.lower_bitfield_extract_to_shifts = true,
.use_interpolated_input_intrinsics = true,
.lower_rotate = true,
+ .lower_to_scalar = true,
};
/* we don't want to lower vertex_id to _zero_based on newer gpus: */
.lower_bitfield_extract_to_shifts = true,
.use_interpolated_input_intrinsics = true,
.lower_rotate = true,
+ .vectorize_io = true,
+ .lower_to_scalar = true,
};
const nir_shader_compiler_options *
OPT_V(s, nir_lower_vars_to_ssa);
progress |= OPT(s, nir_opt_copy_prop_vars);
progress |= OPT(s, nir_opt_dead_write_vars);
- progress |= OPT(s, nir_lower_alu_to_scalar, NULL);
+ progress |= OPT(s, nir_lower_alu_to_scalar, NULL, NULL);
progress |= OPT(s, nir_lower_phis_to_scalar);
progress |= OPT(s, nir_copy_prop);
if (key) {
if (s->info.stage == MESA_SHADER_VERTEX) {
- OPT_V(s, nir_lower_clip_vs, key->ucp_enables, false);
+ OPT_V(s, nir_lower_clip_vs, key->ucp_enables, false, false);
if (key->vclamp_color)
OPT_V(s, nir_lower_clamp_color_outputs);
} else if (s->info.stage == MESA_SHADER_FRAGMENT) {
- OPT_V(s, nir_lower_clip_fs, key->ucp_enables);
+ OPT_V(s, nir_lower_clip_fs, key->ucp_enables, false);
if (key->fclamp_color)
OPT_V(s, nir_lower_clamp_color_outputs);
}
if (ubo_progress || idiv_progress)
ir3_optimize_loop(s);
+ /* Do late algebraic optimization to turn add(a, neg(b)) back into
+ * subs, then the mandatory cleanup after algebraic. Note that it may
+ * produce fnegs, and if so then we need to keep running to squash
+ * fneg(fneg(a)).
+ */
+ bool more_late_algebraic = true;
+ while (more_late_algebraic) {
+ more_late_algebraic = OPT(s, nir_opt_algebraic_late);
+ OPT_V(s, nir_opt_constant_folding);
+ OPT_V(s, nir_copy_prop);
+ OPT_V(s, nir_opt_dce);
+ OPT_V(s, nir_opt_cse);
+ }
+
OPT_V(s, nir_remove_dead_variables, nir_var_function_temp);
OPT_V(s, nir_opt_sink, nir_move_const_undef);