver2 -quiet
-dumpbase
 bug609.cc -auxbase bug609 -O3 -w -version -o /tmp/ccmaYNdH.s
GNU C++14 (GCC) version 11.0.0 20200520 (experimental) (x86_64-pc-linux-gnu)
        compiled by GNU C version Clang 10.0.0 (Fedora 10.0.0-1.fc32), GMP
version 6.1.2, MPFR version 4.0.2-p7, M
PC version 1.1.0, isl version isl-0.16.1-GMP

GGC heuristics: --param ggc-min-expand=30 --param ggc-min-heapsize=4096
ignoring nonexistent directory
"/home/dcb/gcc/results.20200520/lib/gcc/x86_64-pc-linux-gnu/11.0.0/../../../../x86_64-pc-linux-gnu/include"
#include "..." search starts here:
#include <...> search starts here:

/home/dcb/gcc/results.20200520/lib/gcc/x86_64-pc-linux-gnu/11.0.0/../../../../include/c++/11.0.0

/home/dcb/gcc/results.20200520/lib/gcc/x86_64-pc-linux-gnu/11.0.0/../../../../include/c++/11.0.0/x86_64-pc-linux-
gnu

/home/dcb/gcc/results.20200520/lib/gcc/x86_64-pc-linux-gnu/11.0.0/../../../../include/c++/11.0.0/backward
 /home/dcb/gcc/results.20200520/lib/gcc/x86_64-pc-linux-gnu/11.0.0/include
 /usr/local/include
 /home/dcb/gcc/results.20200520/include

/home/dcb/gcc/results.20200520/lib/gcc/x86_64-pc-linux-gnu/11.0.0/include-fixed
 /usr/include
End of search list.
GNU C++14 (GCC) version 11.0.0 20200520 (experimental) (x86_64-pc-linux-gnu)
        compiled by GNU C version Clang 10.0.0 (Fedora 10.0.0-1.fc32), GMP
version 6.1.2, MPFR version 4.0.2-p7, M
PC version 1.1.0, isl version isl-0.16.1-GMP
GGC heuristics: --param ggc-min-expand=30 --param ggc-min-heapsize=4096
Compiler executable checksum: 37a65bc431cc6a404e564314c7e7906a
during RTL pass: expand
LzmaBench.cpp: In member function ‘LONG CEncoderInfo::Decode(UInt32)’:
LzmaBench.cpp:651:1: internal compiler error: Segmentation fault
0xf4bbb9 crash_signal(int)
        ../../trunk.git/gcc/toplev.c:328
0xd5878a mark_jump_label_1(rtx_def*, rtx_insn*, bool, bool)
        ../../trunk.git/gcc/jump.c:1087
0xd588b0 mark_jump_label_1(rtx_def*, rtx_insn*, bool, bool)
        ../../trunk.git/gcc/jump.c:1211
0xd588b0 mark_jump_label_1(rtx_def*, rtx_insn*, bool, bool)
        ../../trunk.git/gcc/jump.c:1211

[Bug middle-end/95229] ice in mark_jump_label_1

[marxin at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=95229

--- Comment #3 from Martin Liška <marxin at gcc dot gnu.org> ---
Reproduced with -march=znver1. It breaks quite some SPEC benchmarks.

[Bug middle-end/95229] [11 Regression] in mark_jump_label_1

[marxin at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=95229

Martin Liška <marxin at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
     Ever confirmed|0                           |1
      Known to fail|                            |11.0
           Keywords|                            |needs-bisection
            Summary|ice in mark_jump_label_1    |[11 Regression] in
                   |                            |mark_jump_label_1
   Last reconfirmed|                            |2020-05-20
   Target Milestone|---                         |11.0
             Status|UNCONFIRMED                 |NEW
             Blocks|                            |26163
      Known to work|                            |10.1.0
            Version|unknown                     |11.0
                 CC|                            |marxin at gcc dot gnu.org


Referenced Bugs:

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=26163
[Bug 26163] [meta-bug] missed optimization in SPEC (2k17, 2k and 2k6 and 95)

[Bug target/95229] [11 Regression] in mark_jump_label_1

[rguenth at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=95229

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Target|                            |x86_64-*-* i?86-*-*
                 CC|                            |uros at gcc dot gnu.org
          Component|middle-end                  |target

--- Comment #4 from Richard Biener <rguenth at gcc dot gnu.org> ---
Program received signal SIGSEGV, Segmentation fault.
0x00000000013747bd in mark_jump_label_1 (x=0x0, insn=0x7ffff6ac8b80, 
    in_mem=false, is_target=false) at ../../src/trunk/gcc/jump.c:1087
1087      RTX_CODE code = GET_CODE (x);
(gdb) p debug_rtx (insn)
(insn 9 8 10 2 (set (reg:V2DI 91 [ vect__3.9 ])
        (zero_extend:V2DI (vec_select:V2SI (nil)
                (parallel [
                        (const_int 0 [0])
                        (const_int 1 [0x1])
                    ])))) "z.c":15:12 -1
     (nil))


the vec_select has a NULL operand somehow.  We're expanding from

  vect__3.9_16 = (vector(2) long unsigned int) vect__2.8_15;

so likely Uros change or it triggers a bogus expand helper or forwprop
does not properly constrain itself.  I'll debug a bit.

[Bug target/95229] [11 Regression] in mark_jump_label_1

[rguenth at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=95229

--- Comment #5 from Richard Biener <rguenth at gcc dot gnu.org> ---
Built by

#5  0x00000000020c4504 in gen_rtx_fmt_ee_stat (code=VEC_SELECT, 
    mode=E_V2SImode, arg0=0x0, arg1=0x7ffff6ad92e0) at ./genrtl.h:49
#6  0x000000000211e242 in gen_sse4_1_zero_extendv2siv2di2 (
    operand0=0x7ffff6ada4f8, operand1=0x0) at insn-emit.c:40857
#7  0x0000000002207aa7 in gen_zero_extendv2siv2di2 (operand0=0x7ffff6ada4f8, 
    operand1=0x7ffff6ada4c8) at ../../src/trunk/gcc/config/i386/sse.md:18012
#8  0x00000000010cfe31 in insn_gen_fn::operator() (
    this=0x3230928 <insn_data+325096>, a0=0x7ffff6ada4f8, a1=0x7ffff6ada4c8)
    at ../../src/trunk/gcc/recog.h:317
#9  0x00000000014aed21 in maybe_gen_insn (icode=CODE_FOR_zero_extendv2siv2di2, 
    nops=2, ops=0x7fffffffa8d0) at ../../src/trunk/gcc/optabs.c:7444
#10 0x00000000014a3c7d in maybe_emit_unop_insn (
    icode=CODE_FOR_zero_extendv2siv2di2, target=0x7ffff6ada4f8, 
--Type <RET> for more, q to quit, c to continue without paging--
    op0=0x7ffff6ada4c8, code=ZERO_EXTEND) at ../../src/trunk/gcc/optabs.c:3597
#11 0x00000000014a3d7f in emit_unop_insn (icode=CODE_FOR_zero_extendv2siv2di2, 
    target=0x7ffff6ada4f8, op0=0x7ffff6ada4c8, code=ZERO_EXTEND)
    at ../../src/trunk/gcc/optabs.c:3621
#12 0x0000000001098d45 in convert_move (to=0x7ffff6ada4f8, 
    from=0x7ffff6ada4c8, unsignedp=1) at ../../src/trunk/gcc/expr.c:260
#13 0x000000000109bc1e in convert_modes (mode=E_V2DImode, oldmode=E_V2SImode, 
    x=0x7ffff6ada4c8, unsignedp=1) at ../../src/trunk/gcc/expr.c:737
#14 0x000000000109b220 in convert_to_mode (mode=E_V2DImode, x=0x7ffff6ada4c8, 
    unsignedp=1) at ../../src/trunk/gcc/expr.c:662
#15 0x00000000010bae75 in expand_expr_real_2 (ops=0x7fffffffb600, target=0x0, 
    tmode=E_VOIDmode, modifier=EXPAND_NORMAL)
    at ../../src/trunk/gcc/expr.c:8665

frame #7 looks OK while #6 is bogus.  Context:

18009     if (!MEM_P (operands[1]))
18010       {
18011         operands[1] = simplify_subreg (V4SImode, operands[1], V2SImode,
0);
18012         emit_insn (gen_sse4_1_<code>v2siv2di2 (operands[0],
operands[1]));
18013         DONE;

simplify_subreg returns NULL.  It is originally (reg:V2SI 90 [ vect__2.8 ]).
Should that be simplify_gen_subreg?

[Bug target/95229] [11 Regression] in mark_jump_label_1

[rguenth at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=95229

--- Comment #6 from Richard Biener <rguenth at gcc dot gnu.org> ---
(In reply to Richard Biener from comment #5)
> Built by
> 
> #5  0x00000000020c4504 in gen_rtx_fmt_ee_stat (code=VEC_SELECT, 
>     mode=E_V2SImode, arg0=0x0, arg1=0x7ffff6ad92e0) at ./genrtl.h:49
> #6  0x000000000211e242 in gen_sse4_1_zero_extendv2siv2di2 (
>     operand0=0x7ffff6ada4f8, operand1=0x0) at insn-emit.c:40857
> #7  0x0000000002207aa7 in gen_zero_extendv2siv2di2 (operand0=0x7ffff6ada4f8, 
>     operand1=0x7ffff6ada4c8) at ../../src/trunk/gcc/config/i386/sse.md:18012
> #8  0x00000000010cfe31 in insn_gen_fn::operator() (
>     this=0x3230928 <insn_data+325096>, a0=0x7ffff6ada4f8, a1=0x7ffff6ada4c8)
>     at ../../src/trunk/gcc/recog.h:317
> #9  0x00000000014aed21 in maybe_gen_insn
> (icode=CODE_FOR_zero_extendv2siv2di2, 
>     nops=2, ops=0x7fffffffa8d0) at ../../src/trunk/gcc/optabs.c:7444
> #10 0x00000000014a3c7d in maybe_emit_unop_insn (
>     icode=CODE_FOR_zero_extendv2siv2di2, target=0x7ffff6ada4f8, 
> --Type <RET> for more, q to quit, c to continue without paging--
>     op0=0x7ffff6ada4c8, code=ZERO_EXTEND) at
> ../../src/trunk/gcc/optabs.c:3597
> #11 0x00000000014a3d7f in emit_unop_insn
> (icode=CODE_FOR_zero_extendv2siv2di2, 
>     target=0x7ffff6ada4f8, op0=0x7ffff6ada4c8, code=ZERO_EXTEND)
>     at ../../src/trunk/gcc/optabs.c:3621
> #12 0x0000000001098d45 in convert_move (to=0x7ffff6ada4f8, 
>     from=0x7ffff6ada4c8, unsignedp=1) at ../../src/trunk/gcc/expr.c:260
> #13 0x000000000109bc1e in convert_modes (mode=E_V2DImode,
> oldmode=E_V2SImode, 
>     x=0x7ffff6ada4c8, unsignedp=1) at ../../src/trunk/gcc/expr.c:737
> #14 0x000000000109b220 in convert_to_mode (mode=E_V2DImode,
> x=0x7ffff6ada4c8, 
>     unsignedp=1) at ../../src/trunk/gcc/expr.c:662
> #15 0x00000000010bae75 in expand_expr_real_2 (ops=0x7fffffffb600,
> target=0x0, 
>     tmode=E_VOIDmode, modifier=EXPAND_NORMAL)
>     at ../../src/trunk/gcc/expr.c:8665
> 
> frame #7 looks OK while #6 is bogus.  Context:
> 
> 18009     if (!MEM_P (operands[1]))
> 18010       {
> 18011         operands[1] = simplify_subreg (V4SImode, operands[1],
> V2SImode, 0);
> 18012         emit_insn (gen_sse4_1_<code>v2siv2di2 (operands[0],
> operands[1]));
> 18013         DONE;
> 
> simplify_subreg returns NULL.  It is originally (reg:V2SI 90 [ vect__2.8 ]).
> Should that be simplify_gen_subreg?

That fixes the testcase.  But simplify_subreg is used in a lot more places
so leaving to Uros to match up with expectations.

[Bug target/95229] [11 Regression] in mark_jump_label_1

[ubizjak at gmail dot com]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=95229

--- Comment #7 from Uroš Bizjak <ubizjak at gmail dot com> ---
(In reply to Richard Biener from comment #6)

> That fixes the testcase.  But simplify_subreg is used in a lot more places
> so leaving to Uros to match up with expectations.

Oh, yes... We don't have hard regs here, so all should be changed to
simplify_gen_subregs. I have a patch.

[Bug target/95229] [11 Regression] in mark_jump_label_1

[cvs-commit at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=95229

--- Comment #8 from CVS 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:3872a519c8fa65318efa1b481d331ef91b3ff044

commit r11-528-g3872a519c8fa65318efa1b481d331ef91b3ff044
Author: Uros Bizjak <ubizjak@gmail.com>
Date:   Wed May 20 19:00:39 2020 +0200

    i386: Fix zero/sign extend expanders [PR95229]

    2020-05-20  Uroš Bizjak  <ubizjak@gmail.com>

    gcc/ChangeLog:
            PR target/95229
            * config/i386/sse.md (<code>v8qiv8hi2): Use
            simplify_gen_subreg instead of simplify_subreg.
            (<code>v8qiv8si2): Ditto.
            (<code>v4qiv4si2): Ditto.
            (<code>v4hiv4si2): Ditto.
            (<code>v8qiv8di2): Ditto.
            (<code>v4qiv4di2): Ditto.
            (<code>v2qiv2di2): Ditto.
            (<code>v4hiv4di2): Ditto.
            (<code>v2hiv2di2): Ditto.
            (<code>v2siv2di2): Ditto.

    gcc/testsuite/ChangeLog:
            PR target/95229
            * g++.target/i386/pr95229.C: New test.

[Bug target/95229] [11 Regression] in mark_jump_label_1

[cvs-commit at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=95229

--- Comment #9 from CVS Commits <cvs-commit at gcc dot gnu.org> ---
The master branch has been updated by Martin Liska <marxin@gcc.gnu.org>:

https://gcc.gnu.org/g:2f20c7040f884b542cadf46fd77188729ab3cb17

commit r11-548-g2f20c7040f884b542cadf46fd77188729ab3cb17
Author: Martin Liska <mliska@suse.cz>
Date:   Thu May 21 14:21:31 2020 +0200

    Fix a test-case warning.

            PR target/95229
            * g++.target/i386/pr95229.C: Fix
            error: unnamed type with no linkage used
            to declare variable â<unnamed class> eâ with linkage
            with -std=gnu++98.

[Bug target/95229] [11 Regression] in mark_jump_label_1

[ubizjak at gmail dot com]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=95229

Uroš Bizjak <ubizjak at gmail dot com> changed:

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

--- Comment #10 from Uroš Bizjak <ubizjak at gmail dot com> ---
Fixed. (thanks Martin for adjusting the testcase!).