[Bug tree-optimization/107672] New: [13 Regression] ICE during GIMPLE pass: slp

[Bug tree-optimization/107672] New: [13 Regression] ICE during GIMPLE pass: slp

[creatorsmithmdt at gmail dot com]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=107672

            Bug ID: 107672
           Summary: [13 Regression] ICE during GIMPLE pass: slp
           Product: gcc
           Version: 13.0
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: tree-optimization
          Assignee: unassigned at gcc dot gnu.org
          Reporter: creatorsmithmdt at gmail dot com
  Target Milestone: ---

The slp GIMPLE pass now ICE's on this code, which it did not previously. While
I am aware GCJ is not currently part of GCC, the GIMPLE it generates is valid
and should not be ICE'ing.

It does not ICE on O1.

This bug occurs on my msterstable branch.
(https://github.com/Zopolis4/gcj/commits/msterstable)

Given that GCJ is not currently part of GCC, and this bug is thus harder to
reproduce as one has to compile my fork, is there a better way for me to report
this? Maybe I could just send the generated GIMPLE that ICE's?

~/g/x/libjava> /bin/bash ./libtool  --tag=GCJ   --mode=compile
/home/zopolis4/gcjbuild/./gcc/gcj
-B/home/zopolis4/gcjbuild/x86_64-pc-linux-gnu/libjava/
 -B/home/zopolis4/gcjbuild/./gcc/ -B/usr/local/x86_64-pc-linux-gnu/bin/
-B/usr/local/x86_64-pc-linux-gnu/lib/ -isystem
/usr/local/x86_64-pc-linux-gnu/include -isystem /usr/
local/x86_64-pc-linux-gnu/sys-include    -fomit-frame-pointer -Usun
-fclasspath=
-fbootclasspath=/home/zopolis4/gcjbuild/../gcj/libjava/classpath/lib
--encoding=UTF-8 -Wno-
deprecated -fbootstrap-classes -g -O2  -c -o gnu/java/security/hash.lo
-fsource-filename=/home/zopolis4/gcjbuild/x86_64-pc-linux-gnu/libjava/classpath/lib/classes
-MT gnu/j
ava/security/hash.lo -MD -MP -MF gnu/java/security/hash.deps
@gnu/java/security/hash.list
libtool: compile:  /home/zopolis4/gcjbuild/./gcc/gcj
-B/home/zopolis4/gcjbuild/x86_64-pc-linux-gnu/libjava/
-B/home/zopolis4/gcjbuild/./gcc/ -B/usr/local/x86_64-pc-linux-gnu/bin/
-B/usr/local/x86_64-pc-linux-gnu/lib/ -isystem
/usr/local/x86_64-pc-linux-gnu/include -isystem
/usr/local/x86_64-pc-linux-gnu/sys-include -fomit-frame-pointer -Usun
-fclasspath=
-fbootclasspath=/home/zopolis4/gcjbuild/../gcj/libjava/classpath/lib
--encoding=UTF-8 -Wno-deprecated -fbootstrap-classes -g -O2 -c
-fsource-filename=/home/zopolis4/gcjbuild/x86_64-pc-linux-gnu/libjava/classpath/lib/classes
-MT gnu/java/security/hash.lo -MD -MP -MF gnu/java/security/hash.deps
@gnu/java/security/hash.list  -fPIC -o gnu/java/security/.libs/hash.o
during GIMPLE pass: slp
/home/zopolis4/gcj/libjava/classpath/gnu/java/security/hash/Whirlpool.java: In
class 'gnu.java.security.hash.Whirlpool':
/home/zopolis4/gcj/libjava/classpath/gnu/java/security/hash/Whirlpool.java: In
method 'gnu.java.security.hash.Whirlpool.transform(byte[],int)':
In file included from
/home/zopolis4/gcj/libjava/classpath/gnu/java/security/hash/Tiger.java:863,
                 from
/home/zopolis4/gcj/libjava/classpath/gnu/java/security/hash/Sha512.java:277,
                 from
/home/zopolis4/gcj/libjava/classpath/gnu/java/security/hash/Sha384.java:275,
                 from
/home/zopolis4/gcj/libjava/classpath/gnu/java/security/hash/Sha256.java:248,
                 from
/home/zopolis4/gcj/libjava/classpath/gnu/java/security/hash/Sha160.java:239,
                 from
/home/zopolis4/gcj/libjava/classpath/gnu/java/security/hash/RipeMD160.java:289,
                 from
/home/zopolis4/gcj/libjava/classpath/gnu/java/security/hash/RipeMD128.java:255,
                 from
/home/zopolis4/gcj/libjava/classpath/gnu/java/security/hash/MD5.java:369,
                 from
/home/zopolis4/gcj/libjava/classpath/gnu/java/security/hash/MD4.java:336,
                 from
/home/zopolis4/gcj/libjava/classpath/gnu/java/security/hash/MD2.java:255,
                 from
/home/zopolis4/gcj/libjava/classpath/gnu/java/security/hash/IMessageDigest.java:31,
                 from
/home/zopolis4/gcj/libjava/classpath/gnu/java/security/hash/Haval.java:805,
                 from
/home/zopolis4/gcj/libjava/classpath/gnu/java/security/hash/HashFactory.java:1764,
                 from
/home/zopolis4/gcj/libjava/classpath/gnu/java/security/hash/BaseHash.java:437,
                 from <built-in>:575:
/home/zopolis4/gcj/libjava/classpath/gnu/java/security/hash/Whirlpool.java:299:
internal compiler error: Segmentation fault
  299 |     n0 = (in[offset++] & 0xFFL) << 56
      |
0xdac2af crash_signal
        /home/zopolis4/gcjbuild/../gcj/gcc/toplev.cc:314
0x1d10ee7 supportable_widening_operation(vec_info*, tree_code, _stmt_vec_info*,
tree_node*, tree_node*, tree_code*, tree_code*, int*, vec<tree_node*, va_heap,
vl_ptr>*)
        /home/zopolis4/gcjbuild/../gcj/gcc/tree-vect-stmts.cc:12196
0x1d3094c vectorizable_conversion
        /home/zopolis4/gcjbuild/../gcj/gcc/tree-vect-stmts.cc:5064
0x1d31a18 vect_analyze_stmt(vec_info*, _stmt_vec_info*, bool*, _slp_tree*,
_slp_instance*, vec<stmt_info_for_cost, va_heap, vl_ptr>*)
        /home/zopolis4/gcjbuild/../gcj/gcc/tree-vect-stmts.cc:11247
0x106344e vect_slp_analyze_node_operations_1
        /home/zopolis4/gcjbuild/../gcj/gcc/tree-vect-slp.cc:5957
0x106344e vect_slp_analyze_node_operations
        /home/zopolis4/gcjbuild/../gcj/gcc/tree-vect-slp.cc:6147
0x106335a vect_slp_analyze_node_operations
        /home/zopolis4/gcjbuild/../gcj/gcc/tree-vect-slp.cc:6126
0x106335a vect_slp_analyze_node_operations
        /home/zopolis4/gcjbuild/../gcj/gcc/tree-vect-slp.cc:6126
0x106335a vect_slp_analyze_node_operations
        /home/zopolis4/gcjbuild/../gcj/gcc/tree-vect-slp.cc:6126
0x106335a vect_slp_analyze_node_operations
        /home/zopolis4/gcjbuild/../gcj/gcc/tree-vect-slp.cc:6126
0x106335a vect_slp_analyze_node_operations
        /home/zopolis4/gcjbuild/../gcj/gcc/tree-vect-slp.cc:6126
0x106335a vect_slp_analyze_node_operations
        /home/zopolis4/gcjbuild/../gcj/gcc/tree-vect-slp.cc:6126
0x106335a vect_slp_analyze_node_operations
        /home/zopolis4/gcjbuild/../gcj/gcc/tree-vect-slp.cc:6126
0x106335a vect_slp_analyze_node_operations
        /home/zopolis4/gcjbuild/../gcj/gcc/tree-vect-slp.cc:6126
0x106335a vect_slp_analyze_node_operations
        /home/zopolis4/gcjbuild/../gcj/gcc/tree-vect-slp.cc:6126
0x1064f0b vect_slp_analyze_operations(vec_info*)
        /home/zopolis4/gcjbuild/../gcj/gcc/tree-vect-slp.cc:6387
0x106b262 vect_slp_analyze_bb_1
        /home/zopolis4/gcjbuild/../gcj/gcc/tree-vect-slp.cc:7372
0x106b262 vect_slp_region
        /home/zopolis4/gcjbuild/../gcj/gcc/tree-vect-slp.cc:7419
0x106cc2c vect_slp_bbs
        /home/zopolis4/gcjbuild/../gcj/gcc/tree-vect-slp.cc:7610
0x106ceb3 vect_slp_function(function*)
        /home/zopolis4/gcjbuild/../gcj/gcc/tree-vect-slp.cc:7698

[Bug tree-optimization/107672] [13 Regression] ICE during GIMPLE pass: slp

[rguenth at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=107672

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
           Keywords|                            |ice-on-valid-code
                 CC|                            |rguenth at gcc dot gnu.org
   Target Milestone|---                         |13.0

--- Comment #1 from Richard Biener <rguenth at gcc dot gnu.org> ---
  for (i = 0; i < MAX_INTERM_CVT_STEPS; i++)
    {
      intermediate_mode = insn_data[icode1].operand[0].mode;
      if (VECTOR_BOOLEAN_TYPE_P (prev_type))
        intermediate_type
          = vect_halve_mask_nunits (prev_type, intermediate_mode);
      else
        intermediate_type
          = lang_hooks.types.type_for_mode (intermediate_mode,
                                            TYPE_UNSIGNED (prev_type));

      if (VECTOR_BOOLEAN_TYPE_P (intermediate_type)

that fails to check for the case the langhook returns NULL.  In fact,
using build_nonstandard_integer_type is prefered, just use
GET_MODE_PRECISION on the intermediate_mode.

Note that's just guessed from the backtrace reported.  If you can fire up
gdb and see what it actually segfaults on that would be helpful.  Likewise
the above would suggest your type_for_mode needs amemdment.

[Bug tree-optimization/107672] [13 Regression] ICE during GIMPLE pass: slp

[rguenth at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=107672

--- Comment #2 from Richard Biener <rguenth at gcc dot gnu.org> ---
It looks like the code looks for a vector type even ...  I'm going to test

diff --git a/gcc/tree-vect-stmts.cc b/gcc/tree-vect-stmts.cc
index bc0ef136f19..b35b986889d 100644
--- a/gcc/tree-vect-stmts.cc
+++ b/gcc/tree-vect-stmts.cc
@@ -12195,9 +12195,15 @@ supportable_widening_operation (vec_info *vinfo,
        intermediate_type
          = vect_halve_mask_nunits (prev_type, intermediate_mode);
       else
-       intermediate_type
-         = lang_hooks.types.type_for_mode (intermediate_mode,
-                                           TYPE_UNSIGNED (prev_type));
+       {
+         gcc_assert (VECTOR_MODE_P (intermediate_mode));
+         tree intermediate_element_type
+           = lang_hooks.types.type_for_mode (GET_MODE_INNER
(intermediate_mode),
+                                             TYPE_UNSIGNED (prev_type));
+         intermediate_type
+           = build_vector_type_for_mode (intermediate_element_type,
+                                         intermediate_mode);
+       }

       if (VECTOR_BOOLEAN_TYPE_P (intermediate_type)
          && VECTOR_BOOLEAN_TYPE_P (prev_type)

can you check if that helps the Java frontend here?

[Bug tree-optimization/107672] [13 Regression] ICE during GIMPLE pass: slp

[rguenth at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=107672

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
           Assignee|unassigned at gcc dot gnu.org      |rguenth at gcc dot gnu.org
             Status|UNCONFIRMED                 |ASSIGNED
     Ever confirmed|0                           |1
   Last reconfirmed|                            |2022-11-22

[Bug tree-optimization/107672] [13 Regression] ICE during GIMPLE pass: slp

[creatorsmithmdt at gmail dot com]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=107672

--- Comment #3 from Zopolis0 <creatorsmithmdt at gmail dot com> ---
I was in the middle of bisecting it, but that solved the issue perfectly!
Thanks!

I haven't tested it yet on a fresh rebase on upstream, but there's no reason
why it wouldn't work there.

[Bug tree-optimization/107672] [13 Regression] ICE during GIMPLE pass: slp

[cvs-commit at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=107672

--- Comment #4 from CVS Commits <cvs-commit at gcc dot gnu.org> ---
The master branch has been updated by Richard Biener <rguenth@gcc.gnu.org>:

https://gcc.gnu.org/g:09b7993ab6d93716f77c03ff9a9165fa4b579f0d

commit r13-4227-g09b7993ab6d93716f77c03ff9a9165fa4b579f0d
Author: Richard Biener <rguenther@suse.de>
Date:   Tue Nov 22 09:03:53 2022 +0100

    tree-optimization/107672 - avoid vector mode type_for_mode call

    The following avoids using type_for_mode on vector modes which might
    not work for all frontends.  Instead we look for the inner mode
    type and use build_vector_type_for_mode instead.

            PR tree-optimization/107672
            * tree-vect-stmts.cc (supportable_widening_operation): Avoid
            type_for_mode on vector modes.

[Bug tree-optimization/107672] [13 Regression] ICE during GIMPLE pass: slp

[rguenth at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=107672

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

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

--- Comment #5 from Richard Biener <rguenth at gcc dot gnu.org> ---
Fixed then.  Thanks for reporting.