[Bug target/99724] New: [11 Regression] CE in in extract_insn, at recog.c:2770

[Bug target/99724] New: [11 Regression] CE in in extract_insn, at recog.c:2770

[marxin at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=99724

            Bug ID: 99724
           Summary: [11 Regression] CE in in extract_insn, at recog.c:2770
           Product: gcc
           Version: 11.0
            Status: UNCONFIRMED
          Keywords: ice-on-valid-code
          Severity: normal
          Priority: P3
         Component: target
          Assignee: unassigned at gcc dot gnu.org
          Reporter: marxin at gcc dot gnu.org
  Target Milestone: ---
              Host: x86_64-linux-gnu
            Target: arm-linux-gnueabi

The following causes an ICE:

$ arm-linux-gnueabi-gcc
/home/marxin/Programming/gcc/gcc/testsuite/gcc.dg/fold-minus-1.c -mcpu=iwmmxt2
/home/marxin/Programming/gcc/gcc/testsuite/gcc.dg/fold-minus-1.c: In function
‘g’:
/home/marxin/Programming/gcc/gcc/testsuite/gcc.dg/fold-minus-1.c:13:1: error:
unrecognizable insn:
   13 | }
      | ^
(insn 10 9 11 2 (set (reg:V2SI 114 [ _2 ])
        (not:V2SI (reg:V2SI 113 [ _1 ])))
"/home/marxin/Programming/gcc/gcc/testsuite/gcc.dg/fold-minus-1.c":10:11 -1
     (nil))
during RTL pass: vregs
/home/marxin/Programming/gcc/gcc/testsuite/gcc.dg/fold-minus-1.c:13:1: internal
compiler error: in extract_insn, at recog.c:2770

[Bug target/99724] [11 Regression] CE in in extract_insn, at recog.c:2770

[rguenth at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=99724

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
   Target Milestone|---                         |11.0

[Bug target/99724] [11 Regression] CE in in extract_insn, at recog.c:2770

[rearnsha at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=99724

Richard Earnshaw <rearnsha at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
   Last reconfirmed|                            |2021-03-23
             Status|UNCONFIRMED                 |NEW
     Ever confirmed|0                           |1

--- Comment #1 from Richard Earnshaw <rearnsha at gcc dot gnu.org> ---
r11-5989 is the likely culprit.  It introduced one_cmpl<mode>2 for neon and mve
under a common framework.

[Bug target/99724] [11 Regression] CE in in extract_insn, at recog.c:2770

[jakub at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=99724

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> ---
Indeed, bisected it to that r11-5989-gfd43603414a9b7bdbac5a822af144dcd559733eb
commit.
And
--- gcc/config/arm/vec-common.md.jj     2021-03-23 10:21:07.096448805 +0100
+++ gcc/config/arm/vec-common.md        2021-03-23 14:18:43.916347335 +0100
@@ -202,7 +202,7 @@ (define_expand "xor<mode>3"
 (define_expand "one_cmpl<mode>2"
   [(set (match_operand:VDQ 0 "s_register_operand")
        (not:VDQ (match_operand:VDQ 1 "s_register_operand")))]
-  "ARM_HAVE_<MODE>_ARITH"
+  "ARM_HAVE_<MODE>_ARITH && !TARGET_REALLY_IWMMXT"
 )

 (define_expand "neg<mode>2"
fixes it.  I just have to wonder how many other expanders have the similar
problem though.

[Bug target/99724] [11 Regression] CE in in extract_insn, at recog.c:2770

[jakub at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=99724

--- Comment #3 from Jakub Jelinek <jakub at gcc dot gnu.org> ---
E.g. I can't find a neg pattern in iwmmxt.md either, so
r11-5990-g4cbb7cab47a3b91a12ad52baab5bbe6e4373ce73
is problematic too.
And r11-6616-g25bef68902f42f414f99626cefb2d3df81de7dc8 too (don't see any
UNSPEC_MISALIGNED_ACCESS in iwmmxt.md either.
and/ior/xor is, so that is likely ok.
So
--- gcc/config/arm/vec-common.md.jj     2021-03-23 10:21:07.096448805 +0100
+++ gcc/config/arm/vec-common.md        2021-03-23 14:36:54.823323029 +0100
@@ -202,13 +202,13 @@ (define_expand "xor<mode>3"
 (define_expand "one_cmpl<mode>2"
   [(set (match_operand:VDQ 0 "s_register_operand")
        (not:VDQ (match_operand:VDQ 1 "s_register_operand")))]
-  "ARM_HAVE_<MODE>_ARITH"
+  "ARM_HAVE_<MODE>_ARITH && !TARGET_REALLY_IWMMXT"
 )

 (define_expand "neg<mode>2"
   [(set (match_operand:VDQWH 0 "s_register_operand" "")
        (neg:VDQWH (match_operand:VDQWH 1 "s_register_operand" "")))]
-  "ARM_HAVE_<MODE>_ARITH"
+  "ARM_HAVE_<MODE>_ARITH && !TARGER_REALLY_IWMMXT"
 )

 (define_expand "cadd<rot><mode>3"
@@ -281,7 +281,8 @@ (define_expand "movmisalign<mode>"
  [(set (match_operand:VDQX 0 "neon_perm_struct_or_reg_operand")
        (unspec:VDQX [(match_operand:VDQX 1 "neon_perm_struct_or_reg_operand")]
         UNSPEC_MISALIGNED_ACCESS))]
- "ARM_HAVE_<MODE>_LDST && !BYTES_BIG_ENDIAN && unaligned_access"
+ "ARM_HAVE_<MODE>_LDST && !BYTES_BIG_ENDIAN
+  && unaligned_access && !TARGET_REALLY_IWMMXT"
 {
  rtx adjust_mem;
  /* This pattern is not permitted to fail during expansion: if both arguments
instead?

[Bug target/99724] [11 Regression] CE in in extract_insn, at recog.c:2770

[clyon at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=99724

--- Comment #4 from Christophe Lyon <clyon at gcc dot gnu.org> ---
Oops sorry, indeed it's likely that several of my other patches in this area
introduced the same problem. I did lots of 'make check' though, too bad iwmmxt
isn't covered well.

[Bug target/99724] [11 Regression] CE in in extract_insn, at recog.c:2770

[jakub at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=99724

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

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

--- Comment #5 from Jakub Jelinek <jakub at gcc dot gnu.org> ---
Created attachment 50465
  --> https://gcc.gnu.org/bugzilla/attachment.cgi?id=50465&action=edit
gcc11-pr99724.patch

Full patch I'm going to bootstrap/regtest.

[Bug target/99724] [11 Regression] CE in in extract_insn, at recog.c:2770

[clyon at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=99724

--- Comment #6 from Christophe Lyon <clyon at gcc dot gnu.org> ---
Looks good to me, thanks

[Bug target/99724] [11 Regression] CE in in extract_insn, at recog.c:2770

[cvs-commit at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=99724

--- Comment #7 from CVS 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:4f992de4f369e17b64a66246ec35bf75f41dc98d

commit r11-7814-g4f992de4f369e17b64a66246ec35bf75f41dc98d
Author: Jakub Jelinek <jakub@redhat.com>
Date:   Wed Mar 24 11:22:35 2021 +0100

    arm: Fix some more vec-common.md patterns for iwmmxt [PR99724]

    The following patch fixes similar issues as in PR98849;
    in older gcc versions, the expanders were present in neon.md guarded
    with TARGET_NEON, but they got moved to vec-common.md and guarded with
    ARM_HAVE_<MODE>_ARITH so that they handle both MVE and Neon.
    The macros are enabled for some modes even for iwmmxt which has some
    vector support for those modes, but only limited.  In particular,
    neither the one_cmpl, nor neg, nor movmisalign patterns are present.
    For some reason I've failed to construct something that ICEs with
    movmisalign, so that is not covered by the testsuite, but both
    one_cmpl and neg ICE.

    2021-03-24  Jakub Jelinek  <jakub@redhat.com>

            PR target/99724
            * config/arm/vec-common.md (one_cmpl<mode>2, neg<mode>2,
            movmisalign<mode>): Disable expanders for TARGET_REALLY_IWMMXT.

            * gcc.target/arm/pr99724.c: New test.

[Bug target/99724] [11 Regression] CE in in extract_insn, at recog.c:2770

[jakub at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=99724

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

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

--- Comment #8 from Jakub Jelinek <jakub at gcc dot gnu.org> ---
Fixed.