[Bug middle-end/102133] New: [12 Regression] ICE in set_rtl building libgcc __muldc3 for 32-bit SPARC

[Bug middle-end/102133] New: [12 Regression] ICE in set_rtl building libgcc __muldc3 for 32-bit SPARC

[jsm28 at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=102133

            Bug ID: 102133
           Summary: [12 Regression] ICE in set_rtl building libgcc
                    __muldc3 for 32-bit SPARC
           Product: gcc
           Version: 12.0
            Status: UNCONFIRMED
          Keywords: ice-on-valid-code
          Severity: normal
          Priority: P3
         Component: middle-end
          Assignee: unassigned at gcc dot gnu.org
          Reporter: jsm28 at gcc dot gnu.org
                CC: liuhongt at gcc dot gnu.org
  Target Milestone: ---
            Target: sparc*-*-*

Created attachment 51380
  --> https://gcc.gnu.org/bugzilla/attachment.cgi?id=51380&action=edit
preprocessed source

commit 7218c2ec365ce95f5a1012a6eb425b0a36aec6bf, "Make sure we're playing with
integral modes before call extract_integral_bit_field.", introduces the
following ICE building libgcc for 32-bit SPARC (32-bit multilib for
sparc64-linux-gnu). Compile the attached test with -m32 -O2.

during RTL pass: expand
/scratch/jmyers/glibc/many12/src/gcc/libgcc/libgcc2.c: In function '__muldc3':
/scratch/jmyers/glibc/many12/src/gcc/libgcc/libgcc2.c:1947:1: internal compiler
error: in set_rtl, at cfgexpand.c:200
 1947 | CONCAT3(__mul,MODE,3) (MTYPE a, MTYPE b, MTYPE c, MTYPE d)
      | ^~~~~~~~
0x7ccbe9 set_rtl
        /scratch/jmyers/glibc/many12/src/gcc/gcc/cfgexpand.c:179
0x7cf028 set_parm_rtl(tree_node*, rtx_def*)
        /scratch/jmyers/glibc/many12/src/gcc/gcc/cfgexpand.c:1401
0x98a451 assign_parm_setup_reg
        /scratch/jmyers/glibc/many12/src/gcc/gcc/function.c:3401
0x98a451 assign_parms
        /scratch/jmyers/glibc/many12/src/gcc/gcc/function.c:3707
0x98de2c expand_function_start(tree_node*)
        /scratch/jmyers/glibc/many12/src/gcc/gcc/function.c:5163
0x7e11b9 execute
        /scratch/jmyers/glibc/many12/src/gcc/gcc/cfgexpand.c:6692
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 middle-end/102133] [12 Regression] ICE in set_rtl building libgcc __muldc3 for 32-bit SPARC

[pinskia at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=102133

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

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

[Bug middle-end/102133] [12 Regression] ICE in set_rtl building libgcc __muldc3 for 32-bit SPARC

[crazylht at gmail dot com]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=102133

Hongtao.liu <crazylht at gmail dot com> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |crazylht at gmail dot com

--- Comment #1 from Hongtao.liu <crazylht at gmail dot com> ---
It also caused regressions on mcore-elf:

FAIL: gcc.dg/torture/builtin-attr-1.c   -O1  (internal compiler error)


Excess errors:
during RTL pass: expand
/home/jlaw/jenkins/workspace/mcore-elf/gcc/gcc/testsuite/gcc.dg/torture/builtin-attr-1.c:147:6: 
internal compiler error: in set_rtl, at cfgexpand.c:179
0x5971b2 set_rtl
         ../../..//gcc/gcc/cfgexpand.c:179 0x89928a set_parm_rtl(tree_node*,
rtx_def*)
         ../../..//gcc/gcc/cfgexpand.c:1401
0xa2b995 assign_parm_setup_reg
         ../../..//gcc/gcc/function.c:3401 0xa2efeb assign_parms
         ../../..//gcc/gcc/function.c:3707 0xa2efeb
expand_function_start(tree_node*)
         ../../..//gcc/gcc/function.c:5163
0x8a7494 execute
         ../../..//gcc/gcc/cfgexpand.c:6692

[Bug middle-end/102133] [12 Regression] ICE in set_rtl building libgcc __muldc3 for 32-bit SPARC

[crazylht at gmail dot com]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=102133

--- Comment #2 from Hongtao.liu <crazylht at gmail dot com> ---
I successfully reproduce error related to 32-bit SPARC libgcc

But failed to configure for target mcore, i didn't find any reference in
https://gcc.gnu.org/install/specific.html

--target=mcore results in
*** Configuration mcore-unknown-none not supported

[Bug middle-end/102133] [12 Regression] ICE in set_rtl building libgcc __muldc3 for 32-bit SPARC

[crazylht at gmail dot com]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=102133

--- Comment #3 from Hongtao.liu <crazylht at gmail dot com> ---
static inline void
set_rtl (tree t, rtx x)
{
  gcc_checking_assert (!x
                       || !(TREE_CODE (t) == SSA_NAME || is_gimple_reg (t))
                       || (use_register_for_decl (t)
                           ? (REG_P (x)
                              || (GET_CODE (x) == CONCAT
                                  && (REG_P (XEXP (x, 0))
                                      || SUBREG_P (XEXP (x, 0)))
                                  && (REG_P (XEXP (x, 1))
                                      || SUBREG_P (XEXP (x, 1))))
                              /* We need to accept PARALLELs for RESUT_DECLs
                                 because of vector types with BLKmode returned
                                 in multiple registers, but they are supposed
                                 to be uncoalesced.  */
                              || (GET_CODE (x) == PARALLEL
                                  && SSAVAR (t)
                                  && TREE_CODE (SSAVAR (t)) == RESULT_DECL
                                  && (GET_MODE (x) == BLKmode
                                      || !flag_tree_coalesce_vars)))
                           : (MEM_P (x) || x == pc_rtx
                              || (GET_CODE (x) == CONCAT
                                  && MEM_P (XEXP (x, 0))
                                  && MEM_P (XEXP (x, 1))))));


x is 
(subreg:DF (reg:DI 154) 0),

t is 
 <ssa_name 0x7fffea4a9a68
    type <real_type 0x7fffea48b1f8 DFtype sizes-gimplified DF
        size <integer_cst 0x7fffea2e2e10 constant 64>
        unit-size <integer_cst 0x7fffea2e2e28 constant 8>
        align:64 warn_if_not_align:0 symtab:0 alias-set -1 canonical-type
0x7fffea2f9498 precision:64 context <translation_unit_decl 0x7fffea49c078
/scratch/jmyers/glibc/many12/src/gcc/libgcc/libgcc2.c>>
    visited var <parm_decl 0x7fffea497b00 d>
    def_stmt GIMPLE_NOP
    version:62>

and it hit REG_P (XEXP (x, 1)), XEXP (x, 1) is invalid for subreg, so set_rtl 
here doesn't accept subreg?

[Bug middle-end/102133] [12 Regression] ICE in set_rtl building libgcc __muldc3 for 32-bit SPARC

[crazylht at gmail dot com]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=102133

--- Comment #4 from Hongtao.liu <crazylht at gmail dot com> ---

> 
> and it hit REG_P (XEXP (x, 1)), XEXP (x, 1) is invalid for subreg, so
> set_rtl  here doesn't accept subreg?

typo, it hit gcc_assert that if X is not REG, it must be CONCAT or PARALLEL,
but here is SUBREG, so gcc_assert failed.

[Bug middle-end/102133] [12 Regression] ICE in set_rtl building libgcc __muldc3 for 32-bit SPARC

[crazylht at gmail dot com]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=102133

--- Comment #5 from Hongtao.liu <crazylht at gmail dot com> ---
(In reply to Hongtao.liu from comment #4)
> > 
> > and it hit REG_P (XEXP (x, 1)), XEXP (x, 1) is invalid for subreg, so
> > set_rtl  here doesn't accept subreg?
> 
> typo, it hit gcc_assert that if X is not REG, it must be CONCAT or PARALLEL,
> but here is SUBREG, so gcc_assert failed.

in extract_bit_bitfield_1
target: (reg/v:DF 153)
op0: (mem/c:DF (plus:SI (reg/f:SI 103 virtual-incoming-args)
        (const_int 24 [0x18])) [1 d+0 S8 A32])


in good case, it will move op0 to target by 2 32-bit loads

---cut from  extract_integral_bit_field-----

  /* Handle fields bigger than a word.  */

  if (bitsize > BITS_PER_WORD)
    {
      /* Here we transfer the words of the field
         in the order least significant first.
         This is because the most significant word is the one which may
         be less than full.  */
--------cut end-------

after my changed, it will return (subreg:DF (reg: DI 0))

[Bug middle-end/102133] [12 Regression] ICE in set_rtl building libgcc __muldc3 for 32-bit SPARC

[crazylht at gmail dot com]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=102133

--- Comment #6 from Hongtao.liu <crazylht at gmail dot com> ---
The difference of insn sequence is like
good one:
(insn 5 4 6 (clobber (reg/v:DF 153))
"/scratch/jmyers/glibc/many12/src/gcc/libgcc/libgcc2.c":1948:1 -1
     (nil))

(insn 6 5 7 (set (subreg:SI (reg/v:DF 153) 4)
        (mem/c:SI (plus:SI (reg/f:SI 103 virtual-incoming-args)
                (const_int 28 [0x1c])) [1 d+4 S4 A32]))
"/scratch/jmyers/glibc/many12/src/gcc/libgcc/libgcc2.c":1948:1 -1
     (nil))

(insn 7 6 0 (set (subreg:SI (reg/v:DF 153) 0)
        (mem/c:SI (plus:SI (reg/f:SI 103 virtual-incoming-args)
                (const_int 24 [0x18])) [1 d+0 S4 A32]))
"/scratch/jmyers/glibc/many12/src/gcc/libgcc/libgcc2.c":1948:1 -1
     (nil))

(reg/v:DF 153)

bad one:

(insn 5 4 6 (clobber (reg:DI 154))
"/scratch/jmyers/glibc/many12/src/gcc/libgcc/libgcc2.c":1948:1 -1
     (nil))

(insn 6 5 7 (set (subreg:SI (reg:DI 154) 4)
        (mem/c:SI (plus:SI (reg/f:SI 103 virtual-incoming-args)
                (const_int 28 [0x1c])) [1 d+4 S4 A32]))
"/scratch/jmyers/glibc/many12/src/gcc/libgcc/libgcc2.c":1948:1 -1
     (nil))

(insn 7 6 0 (set (subreg:SI (reg:DI 154) 0)
        (mem/c:SI (plus:SI (reg/f:SI 103 virtual-incoming-args)
                (const_int 24 [0x18])) [1 d+0 S4 A32]))
"/scratch/jmyers/glibc/many12/src/gcc/libgcc/libgcc2.c":1948:1 -1
     (nil))

(subreg:DF (reg:DI 154) 0)


and subreg hit gcc_assert in set_rtl.

[Bug middle-end/102133] [12 Regression] ICE in set_rtl building libgcc __muldc3 for 32-bit SPARC

[crazylht at gmail dot com]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=102133

--- Comment #7 from Hongtao.liu <crazylht at gmail dot com> ---
Since we also allow something like (concat:(subreg) (subreg)), should we also
allow subreg outside?

   gcc_checking_assert (!x
                       || !(TREE_CODE (t) == SSA_NAME || is_gimple_reg (t))
                       || (use_register_for_decl (t)
-                          ? (REG_P (x)
+                          ? (REG_P (x) || SUBREG_P (x)
                              || (GET_CODE (x) == CONCAT
                                  && (REG_P (XEXP (x, 0))
                                      || SUBREG_P (XEXP (x, 0)))


The below code also looks like it's going to handle SUBREG.

  if (x)
    {
      bool skip = false;
      tree cur = NULL_TREE;
      rtx xm = x;

    retry:
      if (MEM_P (xm))
        cur = MEM_EXPR (xm);
      else if (REG_P (xm))
        cur = REG_EXPR (xm);
      else if (SUBREG_P (xm))
        {
          gcc_assert (subreg_lowpart_p (xm));
          xm = SUBREG_REG (xm);
          goto retry;

[Bug middle-end/102133] [12 Regression] ICE in set_rtl building libgcc __muldc3 for 32-bit SPARC

[crazylht at gmail dot com]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=102133

--- Comment #8 from Hongtao.liu <crazylht at gmail dot com> ---
(In reply to Hongtao.liu from comment #7)
> Since we also allow something like (concat:(subreg) (subreg)), should we
> also allow subreg outside?
> 
>    gcc_checking_assert (!x
>  		       || !(TREE_CODE (t) == SSA_NAME || is_gimple_reg (t))
>  		       || (use_register_for_decl (t)
> -			   ? (REG_P (x)
> +			   ? (REG_P (x) || SUBREG_P (x)
>  			      || (GET_CODE (x) == CONCAT
>  				  && (REG_P (XEXP (x, 0))
>  				      || SUBREG_P (XEXP (x, 0)))
> 
> 

It can also fix FAIL: gcc.dg/torture/builtin-attr-1.c   -O1  (internal compiler
error)

for builtin-attr-1.c assembly is the same as the one built without my, but for
for __muldc3, there's a little bit difference caused by rtl optimization.

[Bug middle-end/102133] [12 Regression] ICE in set_rtl building libgcc __muldc3 for 32-bit SPARC

[rguenth at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=102133

--- Comment #9 from Richard Biener <rguenth at gcc dot gnu.org> ---
SUBREGs are not wanted for DECL_RTL

[Bug middle-end/102133] [12 Regression] ICE in set_rtl building libgcc __muldc3 for 32-bit SPARC

[mikpelinux at gmail dot com]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=102133

--- Comment #10 from Mikael Pettersson <mikpelinux at gmail dot com> ---
(In reply to Hongtao.liu from comment #2)
> But failed to configure for target mcore, i didn't find any reference in
> https://gcc.gnu.org/install/specific.html
> 
> --target=mcore results in
> *** Configuration mcore-unknown-none not supported

It's mcore-elf or mcore-unknown-elf .

[Bug middle-end/102133] [12 Regression] ICE in set_rtl building libgcc __muldc3 for 32-bit SPARC

[cvs-commit at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=102133

--- Comment #11 from CVS Commits <cvs-commit at gcc dot gnu.org> ---
The master branch has been updated by hongtao Liu <liuhongt@gcc.gnu.org>:

https://gcc.gnu.org/g:508fa61b6319377e48cbee98da221aacd475fd10

commit r12-3276-g508fa61b6319377e48cbee98da221aacd475fd10
Author: liuhongt <hongtao.liu@intel.com>
Date:   Tue Aug 31 17:16:08 2021 +0800

    Revert "Make sure we're playing with integral modes before call
extract_integral_bit_field."

    This reverts commit 7218c2ec365ce95f5a1012a6eb425b0a36aec6bf.

         PR middle-end/102133

[Bug middle-end/102133] [12 Regression] ICE in set_rtl building libgcc __muldc3 for 32-bit SPARC

[crazylht at gmail dot com]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=102133

--- Comment #12 from Hongtao.liu <crazylht at gmail dot com> ---
Fixed in GCC12.

[Bug middle-end/102133] [12 Regression] ICE in set_rtl building libgcc __muldc3 for 32-bit SPARC

[rguenth at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=102133

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

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

--- Comment #13 from Richard Biener <rguenth at gcc dot gnu.org> ---
Indeed fixed.