This is the mail archive of the
gcc-bugs@gcc.gnu.org
mailing list for the GCC project.
[Bug target/64617] [5 Regression] ICE: Max. number of generated reload insns per insn is achieved (90) with -ftree-vectorize -mavx512bw -march=slm
- From: "jakub at gcc dot gnu.org" <gcc-bugzilla at gcc dot gnu dot org>
- To: gcc-bugs at gcc dot gnu dot org
- Date: Tue, 20 Jan 2015 16:44:45 +0000
- Subject: [Bug target/64617] [5 Regression] ICE: Max. number of generated reload insns per insn is achieved (90) with -ftree-vectorize -mavx512bw -march=slm
- Auto-submitted: auto-generated
- References: <bug-64617-4 at http dot gcc dot gnu dot org/bugzilla/>
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=64617
Jakub Jelinek <jakub at gcc dot gnu.org> changed:
What |Removed |Added
----------------------------------------------------------------------------
CC| |vmakarov at gcc dot gnu.org
--- Comment #2 from Jakub Jelinek <jakub at gcc dot gnu.org> ---
The patterns in question are
(define_insn "<mask_codefor><avx512>_vec_dup_gpr<mode><mask_name>"
[(set (match_operand:VI12_AVX512VL 0 "register_operand" "=v,v")
(vec_duplicate:VI12_AVX512VL
(match_operand:<ssescalarmode> 1 "nonimmediate_operand" "vm,r")))]
and thus ought to support both HImode MEM as well as GPR.
(insn 60 242 62 4 (set (reg:V32HI 114 [ vect_cst_.38 ])
(vec_duplicate:V32HI (reg/v:HI 169 [ a1 ]))) 4194
{*avx512bw_vec_dup_gprv32hi}
(expr_list:REG_DEAD (reg/v:HI 169 [ a1 ])
(nil)))
(insn 62 60 64 4 (set (reg:V32HI 129 [ vect_cst_.41 ])
(vec_duplicate:V32HI (reg/v:HI 171 [ a2 ]))) 4194
{*avx512bw_vec_dup_gprv32hi}
(expr_list:REG_DEAD (reg/v:HI 171 [ a2 ])
(nil)))
...
Vlad, can you please have a look?