On Tue, 12 Dec 2023, Richard Sandiford wrote:
> Richard Biener <rguent...@suse.de> writes:
> > On Mon, 11 Dec 2023, Tamar Christina wrote:
> >> @@ -5553,6 +5554,83 @@ integer_type_for_mask (tree var, vec_info *vinfo)
> >> return build_nonstandard_integer_type (def_stmt_info->mask_precision,
> >> 1);
> >> }
> >>
> >> +/* Function vect_recog_gcond_pattern
> >> +
> >> + Try to find pattern like following:
> >> +
> >> + if (a op b)
> >> +
> >> + where operator 'op' is not != and convert it to an adjusted boolean
> >> pattern
> >> +
> >> + mask = a op b
> >> + if (mask != 0)
> >> +
> >> + and set the mask type on MASK.
> >> +
> >> + Input:
> >> +
> >> + * STMT_VINFO: The stmt at the end from which the pattern
> >> + search begins, i.e. cast of a bool to
> >> + an integer type.
> >> +
> >> + Output:
> >> +
> >> + * TYPE_OUT: The type of the output of this pattern.
> >> +
> >> + * Return value: A new stmt that will be used to replace the pattern.
> >> */
> >> +
> >> +static gimple *
> >> +vect_recog_gcond_pattern (vec_info *vinfo,
> >> + stmt_vec_info stmt_vinfo, tree *type_out)
> >> +{
> >> + gimple *last_stmt = STMT_VINFO_STMT (stmt_vinfo);
> >> + gcond* cond = NULL;
> >> + if (!(cond = dyn_cast <gcond *> (last_stmt)))
> >> + return NULL;
> >> +
> >> + auto lhs = gimple_cond_lhs (cond);
> >> + auto rhs = gimple_cond_rhs (cond);
> >> + auto code = gimple_cond_code (cond);
> >> +
> >> + tree scalar_type = TREE_TYPE (lhs);
> >> + if (VECTOR_TYPE_P (scalar_type))
> >> + return NULL;
> >> +
> >> + if (code == NE_EXPR && zerop (rhs))
> >
> > I think you need && VECT_SCALAR_BOOLEAN_TYPE_P (scalar_type) here,
> > an integer != 0 would not be an appropriate mask. I guess two
> > relevant testcases would have an early exit like
> >
> > if (here[i] != 0)
> > break;
> >
> > once with a 'bool here[]' and once with a 'int here[]'.
> >
> >> + return NULL;
> >> +
> >> + tree vecitype = get_vectype_for_scalar_type (vinfo, scalar_type);
> >> + if (vecitype == NULL_TREE)
> >> + return NULL;
> >> +
> >> + /* Build a scalar type for the boolean result that when vectorized
> >> matches the
> >> + vector type of the result in size and number of elements. */
> >> + unsigned prec
> >> + = vector_element_size (tree_to_poly_uint64 (TYPE_SIZE (vecitype)),
> >> + TYPE_VECTOR_SUBPARTS (vecitype));
> >> +
> >> + scalar_type
> >> + = build_nonstandard_integer_type (prec, TYPE_UNSIGNED (scalar_type));
> >> +
> >> + vecitype = get_vectype_for_scalar_type (vinfo, scalar_type);
> >> + if (vecitype == NULL_TREE)
> >> + return NULL;
> >> +
> >> + tree vectype = truth_type_for (vecitype);
> >
> > That looks awfully complicated. I guess one complication is that
> > we compute mask_precision & friends before this pattern gets
> > recognized. See vect_determine_mask_precision and its handling
> > of tcc_comparison, see also integer_type_for_mask. For comparisons
> > properly handled during pattern recog the vector type is determined
> > in vect_get_vector_types_for_stmt via
> >
> > else if (vect_use_mask_type_p (stmt_info))
> > {
> > unsigned int precision = stmt_info->mask_precision;
> > scalar_type = build_nonstandard_integer_type (precision, 1);
> > vectype = get_mask_type_for_scalar_type (vinfo, scalar_type,
> > group_size);
> > if (!vectype)
> > return opt_result::failure_at (stmt, "not vectorized: unsupported"
> > " data-type %T\n", scalar_type);
> >
> > Richard, do you have any advice here? I suppose vect_determine_precisions
> > needs to handle the gcond case with bool != 0 somehow and for the
> > extra mask producer we add here we have to emulate what it would have
> > done, right?
>
> How about handling gconds directly in vect_determine_mask_precision?
> In a sense it's not needed, since gconds are always roots, and so we
> could calculate their precision on the fly instead. But handling it in
> vect_determine_mask_precision feels like it should reduce the number
> of special cases.
Yeah, that sounds worth trying.
Richard.
