[Bug tree-optimization/126160] [x86] missed vectorization for __builtin_lrintf

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=126160

--- Comment #5 from Hongtao Liu <liuhongt at gcc dot gnu.org> ---

> Note there's also vectorizable_conversion which might be a better fit
> (that would also in theory support multi-step conversions).
> 
> Confirmed.  The code generation part for WIDEN isn't implemented.

vectorizable_conversion relies on vec_unpack optab to do the widen conversions,
but we don't have vec_unpack_lrint_{lo, hi}, can't reused existed code
vectorizable_conversion, need to manually do vec_unpack. maybe use
BIT_FIELD_REF and check vec_extract_lo/hi optab support

4361    CASE_CONVERT:                        
4362      c1 = VEC_UNPACK_LO_EXPR;           
4363      c2 = VEC_UNPACK_HI_EXPR;           
4364      break;                             
4365                                         
4366    case FLOAT_EXPR:                     
4367      c1 = VEC_UNPACK_FLOAT_LO_EXPR;     
4368      c2 = VEC_UNPACK_FLOAT_HI_EXPR;     
4369      break;                             
4370                                         
4371    case FIX_TRUNC_EXPR:                 
4372      c1 = VEC_UNPACK_FIX_TRUNC_LO_EXPR; 
4373      c2 = VEC_UNPACK_FIX_TRUNC_HI_EXPR; 
4374      break;