[Bug tree-optimization/98066] [11 Regression] ICE: Segmentation fault (in gsi_next)

marxin at gcc dot gnu.org gcc-bugzilla@gcc.gnu.org
Mon Nov 30 10:13:25 GMT 2020


https://gcc.gnu.org/bugzilla/show_bug.cgi?id=98066

--- Comment #3 from Martin Liška <marxin at gcc dot gnu.org> ---
Created attachment 49650
  --> https://gcc.gnu.org/bugzilla/attachment.cgi?id=49650&action=edit
Patch candidate

With the patch I see that the target can't expand it to RTL:

$ ./xgcc -B. pr98066.c -mvsx -O1 -c -fdump-tree-optimized=/dev/stdout
...
  <bb 4> [local count: 1073741824]:
  t7.0_1 = t7;
  _2 = __builtin_altivec_vcmpequw_p (a0_4(D), i1_7(D), t7.0_1);
  if (_2 != 0)
...
during RTL pass: expand
pr98066.c: In function ‘qs’:
pr98066.c:7:12: internal compiler error: in rs6000_expand_vector_set, at
config/rs6000/rs6000.c:7024
    7 |     t7[a0] = a0;
      |     ~~~~~~~^~~~
0x72e396 rs6000_expand_vector_set(rtx_def*, rtx_def*, rtx_def*)
        /home/marxin/Programming/gcc2/gcc/config/rs6000/rs6000.c:7024
0x15a1660 ???
        /home/marxin/Programming/gcc2/gcc/config/rs6000/vector.md:1251
0xc4e448 maybe_expand_insn(insn_code, unsigned int, expand_operand*)
        /home/marxin/Programming/gcc2/gcc/optabs.c:7435
0xaf484a expand_vec_set_optab_fn
        /home/marxin/Programming/gcc2/gcc/internal-fn.c:2879
0xaf484a expand_VEC_SET
        /home/marxin/Programming/gcc2/gcc/internal-fn.def:148
0x8deb17 expand_call_stmt
        /home/marxin/Programming/gcc2/gcc/cfgexpand.c:2740
0x8deb17 expand_gimple_stmt_1
        /home/marxin/Programming/gcc2/gcc/cfgexpand.c:3835
0x8deb17 expand_gimple_stmt
        /home/marxin/Programming/gcc2/gcc/cfgexpand.c:3999
0x8e41ca expand_gimple_basic_block
        /home/marxin/Programming/gcc2/gcc/cfgexpand.c:6040
0x8e5bf6 execute
        /home/marxin/Programming/gcc2/gcc/cfgexpand.c:6724
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.

That's something for Xionghu Luo.


More information about the Gcc-bugs mailing list