https://gcc.gnu.org/bugzilla/show_bug.cgi?id=102847
--- Comment #8 from Kewen Lin <linkw at gcc dot gnu.org> ---
(In reply to Richard Biener from comment #5)
> diff --git a/gcc/tree-vect-stmts.c b/gcc/tree-vect-stmts.c
> index 9cbc1af4cc9..8f527452bd0 100644
> --- a/gcc/tree-vect-stmts.c
> +++ b/gcc/tree-vect-stmts.c
> @@ -1191,6 +1191,13 @@ vect_model_load_cost (vec_info *vinfo,
> ncopies * assumed_nunits,
> scalar_load, stmt_info, 0, vect_body);
> }
> + else if (memory_access_type == VMAT_INVARIANT)
> + {
> + /* Invariant loads will ideally be hoisted and splat to a vector. */
> + prologue_cost += record_stmt_cost (cost_vec, 1,
> + scalar_to_vec, stmt_info, 0,
> + vect_prologue);
> + }
> else
> vect_get_load_cost (vinfo, stmt_info, ncopies,
> alignment_support_scheme, misalignment, first_stmt_p,
>
> should fix that in a better way.
Nice, I confirmed this change can fix the issue. Thanks!
For costing, it seems we need one more scalar_load costing for vect_prologue?
or here we aggressively consider scalar_to_vec is also for load_and_splat?
