[Bug tree-optimization/114792] New: ICE on valid code at -O1 with "-fno-tree-ccp -fno-tree-copy-prop" on x86_64-linux-gnu: in get_loop_body, at cfgloop.cc:903

[Bug tree-optimization/114792] New: ICE on valid code at -O1 with "-fno-tree-ccp -fno-tree-copy-prop" on x86_64-linux-gnu: in get_loop_body, at cfgloop.cc:903

[zhendong.su at inf dot ethz.ch]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=114792

            Bug ID: 114792
           Summary: ICE on valid code at -O1 with "-fno-tree-ccp
                    -fno-tree-copy-prop" on x86_64-linux-gnu: in
                    get_loop_body, at cfgloop.cc:903
           Product: gcc
           Version: unknown
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: tree-optimization
          Assignee: unassigned at gcc dot gnu.org
          Reporter: zhendong.su at inf dot ethz.ch
  Target Milestone: ---

It appears to be a recent regression and does not reproduce with 13.2. 

Compiler Explorer: https://godbolt.org/z/zzdjrb599


[522] % gcctk -v
Using built-in specs.
COLLECT_GCC=gcctk
COLLECT_LTO_WRAPPER=/local/suz-local/software/local/gcc-trunk/libexec/gcc/x86_64-pc-linux-gnu/14.0.1/lto-wrapper
Target: x86_64-pc-linux-gnu
Configured with: ../gcc-trunk/configure --disable-bootstrap
--enable-checking=yes --prefix=/local/suz-local/software/local/gcc-trunk
--enable-sanitizers --enable-languages=c,c++ --disable-werror --enable-multilib
Thread model: posix
Supported LTO compression algorithms: zlib
gcc version 14.0.1 20240421 (experimental) (GCC) 
[523] % 
[523] % gcctk -O1 -fno-tree-ccp -fno-tree-copy-prop small.c
during GIMPLE pass: ch
small.c: In function ‘main’:
small.c:2:5: internal compiler error: in get_loop_body, at cfgloop.cc:903
    2 | int main() {
      |     ^~~~
0x7926fd get_loop_body(loop const*)
        ../../gcc-trunk/gcc/cfgloop.cc:903
0xbfbce9 unloop(loop*, bool*, bitmap_head*)
        ../../gcc-trunk/gcc/cfgloopmanip.cc:1014
0x12ee996 unloop_loops(vec<loop*, va_heap, vl_ptr>&, vec<int, va_heap,
vl_ptr>&, vec<edge_def*, va_heap, vl_ptr>&, bitmap_head*, bool*)
        ../../gcc-trunk/gcc/tree-ssa-loop-ivcanon.cc:688
0x12dfc84 copy_headers
        ../../gcc-trunk/gcc/tree-ssa-loop-ch.cc:1158
0x12e02da execute
        ../../gcc-trunk/gcc/tree-ssa-loop-ch.cc:1179
Please submit a full bug report, with preprocessed source (by using
-freport-bug).
Please include the complete backtrace with any bug report.
See <https://gcc.gnu.org/bugs/> for instructions.
[524] % 
[524] % cat small.c
int a;
int main() {
  int b = -1, c;
  for (; b >= 0; b++) {
    for (c = 0; c; c++) {
    L:
      while (a)
        if (a)
          goto L;
    }
  }
  return 0;
}

[Bug tree-optimization/114792] ICE on valid code at -O1 with "-fno-tree-ccp -fno-tree-copy-prop" on x86_64-linux-gnu: in get_loop_body, at cfgloop.cc:903

[hjl.tools at gmail dot com]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=114792

H.J. Lu <hjl.tools at gmail dot com> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
   Last reconfirmed|                            |2024-04-21
                 CC|                            |jh at suse dot cz
     Ever confirmed|0                           |1
            Version|unknown                     |14.0
             Status|UNCONFIRMED                 |NEW

--- Comment #1 from H.J. Lu <hjl.tools at gmail dot com> ---
It is caused by r14-301.

[Bug tree-optimization/114792] [14 Regression] ICE on valid code at -O1 with "-fno-tree-ccp -fno-tree-copy-prop" on x86_64-linux-gnu: in get_loop_body, at cfgloop.cc:903

[pinskia at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=114792

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |pinskia at gcc dot gnu.org
   Target Milestone|---                         |14.0

[Bug tree-optimization/114792] [14 Regression] ICE on valid code at -O1 with "-fno-tree-ccp -fno-tree-copy-prop" on x86_64-linux-gnu: in get_loop_body, at cfgloop.cc:903

[pinskia at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=114792

--- Comment #2 from Andrew Pinski <pinskia at gcc dot gnu.org> ---
Confirmed.

[Bug tree-optimization/114792] [14 Regression] ICE on valid code at -O1 with "-fno-tree-ccp -fno-tree-copy-prop" on x86_64-linux-gnu: in get_loop_body, at cfgloop.cc:903

[jakub at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=114792

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |jakub at gcc dot gnu.org
           Priority|P3                          |P1

[Bug tree-optimization/114792] [14 Regression] ICE on valid code at -O1 with "-fno-tree-ccp -fno-tree-copy-prop" on x86_64-linux-gnu: in get_loop_body, at cfgloop.cc:903

[rguenth at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=114792

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

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

--- Comment #3 from Richard Biener <rguenth at gcc dot gnu.org> ---
The issue is that unloop_loops doesn't work when to unloop loops are nested. 
The unroller takes great care of working on a single loop level at a time,
cleaning up the CFG and fixing loops inbetween but we fail to ensure this here.

unlooping in CH is somewhat a wrong tool for the wrong task as it doesn't
adhere to constraints present.  There's the possibility this can be salvaged by
making unloop_loops more robust, in particular computing loop bodies (Which is
fragile) up-front.

But it would be much nicer if unlooping would simply add __builtin_unreachable
() as the first stmt into the latch and leave the rest to CFG cleanup.

[Bug tree-optimization/114792] [14 Regression] ICE on valid code at -O1 with "-fno-tree-ccp -fno-tree-copy-prop" on x86_64-linux-gnu: in get_loop_body, at cfgloop.cc:903

[rguenth at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=114792

--- Comment #4 from Richard Biener <rguenth at gcc dot gnu.org> ---
Created attachment 58024
  --> https://gcc.gnu.org/bugzilla/attachment.cgi?id=58024&action=edit
patch

I quickly tried this which works for the testcase but I'm sure it'll break
quickly.

[Bug tree-optimization/114792] [14 Regression] ICE on valid code at -O1 with "-fno-tree-ccp -fno-tree-copy-prop" on x86_64-linux-gnu: in get_loop_body, at cfgloop.cc:903

[rguenth at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=114792

--- Comment #5 from Richard Biener <rguenth at gcc dot gnu.org> ---
(In reply to Richard Biener from comment #4)
> Created attachment 58024 [details]
> patch
> 
> I quickly tried this which works for the testcase but I'm sure it'll break
> quickly.

during GIMPLE pass: cunrolli
../../../../gcc/libgcc/soft-fp/floatsitf.c: In function '__floatsitf':
../../../../gcc/libgcc/soft-fp/floatsitf.c:34:1: internal compiler error: in
unloop_loops, at tree-ssa-loop-ivcanon.cc:721
   34 | __floatsitf (SItype i)
      | ^~~~~~~~~~~
0x76cde9 _start
        ../sysdeps/x86_64/start.S:120
Please submit a full bug report, with preprocessed source (by using
-freport-bug).

[Bug tree-optimization/114792] [14 Regression] ICE on valid code at -O1 with "-fno-tree-ccp -fno-tree-copy-prop" on x86_64-linux-gnu: in get_loop_body, at cfgloop.cc:903

[rguenth at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=114792

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|NEW                         |ASSIGNED

[Bug tree-optimization/114792] [14 Regression] ICE on valid code at -O1 with "-fno-tree-ccp -fno-tree-copy-prop" on x86_64-linux-gnu: in get_loop_body, at cfgloop.cc:903

[rguenth at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=114792

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
  Attachment #58024|0                           |1
        is obsolete|                            |

--- Comment #6 from Richard Biener <rguenth at gcc dot gnu.org> ---
Created attachment 58033
  --> https://gcc.gnu.org/bugzilla/attachment.cgi?id=58033&action=edit
patch

I'm now testing this which is more minimal surgery at least making sure we
process loops in similar order as unrolling does.  I'm not 100% convinced this
is enough but it restores part of the consistency (unroll also doesn't unloop
an outer loop before cleaning up the CFG and rebuilding loops after unlooping
an inner loop).

[Bug tree-optimization/114792] [14 Regression] ICE on valid code at -O1 with "-fno-tree-ccp -fno-tree-copy-prop" on x86_64-linux-gnu: in get_loop_body, at cfgloop.cc:903

[cvs-commit at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=114792

--- Comment #7 from GCC 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:59ff81835fee22a9d4c9a481a4d1814583aae945

commit r14-10120-g59ff81835fee22a9d4c9a481a4d1814583aae945
Author: Richard Biener <rguenther@suse.de>
Date:   Thu Apr 25 08:08:24 2024 +0200

    tree-optimization/114792 - order loops to unloops in CH

    When we use unloop_loops we have to make sure to have loops ordered
    inner to outer as otherwise we can wreck inner loop structure where
    unlooping relies on that being intact.  The following re-sorts the
    vector of to unloop loops after copy-header as that adds to the
    vector in two places and the wrong order.

            PR tree-optimization/114792
            * tree-ssa-loop-ch.cc (ch_order_loops): New function.
            (ch_base::copy_headers): Sort loops to unloop inner-to-outer.

            * gcc.dg/torture/pr114792.c: New testcase.

[Bug tree-optimization/114792] [14 Regression] ICE on valid code at -O1 with "-fno-tree-ccp -fno-tree-copy-prop" on x86_64-linux-gnu: in get_loop_body, at cfgloop.cc:903

[rguenth at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=114792

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

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

--- Comment #8 from Richard Biener <rguenth at gcc dot gnu.org> ---
Fixed.