[Bug target/104004] New: [12 Regression] ICE: in extract_insn, at recog.c:2769 (error: unrecognizable insn)

[Bug target/104004] New: [12 Regression] ICE: in extract_insn, at recog.c:2769 (error: unrecognizable insn)

[asolokha at gmx dot com]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=104004

            Bug ID: 104004
           Summary: [12 Regression] ICE: in extract_insn, at recog.c:2769
                    (error: unrecognizable insn)
           Product: gcc
           Version: 12.0
            Status: UNCONFIRMED
          Keywords: ice-on-invalid-code
          Severity: normal
          Priority: P3
         Component: target
          Assignee: unassigned at gcc dot gnu.org
          Reporter: asolokha at gmx dot com
                CC: linkw at gcc dot gnu.org, wschmidt at gcc dot gnu.org
  Target Milestone: ---
            Target: powerpc-*-linux-gnu

1. gcc 12.0.0 20220109 snapshot (g:49d73c9fb644673323845efebfe6b3106e70af8a)
ICEs when compiling gcc/testsuite/gcc.target/powerpc/mffscrni_p9.c w/
-mcpu=e300c2:

% powerpc-e300c3-linux-gnu-gcc-12.0.0 -mcpu=e300c2 -c
gcc/testsuite/gcc.target/powerpc/mffscrni_p9.c
gcc/testsuite/gcc.target/powerpc/mffscrni_p9.c: In function 'foo':
gcc/testsuite/gcc.target/powerpc/mffscrni_p9.c:9:1: error: unrecognizable insn:
    9 | }
      | ^
(insn 15 14 16 2 (set (reg:DF 119)
        (unspec_volatile:DF [
                (const_int 0 [0])
            ] UNSPECV_MFFS))
"gcc/testsuite/gcc.target/powerpc/mffscrni_p9.c":8:3 -1
     (nil))
during RTL pass: vregs
gcc/testsuite/gcc.target/powerpc/mffscrni_p9.c:9:1: internal compiler error: in
extract_insn, at recog.c:2769
0x6adda1 _fatal_insn(char const*, rtx_def const*, char const*, int, char
const*)
       
/var/tmp/portage/cross-powerpc-e300c3-linux-gnu/gcc-12.0.0_p20220109/work/gcc-12-20220109/gcc/rtl-error.c:108
0x6addc1 _fatal_insn_not_found(rtx_def const*, char const*, int, char const*)
       
/var/tmp/portage/cross-powerpc-e300c3-linux-gnu/gcc-12.0.0_p20220109/work/gcc-12-20220109/gcc/rtl-error.c:116
0x6ac3cf extract_insn(rtx_insn*)
       
/var/tmp/portage/cross-powerpc-e300c3-linux-gnu/gcc-12.0.0_p20220109/work/gcc-12-20220109/gcc/recog.c:2769
0xb23843 instantiate_virtual_regs_in_insn
       
/var/tmp/portage/cross-powerpc-e300c3-linux-gnu/gcc-12.0.0_p20220109/work/gcc-12-20220109/gcc/function.c:1611
0xb23843 instantiate_virtual_regs
       
/var/tmp/portage/cross-powerpc-e300c3-linux-gnu/gcc-12.0.0_p20220109/work/gcc-12-20220109/gcc/function.c:1985
0xb23843 execute
       
/var/tmp/portage/cross-powerpc-e300c3-linux-gnu/gcc-12.0.0_p20220109/work/gcc-12-20220109/gcc/function.c:2034

2. After adding -O1 (in fact, w/ any optimization level) to the command line,
it fails to recognize UNSPECV_MTFSB0 instead:

% powerpc-e300c3-linux-gnu-gcc-12.0.0 -mcpu=e300c2 -O1 -c
gcc/testsuite/gcc.target/powerpc/mffscrni_p9.c
gcc/testsuite/gcc.target/powerpc/mffscrni_p9.c: In function 'foo':
gcc/testsuite/gcc.target/powerpc/mffscrni_p9.c:9:1: error: unrecognizable insn:
    9 | }
      | ^
(insn 5 2 6 2 (unspec_volatile [
            (const_int 31 [0x1f])
        ] UNSPECV_MTFSB0) "gcc/testsuite/gcc.target/powerpc/mffscrni_p9.c":8:3
-1
     (nil))
during RTL pass: vregs
gcc/testsuite/gcc.target/powerpc/mffscrni_p9.c:9:1: internal compiler error: in
extract_insn, at recog.c:2769
0x6adda1 _fatal_insn(char const*, rtx_def const*, char const*, int, char
const*)
       
/var/tmp/portage/cross-powerpc-e300c3-linux-gnu/gcc-12.0.0_p20220109/work/gcc-12-20220109/gcc/rtl-error.c:108
0x6addc1 _fatal_insn_not_found(rtx_def const*, char const*, int, char const*)
       
/var/tmp/portage/cross-powerpc-e300c3-linux-gnu/gcc-12.0.0_p20220109/work/gcc-12-20220109/gcc/rtl-error.c:116
0x6ac3cf extract_insn(rtx_insn*)
       
/var/tmp/portage/cross-powerpc-e300c3-linux-gnu/gcc-12.0.0_p20220109/work/gcc-12-20220109/gcc/recog.c:2769
0xb2309c instantiate_virtual_regs_in_insn
       
/var/tmp/portage/cross-powerpc-e300c3-linux-gnu/gcc-12.0.0_p20220109/work/gcc-12-20220109/gcc/function.c:1660
0xb2309c instantiate_virtual_regs
       
/var/tmp/portage/cross-powerpc-e300c3-linux-gnu/gcc-12.0.0_p20220109/work/gcc-12-20220109/gcc/function.c:1985
0xb2309c execute
       
/var/tmp/portage/cross-powerpc-e300c3-linux-gnu/gcc-12.0.0_p20220109/work/gcc-12-20220109/gcc/function.c:2034

Other builtins related to loading (storing) data from (to) FPSCR are also
affected.

gcc 11.2 rejects the code correctly w/ the following diagnostics:

% powerpc-e300c3-linux-gnu-gcc-11.2.0 -mcpu=e300c2 -O1 -c
gcc/testsuite/gcc.target/powerpc/mffscrni_p9.c
gcc/testsuite/gcc.target/powerpc/mffscrni_p9.c: In function 'foo':
gcc/testsuite/gcc.target/powerpc/mffscrni_p9.c:8:3: error:
'__builtin_set_fpscr_rn' not supported with '-msoft-float'
    8 |   __builtin_set_fpscr_rn (val);
      |   ^~~~~~~~~~~~~~~~~~~~~~~~~~~~

[Bug target/104004] [12 Regression] ICE: in extract_insn, at recog.c:2769 (error: unrecognizable insn)

[rguenth at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=104004

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
   Target Milestone|---                         |12.0

[Bug target/104004] [12 Regression] ICE: in extract_insn, at recog.c:2769 (error: unrecognizable insn)

[linkw at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=104004

Kewen Lin <linkw at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|UNCONFIRMED                 |NEW
     Ever confirmed|0                           |1
   Last reconfirmed|                            |2022-01-13

--- Comment #1 from Kewen Lin <linkw at gcc dot gnu.org> ---
Thanks for reporting!

We have the below check in rs6000_expand_set_fpscr_rn_builtin before:

  if (rs6000_isa_flags & OPTION_MASK_SOFT_FLOAT)
    {
      error ("%<__builtin_set_fpscr_rn%> not supported with %<-msoft-float%>");
      return const0_rtx;
    }

It's supposed to be updated with new attribute nosoft in new bif support.

But grepping in the latest trunk code, there are no bifs using nosoft, I guess
there are something not upstreamed eventually unexpectedly.

The fix seems to add nosoft attribute for __builtin_set_fpscr_rn entry.
Besides, by checking in old bif support, the below ones also require nosoft:

__builtin_set_fpscr_drn (this one needs no32bit)
__builtin_mffsl
__builtin_mtfsb0
__builtin_mtfsb1

[Bug target/104004] [12 Regression] ICE: in extract_insn, at recog.c:2769 (error: unrecognizable insn)

[rguenth at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=104004

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
           Priority|P3                          |P1

[Bug target/104004] [12 Regression] ICE: in extract_insn, at recog.c:2769 (error: unrecognizable insn)

[linkw at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=104004

Kewen Lin <linkw at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|NEW                         |ASSIGNED
           Assignee|unassigned at gcc dot gnu.org      |wschmidt at gcc dot gnu.org
                URL|                            |https://gcc.gnu.org/piperma
                   |                            |il/gcc-patches/2022-January
                   |                            |/589441.html

--- Comment #2 from Kewen Lin <linkw at gcc dot gnu.org> ---
Bill already posted one patch for it.

[Bug target/104004] [12 Regression] ICE: in extract_insn, at recog.c:2769 (error: unrecognizable insn)

[cvs-commit at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=104004

--- Comment #3 from CVS Commits <cvs-commit at gcc dot gnu.org> ---
The master branch has been updated by Segher Boessenkool <segher@gcc.gnu.org>:

https://gcc.gnu.org/g:aaf3a5993ae49f1ae6792800e5161a1d51436ed3

commit r12-7945-gaaf3a5993ae49f1ae6792800e5161a1d51436ed3
Author: Bill Schmidt <wschmidt@linux.ibm.com>
Date:   Fri Jan 28 11:50:26 2022 -0600

    rs6000: Fix some missing built-in attributes [PR104004]

    PR104004 caught some misses on my part in converting to the new built-in
    function infrastructure.  In particular, I forgot to mark all of the
"nosoft"
    built-ins, and one of those should also have been marked "no32bit".

    2022-01-27  Bill Schmidt  <wschmidt@linux.ibm.com>

    gcc/
            PR target/104004
            * config/rs6000/rs6000-builtins.def (MFFSL): Mark nosoft.
            (MTFSB0): Likewise.
            (MTFSB1): Likewise.
            (SET_FPSCR_RN): Likewise.
            (SET_FPSCR_DRN): Mark nosoft and no32bit.

[Bug target/104004] [12 Regression] ICE: in extract_insn, at recog.c:2769 (error: unrecognizable insn)

[segher at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=104004

Segher Boessenkool <segher at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|ASSIGNED                    |RESOLVED
                 CC|                            |segher at gcc dot gnu.org
         Resolution|---                         |FIXED

--- Comment #4 from Segher Boessenkool <segher at gcc dot gnu.org> ---
Fixed now.