On 21/03/16 22:03, Jason Ekstrand wrote: > On Mon, Mar 21, 2016 at 5:05 AM, Samuel Iglesias Gonsálvez < > sigles...@igalia.com> wrote: > >> From: Connor Abbott <connor.w.abb...@intel.com> >> >> v2 (Sam): >> - Keep using nir_op_imov when calling nir_alu_instr_create() at >> rename_variables_block(). nir_op_fmov is not needed anymore. >> >> Signed-off-by: Samuel Iglesias Gonsálvez <sigles...@igalia.com> >> --- >> src/compiler/nir/nir_lower_vars_to_ssa.c | 3 +++ >> 1 file changed, 3 insertions(+) >> >> diff --git a/src/compiler/nir/nir_lower_vars_to_ssa.c >> b/src/compiler/nir/nir_lower_vars_to_ssa.c >> index 2331791..511662e 100644 >> --- a/src/compiler/nir/nir_lower_vars_to_ssa.c >> +++ b/src/compiler/nir/nir_lower_vars_to_ssa.c >> @@ -543,6 +543,8 @@ get_ssa_def_for_block(struct deref_node *node, >> nir_block *block, >> nir_ssa_undef_instr *undef = >> nir_ssa_undef_instr_create(state->shader, >> glsl_get_vector_elements(node->type)); >> + undef->def.bit_size = >> + glsl_get_bit_size(glsl_get_base_type(node->type)); >> > > Can we instead make nir_ssa_undef_instr_create take a bit size? That seems > better than setting it manually. We probably want to do the same for > load_cons. >
Sure, I will do it for both. Sam > >> nir_instr_insert_before_cf_list(&state->impl->body, &undef->instr); >> def_stack_push(node, &undef->def, state); >> return &undef->def; >> @@ -627,6 +629,7 @@ rename_variables_block(nir_block *block, struct >> lower_variables_state *state) >> nir_ssa_undef_instr *undef = >> nir_ssa_undef_instr_create(state->shader, >> intrin->num_components); >> + undef->def.bit_size = intrin->dest.ssa.bit_size; >> >> nir_instr_insert_before(&intrin->instr, &undef->instr); >> nir_instr_remove(&intrin->instr); >> -- >> 2.5.0 >> >> _______________________________________________ >> mesa-dev mailing list >> mesa-dev@lists.freedesktop.org >> https://lists.freedesktop.org/mailman/listinfo/mesa-dev >> > _______________________________________________ mesa-dev mailing list mesa-dev@lists.freedesktop.org https://lists.freedesktop.org/mailman/listinfo/mesa-dev
