};
static bool
-nir_alu_srcs_equal(nir_alu_src src1, nir_alu_src src2, uint8_t read_mask)
+nir_alu_srcs_equal(nir_alu_instr *alu1, nir_alu_instr *alu2, unsigned src1,
+ unsigned src2)
{
- if (src1.abs != src2.abs || src1.negate != src2.negate)
+ if (alu1->src[src1].abs != alu2->src[src2].abs ||
+ alu1->src[src1].negate != alu2->src[src2].negate)
return false;
- for (int i = 0; i < 4; ++i) {
- if (!(read_mask & (1 << i)))
- continue;
-
- if (src1.swizzle[i] != src2.swizzle[i])
+ for (unsigned i = 0; i < nir_ssa_alu_instr_src_components(alu1, src1); i++) {
+ if (alu1->src[src1].swizzle[i] != alu2->src[src2].swizzle[i])
return false;
}
- return nir_srcs_equal(src1.src, src2.src);
+ return nir_srcs_equal(alu1->src[src1].src, alu2->src[src2].src);
}
static bool
if (alu1->op != alu2->op)
return false;
- if (alu1->has_predicate != alu2->has_predicate)
- return false;
-
- if (alu1->has_predicate &&
- !nir_srcs_equal(alu1->predicate, alu2->predicate))
- return false;
-
/* TODO: We can probably acutally do something more inteligent such
* as allowing different numbers and taking a maximum or something
* here */
if (alu1->dest.dest.ssa.num_components != alu2->dest.dest.ssa.num_components)
return false;
- for (unsigned i = 0; i < nir_op_infos[alu1->op].num_inputs; i++) {
- if (!nir_alu_srcs_equal(alu1->src[i], alu2->src[i],
- (1 << alu1->dest.dest.ssa.num_components) - 1))
+ if (nir_op_infos[alu1->op].algebraic_properties & NIR_OP_IS_COMMUTATIVE) {
+ assert(nir_op_infos[alu1->op].num_inputs == 2);
+ return (nir_alu_srcs_equal(alu1, alu2, 0, 0) &&
+ nir_alu_srcs_equal(alu1, alu2, 1, 1)) ||
+ (nir_alu_srcs_equal(alu1, alu2, 0, 1) &&
+ nir_alu_srcs_equal(alu1, alu2, 1, 0));
+ } else {
+ for (unsigned i = 0; i < nir_op_infos[alu1->op].num_inputs; i++) {
+ if (!nir_alu_srcs_equal(alu1, alu2, i, i))
+ return false;
+ }
+ }
+ return true;
+ }
+ case nir_instr_type_tex: {
+ nir_tex_instr *tex1 = nir_instr_as_tex(instr1);
+ nir_tex_instr *tex2 = nir_instr_as_tex(instr2);
+
+ if (tex1->op != tex2->op)
+ return false;
+
+ if (tex1->num_srcs != tex2->num_srcs)
+ return false;
+ for (unsigned i = 0; i < tex1->num_srcs; i++) {
+ if (tex1->src[i].src_type != tex2->src[i].src_type ||
+ !nir_srcs_equal(tex1->src[i].src, tex2->src[i].src)) {
return false;
+ }
+ }
+
+ if (tex1->coord_components != tex2->coord_components ||
+ tex1->sampler_dim != tex2->sampler_dim ||
+ tex1->is_array != tex2->is_array ||
+ tex1->is_shadow != tex2->is_shadow ||
+ tex1->is_new_style_shadow != tex2->is_new_style_shadow ||
+ memcmp(tex1->const_offset, tex2->const_offset,
+ sizeof(tex1->const_offset)) != 0 ||
+ tex1->component != tex2->component ||
+ tex1->sampler_index != tex2->sampler_index ||
+ tex1->sampler_array_size != tex2->sampler_array_size) {
+ return false;
}
+
+ /* Don't support un-lowered sampler derefs currently. */
+ if (tex1->sampler || tex2->sampler)
+ return false;
+
return true;
}
- case nir_instr_type_texture:
- return false;
case nir_instr_type_load_const: {
nir_load_const_instr *load1 = nir_instr_as_load_const(instr1);
nir_load_const_instr *load2 = nir_instr_as_load_const(instr2);
- if (load1->num_components != load2->num_components)
+ if (load1->def.num_components != load2->def.num_components)
return false;
return memcmp(load1->value.f, load2->value.f,
- load1->num_components * sizeof load2->value.f) == 0;
+ load1->def.num_components * sizeof(*load2->value.f)) == 0;
}
case nir_instr_type_phi: {
nir_phi_instr *phi1 = nir_instr_as_phi(instr1);
if (phi1->instr.block != phi2->instr.block)
return false;
- foreach_list_typed(nir_phi_src, src1, node, &phi1->srcs) {
- foreach_list_typed(nir_phi_src, src2, node, &phi2->srcs) {
+ nir_foreach_phi_src(phi1, src1) {
+ nir_foreach_phi_src(phi2, src2) {
if (src1->pred == src2->pred) {
if (!nir_srcs_equal(src1->src, src2->src))
return false;
return true;
}
- case nir_instr_type_intrinsic:
+ case nir_instr_type_intrinsic: {
+ nir_intrinsic_instr *intrinsic1 = nir_instr_as_intrinsic(instr1);
+ nir_intrinsic_instr *intrinsic2 = nir_instr_as_intrinsic(instr2);
+ const nir_intrinsic_info *info =
+ &nir_intrinsic_infos[intrinsic1->intrinsic];
+
+ if (intrinsic1->intrinsic != intrinsic2->intrinsic ||
+ intrinsic1->num_components != intrinsic2->num_components)
+ return false;
+
+ if (info->has_dest && intrinsic1->dest.ssa.num_components !=
+ intrinsic2->dest.ssa.num_components)
+ return false;
+
+ for (unsigned i = 0; i < info->num_srcs; i++) {
+ if (!nir_srcs_equal(intrinsic1->src[i], intrinsic2->src[i]))
+ return false;
+ }
+
+ assert(info->num_variables == 0);
+
+ for (unsigned i = 0; i < info->num_indices; i++) {
+ if (intrinsic1->const_index[i] != intrinsic2->const_index[i])
+ return false;
+ }
+
+ return true;
+ }
case nir_instr_type_call:
case nir_instr_type_jump:
case nir_instr_type_ssa_undef:
static bool
src_is_ssa(nir_src *src, void *data)
{
+ (void) data;
return src->is_ssa;
}
static bool
dest_is_ssa(nir_dest *dest, void *data)
{
+ (void) data;
return dest->is_ssa;
}
static bool
nir_instr_can_cse(nir_instr *instr)
{
+ /* We only handle SSA. */
+ if (!nir_foreach_dest(instr, dest_is_ssa, NULL) ||
+ !nir_foreach_src(instr, src_is_ssa, NULL))
+ return false;
+
switch (instr->type) {
case nir_instr_type_alu:
+ case nir_instr_type_tex:
case nir_instr_type_load_const:
case nir_instr_type_phi:
- return nir_foreach_dest(instr, dest_is_ssa, NULL) &&
- nir_foreach_src(instr, src_is_ssa, NULL);
- case nir_instr_type_texture:
- return false; /* TODO */
- case nir_instr_type_intrinsic:
+ return true;
+ case nir_instr_type_intrinsic: {
+ const nir_intrinsic_info *info =
+ &nir_intrinsic_infos[nir_instr_as_intrinsic(instr)->intrinsic];
+ return (info->flags & NIR_INTRINSIC_CAN_ELIMINATE) &&
+ (info->flags & NIR_INTRINSIC_CAN_REORDER) &&
+ info->num_variables == 0; /* not implemented yet */
+ }
case nir_instr_type_call:
case nir_instr_type_jump:
case nir_instr_type_ssa_undef:
case nir_instr_type_alu:
assert(nir_instr_as_alu(instr)->dest.dest.is_ssa);
return &nir_instr_as_alu(instr)->dest.dest.ssa;
+ case nir_instr_type_tex:
+ assert(nir_instr_as_tex(instr)->dest.is_ssa);
+ return &nir_instr_as_tex(instr)->dest.ssa;
case nir_instr_type_load_const:
- assert(nir_instr_as_load_const(instr)->dest.is_ssa);
- return &nir_instr_as_load_const(instr)->dest.ssa;
+ return &nir_instr_as_load_const(instr)->def;
case nir_instr_type_phi:
assert(nir_instr_as_phi(instr)->dest.is_ssa);
return &nir_instr_as_phi(instr)->dest.ssa;
+ case nir_instr_type_intrinsic:
+ assert(nir_instr_as_intrinsic(instr)->dest.is_ssa);
+ return &nir_instr_as_intrinsic(instr)->dest.ssa;
default:
unreachable("We never ask for any of these");
}
!exec_node_is_head_sentinel(node); node = node->prev) {
nir_instr *other = exec_node_data(nir_instr, node, node);
if (nir_instrs_equal(instr, other)) {
- nir_src other_dest_src = {
- .is_ssa = true,
- .ssa = nir_instr_get_dest_ssa_def(other),
- };
+ nir_ssa_def *other_def = nir_instr_get_dest_ssa_def(other);
nir_ssa_def_rewrite_uses(nir_instr_get_dest_ssa_def(instr),
- other_dest_src, state->mem_ctx);
+ nir_src_for_ssa(other_def),
+ state->mem_ctx);
nir_instr_remove(instr);
state->progress = true;
return;
block != NULL; block = block->imm_dom) {
nir_foreach_instr_reverse(block, other) {
if (nir_instrs_equal(instr, other)) {
- nir_src other_dest_src = {
- .is_ssa = true,
- .ssa = nir_instr_get_dest_ssa_def(other),
- };
+ nir_ssa_def *other_def = nir_instr_get_dest_ssa_def(other);
nir_ssa_def_rewrite_uses(nir_instr_get_dest_ssa_def(instr),
- other_dest_src, state->mem_ctx);
+ nir_src_for_ssa(other_def),
+ state->mem_ctx);
nir_instr_remove(instr);
state->progress = true;
return;
nir_foreach_block(impl, nir_opt_cse_block, &state);
if (state.progress)
- nir_metadata_dirty(impl, nir_metadata_block_index |
- nir_metadata_dominance);
+ nir_metadata_preserve(impl, nir_metadata_block_index |
+ nir_metadata_dominance);
return state.progress;
}
projects / mesa.git / blobdiff