Greedy reg alloc - better splitting

src/compiler/register-allocator.cc

Index: src/compiler/register-allocator.cc
diff --git a/src/compiler/register-allocator.cc b/src/compiler/register-allocator.cc
index 49229d5872cd22b2aac4aaaed8392163f26e5776..7e2d602106f019f4d73db2ddca0eb65f84c75333 100644
--- a/src/compiler/register-allocator.cc
+++ b/src/compiler/register-allocator.cc
@@ -889,6 +889,87 @@ LiveRange* RegisterAllocationData::LiveRangeFor(int index) {
 }
 
 
+#ifdef DEBUG
+void RegisterAllocationData::Print(LiveRange* range) {
+  std::cout << "Range: " << range->id();
+  if (range->is_phi()) std::cout << "phi ";
+  if (range->is_non_loop_phi()) std::cout << "nlphi ";
+
+  std::cout << "{" << std::endl;
+  auto interval = range->first_interval();
+  auto use_pos = range->first_pos();
+  PrintableInstructionOperand pio;
+  pio.register_configuration_ = config();
+  while (use_pos != nullptr) {
+    pio.op_ = *use_pos->operand();
+    std::cout << use_pos->pos() << " " << pio << " ";
+    use_pos = use_pos->next();
+  }
+  std::cout << std::endl;
+
+  while (interval != nullptr) {
+    std::cout << '[' << interval->start() << ", " << interval->end() << ')'
+              << std::endl;
+    interval = interval->next();
+  }
+  std::cout << "}" << std::endl;
+}
+
+
+void RegisterAllocationData::Print(ZoneSet<LiveRange*>& conflicts) {
+  for (LiveRange* range : conflicts) {
+    std::cout << "Range: " << range->id() << std::endl;
+    Print(range);
+    std::cout << std::endl;
+  }
+}
+
+
+void RegisterAllocationData::Print(InstructionSequence* code) {
+  PrintableInstructionSequence prt = {config_, code};
+  std::cout << prt << std::endl;
+}
+
+
+void RegisterAllocationData::Print(Instruction* instr) {
+  PrintableInstruction prt = {config_, instr};
+  std::cout << prt << std::endl;
+}
+
+
+void RegisterAllocationData::Print(InstructionSequence* code,
+                                   InstructionBlock* block) {
+  for (int i = block->first_instruction_index();
+       i < block->last_instruction_index(); i++) {
+    PrintableInstruction prt = {config_, code->InstructionAt(i)};
+    std::cout << prt << std::endl;
+  }
+}
+
+
+void RegisterAllocationData::PrintOperators(LiveRange* range) {
+  PrintableInstructionOperand pio;
+  pio.register_configuration_ = config_;
+  auto use = range->first_pos();
+  while (use != nullptr) {
+    auto op = use->operand();
+    pio.op_ = *op;
+    std::cout << pio;
+    use = use->next();
+    if (use != nullptr) {
+      std::cout << "; ";
+    }
+  }
+  std::cout << std::endl;
+}
+
+
+void RegisterAllocationData::Print(LifetimePosition pos) {
+  std::cout << pos << std::endl;
+}
+#endif
+
+
 MoveOperands* RegisterAllocationData::AddGapMove(
     int index, Instruction::GapPosition position,
     const InstructionOperand& from, const InstructionOperand& to) {
@@ -2399,11 +2480,19 @@ void GreedyAllocator::AssignRangeToRegister(int reg_id, LiveRange* range) {
 
 
 float GreedyAllocator::CalculateSpillWeight(LiveRange* range) {
-  if (range->IsFixed()) return std::numeric_limits<float>::max();
+  InstructionOperand* first_hint = nullptr;
+  if (range->FirstHintPosition() != nullptr) {
+    first_hint = range->FirstHintPosition()->operand();
+  }
 
-  int hint_register;
-  if (range->FirstHintPosition(&hint_register) != nullptr) {
+  if (range->IsFixed()) return std::numeric_limits<float>::max();
+  bool spill;
+  if (!FindProgressingSplitPosition(range, &spill).IsValid())
     return std::numeric_limits<float>::max();
+
+  float multiplier = 1.0;
+  if (first_hint != nullptr && first_hint->IsRegister()) {
+    multiplier = 3.0;
   }
 
   unsigned use_count = 0;
@@ -2413,11 +2502,11 @@ float GreedyAllocator::CalculateSpillWeight(LiveRange* range) {
     pos = pos->next();
   }
 
-  // GetLiveRangeSize is DCHECK-ed to not be 0
   unsigned range_size = GetLiveRangeSize(range);
   DCHECK_NE(0U, range_size);
 
-  return static_cast<float>(use_count) / static_cast<float>(range_size);
+  return multiplier * static_cast<float>(use_count) /
+         static_cast<float>(range_size);
 }
 
 
@@ -2460,32 +2549,48 @@ bool GreedyAllocator::TryAllocatePhysicalRegister(
   return true;
 }
 
+
+int GreedyAllocator::GetHintedRegister(LiveRange* range) {
+  int reg;
+  if (range->FirstHintPosition(&reg) != nullptr) {
+    return reg;
+  }
+  return -1;
+}
+
+
 bool GreedyAllocator::TryAllocate(LiveRange* current,
                                   ZoneSet<LiveRange*>* conflicting) {
-  if (current->HasSpillOperand()) {
-    Spill(current);
-    return true;
-  }
   if (current->IsFixed()) {
     return TryAllocatePhysicalRegister(current->assigned_register(), current,
                                        conflicting);
   }
 
-  if (current->HasRegisterAssigned()) {
-    int reg_id = current->assigned_register();
-    return TryAllocatePhysicalRegister(reg_id, current, conflicting);
+  int hinted_reg_id = GetHintedRegister(current);
+  if (hinted_reg_id >= 0) {
+    if (TryAllocatePhysicalRegister(hinted_reg_id, current, conflicting)) {
+      return true;
+    }
   }
 
+  ZoneSet<LiveRange*> local_conflicts(data()->allocation_zone());

[dcarney, 2015/04/29 18:18:35]

the only things that should be allocated in the allocation_zone are things
surviving phases.  create a local_zone() accessor and use that

[Mircea Trofin, 2015/04/30 18:16:49]

Done.

   for (unsigned candidate_reg = 0; candidate_reg < allocations_.size();
        candidate_reg++) {
-    if (TryAllocatePhysicalRegister(candidate_reg, current, conflicting)) {
+    if (hinted_reg_id >= 0 &&
+        candidate_reg == static_cast<size_t>(hinted_reg_id))
+      continue;
+    local_conflicts.clear();
+    if (TryAllocatePhysicalRegister(candidate_reg, current, &local_conflicts)) {
       conflicting->clear();
       return true;
+    } else {
+      conflicting->insert(local_conflicts.begin(), local_conflicts.end());
     }
   }
   return false;
 }
 
+
 LiveRange* GreedyAllocator::SpillBetweenUntil(LiveRange* range,
                                               LifetimePosition start,
                                               LifetimePosition until,
@@ -2519,12 +2624,11 @@ LiveRange* GreedyAllocator::SpillBetweenUntil(LiveRange* range,
 void GreedyAllocator::Enqueue(LiveRange* range) {
   if (range == nullptr || range->IsEmpty()) return;
   unsigned size = GetLiveRangeSize(range);
+  TRACE("Enqueuing range %d\n", range->id());
   queue_.push(std::make_pair(size, range));
 }
 
 
-// TODO(mtrofin): consolidate with identical code segment in
-// LinearScanAllocator::AllocateRegisters
 bool GreedyAllocator::HandleSpillOperands(LiveRange* range) {
   auto position = range->Start();
   TRACE("Processing interval %d start=%d\n", range->id(), position.value());
@@ -2549,9 +2653,7 @@ bool GreedyAllocator::HandleSpillOperands(LiveRange* range) {
       return true;
     }
   }
-  return false;
-  // TODO(mtrofin): Do we need this?
-  // return (TryReuseSpillForPhi(range));
+  return (TryReuseSpillForPhi(range));
 }
 
 
@@ -2566,7 +2668,7 @@ void GreedyAllocator::AllocateRegisters() {
   }
 
   allocations_.resize(num_registers());
-  auto* zone = data()->allocation_zone();
+  auto* zone = allocations_.get_allocator().zone();

[dcarney, 2015/04/29 18:18:35]

use local_zone() accessor here as well

[Mircea Trofin, 2015/04/30 18:16:49]

Done.

   for (int i = 0; i < num_registers(); i++) {
     allocations_[i] = new (zone) CoallescedLiveRanges(zone);
   }
@@ -2584,25 +2686,36 @@ void GreedyAllocator::AllocateRegisters() {
     auto current_pair = queue_.top();
     queue_.pop();
     auto current = current_pair.second;
-    if (HandleSpillOperands(current)) continue;
+    if (HandleSpillOperands(current)) {
+      continue;
+    }
+    bool spill = false;
     ZoneSet<LiveRange*> conflicting(zone);
     if (!TryAllocate(current, &conflicting)) {
       DCHECK(!conflicting.empty());
-      float this_max = CalculateSpillWeight(current);
-      float max_conflicting = CalculateMaxSpillWeight(conflicting);
-      if (max_conflicting < this_max) {
-        for (auto* conflict : conflicting) {
+      float this_weight = std::numeric_limits<float>::max();
+      LifetimePosition split_pos =
+          FindProgressingSplitPosition(current, &spill);
+      if (split_pos.IsValid()) {
+        this_weight = CalculateSpillWeight(current);
+      }
+
+      bool evicted = false;
+      for (auto* conflict : conflicting) {
+        if (CalculateSpillWeight(conflict) < this_weight) {
           Evict(conflict);
           Enqueue(conflict);
+          evicted = true;
         }
+      }
+      if (evicted) {
         conflicting.clear();
-        bool allocated = TryAllocate(current, &conflicting);
-        CHECK(allocated);
-        DCHECK(conflicting.empty());
-      } else {
+        TryAllocate(current, &conflicting);
+      }
+      if (!conflicting.empty()) {
         DCHECK(!current->IsFixed() || current->CanBeSpilled(current->Start()));
-        bool allocated = AllocateBlockedRange(current, conflicting);
-        CHECK(allocated);
+        DCHECK(split_pos.IsValid());
+        AllocateBlockedRange(current, split_pos, spill);
       }
     }
   }
@@ -2615,20 +2728,157 @@ void GreedyAllocator::AllocateRegisters() {
 }
 
 
-bool GreedyAllocator::AllocateBlockedRange(
-    LiveRange* current, const ZoneSet<LiveRange*>& conflicts) {
-  auto register_use = current->NextRegisterPosition(current->Start());
-  if (register_use == nullptr) {
-    // There is no use in the current live range that requires a register.
-    // We can just spill it.
-    Spill(current);
-    return true;
+LifetimePosition GreedyAllocator::FindProgressingSplitPosition(
+    LiveRange* range, bool* is_spill_pos) {
+  auto start = range->Start();
+  auto end = range->End();
+
+  UsePosition* next_reg_use = range->first_pos();
+  while (next_reg_use != nullptr &&
+         next_reg_use->type() != UsePositionType::kRequiresRegister) {
+    next_reg_use = next_reg_use->next();
   }
 
-  auto second_part = SplitRangeAt(current, register_use->pos());
-  Spill(second_part);
+  if (next_reg_use == nullptr) {
+    *is_spill_pos = true;
+    auto ret = FindOptimalSpillingPos(range, start);
+    DCHECK(ret.IsValid());
+    return ret;
+  }
 
-  return true;
+  *is_spill_pos = false;
+  auto ret = next_reg_use->pos();
+
+  if (start < ret.Prev()) {
+    return ret.Prev();
+  }
+
+  if (end == ret && start.Next() < end) {
+    return end.Prev();
+  }
+
+  auto blocked_pos = start;
+  while (next_reg_use->next() != nullptr &&
+         next_reg_use->next()->pos().Prev() <= next_reg_use->pos()) {
+    next_reg_use = next_reg_use->next();
+    blocked_pos = next_reg_use->pos();
+    ret = blocked_pos.Next();
+  }
+  DCHECK(next_reg_use != nullptr);
+
+  if (ret < end && start < ret) {
+    return ret;
+  }
+  ret = ret.Next();
+
+  if (ret == blocked_pos) {
+    return LifetimePosition::Invalid();
+  }
+
+  if (ret < end && start < ret) {
+    return ret;
+  }
+
+  return LifetimePosition::Invalid();
+}
+
+
+void GreedyAllocator::AllocateBlockedRange(LiveRange* current,
+                                           LifetimePosition pos, bool spill) {
+  auto tail = SplitRangeAt(current, pos);
+  if (spill) {
+    Spill(tail);
+  } else {
+    Enqueue(tail);
+  }
+  if (tail != current) {
+    Enqueue(current);
+  }
+}
+
+
+bool GreedyAllocator::TryReuseSpillForPhi(LiveRange* range) {
+  if (range->IsChild() || !range->is_phi()) return false;
+  DCHECK(!range->HasSpillOperand());
+
+  auto phi_map_value = data()->GetPhiMapValueFor(range->id());
+  auto phi = phi_map_value->phi();
+  auto block = phi_map_value->block();
+  // Count the number of spilled operands.
+  size_t spilled_count = 0;
+  LiveRange* first_op = nullptr;
+  for (size_t i = 0; i < phi->operands().size(); i++) {
+    int op = phi->operands()[i];
+    LiveRange* op_range = LiveRangeFor(op);
+    if (!op_range->HasSpillRange()) continue;
+    auto pred = code()->InstructionBlockAt(block->predecessors()[i]);
+    auto pred_end = LifetimePosition::InstructionFromInstructionIndex(
+        pred->last_instruction_index());
+    while (op_range != nullptr && !op_range->CanCover(pred_end)) {
+      op_range = op_range->next();
+    }
+    if (op_range != nullptr && op_range->spilled()) {
+      spilled_count++;
+      if (first_op == nullptr) {
+        first_op = op_range->TopLevel();
+      }
+    }
+  }
+
+  // Only continue if more than half of the operands are spilled.
+  if (spilled_count * 2 <= phi->operands().size()) {
+    return false;
+  }
+
+  // Try to merge the spilled operands and count the number of merged spilled
+  // operands.
+  DCHECK(first_op != nullptr);
+  auto first_op_spill = first_op->GetSpillRange();
+  size_t num_merged = 1;
+  for (size_t i = 1; i < phi->operands().size(); i++) {
+    int op = phi->operands()[i];
+    auto op_range = LiveRangeFor(op);
+    if (!op_range->HasSpillRange()) continue;
+    auto op_spill = op_range->GetSpillRange();
+    if (op_spill == first_op_spill || first_op_spill->TryMerge(op_spill)) {
+      num_merged++;
+    }
+  }
+
+  // Only continue if enough operands could be merged to the
+  // same spill slot.
+  if (num_merged * 2 <= phi->operands().size() ||
+      AreUseIntervalsIntersecting(first_op_spill->interval(),
+                                  range->first_interval())) {
+    return false;
+  }
+
+  // If the range does not need register soon, spill it to the merged
+  // spill range.
+  auto next_pos = range->Start();
+  if (next_pos.IsGapPosition()) next_pos = next_pos.NextStart();
+  auto pos = range->NextUsePositionRegisterIsBeneficial(next_pos);
+  if (pos == nullptr) {
+    auto spill_range =
+        range->TopLevel()->HasSpillRange()
+            ? range->TopLevel()->GetSpillRange()
+            : data()->AssignSpillRangeToLiveRange(range->TopLevel());
+    bool merged = first_op_spill->TryMerge(spill_range);
+    CHECK(merged);
+    Spill(range);
+    return true;
+  } else if (pos->pos() > range->Start().NextStart()) {
+    auto spill_range =
+        range->TopLevel()->HasSpillRange()
+            ? range->TopLevel()->GetSpillRange()
+            : data()->AssignSpillRangeToLiveRange(range->TopLevel());
+    bool merged = first_op_spill->TryMerge(spill_range);
+    CHECK(merged);
+    Enqueue(
+        SpillBetweenUntil(range, range->Start(), range->Start(), pos->pos()));
+    return true;
+  }
+  return false;
 }
 
 
@@ -2676,8 +2926,9 @@ void OperandAssigner::CommitAssignment() {
       data()->GetPhiMapValueFor(range->id())->CommitAssignment(assigned);
     }
     if (!range->IsChild() && spill_operand != nullptr) {
-      range->CommitSpillsAtDefinition(data()->code(), spill_operand,
-                                      range->has_slot_use());
+      range->CommitSpillsAtDefinition(
+          data()->code(), spill_operand,
+          range->has_slot_use() || range->spilled());
     }
   }
 }