[Bug tree-optimization/106226] New: [13 Regression] ICE in vect_do_peeling, at tree-vect-loop-manip.cc:2702

[Bug tree-optimization/106226] New: [13 Regression] ICE in vect_do_peeling, at tree-vect-loop-manip.cc:2702

[marxin at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=106226

            Bug ID: 106226
           Summary: [13 Regression] ICE in vect_do_peeling, at
                    tree-vect-loop-manip.cc:2702
           Product: gcc
           Version: 13.0
            Status: UNCONFIRMED
          Keywords: ice-on-valid-code
          Severity: normal
          Priority: P3
         Component: tree-optimization
          Assignee: unassigned at gcc dot gnu.org
          Reporter: marxin at gcc dot gnu.org
                CC: rguenth at gcc dot gnu.org
  Target Milestone: ---
              Host: x86_64-linux-gnu
            Target: aarch64-linux-gnu

Still see it with the current master
(g:d89fa97ff318b1f892e2629c5a249313872a01b1)

$ cat x.fppized.f
      SUBROUTINE EFTORD(DM,CHDINT,L4)
      IMPLICIT DOUBLE PRECISION (A-H,O-Z)
      PARAMETER (MXPT=100, MXFRG=50, MXFGPT=MXPT*MXFRG)
      DIMENSION DM(*),CHDINT(L4)
      COMMON /FGRAD / DEF0,DEFT0,TORQ0
     *                ,ATORQ(3,MXFRG)
      COMMON /CSSTV / CX,CY,CZ
     *                EFBTRM(MXFGPT),EFATRM2(MXFGPT),EFBTRM2(MXFGPT),
     *                EFDIP(3,MXFGPT),EFQAD(6,MXFGPT),
     *                EFOCT(10,MXFGPT),FRGNME(MXFGPT)
      IF(NROOTS.EQ.5) CALL ROOT5
      IF(NROOTS.EQ.6) CALL ROOT6
      IF(NROOTS.GE.7) THEN
         CALL ABRT
      END IF
      DO 403 I = 1,IJ
      CHDINT(ICC)=CHDINT(ICC)-DUM*DUMY
      ICC=ICC+1
 403  CONTINUE
      CHDINT(ICC)=CHDINT(ICC)-DUM*DUMY
      DO 550 J=MINJ,MAX
      LJ=LOCJ+J
      IF (LI-LJ) 920,940,940
  920 ID = LJ
      GO TO 960
  940 ID = LI
  960 NN = (ID*(ID-1))/2+JD
      DUM = DM(NN)
      ATORQ(1,INF)=ATORQ(1,INF)-DUM*(CHDINT(ICC+1)*EFDIP(3,IC)
     $           -CHDINT(ICC+2)*EFDIP(2,IC))
      ICC=ICC+1
      ICC=ICC+1
  550 CONTINUE
      END

$ aarch64-linux-gnu-gcc x.fppized.f -O3 -c -march=armv8.2-a
x.fppized.f:23:72:

   23 |       IF (LI-LJ) 920,940,940
      |                                                                       
1
Warning: Fortran 2018 deleted feature: Arithmetic IF statement at (1)
during GIMPLE pass: vect
x.fppized.f:1:23:

    1 |       SUBROUTINE EFTORD(DM,CHDINT,L4)
      |                       ^
internal compiler error: in vect_do_peeling, at tree-vect-loop-manip.cc:2702
0x702052 vect_do_peeling(_loop_vec_info*, tree_node*, tree_node*, tree_node**,
tree_node**, tree_node**, int, bool, bool, tree_node**)
       
/home/marxin/BIG/buildbot/buildworker/marxinbox-gcc-trunk-aarch64/build/gcc/tree-vect-loop-manip.cc:2702
0xf6bb95 vect_transform_loop(_loop_vec_info*, gimple*)
       
/home/marxin/BIG/buildbot/buildworker/marxinbox-gcc-trunk-aarch64/build/gcc/tree-vect-loop.cc:9877
0xf9ddcd vect_transform_loops
       
/home/marxin/BIG/buildbot/buildworker/marxinbox-gcc-trunk-aarch64/build/gcc/tree-vectorizer.cc:1006
0xf9ddcd try_vectorize_loop_1
       
/home/marxin/BIG/buildbot/buildworker/marxinbox-gcc-trunk-aarch64/build/gcc/tree-vectorizer.cc:1136
0xf9ddcd try_vectorize_loop
       
/home/marxin/BIG/buildbot/buildworker/marxinbox-gcc-trunk-aarch64/build/gcc/tree-vectorizer.cc:1165
0xf9e524 execute
       
/home/marxin/BIG/buildbot/buildworker/marxinbox-gcc-trunk-aarch64/build/gcc/tree-vectorizer.cc:1281
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.

==21004== Conditional jump or move depends on uninitialised value(s)
==21004==    at 0xF7630A: vect_do_peeling(_loop_vec_info*, tree_node*,
tree_node*, tree_node**, tree_node**, tree_node**, int, bool, bool,
tree_node**) (tree-vect-loop-manip.cc:2702)
==21004==    by 0xF6BB95: vect_transform_loop(_loop_vec_info*, gimple*)
(tree-vect-loop.cc:9877)
==21004==    by 0xF9DDCD: vect_transform_loops (tree-vectorizer.cc:1006)
==21004==    by 0xF9DDCD: try_vectorize_loop_1 (tree-vectorizer.cc:1136)
==21004==    by 0xF9DDCD: try_vectorize_loop(hash_table<simduid_to_vf, false,
xcallocator>*&, unsigned int*, loop*, function*) (tree-vectorizer.cc:1165)
==21004==    by 0xF9E524: (anonymous
namespace)::pass_vectorize::execute(function*) (tree-vectorizer.cc:1281)
==21004==    by 0xC6DE3A: execute_one_pass(opt_pass*) (passes.cc:2644)
==21004==    by 0xC6E68F: execute_pass_list_1(opt_pass*) (passes.cc:2744)
==21004==    by 0xC6E6A1: execute_pass_list_1(opt_pass*) (passes.cc:2745)
==21004==    by 0xC6E6A1: execute_pass_list_1(opt_pass*) (passes.cc:2745)
==21004==    by 0xC6E6C8: execute_pass_list(function*, opt_pass*)
(passes.cc:2755)
==21004==    by 0x935CA7: expand (cgraphunit.cc:1834)
==21004==    by 0x935CA7: cgraph_node::expand() (cgraphunit.cc:1787)
==21004==    by 0x9371AE: expand_all_functions (cgraphunit.cc:1998)
==21004==    by 0x9371AE: symbol_table::compile() [clone .part.0]
(cgraphunit.cc:2348)
==21004==    by 0x939677: compile (cgraphunit.cc:2261)
==21004==    by 0x939677: symbol_table::finalize_compilation_unit()
(cgraphunit.cc:2529)
==21004== 
during GIMPLE pass: vect
x.fppized.f:1:23:

[Bug tree-optimization/106226] [13 Regression] ICE in vect_do_peeling, at tree-vect-loop-manip.cc:2702

[marxin at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=106226

Martin Liška <marxin at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
   Last reconfirmed|                            |2022-07-07
   Target Milestone|---                         |13.0
     Ever confirmed|0                           |1
             Status|UNCONFIRMED                 |NEW

[Bug tree-optimization/106226] [13 Regression] ICE in vect_do_peeling, at tree-vect-loop-manip.cc:2702

[rguenth at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=106226

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

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

--- Comment #1 from Richard Biener <rguenth at gcc dot gnu.org> ---
I will have a look.

[Bug tree-optimization/106226] [13 Regression] ICE in vect_do_peeling, at tree-vect-loop-manip.cc:2702

[rguenth at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=106226

--- Comment #2 from Richard Biener <rguenth at gcc dot gnu.org> ---
Hm, there's .LOAD_LANES.  OK, so the present spot for updating virtual SSA form
is where we think we need it but that might be when analyzing a totally
different
loop for vectorization.

I'm testing a patch.

[Bug tree-optimization/106226] [13 Regression] ICE in vect_do_peeling, at tree-vect-loop-manip.cc:2702

[cvs-commit at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=106226

--- Comment #3 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:cf3a120084e94614a4917f71940325cd4b537f24

commit r13-1575-gcf3a120084e94614a4917f71940325cd4b537f24
Author: Richard Biener <rguenther@suse.de>
Date:   Fri Jul 8 10:41:59 2022 +0200

    tree-optimization/106226 - move vectorizer virtual SSA update

    When we knowingly have broken virtual SSA form we need to update
    it before we eventually perform slpeel manual updating which will
    call delete_update_ssa.  Currently that's done on-demand but
    communicating whether it's a known unavoidable case is broken
    there.  The following makes that a synchronous operation but
    instead of actually performing the update we instead recod the
    need, clear the update SSA sub-state and force virtual renaming
    at the very end of the vectorization pass.

            PR tree-optimization/106226
            * tree-vect-loop-manip.cc (vect_do_peeling): Assert that
            no SSA update is needed.  Move virtual SSA update ...
            * tree-vectorizer.cc (pass_vectorize::execute): ... here,
            via forced virtual renaming when TODO_update_ssa_only_virtuals
            is queued.
            (vect_transform_loops): Return TODO_update_ssa_only_virtuals
            when virtual SSA update is required.
            (try_vectorize_loop_1): Adjust.
            * tree-vect-stmts.cc (vectorizable_simd_clone_call): Allow
            virtual renaming if the ABI forces an aggregate return
            but the original call did not have a virtual definition.

            * gfortran.dg/pr106226.f: New testcase.

[Bug tree-optimization/106226] [13 Regression] ICE in vect_do_peeling, at tree-vect-loop-manip.cc:2702

[rguenth at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=106226

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

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

--- Comment #4 from Richard Biener <rguenth at gcc dot gnu.org> ---
Fixed.