[Bug target/97701] New: [10/11 Regression] aarch64: ICE in extract_constrain_insn since r10-4447-g095f78c6

[Bug target/97701] New: [10/11 Regression] aarch64: ICE in extract_constrain_insn since r10-4447-g095f78c6

[acoplan at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=97701

            Bug ID: 97701
           Summary: [10/11 Regression] aarch64: ICE in
                    extract_constrain_insn since r10-4447-g095f78c6
           Product: gcc
           Version: 11.0
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: target
          Assignee: unassigned at gcc dot gnu.org
          Reporter: acoplan at gcc dot gnu.org
  Target Milestone: ---

For the following testcase:

extern char a[][12][18][17][17];
extern short b[][12][18][17][17];
extern int c[][2][8][7];
short *d;
void e(signed f, int g, char h, char i, char j) {
  for (int k = 648; k; k += f)
    for (short l; l < j; l += 9)
      for (long m = f + 6LL; m < (h ? h : i); m += 2)
        for (int n = 0; n < 16; n += 3LL) {
          for (int o = g; o; o++)
            a[k][l][m][n][o] = b[k][l][m][n][o] = d[k] ? 2 : 0;
          c[k][l][m][0] = 0;
        }
}

AArch64 GCC ICEs at -O3 since
r10-4447-g095f78c62157124ad479a3f98b6995ced090b807. To reproduce:

$ aarch64-elf-gcc -c test.c -O3
test.c: In function 'e':
test.c:14:1: error: insn does not satisfy its constraints:
   14 | }
      | ^
(insn 2496 2495 2497 5 (set (reg:DI 32 v0 [1415])
        (lo_sum:DI (reg:DI 6 x6 [1416])
            (symbol_ref/u:DI ("*.LC0") [flags 0x82]))) 1007 {add_losym_di}
     (nil))
during RTL pass: reload
test.c:14:1: internal compiler error: in extract_constrain_insn, at
recog.c:2212
0xd1ab2f _fatal_insn(char const*, rtx_def const*, char const*, int, char
const*)
        /home/alecop01/toolchain/src/gcc/gcc/rtl-error.c:108
0xd1ab60 _fatal_insn_not_found(rtx_def const*, char const*, int, char const*)
        /home/alecop01/toolchain/src/gcc/gcc/rtl-error.c:118
0xcec957 extract_constrain_insn(rtx_insn*)
        /home/alecop01/toolchain/src/gcc/gcc/recog.c:2212
0xbbe58d check_rtl
        /home/alecop01/toolchain/src/gcc/gcc/lra.c:2038
0xbc3451 lra(_IO_FILE*)
        /home/alecop01/toolchain/src/gcc/gcc/lra.c:2447
0xb76cff do_reload
        /home/alecop01/toolchain/src/gcc/gcc/ira.c:5725
0xb76cff execute
        /home/alecop01/toolchain/src/gcc/gcc/ira.c:5911
Please submit a full bug report,
with preprocessed source if appropriate.
Please include the complete backtrace with any bug report.
See <https://gcc.gnu.org/bugs/> for instructions.

[Bug target/97701] [10/11 Regression] aarch64: ICE in extract_constrain_insn since r10-4447-g095f78c6

[acoplan at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=97701

Alex Coplan <acoplan at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
   Target Milestone|---                         |10.3
      Known to fail|                            |11.0
             Target|                            |aarch64

[Bug target/97701] [10/11 Regression] aarch64: ICE in extract_constrain_insn since r10-4447-g095f78c6

[rguenth at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=97701

--- Comment #1 from Richard Biener <rguenth at gcc dot gnu.org> ---
Must be a latent problem then I guess.

[Bug target/97701] [10/11 Regression] aarch64: ICE in extract_constrain_insn since r10-4447-g095f78c6

[jakub at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=97701

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

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

--- Comment #2 from Jakub Jelinek <jakub at gcc dot gnu.org> ---
Can reproduce just on the 10 branch (even from today), doesn't reproduce on
trunk for me.

[Bug target/97701] [10/11 Regression] aarch64: ICE in extract_constrain_insn since r10-4447-g095f78c6

[acoplan at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=97701

--- Comment #3 from Alex Coplan <acoplan at gcc dot gnu.org> ---
A bisect in the other direction shows that Jakub's r11-4717-g3e190757f:

commit 3e190757fa332d327bee27495f37beb01155cfab
Author: Jakub Jelinek <jakub@redhat.com>
Date:   Wed Nov 4 10:55:29 2020

    phiopt: Optimize x ? 1024 : 0 to (int) x << 10 [PR97690]

fixes (or perhaps just obscures) this on the trunk.

[Bug target/97701] [10/11 Regression] aarch64: ICE in extract_constrain_insn since r10-4447-g095f78c6

[acoplan at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=97701

--- Comment #4 from Alex Coplan <acoplan at gcc dot gnu.org> ---
Indeed, you can recover the ICE on trunk by adding -fno-ssa-phiopt.

[Bug target/97701] [10/11 Regression] aarch64: ICE in extract_constrain_insn since r10-4447-g095f78c6

[acoplan at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=97701

--- Comment #5 from Alex Coplan <acoplan at gcc dot gnu.org> ---
Perhaps more conveniently, simply changing the 0 to a 1 on the RHS of the
ternary operator, the following testcase still ICEs at -O3 on trunk:

extern char a[][12][18][17][17];
extern short b[][12][18][17][17];
extern int c[][2][8][7];
short *d;
void e(signed f, int g, char h, char i, char j) {
  for (int k = 648; k; k += f)
    for (short l; l < j; l += 9)
      for (long m = f + 6LL; m < (h ? h : i); m += 2)
        for (int n = 0; n < 16; n += 3LL) {
          for (int o = g; o; o++)
            a[k][l][m][n][o] = b[k][l][m][n][o] = d[k] ? 2 : 1;
          c[k][l][m][0] = 0;
        }
}

[Bug target/97701] [10/11 Regression] aarch64: ICE in extract_constrain_insn since r10-4447-g095f78c6

[rguenth at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=97701

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
     Ever confirmed|0                           |1
           Priority|P3                          |P2
             Status|UNCONFIRMED                 |NEW
           Keywords|                            |ice-on-valid-code
   Last reconfirmed|                            |2021-01-14

[Bug target/97701] [10/11 Regression] aarch64: ICE in extract_constrain_insn since r10-4447-g095f78c6

[jakub at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=97701

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |ktkachov at gcc dot gnu.org,
                   |                            |rsandifo at gcc dot gnu.org,
                   |                            |vmakarov at gcc dot gnu.org

--- Comment #6 from Jakub Jelinek <jakub at gcc dot gnu.org> ---
I can still reproduce with current trunk and it seems it was latent before and
is a ra or backend bug.
The problematic insn is created during LRA, and first is:
(insn 2616 2615 0 (set (reg:DI 1467)
        (lo_sum:DI (reg:DI 1468)
            (symbol_ref/u:DI ("*.LC0") [flags 0x82]))) -1
     (nil))
which is fine for:
        (set (match_operand:DI 0 ("register_operand") ("=r"))
            (lo_sum:DI (match_operand:DI 1 ("register_operand") ("r"))
                (match_operand 2 ("aarch64_valid_symref") ("S"))))
insn, but later on LRA changes the destination register to:
(insn 2616 2615 2617 5 (set (reg:DI 32 v0 [1467])
        (lo_sum:DI (reg:DI 1468)
            (symbol_ref/u:DI ("*.LC0") [flags 0x82]))) 1007 {add_losym_di}
     (expr_list:REG_DEAD (reg:DI 1468)
        (nil)))
and that doesn't match anymore, as v0 register is not GENERAL_REGS.

[Bug target/97701] [10/11 Regression] aarch64: ICE in extract_constrain_insn since r10-4447-g095f78c6

[vmakarov at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=97701

--- Comment #7 from Vladimir Makarov <vmakarov at gcc dot gnu.org> ---
I've reproduced it on gcc-10 branch.

For some reason, LRA does not change register class for reload pseudo.

This bug will take some time for a fix as the fix will probably affect very
sensitive part of LRA and will need a lot of testing.  But I guess the patch
will be ready on the next week.

[Bug target/97701] [10/11 Regression] aarch64: ICE in extract_constrain_insn since r10-4447-g095f78c6

[cvs-commit at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=97701

--- Comment #8 from CVS Commits <cvs-commit at gcc dot gnu.org> ---
The master branch has been updated by Vladimir Makarov <vmakarov@gcc.gnu.org>:

https://gcc.gnu.org/g:7f9f83ef300e8734dccb90a7c347997b2787e9e9

commit r11-6986-g7f9f83ef300e8734dccb90a7c347997b2787e9e9
Author: Vladimir N. Makarov <vmakarov@redhat.com>
Date:   Fri Jan 29 11:51:44 2021 -0500

    [PR97701] LRA: Don't narrow class only for REG or MEM.

    Reload pseudos of ALL_REGS class did not narrow class from constraint
    in insn (set (pseudo) (lo_sum ...)) because lo_sum is considered an
    object (OBJECT_P) although the insn is not a classic move.  To permit
    narrowing we are starting to use MEM_P and REG_P instead of OBJECT_P.

    gcc/ChangeLog:

            PR target/97701
            * lra-constraints.c (in_class_p): Don't narrow class only for REG
            or MEM.

    gcc/testsuite/ChangeLog:

            PR target/97701
            * gcc.target/aarch64/pr97701.c: New.

[Bug target/97701] [10/11 Regression] aarch64: ICE in extract_constrain_insn since r10-4447-g095f78c6

[vmakarov at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=97701

--- Comment #9 from Vladimir Makarov <vmakarov at gcc dot gnu.org> ---
I've committed the patch only to the trunk.  I believe the bug on the trunk is
still present but not triggered by the test.

I'll commit a bit modified patch to gcc 10 branch after some time if there is
no problems with the current patch on the trunk.

[Bug target/97701] [10/11 Regression] aarch64: ICE in extract_constrain_insn since r10-4447-g095f78c6

[jakub at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=97701

--- Comment #10 from Jakub Jelinek <jakub at gcc dot gnu.org> ---
(In reply to Vladimir Makarov from comment #9)
> I've committed the patch only to the trunk.  I believe the bug on the trunk
> is still present but not triggered by the test.

See #c5, by using ? 2 : 1 instead of ? 2 : 0 it should trigger it on the trunk
too.

[Bug target/97701] [10/11 Regression] aarch64: ICE in extract_constrain_insn since r10-4447-g095f78c6

[vmakarov at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=97701

--- Comment #11 from Vladimir Makarov <vmakarov at gcc dot gnu.org> ---
(In reply to Jakub Jelinek from comment #10)
> (In reply to Vladimir Makarov from comment #9)
> > I've committed the patch only to the trunk.  I believe the bug on the trunk
> > is still present but not triggered by the test.
> 
> See #c5, by using ? 2 : 1 instead of ? 2 : 0 it should trigger it on the
> trunk too.

Ok. Thank you, Jakub.

I've just checked the modified test with the patch on the trunk.  The test is
successfully compiled.

I'll modify the test on the trunk.

[Bug target/97701] [10/11 Regression] aarch64: ICE in extract_constrain_insn since r10-4447-g095f78c6

[cvs-commit at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=97701

--- Comment #12 from CVS Commits <cvs-commit at gcc dot gnu.org> ---
The master branch has been updated by Vladimir Makarov <vmakarov@gcc.gnu.org>:

https://gcc.gnu.org/g:0202fa3d6359911a9e6d605d33d0ac669e21eaf3

commit r11-6988-g0202fa3d6359911a9e6d605d33d0ac669e21eaf3
Author: Vladimir N. Makarov <vmakarov@redhat.com>
Date:   Fri Jan 29 16:04:03 2021 -0500

    [PR97701] Modify test for trunk

    Original test was for gcc-10.  The modified one for trunk.

    gcc/testsuite/ChangeLog:

            PR target/97701
            * gcc.target/aarch64/pr97701.c: Modify.

[Bug target/97701] [10 Regression] aarch64: ICE in extract_constrain_insn since r10-4447-g095f78c6

[jakub at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=97701

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
            Summary|[10/11 Regression] aarch64: |[10 Regression] aarch64:
                   |ICE in                      |ICE in
                   |extract_constrain_insn      |extract_constrain_insn
                   |since r10-4447-g095f78c6    |since r10-4447-g095f78c6

--- Comment #13 from Jakub Jelinek <jakub at gcc dot gnu.org> ---
Fixed on the trunk.

[Bug target/97701] [10 Regression] aarch64: ICE in extract_constrain_insn since r10-4447-g095f78c6

[cvs-commit at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=97701

--- Comment #14 from CVS Commits <cvs-commit at gcc dot gnu.org> ---
The releases/gcc-10 branch has been updated by Vladimir Makarov
<vmakarov@gcc.gnu.org>:

https://gcc.gnu.org/g:2211aab8e3832356d066915dfb1e081ac94de7e9

commit r10-9346-g2211aab8e3832356d066915dfb1e081ac94de7e9
Author: Vladimir N. Makarov <vmakarov@redhat.com>
Date:   Thu Feb 4 15:57:55 2021 -0500

    [PR97701] LRA: Don't narrow class only for REG or MEM.  A version modified
for gcc-10.

    This is modified version of the patch committed for the trunk.  The
    modification for gcc-10 includes lra-constraint.c code and the test.

    gcc/ChangeLog:

            PR target/97701
            * lra-constraints.c (in_class_p): Don't narrow class only for REG
            or MEM.

    gcc/testsuite/ChangeLog:

            PR target/97701
            * gcc.target/aarch64/pr97701.c: New.

[Bug target/97701] [10 Regression] aarch64: ICE in extract_constrain_insn since r10-4447-g095f78c6

[acoplan at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=97701

--- Comment #15 from Alex Coplan <acoplan at gcc dot gnu.org> ---
So fixed everywhere?

[Bug target/97701] [10 Regression] aarch64: ICE in extract_constrain_insn since r10-4447-g095f78c6

[rguenth at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=97701

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
   Target Milestone|10.3                        |10.4

--- Comment #16 from Richard Biener <rguenth at gcc dot gnu.org> ---
GCC 10.3 is being released, retargeting bugs to GCC 10.4.

[Bug target/97701] [10 Regression] aarch64: ICE in extract_constrain_insn since r10-4447-g095f78c6

[jakub at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=97701

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
   Target Milestone|10.4                        |10.5

--- Comment #17 from Jakub Jelinek <jakub at gcc dot gnu.org> ---
GCC 10.4 is being released, retargeting bugs to GCC 10.5.

[Bug target/97701] [10 Regression] aarch64: ICE in extract_constrain_insn since r10-4447-g095f78c6

[rguenth at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=97701

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
      Known to fail|11.0                        |10.2.0
   Target Milestone|10.5                        |10.3
             Status|NEW                         |RESOLVED
         Resolution|---                         |FIXED
      Known to work|                            |10.3.0, 11.1.0

--- Comment #18 from Richard Biener <rguenth at gcc dot gnu.org> ---
Assuming fixed in GCC 10.3.