https://gcc.gnu.org/bugzilla/show_bug.cgi?id=121699
Hongtao Liu <liuhongt at gcc dot gnu.org> changed:
What |Removed |Added
----------------------------------------------------------------------------
CC| |liuhongt at gcc dot gnu.org
--- Comment #3 from Hongtao Liu <liuhongt at gcc dot gnu.org> ---
It hits
27004; not generated by vectorizer?
27005(define_expand "cond_<insn><mode>"
27006 [(set (match_operand:VI1_AVX512VL 0 "register_operand")
27007 (vec_merge:VI1_AVX512VL
27008 (any_shift:VI1_AVX512VL
27009 (match_operand:VI1_AVX512VL 2 "register_operand")
27010 (match_operand:VI1_AVX512VL 3
"nonimmediate_or_const_vec_dup_operand"))
27011 (match_operand:VI1_AVX512VL 4 "nonimm_or_0_operand")
27012 (match_operand:<avx512fmaskmode> 1 "register_operand")))]
27013 "TARGET_GFNI && TARGET_AVX512F"
27014{
27015 rtx matrix = ix86_vgf2p8affine_shift_matrix (operands[0], operands[2],
<CODE>);
27016 emit_insn (gen_vgf2p8affineqb_<mode>_mask (operands[0], operands[1],
matrix,
27017 const0_rtx, operands[4],
27018 operands[1]));
27019 DONE;
27020})
(match_operand:VI1_AVX512VL 3 "nonimmediate_or_const_vec_dup_operand") it's
better to be const_vec_dup_operand.
