Lines Matching refs:vf
198 poly_uint64 vf;
268 poly_uint64 vf = 1;
319 vf = p->vf;
324 t = build_int_cst (unsigned_type_node, vf);
442 poly_uint64 vf = 1;
449 vf = p->vf;
452 = build_array_type_nelts (TREE_TYPE (TREE_TYPE (decl)), vf);
1020 simduid_to_vf_data->vf = loop_vinfo->vectorization_factor;
1321 /* Fold IFN_GOMP_SIMD_{VF,LANE,LAST_LANE,ORDERED_{START,END}} builtins. */
1423 /* Fold IFN_GOMP_SIMD_{VF,LANE,LAST_LANE,ORDERED_{START,END}} builtins. */
1906 because the main loop's vf is always at least 2x bigger than that
1982 we know that it has a lower cost for at least one runtime VF.
1983 However, we don't know how likely that VF is.
1987 model. E.g. if the estimated VF is also the lowest possible VF,
1989 for the estimated VF, we'd then choose this_loop_vinfo even
1991 other_loop_vinfo for that VF and (b) it would be significantly
1996 estimated other_loop_vinfo VF. For all but trivial loops, this
1998 better than other_loop_vinfo at the estimated VF. */
Indexes created Wed Apr 29 00:23:26 UTC 2026