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;
}
}
