[Bug target/100241] New: internal compiler error: in curr_insn_transform, at lra-constraints.c:4133

[Bug target/100241] New: internal compiler error: in curr_insn_transform, at lra-constraints.c:4133

[mpolacek at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=100241

            Bug ID: 100241
           Summary: internal compiler error: in curr_insn_transform, at
                    lra-constraints.c:4133
           Product: gcc
           Version: 11.0
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: target
          Assignee: unassigned at gcc dot gnu.org
          Reporter: mpolacek at gcc dot gnu.org
  Target Milestone: ---

Currently I only have one test which contains UB:

# cat vp9_blockd.c
extern char num_4x4_blocks_high_lookup, num_4x4_blocks_wide_lookup;
typedef struct {
  char tx_size;
} MODE_INFO;
struct macroblockd_plane {
  int subsampling_y;
};
typedef struct {
  struct macroblockd_plane plane[3];
  MODE_INFO *mi;
  int mb_to_right_edge;
} MACROBLOCKD;
typedef void foreach_transformed_block_visitor();
int vp9_foreach_transformed_block_in_plane_i;
foreach_transformed_block_visitor vp9_foreach_transformed_block_in_plane_visit;

#pragma GCC visibility push(internal)
void vp9_foreach_transformed_block_in_plane(MACROBLOCKD *xd) {
  struct macroblockd_plane pd = xd->plane[0];
  MODE_INFO mi = xd->mi[0];
  char tx_size = mi.tx_size, plane_bsize = pd.subsampling_y;
  int num_4x4_w = num_4x4_blocks_wide_lookup,
      num_4x4_h = num_4x4_blocks_high_lookup, r, c,
      max_blocks_wide = num_4x4_w + xd->mb_to_right_edge,
      max_blocks_high = num_4x4_h, extra_step = max_blocks_wide >> 1;
  for (r = 0; r < max_blocks_high; r += tx_size) {
    for (c = 0; c < max_blocks_wide; c += 1 << tx_size)
      vp9_foreach_transformed_block_in_plane_visit(plane_bsize);
    vp9_foreach_transformed_block_in_plane_i += extra_step;
  }
}
void vp9_encode_sby_pass1(MACROBLOCKD *x) {
vp9_foreach_transformed_block_in_plane(x); }
#pragma GCC visibility pop

struct vpx_codec_iface {
  int i;
  int get_glob_hdrs;
  void (*fp)();
};
void vp9_first_pass ();
void vp9_get_compressed_data() { vp9_first_pass(); }
void encoder_encode() {
  vp9_get_compressed_data();
}

struct vpx_codec_iface vpx_codec_vp9_cx_algo = {1, 0, encoder_encode };

void first_pass_worker_hook()
{
  vp9_first_pass_encode_tile_mb_row();
}

# cat vp9_firstpass.c
typedef long int64_t;
void vp9_encode_sby_pass1();
typedef struct {
  int64_t coded_error;
  int64_t sr_coded_error;
  int64_t frame_noise_energy;
  int64_t intra_error;
} FIRSTPASS_DATA;
typedef struct {
  FIRSTPASS_DATA fp_data;
} TileDataEnc;
void vp9_first_pass_encode_tile_mb_row(int td, FIRSTPASS_DATA *fp_acc_data,
                                       TileDataEnc *tile_data) {
  vp9_encode_sby_pass1(td);
  FIRSTPASS_DATA __trans_tmp_1 = *fp_acc_data;
  TileDataEnc *this_tile = tile_data;
  this_tile->fp_data.coded_error += this_tile->fp_data.sr_coded_error +=
      __trans_tmp_1.sr_coded_error;
  this_tile->fp_data.frame_noise_energy += __trans_tmp_1.frame_noise_energy;
  this_tile->fp_data.intra_error += __trans_tmp_1.intra_error;
}
void launch_enc_workers();
void first_pass_worker_hook();
void vp9_encode_fp_row_mt() { launch_enc_workers(first_pass_worker_hook); }
void vp9_first_pass() { vp9_encode_fp_row_mt(); }

# gcc -flto=auto -ffat-lto-objects -march=armv8-a -fPIC -O3 -Wall -W
vp9_blockd.c vp9_firstpass.c -shared
vp9_blockd.c: In function ‘first_pass_worker_hook’:
vp9_blockd.c:50:3: warning: implicit declaration of function
‘vp9_first_pass_encode_tile_mb_row’ [-Wimplicit-function-declaration]
   50 |   vp9_first_pass_encode_tile_mb_row();
      |   ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
vp9_firstpass.c: In function ‘vp9_first_pass_encode_tile_mb_row’:
vp9_firstpass.c:21:1: error: unable to generate reloads for:
   21 | }
      | ^
(insn 65 68 69 8 (set (reg:V2DI 138)
        (vec_concat:V2DI (mem:DI (plus:DI (reg/v/f:DI 153 [orig:122 fp_acc_data
] [122])
                    (const_int 16 [0x10])) [4
fp_acc_data_12(D)->frame_noise_energy+0 S8 A64])
            (mem:DI (plus:DI (reg/v/f:DI 153 [orig:122 fp_acc_data ] [122])
                    (const_int 24 [0x18])) [4 fp_acc_data_12(D)->intra_error+0
S8 A64]))) "vp9_firstpass.c":17:34 2473 {load_pair_lanesdi}
     (expr_list:REG_DEAD (reg/v/f:DI 122 [ fp_acc_data ])
        (nil)))
during RTL pass: reload
vp9_firstpass.c:21:1: internal compiler error: in curr_insn_transform, at
lra-constraints.c:4133
0x14881df diagnostic_impl(rich_location*, diagnostic_metadata const*, int, char
const*, std::__va_list*, diagnostic_t)
        ???:0
0xc8b7db internal_error(char const*, ...)
        ???:0
0xc8b8cf fancy_abort(char const*, int, char const*)
        ???:0
0x783fab _fatal_insn(char const*, rtx_def const*, char const*, int, char
const*)
        ???:0
0xe152a7 curr_insn_transform(bool)
        ???:0
0xe102d7 lra_constraints(bool)
        ???:0
0x127407b lra(_IO_FILE*)
        ???:0
0x1267e3f (anonymous namespace)::pass_reload::execute(function*)
        ???:0
Please submit a full bug report,
with preprocessed source if appropriate.
Please include the complete backtrace with any bug report.
See <http://bugzilla.redhat.com/bugzilla> for instructions.
make: *** [/tmp/cc1Xqmvm.mk:2: /tmp/ccQcOjSl.ltrans0.ltrans.o] Error 1

[Bug target/100241] internal compiler error: in curr_insn_transform, at lra-constraints.c:4133

[mpolacek at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=100241

Marek Polacek <mpolacek at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
           Keywords|                            |needs-bisection

--- Comment #1 from Marek Polacek <mpolacek at gcc dot gnu.org> ---
I'm reducing this again, the original code was valid I think.

[Bug target/100241] internal compiler error: in curr_insn_transform, at lra-constraints.c:4133

[mpolacek at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=100241

--- Comment #2 from Marek Polacek <mpolacek at gcc dot gnu.org> ---
This is on aarch64:

Using built-in specs.
COLLECT_GCC=gcc
COLLECT_LTO_WRAPPER=/usr/libexec/gcc/aarch64-redhat-linux/11/lto-wrapper
Target: aarch64-redhat-linux
Configured with: ../configure --enable-bootstrap
--enable-languages=c,c++,fortran,lto --prefix=/usr --mandir=/usr/share/man
--infodir=/usr/share/info --with-bugurl=http://bugzilla.redhat.com/bugzilla
--enable-shared --enable-threads=posix --enable-checking=release
--enable-multilib --with-system-zlib --enable-__cxa_atexit
--disable-libunwind-exceptions --enable-gnu-unique-object
--enable-linker-build-id --with-gcc-major-version-only
--with-linker-hash-style=gnu --enable-plugin --enable-initfini-array
--without-isl --enable-gnu-indirect-function --build=aarch64-redhat-linux
Thread model: posix
Supported LTO compression algorithms: zlib zstd
gcc version 11.0.1 20210324 (Red Hat 11.0.1-0) (GCC)

[Bug target/100241] internal compiler error: in curr_insn_transform, at lra-constraints.c:4133

[mpolacek at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=100241

--- Comment #3 from Marek Polacek <mpolacek at gcc dot gnu.org> ---
Now reproduced with a single file.  Reducing that...

[Bug target/100241] internal compiler error: in curr_insn_transform, at lra-constraints.c:4133

[mpolacek at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=100241

--- Comment #4 from Marek Polacek <mpolacek at gcc dot gnu.org> ---
Single, hopefully valid TU, no LTO needed:

$ gcc -march=armv8-a -fPIC -O3 100241.c -fvisibility=internal -c
100241.c: In function ‘vp9_first_pass_encode_tile_mb_row’:
100241.c:32:1: error: unable to generate reloads for:
   32 | }
      | ^
(insn 74 89 90 14 (set (reg:V2DI 136)
        (vec_concat:V2DI (mem:DI (plus:DI (reg/v/f:DI 178 [orig:113 fp_acc_data
] [113])
                    (const_int 16 [0x10])) [3
fp_acc_data_15(D)->frame_noise_energy+0 S8 A64])
            (mem:DI (plus:DI (reg/v/f:DI 178 [orig:113 fp_acc_data ] [113])
                    (const_int 24 [0x18])) [3 fp_acc_data_15(D)->intra_error+0
S8 A64]))) "100241.c":23:63 2473 {load_pair_lanesdi}
     (nil))
during RTL pass: reload
100241.c:32:1: internal compiler error: in curr_insn_transform, at
lra-constraints.c:4133
0x152862f diagnostic_impl(rich_location*, diagnostic_metadata const*, int, char
const*, std::__va_list*, diagnostic_t)
        ???:0
0xce52ff internal_error(char const*, ...)
        ???:0
0xce53f3 fancy_abort(char const*, int, char const*)
        ???:0
0x7db1d7 _fatal_insn(char const*, rtx_def const*, char const*, int, char
const*)
        ???:0
0xea8d57 curr_insn_transform(bool)
        ???:0
0xea3d87 lra_constraints(bool)
        ???:0
0x131523b lra(_IO_FILE*)
        ???:0
0x1308fff (anonymous namespace)::pass_reload::execute(function*)
        ???:0



extern char num_4x4_blocks_high_lookup, num_4x4_blocks_wide_lookup,
    vp9_foreach_transformed_block_in_plane_xd_0_0_0;
int vp9_first_pass_encode_tile_mb_row_cpi_0,
    vp9_first_pass_encode_tile_mb_row_tile_data_0_0, vp9_encode_sby_pass1_x;
typedef void foreach_transformed_block_visitor();
void vp9_encode_sby_pass1();
typedef struct {
  long coded_error;
  long sr_coded_error;
  long frame_noise_energy;
  long intra_error;
} FIRSTPASS_DATA;
typedef struct {
  FIRSTPASS_DATA fp_data;
} TileDataEnc;
TileDataEnc accumulate_fp_mb_row_stat_this_tile;
void vp9_first_pass_encode_tile_mb_row(FIRSTPASS_DATA *fp_acc_data) {
  int mb_col_end = vp9_first_pass_encode_tile_mb_row_tile_data_0_0;
  for (; mb_col_end;) {
    vp9_encode_sby_pass1();
    if (vp9_first_pass_encode_tile_mb_row_cpi_0) {
      FIRSTPASS_DATA __trans_tmp_1 = *fp_acc_data;
      accumulate_fp_mb_row_stat_this_tile.fp_data.coded_error +=
          accumulate_fp_mb_row_stat_this_tile.fp_data.sr_coded_error +=
          __trans_tmp_1.sr_coded_error;
      accumulate_fp_mb_row_stat_this_tile.fp_data.frame_noise_energy +=
          __trans_tmp_1.frame_noise_energy;
      accumulate_fp_mb_row_stat_this_tile.fp_data.intra_error +=
          __trans_tmp_1.intra_error;
    }
  }
}
foreach_transformed_block_visitor vp9_foreach_transformed_block_in_plane_visit;
void vp9_foreach_transformed_block_in_plane(void *arg) {
  int c, max_blocks_wide = num_4x4_blocks_wide_lookup;
  for (; num_4x4_blocks_high_lookup;) {
    c = 0;
    for (; c < max_blocks_wide;
         c += vp9_foreach_transformed_block_in_plane_xd_0_0_0)
      vp9_foreach_transformed_block_in_plane_visit(arg);
  }
}
void vp9_encode_sby_pass1() {
  vp9_foreach_transformed_block_in_plane(&vp9_encode_sby_pass1_x);
}

[Bug target/100241] internal compiler error: in curr_insn_transform, at lra-constraints.c:4133

[mpolacek at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=100241

--- Comment #5 from Marek Polacek <mpolacek at gcc dot gnu.org> ---
I can't reproduce this with latest trunk :-(.  Bisecting...

[Bug target/100241] internal compiler error: in curr_insn_transform, at lra-constraints.c:4133

[mpolacek at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=100241

--- Comment #6 from Marek Polacek <mpolacek at gcc dot gnu.org> ---
Fixed by r11-7864.  Not sure if we want this test.  It doesn't contain __fp16
like the test in r11-7864.

[Bug target/100241] internal compiler error: in curr_insn_transform, at lra-constraints.c:4133

[pinskia at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=100241

--- Comment #7 from Andrew Pinski <pinskia at gcc dot gnu.org> ---
(In reply to Marek Polacek from comment #6)
> Fixed by r11-7864.  Not sure if we want this test.  It doesn't contain
> __fp16 like the test in r11-7864.

PR 99787 also has a testcase that could be done too.

I think we should add a testcase for both bugs.

[Bug target/100241] internal compiler error: in curr_insn_transform, at lra-constraints.c:4133

[pinskia at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=100241

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
           See Also|                            |https://gcc.gnu.org/bugzill
                   |                            |a/show_bug.cgi?id=99787
             Status|UNCONFIRMED                 |ASSIGNED
     Ever confirmed|0                           |1
           Assignee|unassigned at gcc dot gnu.org      |pinskia at gcc dot gnu.org
   Last reconfirmed|                            |2022-01-01

--- Comment #8 from Andrew Pinski <pinskia at gcc dot gnu.org> ---
Let me add the testcase for this one and PR 99787

[Bug target/100241] internal compiler error: in curr_insn_transform, at lra-constraints.c:4133

[cvs-commit at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=100241

--- Comment #9 from CVS Commits <cvs-commit at gcc dot gnu.org> ---
The trunk branch has been updated by Andrew Pinski <pinskia@gcc.gnu.org>:

https://gcc.gnu.org/g:5fa4f982636e7e66eee6a9b45cc0939ae95b4659

commit r12-6164-g5fa4f982636e7e66eee6a9b45cc0939ae95b4659
Author: Andrew Pinski <apinski@marvell.com>
Date:   Sat Jan 1 08:44:48 2022 +0000

    Committed: Add testcases for a few PRs

    These were fixed as part of the fix for PR 99766,
    I thought it would be useful to add a few testcases
    for the other cases that were failing.

    Committed as obvious after running the tests to make
    sure they work.

            PR rtl-optimization/100241
            PR rtl-optimization/99787

    gcc/testsuite/ChangeLog:

            * gcc.c-torture/compile/pr100241-1.c: New test.
            * gcc.c-torture/compile/pr99787-1.c: New test.

[Bug target/100241] internal compiler error: in curr_insn_transform, at lra-constraints.c:4133

[pinskia at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=100241

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

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

--- Comment #10 from Andrew Pinski <pinskia at gcc dot gnu.org> ---
testcases added so marking this as a dup of bug 99766.

*** This bug has been marked as a duplicate of bug 99766 ***