/**
* Emit a typed surface read opcode. \p dims determines the number of
* components of the address and \p size the number of components of the
* returned value.
*/
fs_reg
emit_typed_read(const fs_builder &bld, const fs_reg &surface,
const fs_reg &addr, unsigned dims, unsigned size)
{
return emit_send(bld, SHADER_OPCODE_TYPED_SURFACE_READ_LOGICAL,
addr, fs_reg(), surface, dims, size, size);
}
void
brw_blorp_eu_emitter::emit_texture_lookup(const struct brw_reg &dst,
enum opcode op,
unsigned base_mrf,
unsigned msg_length)
{
fs_inst *inst = new (mem_ctx) fs_inst(op, dst, brw_message_reg(base_mrf),
fs_reg(0u));
inst->base_mrf = base_mrf;
inst->mlen = msg_length;
inst->header_size = 0;
insts.push_tail(inst);
}
bool
fs_visitor::opt_cse_local(bblock_t *block, exec_list *aeb)
{
bool progress = false;
void *mem_ctx = ralloc_context(this->mem_ctx);
for (fs_inst *inst = (fs_inst *)block->start;
inst != block->end->next;
inst = (fs_inst *) inst->next) {
/* Skip some cases. */
if (is_expression(inst) && !inst->predicate && inst->mlen == 0 &&
!inst->force_uncompressed && !inst->force_sechalf &&
!inst->conditional_mod)
{
bool found = false;
aeb_entry *entry;
foreach_list(entry_node, aeb) {
entry = (aeb_entry *) entry_node;
/* Match current instruction's expression against those in AEB. */
if (inst->opcode == entry->generator->opcode &&
inst->saturate == entry->generator->saturate &&
operands_match(entry->generator->src, inst->src)) {
found = true;
progress = true;
break;
}
}
if (!found) {
/* Our first sighting of this expression. Create an entry. */
aeb_entry *entry = ralloc(mem_ctx, aeb_entry);
entry->tmp = reg_undef;
entry->generator = inst;
aeb->push_tail(entry);
} else {
/* This is at least our second sighting of this expression.
* If we don't have a temporary already, make one.
*/
bool no_existing_temp = entry->tmp.file == BAD_FILE;
if (no_existing_temp) {
entry->tmp = fs_reg(this, glsl_type::float_type);
entry->tmp.type = inst->dst.type;
fs_inst *copy = new(ralloc_parent(inst))
fs_inst(BRW_OPCODE_MOV, entry->generator->dst, entry->tmp);
entry->generator->insert_after(copy);
entry->generator->dst = entry->tmp;
}
/* dest <- temp */
fs_inst *copy = new(ralloc_parent(inst))
fs_inst(BRW_OPCODE_MOV, inst->dst, entry->tmp);
inst->replace_with(copy);
/* Appending an instruction may have changed our bblock end. */
if (inst == block->end) {
block->end = copy;
}
/* Continue iteration with copy->next */
inst = copy;
}
}
bool
fs_visitor::dead_code_eliminate()
{
bool progress = false;
calculate_live_intervals();
int num_vars = live_intervals->num_vars;
BITSET_WORD *live = ralloc_array(NULL, BITSET_WORD, BITSET_WORDS(num_vars));
foreach_block (block, cfg) {
memcpy(live, live_intervals->bd[block->num].liveout,
sizeof(BITSET_WORD) * BITSET_WORDS(num_vars));
foreach_inst_in_block_reverse(fs_inst, inst, block) {
if (inst->dst.file == GRF &&
!inst->has_side_effects() &&
!inst->writes_flag()) {
bool result_live = false;
if (inst->regs_written == 1) {
int var = live_intervals->var_from_reg(&inst->dst);
result_live = BITSET_TEST(live, var);
} else {
int var = live_intervals->var_from_vgrf[inst->dst.reg];
for (int i = 0; i < inst->regs_written; i++) {
result_live = result_live || BITSET_TEST(live, var + i);
}
}
if (!result_live) {
progress = true;
if (inst->writes_accumulator) {
inst->dst = fs_reg(retype(brw_null_reg(), inst->dst.type));
} else {
inst->opcode = BRW_OPCODE_NOP;
continue;
}
}
}
if (inst->dst.file == GRF) {
if (!inst->is_partial_write()) {
int var = live_intervals->var_from_vgrf[inst->dst.reg];
for (int i = 0; i < inst->regs_written; i++) {
BITSET_CLEAR(live, var + inst->dst.reg_offset + i);
}
}
}
for (int i = 0; i < inst->sources; i++) {
if (inst->src[i].file == GRF) {
int var = live_intervals->var_from_vgrf[inst->src[i].reg];
for (int j = 0; j < inst->regs_read(this, i); j++) {
BITSET_SET(live, var + inst->src[i].reg_offset + j);
}
}
}
}
}
bool
fs_visitor::opt_cse_local(bblock_t *block, exec_list *aeb)
{
bool progress = false;
void *cse_ctx = ralloc_context(NULL);
int ip = block->start_ip;
for (fs_inst *inst = (fs_inst *)block->start;
inst != block->end->next;
inst = (fs_inst *) inst->next) {
/* Skip some cases. */
if (is_expression(inst) && !inst->is_partial_write() &&
(inst->dst.file != HW_REG || inst->dst.is_null()))
{
bool found = false;
aeb_entry *entry;
foreach_list(entry_node, aeb) {
entry = (aeb_entry *) entry_node;
/* Match current instruction's expression against those in AEB. */
if (instructions_match(inst, entry->generator)) {
found = true;
progress = true;
break;
}
}
if (!found) {
/* Our first sighting of this expression. Create an entry. */
aeb_entry *entry = ralloc(cse_ctx, aeb_entry);
entry->tmp = reg_undef;
entry->generator = inst;
aeb->push_tail(entry);
} else {
/* This is at least our second sighting of this expression.
* If we don't have a temporary already, make one.
*/
bool no_existing_temp = entry->tmp.file == BAD_FILE;
if (no_existing_temp && !entry->generator->dst.is_null()) {
int written = entry->generator->regs_written;
fs_reg orig_dst = entry->generator->dst;
fs_reg tmp = fs_reg(GRF, virtual_grf_alloc(written),
orig_dst.type);
entry->tmp = tmp;
entry->generator->dst = tmp;
fs_inst *copy;
if (written > 1) {
fs_reg *sources = ralloc_array(mem_ctx, fs_reg, written);
for (int i = 0; i < written; i++) {
sources[i] = tmp;
sources[i].reg_offset = i;
}
copy = LOAD_PAYLOAD(orig_dst, sources, written);
} else {
copy = MOV(orig_dst, tmp);
copy->force_writemask_all =
entry->generator->force_writemask_all;
}
entry->generator->insert_after(copy);
}
/* dest <- temp */
if (!inst->dst.is_null()) {
int written = inst->regs_written;
assert(written == entry->generator->regs_written);
assert(inst->dst.type == entry->tmp.type);
fs_reg dst = inst->dst;
fs_reg tmp = entry->tmp;
fs_inst *copy;
if (written > 1) {
fs_reg *sources = ralloc_array(mem_ctx, fs_reg, written);
for (int i = 0; i < written; i++) {
sources[i] = tmp;
sources[i].reg_offset = i;
}
copy = LOAD_PAYLOAD(dst, sources, written);
} else {
copy = MOV(dst, tmp);
copy->force_writemask_all = inst->force_writemask_all;
}
inst->insert_before(copy);
}
/* Set our iterator so that next time through the loop inst->next
* will get the instruction in the basic block after the one we've
* removed.
*/
fs_inst *prev = (fs_inst *)inst->prev;
inst->remove();
/* Appending an instruction may have changed our bblock end. */
if (inst == block->end) {
block->end = prev;
}
inst = prev;
}
}
bool
fs_visitor::opt_cse_local(bblock_t *block, exec_list *aeb)
{
bool progress = false;
void *mem_ctx = ralloc_context(this->mem_ctx);
int ip = block->start_ip;
for (fs_inst *inst = (fs_inst *)block->start;
inst != block->end->next;
inst = (fs_inst *) inst->next) {
/* Skip some cases. */
if (is_expression(inst) &&
!inst->predicate &&
!inst->is_partial_write() &&
!inst->conditional_mod)
{
bool found = false;
aeb_entry *entry;
foreach_list(entry_node, aeb) {
entry = (aeb_entry *) entry_node;
/* Match current instruction's expression against those in AEB. */
if (inst->opcode == entry->generator->opcode &&
inst->saturate == entry->generator->saturate &&
inst->dst.type == entry->generator->dst.type &&
operands_match(entry->generator->src, inst->src)) {
found = true;
progress = true;
break;
}
}
if (!found) {
/* Our first sighting of this expression. Create an entry. */
aeb_entry *entry = ralloc(mem_ctx, aeb_entry);
entry->tmp = reg_undef;
entry->generator = inst;
aeb->push_tail(entry);
} else {
/* This is at least our second sighting of this expression.
* If we don't have a temporary already, make one.
*/
bool no_existing_temp = entry->tmp.file == BAD_FILE;
if (no_existing_temp) {
int written = entry->generator->regs_written;
fs_reg orig_dst = entry->generator->dst;
fs_reg tmp = fs_reg(GRF, virtual_grf_alloc(written),
orig_dst.type);
entry->tmp = tmp;
entry->generator->dst = tmp;
for (int i = 0; i < written; i++) {
fs_inst *copy = MOV(orig_dst, tmp);
copy->force_writemask_all =
entry->generator->force_writemask_all;
entry->generator->insert_after(copy);
orig_dst.reg_offset++;
tmp.reg_offset++;
}
}
/* dest <- temp */
int written = inst->regs_written;
assert(written == entry->generator->regs_written);
assert(inst->dst.type == entry->tmp.type);
fs_reg dst = inst->dst;
fs_reg tmp = entry->tmp;
fs_inst *copy = NULL;
for (int i = 0; i < written; i++) {
copy = MOV(dst, tmp);
copy->force_writemask_all = inst->force_writemask_all;
inst->insert_before(copy);
dst.reg_offset++;
tmp.reg_offset++;
}
inst->remove();
/* Appending an instruction may have changed our bblock end. */
if (inst == block->end) {
block->end = copy;
}
/* Continue iteration with copy->next */
inst = copy;
}
}
