* dataflow analysis on the copies available at the end of each block to re-do
* local copy propagation with more copies available.
*
- * See Muchnik's Advanced Compiler Design and Implementation, section
+ * See Muchnick's Advanced Compiler Design and Implementation, section
* 12.5 (p356).
*/
#define ACP_HASH_SIZE 16
-#include "main/bitset.h"
+#include "util/bitset.h"
#include "brw_fs.h"
#include "brw_cfg.h"
struct acp_entry : public exec_node {
fs_reg dst;
fs_reg src;
+ uint8_t regs_written;
+ enum opcode opcode;
+ bool saturate;
};
struct block_data {
void setup_initial_values();
void run();
+ void dump_block_data() const;
+
void *mem_ctx;
cfg_t *cfg;
bd = rzalloc_array(mem_ctx, struct block_data, cfg->num_blocks);
num_acp = 0;
- for (int b = 0; b < cfg->num_blocks; b++) {
+ foreach_block (block, cfg) {
for (int i = 0; i < ACP_HASH_SIZE; i++) {
- foreach_list(entry_node, &out_acp[b][i]) {
- num_acp++;
- }
+ num_acp += out_acp[block->num][i].length();
}
}
bitset_words = BITSET_WORDS(num_acp);
int next_acp = 0;
- for (int b = 0; b < cfg->num_blocks; b++) {
- bd[b].livein = rzalloc_array(bd, BITSET_WORD, bitset_words);
- bd[b].liveout = rzalloc_array(bd, BITSET_WORD, bitset_words);
- bd[b].copy = rzalloc_array(bd, BITSET_WORD, bitset_words);
- bd[b].kill = rzalloc_array(bd, BITSET_WORD, bitset_words);
+ foreach_block (block, cfg) {
+ bd[block->num].livein = rzalloc_array(bd, BITSET_WORD, bitset_words);
+ bd[block->num].liveout = rzalloc_array(bd, BITSET_WORD, bitset_words);
+ bd[block->num].copy = rzalloc_array(bd, BITSET_WORD, bitset_words);
+ bd[block->num].kill = rzalloc_array(bd, BITSET_WORD, bitset_words);
for (int i = 0; i < ACP_HASH_SIZE; i++) {
- foreach_list(entry_node, &out_acp[b][i]) {
- acp_entry *entry = (acp_entry *)entry_node;
-
+ foreach_in_list(acp_entry, entry, &out_acp[block->num][i]) {
acp[next_acp] = entry;
/* opt_copy_propagate_local populates out_acp with copies created
* in a block which are still live at the end of the block. This
* is exactly what we want in the COPY set.
*/
- BITSET_SET(bd[b].copy, next_acp);
+ BITSET_SET(bd[block->num].copy, next_acp);
next_acp++;
}
fs_copy_prop_dataflow::setup_initial_values()
{
/* Initialize the COPY and KILL sets. */
- for (int b = 0; b < cfg->num_blocks; b++) {
- bblock_t *block = cfg->blocks[b];
-
- for (fs_inst *inst = (fs_inst *)block->start;
- inst != block->end->next;
- inst = (fs_inst *)inst->next) {
+ foreach_block (block, cfg) {
+ foreach_inst_in_block(fs_inst, inst, block) {
if (inst->dst.file != GRF)
continue;
for (int i = 0; i < num_acp; i++) {
if (inst->overwrites_reg(acp[i]->dst) ||
inst->overwrites_reg(acp[i]->src)) {
- BITSET_SET(bd[b].kill, i);
+ BITSET_SET(bd[block->num].kill, i);
}
}
}
* For the others, set liveout to 0 (the empty set) and livein to ~0
* (the universal set).
*/
- for (int b = 0; b < cfg->num_blocks; b++) {
- bblock_t *block = cfg->blocks[b];
+ foreach_block (block, cfg) {
if (block->parents.is_empty()) {
for (int i = 0; i < bitset_words; i++) {
- bd[b].livein[i] = 0u;
- bd[b].liveout[i] = bd[b].copy[i];
+ bd[block->num].livein[i] = 0u;
+ bd[block->num].liveout[i] = bd[block->num].copy[i];
}
} else {
for (int i = 0; i < bitset_words; i++) {
- bd[b].liveout[i] = 0u;
- bd[b].livein[i] = ~0u;
+ bd[block->num].liveout[i] = 0u;
+ bd[block->num].livein[i] = ~0u;
}
}
}
progress = false;
/* Update liveout for all blocks. */
- for (int b = 0; b < cfg->num_blocks; b++) {
- if (cfg->blocks[b]->parents.is_empty())
+ foreach_block (block, cfg) {
+ if (block->parents.is_empty())
continue;
for (int i = 0; i < bitset_words; i++) {
- const BITSET_WORD old_liveout = bd[b].liveout[i];
+ const BITSET_WORD old_liveout = bd[block->num].liveout[i];
- bd[b].liveout[i] |=
- bd[b].copy[i] | (bd[b].livein[i] & ~bd[b].kill[i]);
+ bd[block->num].liveout[i] =
+ bd[block->num].copy[i] | (bd[block->num].livein[i] &
+ ~bd[block->num].kill[i]);
- if (old_liveout != bd[b].liveout[i])
+ if (old_liveout != bd[block->num].liveout[i])
progress = true;
}
}
/* Update livein for all blocks. If a copy is live out of all parent
* blocks, it's live coming in to this block.
*/
- for (int b = 0; b < cfg->num_blocks; b++) {
- if (cfg->blocks[b]->parents.is_empty())
+ foreach_block (block, cfg) {
+ if (block->parents.is_empty())
continue;
for (int i = 0; i < bitset_words; i++) {
- BITSET_WORD new_livein = ~bd[b].livein[i];
- foreach_list(block_node, &cfg->blocks[b]->parents) {
- bblock_link *link = (bblock_link *)block_node;
- bblock_t *block = link->block;
- new_livein &= bd[block->block_num].liveout[i];
- if (!new_livein)
- break;
+ const BITSET_WORD old_livein = bd[block->num].livein[i];
+
+ bd[block->num].livein[i] = ~0u;
+ foreach_list_typed(bblock_link, parent_link, link, &block->parents) {
+ bblock_t *parent = parent_link->block;
+ bd[block->num].livein[i] &= bd[parent->num].liveout[i];
}
- if (new_livein) {
- bd[b].livein[i] |= new_livein;
+
+ if (old_livein != bd[block->num].livein[i])
progress = true;
- }
}
}
} while (progress);
}
+void
+fs_copy_prop_dataflow::dump_block_data() const
+{
+ foreach_block (block, cfg) {
+ fprintf(stderr, "Block %d [%d, %d] (parents ", block->num,
+ block->start_ip, block->end_ip);
+ foreach_list_typed(bblock_link, link, link, &block->parents) {
+ bblock_t *parent = link->block;
+ fprintf(stderr, "%d ", parent->num);
+ }
+ fprintf(stderr, "):\n");
+ fprintf(stderr, " livein = 0x");
+ for (int i = 0; i < bitset_words; i++)
+ fprintf(stderr, "%08x", bd[block->num].livein[i]);
+ fprintf(stderr, ", liveout = 0x");
+ for (int i = 0; i < bitset_words; i++)
+ fprintf(stderr, "%08x", bd[block->num].liveout[i]);
+ fprintf(stderr, ",\n copy = 0x");
+ for (int i = 0; i < bitset_words; i++)
+ fprintf(stderr, "%08x", bd[block->num].copy[i]);
+ fprintf(stderr, ", kill = 0x");
+ for (int i = 0; i < bitset_words; i++)
+ fprintf(stderr, "%08x", bd[block->num].kill[i]);
+ fprintf(stderr, "\n");
+ }
+}
+
+static bool
+is_logic_op(enum opcode opcode)
+{
+ return (opcode == BRW_OPCODE_AND ||
+ opcode == BRW_OPCODE_OR ||
+ opcode == BRW_OPCODE_XOR ||
+ opcode == BRW_OPCODE_NOT);
+}
+
bool
fs_visitor::try_copy_propagate(fs_inst *inst, int arg, acp_entry *entry)
{
+ if (inst->src[arg].file != GRF)
+ return false;
+
if (entry->src.file == IMM)
return false;
+ assert(entry->src.file == GRF || entry->src.file == UNIFORM);
- if (inst->src[arg].file != entry->dst.file ||
- inst->src[arg].reg != entry->dst.reg ||
- inst->src[arg].reg_offset != entry->dst.reg_offset) {
+ if (entry->opcode == SHADER_OPCODE_LOAD_PAYLOAD &&
+ inst->opcode == SHADER_OPCODE_LOAD_PAYLOAD)
+ return false;
+
+ assert(entry->dst.file == GRF);
+ if (inst->src[arg].reg != entry->dst.reg)
return false;
- }
- /* See resolve_ud_negate() and comment in brw_fs_emit.cpp. */
- if (inst->conditional_mod &&
- inst->src[arg].type == BRW_REGISTER_TYPE_UD &&
+ /* Bail if inst is reading a range that isn't contained in the range
+ * that entry is writing.
+ */
+ if (inst->src[arg].reg_offset < entry->dst.reg_offset ||
+ (inst->src[arg].reg_offset * 32 + inst->src[arg].subreg_offset +
+ inst->regs_read(arg) * inst->src[arg].stride * 32) >
+ (entry->dst.reg_offset + entry->regs_written) * 32)
+ return false;
+
+ /* we can't generally copy-propagate UD negations because we
+ * can end up accessing the resulting values as signed integers
+ * instead. See also resolve_ud_negate() and comment in
+ * fs_generator::generate_code.
+ */
+ if (inst->src[arg].type == BRW_REGISTER_TYPE_UD &&
entry->src.negate)
return false;
bool has_source_modifiers = entry->src.abs || entry->src.negate;
if ((has_source_modifiers || entry->src.file == UNIFORM ||
- entry->src.smear != -1) && !can_do_source_mods(inst))
+ !entry->src.is_contiguous()) &&
+ !inst->can_do_source_mods(brw))
+ return false;
+
+ if (has_source_modifiers &&
+ inst->opcode == SHADER_OPCODE_GEN4_SCRATCH_WRITE)
+ return false;
+
+ /* Bail if the result of composing both strides would exceed the
+ * hardware limit.
+ */
+ if (entry->src.stride * inst->src[arg].stride > 4)
+ return false;
+
+ /* Bail if the result of composing both strides cannot be expressed
+ * as another stride. This avoids, for example, trying to transform
+ * this:
+ *
+ * MOV (8) rX<1>UD rY<0;1,0>UD
+ * FOO (8) ... rX<8;8,1>UW
+ *
+ * into this:
+ *
+ * FOO (8) ... rY<0;1,0>UW
+ *
+ * Which would have different semantics.
+ */
+ if (entry->src.stride != 1 &&
+ (inst->src[arg].stride *
+ type_sz(inst->src[arg].type)) % type_sz(entry->src.type) != 0)
return false;
if (has_source_modifiers && entry->dst.type != inst->src[arg].type)
return false;
+ if (brw->gen >= 8 && (entry->src.negate || entry->src.abs) &&
+ is_logic_op(inst->opcode)) {
+ return false;
+ }
+
+ if (entry->saturate) {
+ switch(inst->opcode) {
+ case BRW_OPCODE_SEL:
+ if (inst->src[1].file != IMM ||
+ inst->src[1].fixed_hw_reg.dw1.f < 0.0 ||
+ inst->src[1].fixed_hw_reg.dw1.f > 1.0) {
+ return false;
+ }
+ break;
+ default:
+ return false;
+ }
+ }
+
inst->src[arg].file = entry->src.file;
inst->src[arg].reg = entry->src.reg;
- inst->src[arg].reg_offset = entry->src.reg_offset;
- if (entry->src.smear != -1)
- inst->src[arg].smear = entry->src.smear;
+ inst->src[arg].stride *= entry->src.stride;
+ inst->saturate = inst->saturate || entry->saturate;
+
+ switch (entry->src.file) {
+ case UNIFORM:
+ assert(entry->src.width == 1);
+ case BAD_FILE:
+ case HW_REG:
+ inst->src[arg].width = entry->src.width;
+ inst->src[arg].reg_offset = entry->src.reg_offset;
+ inst->src[arg].subreg_offset = entry->src.subreg_offset;
+ break;
+ case GRF:
+ {
+ assert(entry->src.width % inst->src[arg].width == 0);
+ /* In this case, we'll just leave the width alone. The source
+ * register could have different widths depending on how it is
+ * being used. For instance, if only half of the register was
+ * used then we want to preserve that and continue to only use
+ * half.
+ *
+ * Also, we have to deal with mapping parts of vgrfs to other
+ * parts of vgrfs so we have to do some reg_offset magic.
+ */
+
+ /* Compute the offset of inst->src[arg] relative to inst->dst */
+ assert(entry->dst.subreg_offset == 0);
+ int rel_offset = inst->src[arg].reg_offset - entry->dst.reg_offset;
+ int rel_suboffset = inst->src[arg].subreg_offset;
+
+ /* Compute the final register offset (in bytes) */
+ int offset = entry->src.reg_offset * 32 + entry->src.subreg_offset;
+ offset += rel_offset * 32 + rel_suboffset;
+ inst->src[arg].reg_offset = offset / 32;
+ inst->src[arg].subreg_offset = offset % 32;
+ }
+ break;
+ default:
+ unreachable("Invalid register file");
+ break;
+ }
if (!inst->src[arg].abs) {
inst->src[arg].abs = entry->src.abs;
if (entry->src.file != IMM)
return false;
+ if (entry->saturate)
+ return false;
+
+ for (int i = inst->sources - 1; i >= 0; i--) {
+ if (inst->src[i].file != GRF)
+ continue;
- for (int i = 2; i >= 0; i--) {
- if (inst->src[i].file != entry->dst.file ||
- inst->src[i].reg != entry->dst.reg ||
- inst->src[i].reg_offset != entry->dst.reg_offset)
+ assert(entry->dst.file == GRF);
+ if (inst->src[i].reg != entry->dst.reg)
continue;
- /* Don't bother with cases that should have been taken care of by the
- * GLSL compiler's constant folding pass.
+ /* Bail if inst is reading a range that isn't contained in the range
+ * that entry is writing.
*/
- if (inst->src[i].negate || inst->src[i].abs)
+ if (inst->src[i].reg_offset < entry->dst.reg_offset ||
+ (inst->src[i].reg_offset * 32 + inst->src[i].subreg_offset +
+ inst->regs_read(i) * inst->src[i].stride * 32) >
+ (entry->dst.reg_offset + entry->regs_written) * 32)
continue;
+ fs_reg val = entry->src;
+ val.effective_width = inst->src[i].effective_width;
+ val.type = inst->src[i].type;
+
+ if (inst->src[i].abs) {
+ if ((brw->gen >= 8 && is_logic_op(inst->opcode)) ||
+ !brw_abs_immediate(val.type, &val.fixed_hw_reg)) {
+ continue;
+ }
+ }
+
+ if (inst->src[i].negate) {
+ if ((brw->gen >= 8 && is_logic_op(inst->opcode)) ||
+ !brw_negate_immediate(val.type, &val.fixed_hw_reg)) {
+ continue;
+ }
+ }
+
switch (inst->opcode) {
case BRW_OPCODE_MOV:
- inst->src[i] = entry->src;
+ case SHADER_OPCODE_LOAD_PAYLOAD:
+ inst->src[i] = val;
progress = true;
break;
+ case SHADER_OPCODE_POW:
+ case SHADER_OPCODE_INT_QUOTIENT:
+ case SHADER_OPCODE_INT_REMAINDER:
+ if (brw->gen < 8)
+ break;
+ /* fallthrough */
+ case BRW_OPCODE_BFI1:
+ case BRW_OPCODE_ASR:
+ case BRW_OPCODE_SHL:
+ case BRW_OPCODE_SHR:
+ case BRW_OPCODE_SUBB:
+ if (i == 1) {
+ inst->src[i] = val;
+ progress = true;
+ }
+ break;
+
case BRW_OPCODE_MACH:
case BRW_OPCODE_MUL:
case BRW_OPCODE_ADD:
+ case BRW_OPCODE_OR:
+ case BRW_OPCODE_AND:
+ case BRW_OPCODE_XOR:
+ case BRW_OPCODE_ADDC:
if (i == 1) {
- inst->src[i] = entry->src;
+ inst->src[i] = val;
progress = true;
} else if (i == 0 && inst->src[1].file != IMM) {
/* Fit this constant in by commuting the operands.
inst->src[1].type == BRW_REGISTER_TYPE_UD))
break;
inst->src[0] = inst->src[1];
- inst->src[1] = entry->src;
+ inst->src[1] = val;
progress = true;
}
break;
case BRW_OPCODE_CMP:
case BRW_OPCODE_IF:
if (i == 1) {
- inst->src[i] = entry->src;
+ inst->src[i] = val;
progress = true;
} else if (i == 0 && inst->src[1].file != IMM) {
- uint32_t new_cmod;
+ enum brw_conditional_mod new_cmod;
new_cmod = brw_swap_cmod(inst->conditional_mod);
- if (new_cmod != ~0u) {
+ if (new_cmod != BRW_CONDITIONAL_NONE) {
/* Fit this constant in by swapping the operands and
* flipping the test
*/
inst->src[0] = inst->src[1];
- inst->src[1] = entry->src;
+ inst->src[1] = val;
inst->conditional_mod = new_cmod;
progress = true;
}
case BRW_OPCODE_SEL:
if (i == 1) {
- inst->src[i] = entry->src;
+ inst->src[i] = val;
progress = true;
} else if (i == 0 && inst->src[1].file != IMM) {
inst->src[0] = inst->src[1];
- inst->src[1] = entry->src;
+ inst->src[1] = val;
/* If this was predicated, flipping operands means
* we also need to flip the predicate.
* anyway.
*/
assert(i == 0);
- if (inst->src[0].imm.f != 0.0f) {
+ if (inst->src[0].fixed_hw_reg.dw1.f != 0.0f) {
inst->opcode = BRW_OPCODE_MOV;
- inst->src[0] = entry->src;
- inst->src[0].imm.f = 1.0f / inst->src[0].imm.f;
+ inst->src[0] = val;
+ inst->src[0].fixed_hw_reg.dw1.f = 1.0f / inst->src[0].fixed_hw_reg.dw1.f;
progress = true;
}
break;
case FS_OPCODE_UNIFORM_PULL_CONSTANT_LOAD:
- inst->src[i] = entry->src;
+ inst->src[i] = val;
+ progress = true;
+ break;
+
+ case BRW_OPCODE_MAD:
+ case BRW_OPCODE_LRP:
+ inst->src[i] = val;
progress = true;
break;
return progress;
}
+
+static bool
+can_propagate_from(fs_inst *inst)
+{
+ return (inst->opcode == BRW_OPCODE_MOV &&
+ inst->dst.file == GRF &&
+ ((inst->src[0].file == GRF &&
+ (inst->src[0].reg != inst->dst.reg ||
+ inst->src[0].reg_offset != inst->dst.reg_offset)) ||
+ inst->src[0].file == UNIFORM ||
+ inst->src[0].file == IMM) &&
+ inst->src[0].type == inst->dst.type &&
+ !inst->is_partial_write());
+}
+
/* Walks a basic block and does copy propagation on it using the acp
* list.
*/
bool
-fs_visitor::opt_copy_propagate_local(void *mem_ctx, bblock_t *block,
+fs_visitor::opt_copy_propagate_local(void *copy_prop_ctx, bblock_t *block,
exec_list *acp)
{
bool progress = false;
- for (fs_inst *inst = (fs_inst *)block->start;
- inst != block->end->next;
- inst = (fs_inst *)inst->next) {
-
+ foreach_inst_in_block(fs_inst, inst, block) {
/* Try propagating into this instruction. */
- for (int i = 0; i < 3; i++) {
+ for (int i = 0; i < inst->sources; i++) {
if (inst->src[i].file != GRF)
continue;
- foreach_list(entry_node, &acp[inst->src[i].reg % ACP_HASH_SIZE]) {
- acp_entry *entry = (acp_entry *)entry_node;
-
+ foreach_in_list(acp_entry, entry, &acp[inst->src[i].reg % ACP_HASH_SIZE]) {
if (try_constant_propagate(inst, entry))
progress = true;
/* kill the destination from the ACP */
if (inst->dst.file == GRF) {
- foreach_list_safe(entry_node, &acp[inst->dst.reg % ACP_HASH_SIZE]) {
- acp_entry *entry = (acp_entry *)entry_node;
-
+ foreach_in_list_safe(acp_entry, entry, &acp[inst->dst.reg % ACP_HASH_SIZE]) {
if (inst->overwrites_reg(entry->dst)) {
entry->remove();
}
* the source, so walk across the entire table.
*/
for (int i = 0; i < ACP_HASH_SIZE; i++) {
- foreach_list_safe(entry_node, &acp[i]) {
- acp_entry *entry = (acp_entry *)entry_node;
+ foreach_in_list_safe(acp_entry, entry, &acp[i]) {
if (inst->overwrites_reg(entry->src))
entry->remove();
}
/* If this instruction's source could potentially be folded into the
* operand of another instruction, add it to the ACP.
*/
- if (inst->opcode == BRW_OPCODE_MOV &&
- inst->dst.file == GRF &&
- ((inst->src[0].file == GRF &&
- (inst->src[0].reg != inst->dst.reg ||
- inst->src[0].reg_offset != inst->dst.reg_offset)) ||
- inst->src[0].file == UNIFORM ||
- inst->src[0].file == IMM) &&
- inst->src[0].type == inst->dst.type &&
- !inst->saturate &&
- !inst->is_partial_write()) {
- acp_entry *entry = ralloc(mem_ctx, acp_entry);
+ if (can_propagate_from(inst)) {
+ acp_entry *entry = ralloc(copy_prop_ctx, acp_entry);
entry->dst = inst->dst;
entry->src = inst->src[0];
+ entry->regs_written = inst->regs_written;
+ entry->opcode = inst->opcode;
+ entry->saturate = inst->saturate;
acp[entry->dst.reg % ACP_HASH_SIZE].push_tail(entry);
+ } else if (inst->opcode == SHADER_OPCODE_LOAD_PAYLOAD &&
+ inst->dst.file == GRF) {
+ int offset = 0;
+ for (int i = 0; i < inst->sources; i++) {
+ int regs_written = ((inst->src[i].effective_width *
+ type_sz(inst->src[i].type)) + 31) / 32;
+ if (inst->src[i].file == GRF) {
+ acp_entry *entry = ralloc(copy_prop_ctx, acp_entry);
+ entry->dst = inst->dst;
+ entry->dst.reg_offset = offset;
+ entry->dst.width = inst->src[i].effective_width;
+ entry->src = inst->src[i];
+ entry->regs_written = regs_written;
+ entry->opcode = inst->opcode;
+ if (!entry->dst.equals(inst->src[i])) {
+ acp[entry->dst.reg % ACP_HASH_SIZE].push_tail(entry);
+ } else {
+ ralloc_free(entry);
+ }
+ }
+ offset += regs_written;
+ }
}
}
fs_visitor::opt_copy_propagate()
{
bool progress = false;
- void *mem_ctx = ralloc_context(this->mem_ctx);
- cfg_t cfg(this);
- exec_list *out_acp[cfg.num_blocks];
- for (int i = 0; i < cfg.num_blocks; i++)
+ void *copy_prop_ctx = ralloc_context(NULL);
+ exec_list *out_acp[cfg->num_blocks];
+
+ for (int i = 0; i < cfg->num_blocks; i++)
out_acp[i] = new exec_list [ACP_HASH_SIZE];
/* First, walk through each block doing local copy propagation and getting
* the set of copies available at the end of the block.
*/
- for (int b = 0; b < cfg.num_blocks; b++) {
- bblock_t *block = cfg.blocks[b];
-
- progress = opt_copy_propagate_local(mem_ctx, block,
- out_acp[b]) || progress;
+ foreach_block (block, cfg) {
+ progress = opt_copy_propagate_local(copy_prop_ctx, block,
+ out_acp[block->num]) || progress;
}
- #if 0
/* Do dataflow analysis for those available copies. */
- fs_copy_prop_dataflow dataflow(mem_ctx, &cfg, out_acp);
+ fs_copy_prop_dataflow dataflow(copy_prop_ctx, cfg, out_acp);
/* Next, re-run local copy propagation, this time with the set of copies
* provided by the dataflow analysis available at the start of a block.
*/
- for (int b = 0; b < cfg.num_blocks; b++) {
- bblock_t *block = cfg.blocks[b];
+ foreach_block (block, cfg) {
exec_list in_acp[ACP_HASH_SIZE];
for (int i = 0; i < dataflow.num_acp; i++) {
- if (BITSET_TEST(dataflow.bd[b].livein, i)) {
+ if (BITSET_TEST(dataflow.bd[block->num].livein, i)) {
struct acp_entry *entry = dataflow.acp[i];
in_acp[entry->dst.reg % ACP_HASH_SIZE].push_tail(entry);
}
}
- progress = opt_copy_propagate_local(mem_ctx, block, in_acp) || progress;
+ progress = opt_copy_propagate_local(copy_prop_ctx, block, in_acp) || progress;
}
- #endif
- for (int i = 0; i < cfg.num_blocks; i++)
+ for (int i = 0; i < cfg->num_blocks; i++)
delete [] out_acp[i];
- ralloc_free(mem_ctx);
+ ralloc_free(copy_prop_ctx);
if (progress)
- live_intervals_valid = false;
+ invalidate_live_intervals();
return progress;
}