[Bug target/108484] New: [13 Regression] ICE building glibc for ia64 in cselib_subst_to_values, at cselib.cc:2148

["jsm28 at gcc dot gnu.org" <gcc-bugzilla@gcc.gnu.org>]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=108484

            Bug ID: 108484
           Summary: [13 Regression] ICE building glibc for ia64 in
                    cselib_subst_to_values, at cselib.cc:2148
           Product: gcc
           Version: 13.0
            Status: UNCONFIRMED
          Keywords: ice-on-valid-code
          Severity: normal
          Priority: P3
         Component: target
          Assignee: unassigned at gcc dot gnu.org
          Reporter: jsm28 at gcc dot gnu.org
                CC: aoliva at gcc dot gnu.org
  Target Milestone: ---
            Target: ia64*-*-*

Created attachment 54319
  --> https://gcc.gnu.org/bugzilla/attachment.cgi?id=54319&action=edit
preprocessed source

Compile the attached file (from glibc) with -g -O2 for ia64-linux-gnu. This
produces the following ICE:

during RTL pass: mach
In file included from ../sysdeps/ia64/unwind-resume.c:18:
../sysdeps/generic/unwind-resume.c: In function '_Unwind_Resume':
../sysdeps/generic/unwind-resume.c:38:1: internal compiler error: in
cselib_subst_to_values, at cselib.cc:2148
0x5fd409 cselib_subst_to_values(rtx_def*, machine_mode)
        /scratch/jmyers/glibc/many13/src/gcc/gcc/cselib.cc:2148
0x98c873 cselib_subst_to_values(rtx_def*, machine_mode)
        /scratch/jmyers/glibc/many13/src/gcc/gcc/cselib.cc:2193
0x98f761 cselib_subst_to_values_from_insn(rtx_def*, machine_mode, rtx_insn*)
        /scratch/jmyers/glibc/many13/src/gcc/gcc/cselib.cc:2264
0x18172aa add_insn_mem_dependence
        /scratch/jmyers/glibc/many13/src/gcc/gcc/sched-deps.cc:1741
0x181b230 sched_analyze_2
        /scratch/jmyers/glibc/many13/src/gcc/gcc/sched-deps.cc:2688
0x181b267 sched_analyze_2
        /scratch/jmyers/glibc/many13/src/gcc/gcc/sched-deps.cc:2806
0x181b9eb sched_analyze_insn
        /scratch/jmyers/glibc/many13/src/gcc/gcc/sched-deps.cc:2960
0x181ddb0 deps_analyze_insn(deps_desc*, rtx_insn*)
        /scratch/jmyers/glibc/many13/src/gcc/gcc/sched-deps.cc:3683
0x181e40e sched_analyze(deps_desc*, rtx_insn*, rtx_insn*)
        /scratch/jmyers/glibc/many13/src/gcc/gcc/sched-deps.cc:3836
0x1821870 schedule_ebb(rtx_insn*, rtx_insn*, bool)
        /scratch/jmyers/glibc/many13/src/gcc/gcc/sched-ebb.cc:505
0x1821da2 schedule_ebbs()
        /scratch/jmyers/glibc/many13/src/gcc/gcc/sched-ebb.cc:655
0x11b3346 ia64_reorg
        /scratch/jmyers/glibc/many13/src/gcc/gcc/config/ia64/ia64.cc:9862
0xdb527d execute
        /scratch/jmyers/glibc/many13/src/gcc/gcc/reorg.cc:3927
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.

Introduced by g:3c99493bf39a7fef9213e6f5af94b78bb15fcfdc

commit 3c99493bf39a7fef9213e6f5af94b78bb15fcfdc
Author: Alexandre Oliva <oliva@adacore.com>
Date:   Thu Jan 19 01:09:15 2023 -0300

    [PR106746] drop cselib addr lookup in debug insn mem

    The testcase used to get scheduled differently depending on the
    presence of debug insns with MEMs.  It's not clear to me why those
    MEMs affected scheduling, but the cselib pre-canonicalization of the
    MEM address is not used at all when analyzing debug insns, so the
    memory allocation and lookup are pure waste.  Somehow, avoiding that
    waste fixes the problem, or makes it go latent.


    for  gcc/ChangeLog

            PR debug/106746
            * sched-deps.cc (sched_analyze_2): Skip cselib address lookup
            within debug insns.

    for  gcc/testsuite/ChangeLog

            PR debug/106746
            * gcc.target/i386/pr106746.c: New.