[Bug tree-optimization/103321] New: [12 regression] ICE at tree-ssa.c:1211 after r12-5300

[Bug tree-optimization/103321] New: [12 regression] ICE at tree-ssa.c:1211 after r12-5300

[seurer at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=103321

            Bug ID: 103321
           Summary: [12 regression] ICE at tree-ssa.c:1211 after r12-5300
           Product: gcc
           Version: 12.0
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: tree-optimization
          Assignee: unassigned at gcc dot gnu.org
          Reporter: seurer at gcc dot gnu.org
  Target Milestone: ---

g:f98f373dd822b35c52356b753d528924e9f89678, r12-5300 

I am seeing an ICE when building the spec2017 test case 445.gobmk

/home/seurer/gcc/git/install/gcc-test/bin/gcc -c -o engine/value_moves.o
-DSPEC_CPU -DNDEBUG -DHAVE_CONFIG_H -I. -I.. -I../include -I./include  -m64 -O3
-mcpu=power8 -ffast-math -funroll-loops -fpeel-loops -fvect-cost-model
-mpopcntd -mrecip=rsqrt                 engine/value_moves.c
engine/value_moves.c: In function 'value_move_reasons':
engine/value_moves.c:3030:1: error: definition in block 38 does not dominate
use in block 37
 3030 | }
      | ^
for SSA_NAME: iftmp.54_199 in statement:
iftmp.54_143 = PHI <iftmp.54_199(37), _89(38)>
PHI argument
iftmp.54_199
for PHI node
iftmp.54_143 = PHI <iftmp.54_199(37), _89(38)>
during GIMPLE pass: phiopt
engine/value_moves.c:3030:1: internal compiler error: verify_ssa failed
0x10f15523 verify_ssa(bool, bool)
        /home/seurer/gcc/git/gcc-test/gcc/tree-ssa.c:1211
0x10a6a65f execute_function_todo
        /home/seurer/gcc/git/gcc-test/gcc/passes.c:2049
0x10a6b8eb do_per_function
        /home/seurer/gcc/git/gcc-test/gcc/passes.c:1687
0x10a6bb0b execute_todo
        /home/seurer/gcc/git/gcc-test/gcc/passes.c:2096
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.
specmake: *** [engine/value_moves.o] Error 1
Error with make 'specmake -j 1 build': check file
'/home/seurer/gcc/cpu2006/benchspec/CPU2006/445.gobmk/build/build_base_ppc64.0000/make.err'
  Command returned exit code 2
  Error with make!
*** Error building 445.gobmk


commit f98f373dd822b35c52356b753d528924e9f89678
Author: Richard Biener <rguenther@suse.de>
Date:   Mon Nov 15 15:19:36 2021 +0100

    tree-optimization/102880 - make PHI-OPT recognize more CFGs

[Bug tree-optimization/103321] [12 regression] ICE at tree-ssa.c:1211 after r12-5300

[pinskia at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=103321

--- Comment #1 from Andrew Pinski <pinskia at gcc dot gnu.org> ---
https://gcc.gnu.org/pipermail/gcc-cvs/2021-November/356905.html

[Bug tree-optimization/103321] [12 regression] ICE at tree-ssa.c:1211 after r12-5300

[pinskia at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=103321

--- Comment #2 from Andrew Pinski <pinskia at gcc dot gnu.org> ---
I don't have SPEC CPU 2006 setup, can you attach the preprocessed source?

[Bug tree-optimization/103321] [12 regression] ICE at tree-ssa.c:1211 after r12-5300

[pinskia at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=103321

--- Comment #3 from Andrew Pinski <pinskia at gcc dot gnu.org> ---
(In reply to Andrew Pinski from comment #2)
> I don't have SPEC CPU 2006 setup, can you attach the preprocessed source?

Also can you test if https://gcc.gnu.org/bugzilla/attachment.cgi?id=51834 fixes
the problem?

[Bug tree-optimization/103321] [12 regression] ICE at tree-ssa.c:1211 after r12-5300

[vsevolod.livinskij at frtk dot ru]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=103321

Vsevolod Livinskiy <vsevolod.livinskij at frtk dot ru> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |vsevolod.livinskij at frtk dot ru

--- Comment #4 from Vsevolod Livinskiy <vsevolod.livinskij at frtk dot ru> ---
The same error can be triggered for X86.
Here is the link to the Compiler Explorer: https://godbolt.org/z/8zMhfhs5o

This reproducer was found with YARPGen.

Reproducer:
long a;
short b;
extern unsigned long d[];
const unsigned long long &e(const unsigned long long &f,
                            const unsigned long long &g) {
  if (f < g)
    return g;
  return f;
}
const unsigned long long &h(const unsigned long long &f,
                            const unsigned long long &g) {
  return f < g ? f : g;
}
void j() {
  for (int i = 0;; i = 8)
    d[i] = h(e(a, c[i] && b), 7);
}

Error:
>$ g++ -O2 -c func.cpp
func.cpp: In function 'void j()':
func.cpp:15:6: error: definition in block 5 does not dominate use in block 3
   15 | void j() {
      |      ^
for SSA_NAME: _23 in statement:
_6 = PHI <_4(5), _23(3), _23(4)>
PHI argument
_23
for PHI node
_6 = PHI <_4(5), _23(3), _23(4)>
during GIMPLE pass: phiopt
func.cpp:15:6: internal compiler error: verify_ssa failed
0x141284f verify_ssa(bool, bool)
        /testing/gcc/gcc_src_master/gcc/tree-ssa.c:1211
0x10e86d5 execute_function_todo
        /testing/gcc/gcc_src_master/gcc/passes.c:2049
0x10e901b execute_todo
        /testing/gcc/gcc_src_master/gcc/passes.c:2096
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.

gcc version 12.0.0 20211118 (d6ec661e3931773e2f571ed4f6dd8b0402d8687d) (GCC)

[Bug tree-optimization/103321] [12 regression] ICE at tree-ssa.c:1211 after r12-5300

[pinskia at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=103321

--- Comment #5 from Andrew Pinski <pinskia at gcc dot gnu.org> ---
(In reply to Vsevolod Livinskiy from comment #4)
> The same error can be triggered for X86.
> Here is the link to the Compiler Explorer: https://godbolt.org/z/8zMhfhs5o
> 
> This reproducer was found with YARPGen.

This one is specifically the same as PR 103317 where I submitted a patch:
https://gcc.gnu.org/pipermail/gcc-patches/2021-November/584926.html

I don't know if the original testcase is the same though.

[Bug tree-optimization/103321] [12 regression] ICE at tree-ssa.c:1211 after r12-5300

[pinskia at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=103321

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
   Last reconfirmed|                            |2021-11-19
             Status|UNCONFIRMED                 |WAITING
     Ever confirmed|0                           |1
           Keywords|                            |ice-on-valid-code
   Target Milestone|---                         |12.0

--- Comment #6 from Andrew Pinski <pinskia at gcc dot gnu.org> ---
Can you test this after r12-5393 ?

[Bug tree-optimization/103321] [12 regression] ICE at tree-ssa.c:1211 after r12-5300

[marxin at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=103321

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

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

--- Comment #7 from Martin Liška <marxin at gcc dot gnu.org> ---
Fixed on master.

[Bug tree-optimization/103321] [12 regression] ICE at tree-ssa.c:1211 after r12-5300

[pinskia at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=103321

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
         Resolution|FIXED                       |DUPLICATE

--- Comment #8 from Andrew Pinski <pinskia at gcc dot gnu.org> ---
(In reply to Martin Liška from comment #7)
> Fixed on master.

So marking as a dup of bug 103317.

*** This bug has been marked as a duplicate of bug 103317 ***