c1_LIRGenerator.cpp

/*
 * Copyright (c) 2005, 2011, Oracle and/or its affiliates. All rights reserved.
 * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
 *
 * This code is free software; you can redistribute it and/or modify it
 * under the terms of the GNU General Public License version 2 only, as
 * published by the Free Software Foundation.
 *
 * This code is distributed in the hope that it will be useful, but WITHOUT
 * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
 * FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License
 * version 2 for more details (a copy is included in the LICENSE file that
 * accompanied this code).
 *
 * You should have received a copy of the GNU General Public License version
 * 2 along with this work; if not, write to the Free Software Foundation,
 * Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA.
 *
 * Please contact Oracle, 500 Oracle Parkway, Redwood Shores, CA 94065 USA
 * or visit www.oracle.com if you need additional information or have any
 * questions.
 *
 */

#include "precompiled.hpp"
#include "c1/c1_Compilation.hpp"
#include "c1/c1_FrameMap.hpp"
#include "c1/c1_Instruction.hpp"
#include "c1/c1_LIRAssembler.hpp"
#include "c1/c1_LIRGenerator.hpp"
#include "c1/c1_ValueStack.hpp"
#include "ci/ciArrayKlass.hpp"
#include "ci/ciCPCache.hpp"
#include "ci/ciInstance.hpp"
#include "runtime/sharedRuntime.hpp"
#include "runtime/stubRoutines.hpp"
#include "utilities/bitMap.inline.hpp"
#ifndef SERIALGC
#include "gc_implementation/g1/heapRegion.hpp"
#endif

#ifdef ASSERT
#define __ gen()->lir(__FILE__, __LINE__)->
#else
#define __ gen()->lir()->
#endif

// TODO: ARM - Use some recognizable constant which still fits architectural constraints
#ifdef ARM
#define PATCHED_ADDR  (204)
#else
#define PATCHED_ADDR  (max_jint)
#endif

void PhiResolverState::reset(int max_vregs) {
  // Initialize array sizes
  _virtual_operands.at_put_grow(max_vregs - 1, NULL, NULL);
  _virtual_operands.trunc_to(0);
  _other_operands.at_put_grow(max_vregs - 1, NULL, NULL);
  _other_operands.trunc_to(0);
  _vreg_table.at_put_grow(max_vregs - 1, NULL, NULL);
  _vreg_table.trunc_to(0);
}



//--------------------------------------------------------------
// PhiResolver

// Resolves cycles:
//
//  r1 := r2  becomes  temp := r1
//  r2 := r1           r1 := r2
//                     r2 := temp
// and orders moves:
//
//  r2 := r3  becomes  r1 := r2
//  r1 := r2           r2 := r3

PhiResolver::PhiResolver(LIRGenerator* gen, int max_vregs)
 : _gen(gen)
 , _state(gen->resolver_state())
 , _temp(LIR_OprFact::illegalOpr)
{
  // reinitialize the shared state arrays
  _state.reset(max_vregs);
}


void PhiResolver::emit_move(LIR_Opr src, LIR_Opr dest) {
  assert(src->is_valid(), "");
  assert(dest->is_valid(), "");
  __ move(src, dest);
}


void PhiResolver::move_temp_to(LIR_Opr dest) {
  assert(_temp->is_valid(), "");
  emit_move(_temp, dest);
  NOT_PRODUCT(_temp = LIR_OprFact::illegalOpr);
}


void PhiResolver::move_to_temp(LIR_Opr src) {
  assert(_temp->is_illegal(), "");
  _temp = _gen->new_register(src->type());
  emit_move(src, _temp);
}


// Traverse assignment graph in depth first order and generate moves in post order
// ie. two assignments: b := c, a := b start with node c:
// Call graph: move(NULL, c) -> move(c, b) -> move(b, a)
// Generates moves in this order: move b to a and move c to b
// ie. cycle a := b, b := a start with node a
// Call graph: move(NULL, a) -> move(a, b) -> move(b, a)
// Generates moves in this order: move b to temp, move a to b, move temp to a
void PhiResolver::move(ResolveNode* src, ResolveNode* dest) {
  if (!dest->visited()) {
    dest->set_visited();
    for (int i = dest->no_of_destinations()-1; i >= 0; i --) {
      move(dest, dest->destination_at(i));
    }
  } else if (!dest->start_node()) {
    // cylce in graph detected
    assert(_loop == NULL, "only one loop valid!");
    _loop = dest;
    move_to_temp(src->operand());
    return;
  } // else dest is a start node

  if (!dest->assigned()) {
    if (_loop == dest) {
      move_temp_to(dest->operand());
      dest->set_assigned();
    } else if (src != NULL) {
      emit_move(src->operand(), dest->operand());
      dest->set_assigned();
    }
  }
}


PhiResolver::~PhiResolver() {
  int i;
  // resolve any cycles in moves from and to virtual registers
  for (i = virtual_operands().length() - 1; i >= 0; i --) {
    ResolveNode* node = virtual_operands()[i];
    if (!node->visited()) {
      _loop = NULL;
      move(NULL, node);
      node->set_start_node();
      assert(_temp->is_illegal(), "move_temp_to() call missing");
    }
  }

  // generate move for move from non virtual register to abitrary destination
  for (i = other_operands().length() - 1; i >= 0; i --) {
    ResolveNode* node = other_operands()[i];
    for (int j = node->no_of_destinations() - 1; j >= 0; j --) {
      emit_move(node->operand(), node->destination_at(j)->operand());
    }
  }
}


ResolveNode* PhiResolver::create_node(LIR_Opr opr, bool source) {
  ResolveNode* node;
  if (opr->is_virtual()) {
    int vreg_num = opr->vreg_number();
    node = vreg_table().at_grow(vreg_num, NULL);
    assert(node == NULL || node->operand() == opr, "");
    if (node == NULL) {
      node = new ResolveNode(opr);
      vreg_table()[vreg_num] = node;
    }
    // Make sure that all virtual operands show up in the list when
    // they are used as the source of a move.
    if (source && !virtual_operands().contains(node)) {
      virtual_operands().append(node);
    }
  } else {
    assert(source, "");
    node = new ResolveNode(opr);
    other_operands().append(node);
  }
  return node;
}


void PhiResolver::move(LIR_Opr src, LIR_Opr dest) {
  assert(dest->is_virtual(), "");
  // tty->print("move "); src->print(); tty->print(" to "); dest->print(); tty->cr();
  assert(src->is_valid(), "");
  assert(dest->is_valid(), "");
  ResolveNode* source = source_node(src);
  source->append(destination_node(dest));
}


//--------------------------------------------------------------
// LIRItem

void LIRItem::set_result(LIR_Opr opr) {
  assert(value()->operand()->is_illegal() || value()->operand()->is_constant(), "operand should never change");
  value()->set_operand(opr);

  if (opr->is_virtual()) {
    _gen->_instruction_for_operand.at_put_grow(opr->vreg_number(), value(), NULL);
  }

  _result = opr;
}

void LIRItem::load_item() {
  if (result()->is_illegal()) {
    // update the items result
    _result = value()->operand();
  }
  if (!result()->is_register()) {
    LIR_Opr reg = _gen->new_register(value()->type());
    __ move(result(), reg);
    if (result()->is_constant()) {
      _result = reg;
    } else {
      set_result(reg);
    }
  }
}


void LIRItem::load_for_store(BasicType type) {
  if (_gen->can_store_as_constant(value(), type)) {
    _result = value()->operand();
    if (!_result->is_constant()) {
      _result = LIR_OprFact::value_type(value()->type());
    }
  } else if (type == T_BYTE || type == T_BOOLEAN) {
    load_byte_item();
  } else {
    load_item();
  }
}

void LIRItem::load_item_force(LIR_Opr reg) {
  LIR_Opr r = result();
  if (r != reg) {
#if !defined(ARM) && !defined(E500V2)
    if (r->type() != reg->type()) {
      // moves between different types need an intervening spill slot
      r = _gen->force_to_spill(r, reg->type());
    }
#endif
    __ move(r, reg);
    _result = reg;
  }
}

ciObject* LIRItem::get_jobject_constant() const {
  ObjectType* oc = type()->as_ObjectType();
  if (oc) {
    return oc->constant_value();
  }
  return NULL;
}


jint LIRItem::get_jint_constant() const {
  assert(is_constant() && value() != NULL, "");
  assert(type()->as_IntConstant() != NULL, "type check");
  return type()->as_IntConstant()->value();
}


jint LIRItem::get_address_constant() const {
  assert(is_constant() && value() != NULL, "");
  assert(type()->as_AddressConstant() != NULL, "type check");
  return type()->as_AddressConstant()->value();
}


jfloat LIRItem::get_jfloat_constant() const {
  assert(is_constant() && value() != NULL, "");
  assert(type()->as_FloatConstant() != NULL, "type check");
  return type()->as_FloatConstant()->value();
}


jdouble LIRItem::get_jdouble_constant() const {
  assert(is_constant() && value() != NULL, "");
  assert(type()->as_DoubleConstant() != NULL, "type check");
  return type()->as_DoubleConstant()->value();
}


jlong LIRItem::get_jlong_constant() const {
  assert(is_constant() && value() != NULL, "");
  assert(type()->as_LongConstant() != NULL, "type check");
  return type()->as_LongConstant()->value();
}



//--------------------------------------------------------------


void LIRGenerator::init() {
  _bs = Universe::heap()->barrier_set();
}


void LIRGenerator::block_do_prolog(BlockBegin* block) {
#ifndef PRODUCT
  if (PrintIRWithLIR) {
    block->print();
  }
#endif

  // set up the list of LIR instructions
  assert(block->lir() == NULL, "LIR list already computed for this block");
  _lir = new LIR_List(compilation(), block);
  block->set_lir(_lir);

  __ branch_destination(block->label());

  if (LIRTraceExecution &&
      Compilation::current()->hir()->start()->block_id() != block->block_id() &&
      !block->is_set(BlockBegin::exception_entry_flag)) {
    assert(block->lir()->instructions_list()->length() == 1, "should come right after br_dst");
    trace_block_entry(block);
  }
}


void LIRGenerator::block_do_epilog(BlockBegin* block) {
#ifndef PRODUCT
  if (PrintIRWithLIR) {
    tty->cr();
  }
#endif

  // LIR_Opr for unpinned constants shouldn't be referenced by other
  // blocks so clear them out after processing the block.
  for (int i = 0; i < _unpinned_constants.length(); i++) {
    _unpinned_constants.at(i)->clear_operand();
  }
  _unpinned_constants.trunc_to(0);

  // clear our any registers for other local constants
  _constants.trunc_to(0);
  _reg_for_constants.trunc_to(0);
}


void LIRGenerator::block_do(BlockBegin* block) {
  CHECK_BAILOUT();

  block_do_prolog(block);
  set_block(block);

  for (Instruction* instr = block; instr != NULL; instr = instr->next()) {
    if (instr->is_pinned()) do_root(instr);
  }

  set_block(NULL);
  block_do_epilog(block);
}


//-------------------------LIRGenerator-----------------------------

// This is where the tree-walk starts; instr must be root;
void LIRGenerator::do_root(Value instr) {
  CHECK_BAILOUT();

  InstructionMark im(compilation(), instr);

  assert(instr->is_pinned(), "use only with roots");
  assert(instr->subst() == instr, "shouldn't have missed substitution");

  instr->visit(this);

  assert(!instr->has_uses() || instr->operand()->is_valid() ||
         instr->as_Constant() != NULL || bailed_out(), "invalid item set");
}


// This is called for each node in tree; the walk stops if a root is reached
void LIRGenerator::walk(Value instr) {
  InstructionMark im(compilation(), instr);
  //stop walk when encounter a root
  if (instr->is_pinned() && instr->as_Phi() == NULL || instr->operand()->is_valid()) {
    assert(instr->operand() != LIR_OprFact::illegalOpr || instr->as_Constant() != NULL, "this root has not yet been visited");
  } else {
    assert(instr->subst() == instr, "shouldn't have missed substitution");
    instr->visit(this);
    // assert(instr->use_count() > 0 || instr->as_Phi() != NULL, "leaf instruction must have a use");
  }
}


CodeEmitInfo* LIRGenerator::state_for(Instruction* x, ValueStack* state, bool ignore_xhandler) {
  assert(state != NULL, "state must be defined");

  ValueStack* s = state;
  for_each_state(s) {
    if (s->kind() == ValueStack::EmptyExceptionState) {
      assert(s->stack_size() == 0 && s->locals_size() == 0 && (s->locks_size() == 0 || s->locks_size() == 1), "state must be empty");
      continue;
    }

    int index;
    Value value;
    for_each_stack_value(s, index, value) {
      assert(value->subst() == value, "missed substitution");
      if (!value->is_pinned() && value->as_Constant() == NULL && value->as_Local() == NULL) {
        walk(value);
        assert(value->operand()->is_valid(), "must be evaluated now");
      }
    }

    int bci = s->bci();
    IRScope* scope = s->scope();
    ciMethod* method = scope->method();

    MethodLivenessResult liveness = method->liveness_at_bci(bci);
    if (bci == SynchronizationEntryBCI) {
      if (x->as_ExceptionObject() || x->as_Throw()) {
        // all locals are dead on exit from the synthetic unlocker
        liveness.clear();
      } else {
        assert(x->as_MonitorEnter(), "only other case is MonitorEnter");
      }
    }
    if (!liveness.is_valid()) {
      // Degenerate or breakpointed method.
      bailout("Degenerate or breakpointed method");
    } else {
      assert((int)liveness.size() == s->locals_size(), "error in use of liveness");
      for_each_local_value(s, index, value) {
        assert(value->subst() == value, "missed substition");
        if (liveness.at(index) && !value->type()->is_illegal()) {
          if (!value->is_pinned() && value->as_Constant() == NULL && value->as_Local() == NULL) {
            walk(value);
            assert(value->operand()->is_valid(), "must be evaluated now");
          }
        } else {
          // NULL out this local so that linear scan can assume that all non-NULL values are live.
          s->invalidate_local(index);
        }
      }
    }
  }

  return new CodeEmitInfo(state, ignore_xhandler ? NULL : x->exception_handlers());
}


CodeEmitInfo* LIRGenerator::state_for(Instruction* x) {
  return state_for(x, x->exception_state());
}


void LIRGenerator::jobject2reg_with_patching(LIR_Opr r, ciObject* obj, CodeEmitInfo* info) {
  if (!obj->is_loaded() || PatchALot) {
    assert(info != NULL, "info must be set if class is not loaded");
    __ oop2reg_patch(NULL, r, info);
  } else {
    // no patching needed
    __ oop2reg(obj->constant_encoding(), r);
  }
}


void LIRGenerator::array_range_check(LIR_Opr array, LIR_Opr index,
                                    CodeEmitInfo* null_check_info, CodeEmitInfo* range_check_info) {
  CodeStub* stub = new RangeCheckStub(range_check_info, index);
  if (index->is_constant()) {
    cmp_mem_int(lir_cond_belowEqual, array, arrayOopDesc::length_offset_in_bytes(),
                index->as_jint(), null_check_info);
    __ branch(lir_cond_belowEqual, T_INT, stub); // forward branch
  } else {
    cmp_reg_mem(lir_cond_aboveEqual, index, array,
                arrayOopDesc::length_offset_in_bytes(), T_INT, null_check_info);
    __ branch(lir_cond_aboveEqual, T_INT, stub); // forward branch
  }
}


void LIRGenerator::nio_range_check(LIR_Opr buffer, LIR_Opr index, LIR_Opr result, CodeEmitInfo* info) {
  CodeStub* stub = new RangeCheckStub(info, index, true);
  if (index->is_constant()) {
    cmp_mem_int(lir_cond_belowEqual, buffer, java_nio_Buffer::limit_offset(), index->as_jint(), info);
    __ branch(lir_cond_belowEqual, T_INT, stub); // forward branch
  } else {
    cmp_reg_mem(lir_cond_aboveEqual, index, buffer,
                java_nio_Buffer::limit_offset(), T_INT, info);
    __ branch(lir_cond_aboveEqual, T_INT, stub); // forward branch
  }
  __ move(index, result);
}



void LIRGenerator::arithmetic_op(Bytecodes::Code code, LIR_Opr result, LIR_Opr left, LIR_Opr right, bool is_strictfp, LIR_Opr tmp_op, CodeEmitInfo* info) {
  LIR_Opr result_op = result;
  LIR_Opr left_op   = left;
  LIR_Opr right_op  = right;

  if (TwoOperandLIRForm && left_op != result_op) {
    assert(right_op != result_op, "malformed");
    __ move(left_op, result_op);
    left_op = result_op;
  }

  switch(code) {
    case Bytecodes::_dadd:
    case Bytecodes::_fadd:
    case Bytecodes::_ladd:
    case Bytecodes::_iadd:  __ add(left_op, right_op, result_op); break;
    case Bytecodes::_fmul:
    case Bytecodes::_lmul:  __ mul(left_op, right_op, result_op); break;

    case Bytecodes::_dmul:
      {
        if (is_strictfp) {
          __ mul_strictfp(left_op, right_op, result_op, tmp_op); break;
        } else {
          __ mul(left_op, right_op, result_op); break;
        }
      }
      break;

    case Bytecodes::_imul:
      {
        bool    did_strength_reduce = false;

        if (right->is_constant()) {
          int c = right->as_jint();
          if (is_power_of_2(c)) {
            // do not need tmp here
            __ shift_left(left_op, exact_log2(c), result_op);
            did_strength_reduce = true;
          } else {
            did_strength_reduce = strength_reduce_multiply(left_op, c, result_op, tmp_op);
          }
        }
        // we couldn't strength reduce so just emit the multiply
        if (!did_strength_reduce) {
          __ mul(left_op, right_op, result_op);
        }
      }
      break;

    case Bytecodes::_dsub:
    case Bytecodes::_fsub:
    case Bytecodes::_lsub:
    case Bytecodes::_isub: __ sub(left_op, right_op, result_op); break;

    case Bytecodes::_fdiv: __ div (left_op, right_op, result_op); break;
    // ldiv and lrem are implemented with a direct runtime call

    case Bytecodes::_ddiv:
      {
        if (is_strictfp) {
          __ div_strictfp (left_op, right_op, result_op, tmp_op); break;
        } else {
          __ div (left_op, right_op, result_op); break;
        }
      }
      break;

    case Bytecodes::_drem:
    case Bytecodes::_frem: __ rem (left_op, right_op, result_op); break;

    default: ShouldNotReachHere();
  }
}


void LIRGenerator::arithmetic_op_int(Bytecodes::Code code, LIR_Opr result, LIR_Opr left, LIR_Opr right, LIR_Opr tmp) {
  arithmetic_op(code, result, left, right, false, tmp);
}


void LIRGenerator::arithmetic_op_long(Bytecodes::Code code, LIR_Opr result, LIR_Opr left, LIR_Opr right, CodeEmitInfo* info) {
  arithmetic_op(code, result, left, right, false, LIR_OprFact::illegalOpr, info);
}


void LIRGenerator::arithmetic_op_fpu(Bytecodes::Code code, LIR_Opr result, LIR_Opr left, LIR_Opr right, bool is_strictfp, LIR_Opr tmp) {
  arithmetic_op(code, result, left, right, is_strictfp, tmp);
}


void LIRGenerator::shift_op(Bytecodes::Code code, LIR_Opr result_op, LIR_Opr value, LIR_Opr count, LIR_Opr tmp) {
  if (TwoOperandLIRForm && value != result_op) {
    assert(count != result_op, "malformed");
    __ move(value, result_op);
    value = result_op;
  }

  assert(count->is_constant() || count->is_register(), "must be");
  switch(code) {
  case Bytecodes::_ishl:
  case Bytecodes::_lshl: __ shift_left(value, count, result_op, tmp); break;
  case Bytecodes::_ishr:
  case Bytecodes::_lshr: __ shift_right(value, count, result_op, tmp); break;
  case Bytecodes::_iushr:
  case Bytecodes::_lushr: __ unsigned_shift_right(value, count, result_op, tmp); break;
  default: ShouldNotReachHere();
  }
}


void LIRGenerator::logic_op (Bytecodes::Code code, LIR_Opr result_op, LIR_Opr left_op, LIR_Opr right_op) {
  if (TwoOperandLIRForm && left_op != result_op) {
    assert(right_op != result_op, "malformed");
    __ move(left_op, result_op);
    left_op = result_op;
  }

  switch(code) {
    case Bytecodes::_iand:
    case Bytecodes::_land:  __ logical_and(left_op, right_op, result_op); break;

    case Bytecodes::_ior:
    case Bytecodes::_lor:   __ logical_or(left_op, right_op, result_op);  break;

    case Bytecodes::_ixor:
    case Bytecodes::_lxor:  __ logical_xor(left_op, right_op, result_op); break;

    default: ShouldNotReachHere();
  }
}


void LIRGenerator::monitor_enter(LIR_Opr object, LIR_Opr lock, LIR_Opr hdr, LIR_Opr scratch, int monitor_no, CodeEmitInfo* info_for_exception, CodeEmitInfo* info) {
  if (!GenerateSynchronizationCode) return;
  // for slow path, use debug info for state after successful locking
  CodeStub* slow_path = new MonitorEnterStub(object, lock, info);
  __ load_stack_address_monitor(monitor_no, lock);
  // for handling NullPointerException, use debug info representing just the lock stack before this monitorenter
  __ lock_object(hdr, object, lock, scratch, slow_path, info_for_exception);
}


void LIRGenerator::monitor_exit(LIR_Opr object, LIR_Opr lock, LIR_Opr new_hdr, LIR_Opr scratch, int monitor_no) {
  if (!GenerateSynchronizationCode) return;
  // setup registers
  LIR_Opr hdr = lock;
  lock = new_hdr;
  CodeStub* slow_path = new MonitorExitStub(lock, UseFastLocking, monitor_no);
  __ load_stack_address_monitor(monitor_no, lock);
  __ unlock_object(hdr, object, lock, scratch, slow_path);
}


void LIRGenerator::new_instance(LIR_Opr dst, ciInstanceKlass* klass, LIR_Opr scratch1, LIR_Opr scratch2, LIR_Opr scratch3, LIR_Opr scratch4, LIR_Opr klass_reg, CodeEmitInfo* info) {
  jobject2reg_with_patching(klass_reg, klass, info);
  // If klass is not loaded we do not know if the klass has finalizers:
  if (UseFastNewInstance && klass->is_loaded()
      && !Klass::layout_helper_needs_slow_path(klass->layout_helper())) {

    Runtime1::StubID stub_id = klass->is_initialized() ? Runtime1::fast_new_instance_id : Runtime1::fast_new_instance_init_check_id;

    CodeStub* slow_path = new NewInstanceStub(klass_reg, dst, klass, info, stub_id);

    assert(klass->is_loaded(), "must be loaded");
    // allocate space for instance
    assert(klass->size_helper() >= 0, "illegal instance size");
    const int instance_size = align_object_size(klass->size_helper());
    __ allocate_object(dst, scratch1, scratch2, scratch3, scratch4,
                       oopDesc::header_size(), instance_size, klass_reg, !klass->is_initialized(), slow_path);
  } else {
    CodeStub* slow_path = new NewInstanceStub(klass_reg, dst, klass, info, Runtime1::new_instance_id);
    __ branch(lir_cond_always, T_ILLEGAL, slow_path);
    __ branch_destination(slow_path->continuation());
  }
}


static bool is_constant_zero(Instruction* inst) {
  IntConstant* c = inst->type()->as_IntConstant();
  if (c) {
    return (c->value() == 0);
  }
  return false;
}


static bool positive_constant(Instruction* inst) {
  IntConstant* c = inst->type()->as_IntConstant();
  if (c) {
    return (c->value() >= 0);
  }
  return false;
}


static ciArrayKlass* as_array_klass(ciType* type) {
  if (type != NULL && type->is_array_klass() && type->is_loaded()) {
    return (ciArrayKlass*)type;
  } else {
    return NULL;
  }
}

void LIRGenerator::arraycopy_helper(Intrinsic* x, int* flagsp, ciArrayKlass** expected_typep) {
  Instruction* src     = x->argument_at(0);
  Instruction* src_pos = x->argument_at(1);
  Instruction* dst     = x->argument_at(2);
  Instruction* dst_pos = x->argument_at(3);
  Instruction* length  = x->argument_at(4);

  // first try to identify the likely type of the arrays involved
  ciArrayKlass* expected_type = NULL;
  bool is_exact = false;
  {
    ciArrayKlass* src_exact_type    = as_array_klass(src->exact_type());
    ciArrayKlass* src_declared_type = as_array_klass(src->declared_type());
    ciArrayKlass* dst_exact_type    = as_array_klass(dst->exact_type());
    ciArrayKlass* dst_declared_type = as_array_klass(dst->declared_type());
    if (src_exact_type != NULL && src_exact_type == dst_exact_type) {
      // the types exactly match so the type is fully known
      is_exact = true;
      expected_type = src_exact_type;
    } else if (dst_exact_type != NULL && dst_exact_type->is_obj_array_klass()) {
      ciArrayKlass* dst_type = (ciArrayKlass*) dst_exact_type;
      ciArrayKlass* src_type = NULL;
      if (src_exact_type != NULL && src_exact_type->is_obj_array_klass()) {
        src_type = (ciArrayKlass*) src_exact_type;
      } else if (src_declared_type != NULL && src_declared_type->is_obj_array_klass()) {
        src_type = (ciArrayKlass*) src_declared_type;
      }
      if (src_type != NULL) {
        if (src_type->element_type()->is_subtype_of(dst_type->element_type())) {
          is_exact = true;
          expected_type = dst_type;
        }
      }
    }
    // at least pass along a good guess
    if (expected_type == NULL) expected_type = dst_exact_type;
    if (expected_type == NULL) expected_type = src_declared_type;
    if (expected_type == NULL) expected_type = dst_declared_type;
  }

  // if a probable array type has been identified, figure out if any
  // of the required checks for a fast case can be elided.
  int flags = LIR_OpArrayCopy::all_flags;
  if (expected_type != NULL) {
    // try to skip null checks
    if (src->as_NewArray() != NULL)
      flags &= ~LIR_OpArrayCopy::src_null_check;
    if (dst->as_NewArray() != NULL)
      flags &= ~LIR_OpArrayCopy::dst_null_check;

    // check from incoming constant values
    if (positive_constant(src_pos))
      flags &= ~LIR_OpArrayCopy::src_pos_positive_check;
    if (positive_constant(dst_pos))
      flags &= ~LIR_OpArrayCopy::dst_pos_positive_check;
    if (positive_constant(length))
      flags &= ~LIR_OpArrayCopy::length_positive_check;

    // see if the range check can be elided, which might also imply
    // that src or dst is non-null.
    ArrayLength* al = length->as_ArrayLength();
    if (al != NULL) {
      if (al->array() == src) {
        // it's the length of the source array
        flags &= ~LIR_OpArrayCopy::length_positive_check;
        flags &= ~LIR_OpArrayCopy::src_null_check;
        if (is_constant_zero(src_pos))
          flags &= ~LIR_OpArrayCopy::src_range_check;
      }
      if (al->array() == dst) {
        // it's the length of the destination array
        flags &= ~LIR_OpArrayCopy::length_positive_check;
        flags &= ~LIR_OpArrayCopy::dst_null_check;
        if (is_constant_zero(dst_pos))
          flags &= ~LIR_OpArrayCopy::dst_range_check;
      }
    }
    if (is_exact) {
      flags &= ~LIR_OpArrayCopy::type_check;
    }
  }

  if (src == dst) {
    // moving within a single array so no type checks are needed
    if (flags & LIR_OpArrayCopy::type_check) {
      flags &= ~LIR_OpArrayCopy::type_check;
    }
  }
  *flagsp = flags;
  *expected_typep = (ciArrayKlass*)expected_type;
}


LIR_Opr LIRGenerator::round_item(LIR_Opr opr) {
  assert(opr->is_register(), "why spill if item is not register?");

  if (RoundFPResults && UseSSE < 1 && opr->is_single_fpu()) {
    LIR_Opr result = new_register(T_FLOAT);
    set_vreg_flag(result, must_start_in_memory);
    assert(opr->is_register(), "only a register can be spilled");
    assert(opr->value_type()->is_float(), "rounding only for floats available");
    __ roundfp(opr, LIR_OprFact::illegalOpr, result);
    return result;
  }
  return opr;
}


LIR_Opr LIRGenerator::force_to_spill(LIR_Opr value, BasicType t) {
  assert(type2size[t] == type2size[value->type()], "size mismatch");
  if (!value->is_register()) {
    // force into a register
    LIR_Opr r = new_register(value->type());
    __ move(value, r);
    value = r;
  }

  // create a spill location
  LIR_Opr tmp = new_register(t);
  set_vreg_flag(tmp, LIRGenerator::must_start_in_memory);

  // move from register to spill
  __ move(value, tmp);
  return tmp;
}

void LIRGenerator::profile_branch(If* if_instr, If::Condition cond) {
  if (if_instr->should_profile()) {
    ciMethod* method = if_instr->profiled_method();
    assert(method != NULL, "method should be set if branch is profiled");
    ciMethodData* md = method->method_data_or_null();
    assert(md != NULL, "Sanity");
    ciProfileData* data = md->bci_to_data(if_instr->profiled_bci());
    assert(data != NULL, "must have profiling data");
    assert(data->is_BranchData(), "need BranchData for two-way branches");
    int taken_count_offset     = md->byte_offset_of_slot(data, BranchData::taken_offset());
    int not_taken_count_offset = md->byte_offset_of_slot(data, BranchData::not_taken_offset());
    if (if_instr->is_swapped()) {
      int t = taken_count_offset;
      taken_count_offset = not_taken_count_offset;
      not_taken_count_offset = t;
    }

    LIR_Opr md_reg = new_register(T_OBJECT);
    __ oop2reg(md->constant_encoding(), md_reg);

    LIR_Opr data_offset_reg = new_pointer_register();
    __ cmove(lir_cond(cond),
             LIR_OprFact::intptrConst(taken_count_offset),
             LIR_OprFact::intptrConst(not_taken_count_offset),
             data_offset_reg, as_BasicType(if_instr->x()->type()));

    // MDO cells are intptr_t, so the data_reg width is arch-dependent.
    LIR_Opr data_reg = new_pointer_register();
    LIR_Address* data_addr = new LIR_Address(md_reg, data_offset_reg, data_reg->type());
    __ move(data_addr, data_reg);
    // Use leal instead of add to avoid destroying condition codes on x86
    LIR_Address* fake_incr_value = new LIR_Address(data_reg, DataLayout::counter_increment, T_INT);
    __ leal(LIR_OprFact::address(fake_incr_value), data_reg);
    __ move(data_reg, data_addr);
  }
}

// Phi technique:
// This is about passing live values from one basic block to the other.
// In code generated with Java it is rather rare that more than one
// value is on the stack from one basic block to the other.
// We optimize our technique for efficient passing of one value
// (of type long, int, double..) but it can be extended.
// When entering or leaving a basic block, all registers and all spill
// slots are release and empty. We use the released registers
// and spill slots to pass the live values from one block
// to the other. The topmost value, i.e., the value on TOS of expression
// stack is passed in registers. All other values are stored in spilling
// area. Every Phi has an index which designates its spill slot
// At exit of a basic block, we fill the register(s) and spill slots.
// At entry of a basic block, the block_prolog sets up the content of phi nodes
// and locks necessary registers and spilling slots.


// move current value to referenced phi function
void LIRGenerator::move_to_phi(PhiResolver* resolver, Value cur_val, Value sux_val) {
  Phi* phi = sux_val->as_Phi();
  // cur_val can be null without phi being null in conjunction with inlining
  if (phi != NULL && cur_val != NULL && cur_val != phi && !phi->is_illegal()) {
    LIR_Opr operand = cur_val->operand();
    if (cur_val->operand()->is_illegal()) {
      assert(cur_val->as_Constant() != NULL || cur_val->as_Local() != NULL,
             "these can be produced lazily");
      operand = operand_for_instruction(cur_val);
    }
    resolver->move(operand, operand_for_instruction(phi));
  }
}


// Moves all stack values into their PHI position
void LIRGenerator::move_to_phi(ValueStack* cur_state) {
  BlockBegin* bb = block();
  if (bb->number_of_sux() == 1) {
    BlockBegin* sux = bb->sux_at(0);
    assert(sux->number_of_preds() > 0, "invalid CFG");

    // a block with only one predecessor never has phi functions
    if (sux->number_of_preds() > 1) {
      int max_phis = cur_state->stack_size() + cur_state->locals_size();
      PhiResolver resolver(this, _virtual_register_number + max_phis * 2);

      ValueStack* sux_state = sux->state();
      Value sux_value;
      int index;

      assert(cur_state->scope() == sux_state->scope(), "not matching");
      assert(cur_state->locals_size() == sux_state->locals_size(), "not matching");
      assert(cur_state->stack_size() == sux_state->stack_size(), "not matching");

      for_each_stack_value(sux_state, index, sux_value) {
        move_to_phi(&resolver, cur_state->stack_at(index), sux_value);
      }

      for_each_local_value(sux_state, index, sux_value) {
        move_to_phi(&resolver, cur_state->local_at(index), sux_value);
      }

      assert(cur_state->caller_state() == sux_state->caller_state(), "caller states must be equal");
    }
  }
}


LIR_Opr LIRGenerator::new_register(BasicType type) {
  int vreg = _virtual_register_number;
  // add a little fudge factor for the bailout, since the bailout is
  // only checked periodically.  This gives a few extra registers to
  // hand out before we really run out, which helps us keep from
  // tripping over assertions.
  if (vreg + 20 >= LIR_OprDesc::vreg_max) {
    bailout("out of virtual registers");
    if (vreg + 2 >= LIR_OprDesc::vreg_max) {
      // wrap it around
      _virtual_register_number = LIR_OprDesc::vreg_base;
    }
  }
  _virtual_register_number += 1;
  return LIR_OprFact::virtual_register(vreg, type);
}


// Try to lock using register in hint
LIR_Opr LIRGenerator::rlock(Value instr) {
  return new_register(instr->type());
}


// does an rlock and sets result
LIR_Opr LIRGenerator::rlock_result(Value x) {
  LIR_Opr reg = rlock(x);
  set_result(x, reg);
  return reg;
}


// does an rlock and sets result
LIR_Opr LIRGenerator::rlock_result(Value x, BasicType type) {
  LIR_Opr reg;
  switch (type) {
  case T_BYTE:
  case T_BOOLEAN:
    reg = rlock_byte(type);
    break;
  default:
    reg = rlock(x);
    break;
  }

  set_result(x, reg);
  return reg;
}


//---------------------------------------------------------------------
ciObject* LIRGenerator::get_jobject_constant(Value value) {
  ObjectType* oc = value->type()->as_ObjectType();
  if (oc) {
    return oc->constant_value();
  }
  return NULL;
}


void LIRGenerator::do_ExceptionObject(ExceptionObject* x) {
  assert(block()->is_set(BlockBegin::exception_entry_flag), "ExceptionObject only allowed in exception handler block");
  assert(block()->next() == x, "ExceptionObject must be first instruction of block");

  // no moves are created for phi functions at the begin of exception
  // handlers, so assign operands manually here
  for_each_phi_fun(block(), phi,
                   operand_for_instruction(phi));

  LIR_Opr thread_reg = getThreadPointer();
  __ move_wide(new LIR_Address(thread_reg, in_bytes(JavaThread::exception_oop_offset()), T_OBJECT),
               exceptionOopOpr());
  __ move_wide(LIR_OprFact::oopConst(NULL),
               new LIR_Address(thread_reg, in_bytes(JavaThread::exception_oop_offset()), T_OBJECT));
  __ move_wide(LIR_OprFact::oopConst(NULL),
               new LIR_Address(thread_reg, in_bytes(JavaThread::exception_pc_offset()), T_OBJECT));

  LIR_Opr result = new_register(T_OBJECT);
  __ move(exceptionOopOpr(), result);
  set_result(x, result);
}


//----------------------------------------------------------------------
//----------------------------------------------------------------------
//----------------------------------------------------------------------
//----------------------------------------------------------------------
//                        visitor functions
//----------------------------------------------------------------------
//----------------------------------------------------------------------
//----------------------------------------------------------------------
//----------------------------------------------------------------------

void LIRGenerator::do_Phi(Phi* x) {
  // phi functions are never visited directly
  ShouldNotReachHere();
}


// Code for a constant is generated lazily unless the constant is frequently used and can't be inlined.
void LIRGenerator::do_Constant(Constant* x) {
  if (x->state_before() != NULL) {
    // Any constant with a ValueStack requires patching so emit the patch here
    LIR_Opr reg = rlock_result(x);
    CodeEmitInfo* info = state_for(x, x->state_before());
    __ oop2reg_patch(NULL, reg, info);
  } else if (x->use_count() > 1 && !can_inline_as_constant(x)) {
    if (!x->is_pinned()) {
      // unpinned constants are handled specially so that they can be
      // put into registers when they are used multiple times within a
      // block.  After the block completes their operand will be
      // cleared so that other blocks can't refer to that register.
      set_result(x, load_constant(x));
    } else {
      LIR_Opr res = x->operand();
      if (!res->is_valid()) {
        res = LIR_OprFact::value_type(x->type());
      }
      if (res->is_constant()) {
        LIR_Opr reg = rlock_result(x);
        __ move(res, reg);
      } else {
        set_result(x, res);
      }
    }
  } else {
    set_result(x, LIR_OprFact::value_type(x->type()));
  }
}


void LIRGenerator::do_Local(Local* x) {
  // operand_for_instruction has the side effect of setting the result
  // so there's no need to do it here.
  operand_for_instruction(x);
}


void LIRGenerator::do_IfInstanceOf(IfInstanceOf* x) {
  Unimplemented();
}


void LIRGenerator::do_Return(Return* x) {
  if (compilation()->env()->dtrace_method_probes()) {
    BasicTypeList signature;
    signature.append(LP64_ONLY(T_LONG) NOT_LP64(T_INT));    // thread
    signature.append(T_OBJECT); // methodOop
    LIR_OprList* args = new LIR_OprList();
    args->append(getThreadPointer());
    LIR_Opr meth = new_register(T_OBJECT);
    __ oop2reg(method()->constant_encoding(), meth);
    args->append(meth);
    call_runtime(&signature, args, CAST_FROM_FN_PTR(address, SharedRuntime::dtrace_method_exit), voidType, NULL);
  }

  if (x->type()->is_void()) {
    __ return_op(LIR_OprFact::illegalOpr);
  } else {
    LIR_Opr reg = result_register_for(x->type(), /*callee=*/true);
    LIRItem result(x->result(), this);

    result.load_item_force(reg);
    __ return_op(result.result());
  }
  set_no_result(x);
}

// Examble: ref.get()
// Combination of LoadField and g1 pre-write barrier
void LIRGenerator::do_Reference_get(Intrinsic* x) {

  const int referent_offset = java_lang_ref_Reference::referent_offset;
  guarantee(referent_offset > 0, "referent offset not initialized");

  assert(x->number_of_arguments() == 1, "wrong type");

  LIRItem reference(x->argument_at(0), this);
  reference.load_item();

  // need to perform the null check on the reference objecy
  CodeEmitInfo* info = NULL;
  if (x->needs_null_check()) {
    info = state_for(x);
  }

  LIR_Address* referent_field_adr =
    new LIR_Address(reference.result(), referent_offset, T_OBJECT);

  LIR_Opr result = rlock_result(x);

  __ load(referent_field_adr, result, info);

  // Register the value in the referent field with the pre-barrier
  pre_barrier(LIR_OprFact::illegalOpr /* addr_opr */,
              result /* pre_val */,
              false  /* do_load */,
              false  /* patch */,
              NULL   /* info */);
}

// Example: object.getClass ()
void LIRGenerator::do_getClass(Intrinsic* x) {
  assert(x->number_of_arguments() == 1, "wrong type");

  LIRItem rcvr(x->argument_at(0), this);
  rcvr.load_item();
  LIR_Opr result = rlock_result(x);

  // need to perform the null check on the rcvr
  CodeEmitInfo* info = NULL;
  if (x->needs_null_check()) {
    info = state_for(x);
  }
  __ move(new LIR_Address(rcvr.result(), oopDesc::klass_offset_in_bytes(), T_OBJECT), result, info);
  __ move_wide(new LIR_Address(result, Klass::java_mirror_offset_in_bytes() +
                               klassOopDesc::klass_part_offset_in_bytes(), T_OBJECT), result);
}


// Example: Thread.currentThread()
void LIRGenerator::do_currentThread(Intrinsic* x) {
  assert(x->number_of_arguments() == 0, "wrong type");
  LIR_Opr reg = rlock_result(x);
  __ move_wide(new LIR_Address(getThreadPointer(), in_bytes(JavaThread::threadObj_offset()), T_OBJECT), reg);
}


void LIRGenerator::do_RegisterFinalizer(Intrinsic* x) {
  assert(x->number_of_arguments() == 1, "wrong type");
  LIRItem receiver(x->argument_at(0), this);

  receiver.load_item();
  BasicTypeList signature;
  signature.append(T_OBJECT); // receiver
  LIR_OprList* args = new LIR_OprList();
  args->append(receiver.result());
  CodeEmitInfo* info = state_for(x, x->state());
  call_runtime(&signature, args,
               CAST_FROM_FN_PTR(address, Runtime1::entry_for(Runtime1::register_finalizer_id)),
               voidType, info);

  set_no_result(x);
}


//------------------------local access--------------------------------------

LIR_Opr LIRGenerator::operand_for_instruction(Instruction* x) {
  if (x->operand()->is_illegal()) {
    Constant* c = x->as_Constant();
    if (c != NULL) {
      x->set_operand(LIR_OprFact::value_type(c->type()));
    } else {
      assert(x->as_Phi() || x->as_Local() != NULL, "only for Phi and Local");
      // allocate a virtual register for this local or phi
      x->set_operand(rlock(x));
      _instruction_for_operand.at_put_grow(x->operand()->vreg_number(), x, NULL);
    }
  }
  return x->operand();
}


Instruction* LIRGenerator::instruction_for_opr(LIR_Opr opr) {
  if (opr->is_virtual()) {
    return instruction_for_vreg(opr->vreg_number());
  }
  return NULL;
}


Instruction* LIRGenerator::instruction_for_vreg(int reg_num) {
  if (reg_num < _instruction_for_operand.length()) {
    return _instruction_for_operand.at(reg_num);
  }
  return NULL;
}


void LIRGenerator::set_vreg_flag(int vreg_num, VregFlag f) {
  if (_vreg_flags.size_in_bits() == 0) {
    BitMap2D temp(100, num_vreg_flags);
    temp.clear();
    _vreg_flags = temp;
  }
  _vreg_flags.at_put_grow(vreg_num, f, true);
}

bool LIRGenerator::is_vreg_flag_set(int vreg_num, VregFlag f) {
  if (!_vreg_flags.is_valid_index(vreg_num, f)) {
    return false;
  }
  return _vreg_flags.at(vreg_num, f);
}


// Block local constant handling.  This code is useful for keeping
// unpinned constants and constants which aren't exposed in the IR in
// registers.  Unpinned Constant instructions have their operands
// cleared when the block is finished so that other blocks can't end
// up referring to their registers.

LIR_Opr LIRGenerator::load_constant(Constant* x) {
  assert(!x->is_pinned(), "only for unpinned constants");
  _unpinned_constants.append(x);
  return load_constant(LIR_OprFact::value_type(x->type())->as_constant_ptr());
}


LIR_Opr LIRGenerator::load_constant(LIR_Const* c) {
  BasicType t = c->type();
  for (int i = 0; i < _constants.length(); i++) {
    LIR_Const* other = _constants.at(i);
    if (t == other->type()) {
      switch (t) {
      case T_INT:
      case T_FLOAT:
        if (c->as_jint_bits() != other->as_jint_bits()) continue;
        break;
      case T_LONG:
      case T_DOUBLE:
        if (c->as_jint_hi_bits() != other->as_jint_hi_bits()) continue;
        if (c->as_jint_lo_bits() != other->as_jint_lo_bits()) continue;
        break;
      case T_OBJECT:
        if (c->as_jobject() != other->as_jobject()) continue;
        break;
      }
      return _reg_for_constants.at(i);
    }
  }

  LIR_Opr result = new_register(t);
  __ move((LIR_Opr)c, result);
  _constants.append(c);
  _reg_for_constants.append(result);
  return result;
}

// Various barriers

void LIRGenerator::pre_barrier(LIR_Opr addr_opr, LIR_Opr pre_val,
                               bool do_load, bool patch, CodeEmitInfo* info) {
  // Do the pre-write barrier, if any.
  switch (_bs->kind()) {
#ifndef SERIALGC
    case BarrierSet::G1SATBCT:
    case BarrierSet::G1SATBCTLogging:
      G1SATBCardTableModRef_pre_barrier(addr_opr, pre_val, do_load, patch, info);
      break;
#endif // SERIALGC
    case BarrierSet::CardTableModRef:
    case BarrierSet::CardTableExtension:
      // No pre barriers
      break;
    case BarrierSet::ModRef:
    case BarrierSet::Other:
      // No pre barriers
      break;
    default      :
      ShouldNotReachHere();

  }
}

void LIRGenerator::post_barrier(LIR_OprDesc* addr, LIR_OprDesc* new_val) {
  switch (_bs->kind()) {
#ifndef SERIALGC
    case BarrierSet::G1SATBCT:
    case BarrierSet::G1SATBCTLogging:
      G1SATBCardTableModRef_post_barrier(addr,  new_val);
      break;
#endif // SERIALGC
    case BarrierSet::CardTableModRef:
    case BarrierSet::CardTableExtension:
      CardTableModRef_post_barrier(addr,  new_val);
      break;
    case BarrierSet::ModRef:
    case BarrierSet::Other:
      // No post barriers
      break;
    default      :
      ShouldNotReachHere();
    }
}

////////////////////////////////////////////////////////////////////////
#ifndef SERIALGC

void LIRGenerator::G1SATBCardTableModRef_pre_barrier(LIR_Opr addr_opr, LIR_Opr pre_val,
                                                     bool do_load, bool patch, CodeEmitInfo* info) {
  // First we test whether marking is in progress.
  BasicType flag_type;
  if (in_bytes(PtrQueue::byte_width_of_active()) == 4) {
    flag_type = T_INT;
  } else {
    guarantee(in_bytes(PtrQueue::byte_width_of_active()) == 1,
              "Assumption");
    flag_type = T_BYTE;
  }
  LIR_Opr thrd = getThreadPointer();
  LIR_Address* mark_active_flag_addr =
    new LIR_Address(thrd,
                    in_bytes(JavaThread::satb_mark_queue_offset() +
                             PtrQueue::byte_offset_of_active()),
                    flag_type);
  // Read the marking-in-progress flag.
  LIR_Opr flag_val = new_register(T_INT);
  __ load(mark_active_flag_addr, flag_val);
  __ cmp(lir_cond_notEqual, flag_val, LIR_OprFact::intConst(0));

  LIR_PatchCode pre_val_patch_code = lir_patch_none;

  CodeStub* slow;

  if (do_load) {
    assert(pre_val == LIR_OprFact::illegalOpr, "sanity");
    assert(addr_opr != LIR_OprFact::illegalOpr, "sanity");

    if (patch)
      pre_val_patch_code = lir_patch_normal;

    pre_val = new_register(T_OBJECT);

    if (!addr_opr->is_address()) {
      assert(addr_opr->is_register(), "must be");
      addr_opr = LIR_OprFact::address(new LIR_Address(addr_opr, T_OBJECT));
    }
    slow = new G1PreBarrierStub(addr_opr, pre_val, pre_val_patch_code, info);
  } else {
    assert(addr_opr == LIR_OprFact::illegalOpr, "sanity");
    assert(pre_val->is_register(), "must be");
    assert(pre_val->type() == T_OBJECT, "must be an object");
    assert(info == NULL, "sanity");

    slow = new G1PreBarrierStub(pre_val);
  }

  __ branch(lir_cond_notEqual, T_INT, slow);
  __ branch_destination(slow->continuation());
}

void LIRGenerator::G1SATBCardTableModRef_post_barrier(LIR_OprDesc* addr, LIR_OprDesc* new_val) {
  // If the "new_val" is a constant NULL, no barrier is necessary.
  if (new_val->is_constant() &&
      new_val->as_constant_ptr()->as_jobject() == NULL) return;

  if (!new_val->is_register()) {
    LIR_Opr new_val_reg = new_register(T_OBJECT);
    if (new_val->is_constant()) {
      __ move(new_val, new_val_reg);
    } else {
      __ leal(new_val, new_val_reg);
    }
    new_val = new_val_reg;
  }
  assert(new_val->is_register(), "must be a register at this point");

  if (addr->is_address()) {
    LIR_Address* address = addr->as_address_ptr();
    LIR_Opr ptr = new_register(T_OBJECT);
    if (!address->index()->is_valid() && address->disp() == 0) {
      __ move(address->base(), ptr);
    } else {
      assert(address->disp() != max_jint, "lea doesn't support patched addresses!");
      __ leal(addr, ptr);
    }
    addr = ptr;
  }
  assert(addr->is_register(), "must be a register at this point");

  LIR_Opr xor_res = new_pointer_register();
  LIR_Opr xor_shift_res = new_pointer_register();
  if (TwoOperandLIRForm ) {
    __ move(addr, xor_res);
    __ logical_xor(xor_res, new_val, xor_res);
    __ move(xor_res, xor_shift_res);
    __ unsigned_shift_right(xor_shift_res,
                            LIR_OprFact::intConst(HeapRegion::LogOfHRGrainBytes),
                            xor_shift_res,
                            LIR_OprDesc::illegalOpr());
  } else {
    __ logical_xor(addr, new_val, xor_res);
    __ unsigned_shift_right(xor_res,
                            LIR_OprFact::intConst(HeapRegion::LogOfHRGrainBytes),
                            xor_shift_res,
                            LIR_OprDesc::illegalOpr());
  }

  if (!new_val->is_register()) {
    LIR_Opr new_val_reg = new_register(T_OBJECT);
    __ leal(new_val, new_val_reg);
    new_val = new_val_reg;
  }
  assert(new_val->is_register(), "must be a register at this point");

  __ cmp(lir_cond_notEqual, xor_shift_res, LIR_OprFact::intptrConst(NULL_WORD));

  CodeStub* slow = new G1PostBarrierStub(addr, new_val);
  __ branch(lir_cond_notEqual, LP64_ONLY(T_LONG) NOT_LP64(T_INT), slow);
  __ branch_destination(slow->continuation());
}

#endif // SERIALGC
////////////////////////////////////////////////////////////////////////

void LIRGenerator::CardTableModRef_post_barrier(LIR_OprDesc* addr, LIR_OprDesc* new_val) {

  assert(sizeof(*((CardTableModRefBS*)_bs)->byte_map_base) == sizeof(jbyte), "adjust this code");
  LIR_Const* card_table_base = new LIR_Const(((CardTableModRefBS*)_bs)->byte_map_base);
  if (addr->is_address()) {
    LIR_Address* address = addr->as_address_ptr();
    LIR_Opr ptr = new_register(T_OBJECT);
    if (!address->index()->is_valid() && address->disp() == 0) {
      __ move(address->base(), ptr);
    } else {
      assert(address->disp() != max_jint, "lea doesn't support patched addresses!");
      __ leal(addr, ptr);
    }
    addr = ptr;
  }
  assert(addr->is_register(), "must be a register at this point");

#ifdef ARM
  // TODO: ARM - move to platform-dependent code
  LIR_Opr tmp = FrameMap::R14_opr;
  if (VM_Version::supports_movw()) {
    __ move((LIR_Opr)card_table_base, tmp);
  } else {
    __ move(new LIR_Address(FrameMap::Rthread_opr, in_bytes(JavaThread::card_table_base_offset()), T_ADDRESS), tmp);
  }

  CardTableModRefBS* ct = (CardTableModRefBS*)_bs;
  LIR_Address *card_addr = new LIR_Address(tmp, addr, (LIR_Address::Scale) -CardTableModRefBS::card_shift, 0, T_BYTE);
  if(((int)ct->byte_map_base & 0xff) == 0) {
    __ move(tmp, card_addr);
  } else {
    LIR_Opr tmp_zero = new_register(T_INT);
    __ move(LIR_OprFact::intConst(0), tmp_zero);
    __ move(tmp_zero, card_addr);
  }
#else // ARM
  LIR_Opr tmp = new_pointer_register();
  if (TwoOperandLIRForm) {
    __ move(addr, tmp);
    __ unsigned_shift_right(tmp, CardTableModRefBS::card_shift, tmp);
  } else {
    __ unsigned_shift_right(addr, CardTableModRefBS::card_shift, tmp);
  }
  if (can_inline_as_constant(card_table_base)) {
    __ move(LIR_OprFact::intConst(0),
              new LIR_Address(tmp, card_table_base->as_jint(), T_BYTE));
  } else {
    __ move(LIR_OprFact::intConst(0),
              new LIR_Address(tmp, load_constant(card_table_base),
                              T_BYTE));
  }
#endif // ARM
}


//------------------------field access--------------------------------------

// Comment copied form templateTable_i486.cpp
// ----------------------------------------------------------------------------
// Volatile variables demand their effects be made known to all CPU's in
// order.  Store buffers on most chips allow reads & writes to reorder; the
// JMM's ReadAfterWrite.java test fails in -Xint mode without some kind of
// memory barrier (i.e., it's not sufficient that the interpreter does not
// reorder volatile references, the hardware also must not reorder them).
//
// According to the new Java Memory Model (JMM):
// (1) All volatiles are serialized wrt to each other.
// ALSO reads & writes act as aquire & release, so:
// (2) A read cannot let unrelated NON-volatile memory refs that happen after
// the read float up to before the read.  It's OK for non-volatile memory refs
// that happen before the volatile read to float down below it.
// (3) Similar a volatile write cannot let unrelated NON-volatile memory refs
// that happen BEFORE the write float down to after the write.  It's OK for
// non-volatile memory refs that happen after the volatile write to float up
// before it.
//
// We only put in barriers around volatile refs (they are expensive), not
// _between_ memory refs (that would require us to track the flavor of the
// previous memory refs).  Requirements (2) and (3) require some barriers
// before volatile stores and after volatile loads.  These nearly cover
// requirement (1) but miss the volatile-store-volatile-load case.  This final
// case is placed after volatile-stores although it could just as well go
// before volatile-loads.


void LIRGenerator::do_StoreField(StoreField* x) {
  bool needs_patching = x->needs_patching();
  bool is_volatile = x->field()->is_volatile();
  BasicType field_type = x->field_type();
  bool is_oop = (field_type == T_ARRAY || field_type == T_OBJECT);

  CodeEmitInfo* info = NULL;
  if (needs_patching) {
    assert(x->explicit_null_check() == NULL, "can't fold null check into patching field access");
    info = state_for(x, x->state_before());
  } else if (x->needs_null_check()) {
    NullCheck* nc = x->explicit_null_check();
    if (nc == NULL) {
      info = state_for(x);
    } else {
      info = state_for(nc);
    }
  }


  LIRItem object(x->obj(), this);
  LIRItem value(x->value(),  this);

  object.load_item();

  if (is_volatile || needs_patching) {
    // load item if field is volatile (fewer special cases for volatiles)
    // load item if field not initialized
    // load item if field not constant
    // because of code patching we cannot inline constants
    if (field_type == T_BYTE || field_type == T_BOOLEAN) {
      value.load_byte_item();
    } else  {
      value.load_item();
    }
  } else {
    value.load_for_store(field_type);
  }

  set_no_result(x);

#ifndef PRODUCT
  if (PrintNotLoaded && needs_patching) {
    tty->print_cr("   ###class not loaded at store_%s bci %d",
                  x->is_static() ?  "static" : "field", x->printable_bci());
  }
#endif

  if (x->needs_null_check() &&
      (needs_patching ||
       MacroAssembler::needs_explicit_null_check(x->offset()))) {
    // emit an explicit null check because the offset is too large
    __ null_check(object.result(), new CodeEmitInfo(info));
  }

  LIR_Address* address;
  if (needs_patching) {
    // we need to patch the offset in the instruction so don't allow
    // generate_address to try to be smart about emitting the -1.
    // Otherwise the patching code won't know how to find the
    // instruction to patch.
    address = new LIR_Address(object.result(), PATCHED_ADDR, field_type);
  } else {
    address = generate_address(object.result(), x->offset(), field_type);
  }

  if (is_volatile && os::is_MP()) {
    __ membar_release();
  }

  if (is_oop) {
    // Do the pre-write barrier, if any.
    pre_barrier(LIR_OprFact::address(address),
                LIR_OprFact::illegalOpr /* pre_val */,
                true /* do_load*/,
                needs_patching,
                (info ? new CodeEmitInfo(info) : NULL));
  }

  if (is_volatile && !needs_patching) {
    volatile_field_store(value.result(), address, info);
  } else {
    LIR_PatchCode patch_code = needs_patching ? lir_patch_normal : lir_patch_none;
    __ store(value.result(), address, info, patch_code);
  }

  if (is_oop) {
    // Store to object so mark the card of the header
    post_barrier(object.result(), value.result());
  }

  if (is_volatile && os::is_MP()) {
    __ membar();
  }
}


void LIRGenerator::do_LoadField(LoadField* x) {
  bool needs_patching = x->needs_patching();
  bool is_volatile = x->field()->is_volatile();
  BasicType field_type = x->field_type();

  CodeEmitInfo* info = NULL;
  if (needs_patching) {
    assert(x->explicit_null_check() == NULL, "can't fold null check into patching field access");
    info = state_for(x, x->state_before());
  } else if (x->needs_null_check()) {
    NullCheck* nc = x->explicit_null_check();
    if (nc == NULL) {
      info = state_for(x);
    } else {
      info = state_for(nc);
    }
  }

  LIRItem object(x->obj(), this);

  object.load_item();

#ifndef PRODUCT
  if (PrintNotLoaded && needs_patching) {
    tty->print_cr("   ###class not loaded at load_%s bci %d",
                  x->is_static() ?  "static" : "field", x->printable_bci());
  }
#endif

  if (x->needs_null_check() &&
      (needs_patching ||
       MacroAssembler::needs_explicit_null_check(x->offset()))) {
    // emit an explicit null check because the offset is too large
    __ null_check(object.result(), new CodeEmitInfo(info));
  }

  LIR_Opr reg = rlock_result(x, field_type);
  LIR_Address* address;
  if (needs_patching) {
    // we need to patch the offset in the instruction so don't allow
    // generate_address to try to be smart about emitting the -1.
    // Otherwise the patching code won't know how to find the
    // instruction to patch.
    address = new LIR_Address(object.result(), PATCHED_ADDR, field_type);
  } else {
    address = generate_address(object.result(), x->offset(), field_type);
  }

  if (is_volatile && !needs_patching) {
    volatile_field_load(address, reg, info);
  } else {
    LIR_PatchCode patch_code = needs_patching ? lir_patch_normal : lir_patch_none;
    __ load(address, reg, info, patch_code);
  }

  if (is_volatile && os::is_MP()) {
    __ membar_acquire();
  }
}


//------------------------java.nio.Buffer.checkIndex------------------------

// int java.nio.Buffer.checkIndex(int)
void LIRGenerator::do_NIOCheckIndex(Intrinsic* x) {
  // NOTE: by the time we are in checkIndex() we are guaranteed that
  // the buffer is non-null (because checkIndex is package-private and
  // only called from within other methods in the buffer).
  assert(x->number_of_arguments() == 2, "wrong type");
  LIRItem buf  (x->argument_at(0), this);
  LIRItem index(x->argument_at(1), this);
  buf.load_item();
  index.load_item();

  LIR_Opr result = rlock_result(x);
  if (GenerateRangeChecks) {
    CodeEmitInfo* info = state_for(x);
    CodeStub* stub = new RangeCheckStub(info, index.result(), true);
    if (index.result()->is_constant()) {
      cmp_mem_int(lir_cond_belowEqual, buf.result(), java_nio_Buffer::limit_offset(), index.result()->as_jint(), info);
      __ branch(lir_cond_belowEqual, T_INT, stub);
    } else {
      cmp_reg_mem(lir_cond_aboveEqual, index.result(), buf.result(),
                  java_nio_Buffer::limit_offset(), T_INT, info);
      __ branch(lir_cond_aboveEqual, T_INT, stub);
    }
    __ move(index.result(), result);
  } else {
    // Just load the index into the result register
    __ move(index.result(), result);
  }
}


//------------------------array access--------------------------------------


void LIRGenerator::do_ArrayLength(ArrayLength* x) {
  LIRItem array(x->array(), this);
  array.load_item();
  LIR_Opr reg = rlock_result(x);

  CodeEmitInfo* info = NULL;
  if (x->needs_null_check()) {
    NullCheck* nc = x->explicit_null_check();
    if (nc == NULL) {
      info = state_for(x);
    } else {
      info = state_for(nc);
    }
  }
  __ load(new LIR_Address(array.result(), arrayOopDesc::length_offset_in_bytes(), T_INT), reg, info, lir_patch_none);
}


void LIRGenerator::do_LoadIndexed(LoadIndexed* x) {
  bool use_length = x->length() != NULL;
  LIRItem array(x->array(), this);
  LIRItem index(x->index(), this);
  LIRItem length(this);
  bool needs_range_check = true;

  if (use_length) {
    needs_range_check = x->compute_needs_range_check();
    if (needs_range_check) {
      length.set_instruction(x->length());
      length.load_item();
    }
  }

  array.load_item();
  if (index.is_constant() && can_inline_as_constant(x->index())) {
    // let it be a constant
    index.dont_load_item();
  } else {
    index.load_item();
  }

  CodeEmitInfo* range_check_info = state_for(x);
  CodeEmitInfo* null_check_info = NULL;
  if (x->needs_null_check()) {
    NullCheck* nc = x->explicit_null_check();
    if (nc != NULL) {
      null_check_info = state_for(nc);
    } else {
      null_check_info = range_check_info;
    }
  }

  // emit array address setup early so it schedules better
  LIR_Address* array_addr = emit_array_address(array.result(), index.result(), x->elt_type(), false);

  if (GenerateRangeChecks && needs_range_check) {
    if (use_length) {
      // TODO: use a (modified) version of array_range_check that does not require a
      //       constant length to be loaded to a register
      __ cmp(lir_cond_belowEqual, length.result(), index.result());
      __ branch(lir_cond_belowEqual, T_INT, new RangeCheckStub(range_check_info, index.result()));
    } else {
      array_range_check(array.result(), index.result(), null_check_info, range_check_info);
      // The range check performs the null check, so clear it out for the load
      null_check_info = NULL;
    }
  }

  __ move(array_addr, rlock_result(x, x->elt_type()), null_check_info);
}


void LIRGenerator::do_NullCheck(NullCheck* x) {
  if (x->can_trap()) {
    LIRItem value(x->obj(), this);
    value.load_item();
    CodeEmitInfo* info = state_for(x);
    __ null_check(value.result(), info);
  }
}


void LIRGenerator::do_Throw(Throw* x) {
  LIRItem exception(x->exception(), this);
  exception.load_item();
  set_no_result(x);
  LIR_Opr exception_opr = exception.result();
  CodeEmitInfo* info = state_for(x, x->state());

#ifndef PRODUCT
  if (PrintC1Statistics) {
    increment_counter(Runtime1::throw_count_address(), T_INT);
  }
#endif

  // check if the instruction has an xhandler in any of the nested scopes
  bool unwind = false;
  if (info->exception_handlers()->length() == 0) {
    // this throw is not inside an xhandler
    unwind = true;
  } else {
    // get some idea of the throw type
    bool type_is_exact = true;
    ciType* throw_type = x->exception()->exact_type();
    if (throw_type == NULL) {
      type_is_exact = false;
      throw_type = x->exception()->declared_type();
    }
    if (throw_type != NULL && throw_type->is_instance_klass()) {
      ciInstanceKlass* throw_klass = (ciInstanceKlass*)throw_type;
      unwind = !x->exception_handlers()->could_catch(throw_klass, type_is_exact);
    }
  }

  // do null check before moving exception oop into fixed register
  // to avoid a fixed interval with an oop during the null check.
  // Use a copy of the CodeEmitInfo because debug information is
  // different for null_check and throw.
  if (GenerateCompilerNullChecks &&
      (x->exception()->as_NewInstance() == NULL && x->exception()->as_ExceptionObject() == NULL)) {
    // if the exception object wasn't created using new then it might be null.
    __ null_check(exception_opr, new CodeEmitInfo(info, x->state()->copy(ValueStack::ExceptionState, x->state()->bci())));
  }

  if (compilation()->env()->jvmti_can_post_on_exceptions()) {
    // we need to go through the exception lookup path to get JVMTI
    // notification done
    unwind = false;
  }

  // move exception oop into fixed register
  __ move(exception_opr, exceptionOopOpr());

  if (unwind) {
    __ unwind_exception(exceptionOopOpr());
  } else {
    __ throw_exception(exceptionPcOpr(), exceptionOopOpr(), info);
  }
}


void LIRGenerator::do_RoundFP(RoundFP* x) {
  LIRItem input(x->input(), this);
  input.load_item();
  LIR_Opr input_opr = input.result();
  assert(input_opr->is_register(), "why round if value is not in a register?");
  assert(input_opr->is_single_fpu() || input_opr->is_double_fpu(), "input should be floating-point value");
  if (input_opr->is_single_fpu()) {
    set_result(x, round_item(input_opr)); // This code path not currently taken
  } else {
    LIR_Opr result = new_register(T_DOUBLE);
    set_vreg_flag(result, must_start_in_memory);
    __ roundfp(input_opr, LIR_OprFact::illegalOpr, result);
    set_result(x, result);
  }
}

void LIRGenerator::do_UnsafeGetRaw(UnsafeGetRaw* x) {
  LIRItem base(x->base(), this);
  LIRItem idx(this);

  base.load_item();
  if (x->has_index()) {
    idx.set_instruction(x->index());
    idx.load_nonconstant();
  }

  LIR_Opr reg = rlock_result(x, x->basic_type());

  int   log2_scale = 0;
  if (x->has_index()) {
    assert(x->index()->type()->tag() == intTag, "should not find non-int index");
    log2_scale = x->log2_scale();
  }

  assert(!x->has_index() || idx.value() == x->index(), "should match");

  LIR_Opr base_op = base.result();
#ifndef _LP64
  if (x->base()->type()->tag() == longTag) {
    base_op = new_register(T_INT);
    __ convert(Bytecodes::_l2i, base.result(), base_op);
  } else {
    assert(x->base()->type()->tag() == intTag, "must be");
  }
#endif

  BasicType dst_type = x->basic_type();
  LIR_Opr index_op = idx.result();

  LIR_Address* addr;
  if (index_op->is_constant()) {
    assert(log2_scale == 0, "must not have a scale");
    addr = new LIR_Address(base_op, index_op->as_jint(), dst_type);
  } else {
#ifdef X86
#ifdef _LP64
    if (!index_op->is_illegal() && index_op->type() == T_INT) {
      LIR_Opr tmp = new_pointer_register();
      __ convert(Bytecodes::_i2l, index_op, tmp);
      index_op = tmp;
    }
#endif
    addr = new LIR_Address(base_op, index_op, LIR_Address::Scale(log2_scale), 0, dst_type);
#elif defined(ARM)
    addr = generate_address(base_op, index_op, log2_scale, 0, dst_type);
#else
    if (index_op->is_illegal() || log2_scale == 0) {
#ifdef _LP64
      if (!index_op->is_illegal() && index_op->type() == T_INT) {
        LIR_Opr tmp = new_pointer_register();
        __ convert(Bytecodes::_i2l, index_op, tmp);
        index_op = tmp;
      }
#endif
      addr = new LIR_Address(base_op, index_op, dst_type);
    } else {
      LIR_Opr tmp = new_pointer_register();
      __ shift_left(index_op, log2_scale, tmp);
      addr = new LIR_Address(base_op, tmp, dst_type);
    }
#endif
  }

  if (x->may_be_unaligned() && (dst_type == T_LONG || dst_type == T_DOUBLE)) {
    __ unaligned_move(addr, reg);
  } else {
    if (dst_type == T_OBJECT && x->is_wide()) {
      __ move_wide(addr, reg);
    } else {
      __ move(addr, reg);
    }
  }
}


void LIRGenerator::do_UnsafePutRaw(UnsafePutRaw* x) {
  int  log2_scale = 0;
  BasicType type = x->basic_type();

  if (x->has_index()) {
    assert(x->index()->type()->tag() == intTag, "should not find non-int index");
    log2_scale = x->log2_scale();
  }

  LIRItem base(x->base(), this);
  LIRItem value(x->value(), this);
  LIRItem idx(this);

  base.load_item();
  if (x->has_index()) {
    idx.set_instruction(x->index());
    idx.load_item();
  }

  if (type == T_BYTE || type == T_BOOLEAN) {
    value.load_byte_item();
  } else {
    value.load_item();
  }

  set_no_result(x);

  LIR_Opr base_op = base.result();
#ifndef _LP64
  if (x->base()->type()->tag() == longTag) {
    base_op = new_register(T_INT);
    __ convert(Bytecodes::_l2i, base.result(), base_op);
  } else {
    assert(x->base()->type()->tag() == intTag, "must be");
  }
#endif

  LIR_Opr index_op = idx.result();
  if (log2_scale != 0) {
    // temporary fix (platform dependent code without shift on Intel would be better)
    index_op = new_pointer_register();
#ifdef _LP64
    if(idx.result()->type() == T_INT) {
      __ convert(Bytecodes::_i2l, idx.result(), index_op);
    } else {
#endif
      // TODO: ARM also allows embedded shift in the address
      __ move(idx.result(), index_op);
#ifdef _LP64
    }
#endif
    __ shift_left(index_op, log2_scale, index_op);
  }
#ifdef _LP64
  else if(!index_op->is_illegal() && index_op->type() == T_INT) {
    LIR_Opr tmp = new_pointer_register();
    __ convert(Bytecodes::_i2l, index_op, tmp);
    index_op = tmp;
  }
#endif

  LIR_Address* addr = new LIR_Address(base_op, index_op, x->basic_type());
  __ move(value.result(), addr);
}


void LIRGenerator::do_UnsafeGetObject(UnsafeGetObject* x) {
  BasicType type = x->basic_type();
  LIRItem src(x->object(), this);
  LIRItem off(x->offset(), this);

  off.load_item();
  src.load_item();

  LIR_Opr reg = rlock_result(x, x->basic_type());

  get_Object_unsafe(reg, src.result(), off.result(), type, x->is_volatile());

#ifndef SERIALGC
  // We might be reading the value of the referent field of a
  // Reference object in order to attach it back to the live
  // object graph. If G1 is enabled then we need to record
  // the value that is being returned in an SATB log buffer.
  //
  // We need to generate code similar to the following...
  //
  // if (offset == java_lang_ref_Reference::referent_offset) {
  //   if (src != NULL) {
  //     if (klass(src)->reference_type() != REF_NONE) {
  //       pre_barrier(..., reg, ...);
  //     }
  //   }
  // }
  //
  // The first non-constant check of either the offset or
  // the src operand will be done here; the remainder
  // will take place in the generated code stub.

  if (UseG1GC && type == T_OBJECT) {
    bool gen_code_stub = true;       // Assume we need to generate the slow code stub.
    bool gen_offset_check = true;       // Assume the code stub has to generate the offset guard.
    bool gen_source_check = true;       // Assume the code stub has to check the src object for null.

    if (off.is_constant()) {
      jint off_con = off.get_jint_constant();

      if (off_con != java_lang_ref_Reference::referent_offset) {
        // The constant offset is something other than referent_offset.
        // We can skip generating/checking the remaining guards and
        // skip generation of the code stub.
        gen_code_stub = false;
      } else {
        // The constant offset is the same as referent_offset -
        // we do not need to generate a runtime offset check.
        gen_offset_check = false;
      }
    }

    // We don't need to generate stub if the source object is an array
    if (gen_code_stub && src.type()->is_array()) {
      gen_code_stub = false;
    }

    if (gen_code_stub) {
      // We still need to continue with the checks.
      if (src.is_constant()) {
        ciObject* src_con = src.get_jobject_constant();

        if (src_con->is_null_object()) {
          // The constant src object is null - We can skip
          // generating the code stub.
          gen_code_stub = false;
        } else {
          // Non-null constant source object. We still have to generate
          // the slow stub - but we don't need to generate the runtime
          // null object check.
          gen_source_check = false;
        }
      }
    }

    if (gen_code_stub) {
      // Temoraries.
      LIR_Opr src_klass = new_register(T_OBJECT);

      // Get the thread pointer for the pre-barrier
      LIR_Opr thread = getThreadPointer();

      CodeStub* stub;

      // We can have generate one runtime check here. Let's start with
      // the offset check.
      if (gen_offset_check) {
        // if (offset == referent_offset) -> slow code stub
        __ cmp(lir_cond_equal, off.result(),
               LIR_OprFact::intConst(java_lang_ref_Reference::referent_offset));

        // Optionally generate "src == null" check.
        stub = new G1UnsafeGetObjSATBBarrierStub(reg, src.result(),
                                                    src_klass, thread,
                                                    gen_source_check);

        __ branch(lir_cond_equal, T_INT, stub);
      } else {
        if (gen_source_check) {
          // offset is a const and equals referent offset
          // if (source != null) -> slow code stub
          __ cmp(lir_cond_notEqual, src.result(), LIR_OprFact::oopConst(NULL));

          // Since we are generating the "if src == null" guard here,
          // there is no need to generate the "src == null" check again.
          stub = new G1UnsafeGetObjSATBBarrierStub(reg, src.result(),
                                                    src_klass, thread,
                                                    false);

          __ branch(lir_cond_notEqual, T_OBJECT, stub);
        } else {
          // We have statically determined that offset == referent_offset
          // && src != null so we unconditionally branch to code stub
          // to perform the guards and record reg in the SATB log buffer.

          stub = new G1UnsafeGetObjSATBBarrierStub(reg, src.result(),
                                                    src_klass, thread,
                                                    false);

          __ branch(lir_cond_always, T_ILLEGAL, stub);
        }
      }

      // Continuation point
      __ branch_destination(stub->continuation());
    }
  }
#endif // SERIALGC

  if (x->is_volatile() && os::is_MP()) __ membar_acquire();
}


void LIRGenerator::do_UnsafePutObject(UnsafePutObject* x) {
  BasicType type = x->basic_type();
  LIRItem src(x->object(), this);
  LIRItem off(x->offset(), this);
  LIRItem data(x->value(), this);

  src.load_item();
  if (type == T_BOOLEAN || type == T_BYTE) {
    data.load_byte_item();
  } else {
    data.load_item();
  }
  off.load_item();

  set_no_result(x);

  if (x->is_volatile() && os::is_MP()) __ membar_release();
  put_Object_unsafe(src.result(), off.result(), data.result(), type, x->is_volatile());
  if (x->is_volatile() && os::is_MP()) __ membar();
}


void LIRGenerator::do_UnsafePrefetch(UnsafePrefetch* x, bool is_store) {
  LIRItem src(x->object(), this);
  LIRItem off(x->offset(), this);

  src.load_item();
  if (off.is_constant() && can_inline_as_constant(x->offset())) {
    // let it be a constant
    off.dont_load_item();
  } else {
    off.load_item();
  }

  set_no_result(x);

  LIR_Address* addr = generate_address(src.result(), off.result(), 0, 0, T_BYTE);
  __ prefetch(addr, is_store);
}


void LIRGenerator::do_UnsafePrefetchRead(UnsafePrefetchRead* x) {
  do_UnsafePrefetch(x, false);
}


void LIRGenerator::do_UnsafePrefetchWrite(UnsafePrefetchWrite* x) {
  do_UnsafePrefetch(x, true);
}


void LIRGenerator::do_SwitchRanges(SwitchRangeArray* x, LIR_Opr value, BlockBegin* default_sux) {
  int lng = x->length();

  for (int i = 0; i < lng; i++) {
    SwitchRange* one_range = x->at(i);
    int low_key = one_range->low_key();
    int high_key = one_range->high_key();
    BlockBegin* dest = one_range->sux();
    if (low_key == high_key) {
      __ cmp(lir_cond_equal, value, low_key);
      __ branch(lir_cond_equal, T_INT, dest);
    } else if (high_key - low_key == 1) {
      __ cmp(lir_cond_equal, value, low_key);
      __ branch(lir_cond_equal, T_INT, dest);
      __ cmp(lir_cond_equal, value, high_key);
      __ branch(lir_cond_equal, T_INT, dest);
    } else {
      LabelObj* L = new LabelObj();
      __ cmp(lir_cond_less, value, low_key);
      __ branch(lir_cond_less, L->label());
      __ cmp(lir_cond_lessEqual, value, high_key);
      __ branch(lir_cond_lessEqual, T_INT, dest);
      __ branch_destination(L->label());
    }
  }
  __ jump(default_sux);
}


SwitchRangeArray* LIRGenerator::create_lookup_ranges(TableSwitch* x) {
  SwitchRangeList* res = new SwitchRangeList();
  int len = x->length();
  if (len > 0) {
    BlockBegin* sux = x->sux_at(0);
    int key = x->lo_key();
    BlockBegin* default_sux = x->default_sux();
    SwitchRange* range = new SwitchRange(key, sux);
    for (int i = 0; i < len; i++, key++) {
      BlockBegin* new_sux = x->sux_at(i);
      if (sux == new_sux) {
        // still in same range
        range->set_high_key(key);
      } else {
        // skip tests which explicitly dispatch to the default
        if (sux != default_sux) {
          res->append(range);
        }
        range = new SwitchRange(key, new_sux);
      }
      sux = new_sux;
    }
    if (res->length() == 0 || res->last() != range)  res->append(range);
  }
  return res;
}


// we expect the keys to be sorted by increasing value
SwitchRangeArray* LIRGenerator::create_lookup_ranges(LookupSwitch* x) {
  SwitchRangeList* res = new SwitchRangeList();
  int len = x->length();
  if (len > 0) {
    BlockBegin* default_sux = x->default_sux();
    int key = x->key_at(0);
    BlockBegin* sux = x->sux_at(0);
    SwitchRange* range = new SwitchRange(key, sux);
    for (int i = 1; i < len; i++) {
      int new_key = x->key_at(i);
      BlockBegin* new_sux = x->sux_at(i);
      if (key+1 == new_key && sux == new_sux) {
        // still in same range
        range->set_high_key(new_key);
      } else {
        // skip tests which explicitly dispatch to the default
        if (range->sux() != default_sux) {
          res->append(range);
        }
        range = new SwitchRange(new_key, new_sux);
      }
      key = new_key;
      sux = new_sux;
    }
    if (res->length() == 0 || res->last() != range)  res->append(range);
  }
  return res;
}


void LIRGenerator::do_TableSwitch(TableSwitch* x) {
  LIRItem tag(x->tag(), this);
  tag.load_item();
  set_no_result(x);

  if (x->is_safepoint()) {
    __ safepoint(safepoint_poll_register(), state_for(x, x->state_before()));
  }

  // move values into phi locations
  move_to_phi(x->state());

  int lo_key = x->lo_key();
  int hi_key = x->hi_key();
  int len = x->length();
  LIR_Opr value = tag.result();
  if (UseTableRanges) {
    do_SwitchRanges(create_lookup_ranges(x), value, x->default_sux());
  } else {
    for (int i = 0; i < len; i++) {
      __ cmp(lir_cond_equal, value, i + lo_key);
      __ branch(lir_cond_equal, T_INT, x->sux_at(i));
    }
    __ jump(x->default_sux());
  }
}


void LIRGenerator::do_LookupSwitch(LookupSwitch* x) {
  LIRItem tag(x->tag(), this);
  tag.load_item();
  set_no_result(x);

  if (x->is_safepoint()) {
    __ safepoint(safepoint_poll_register(), state_for(x, x->state_before()));
  }

  // move values into phi locations
  move_to_phi(x->state());

  LIR_Opr value = tag.result();
  if (UseTableRanges) {
    do_SwitchRanges(create_lookup_ranges(x), value, x->default_sux());
  } else {
    int len = x->length();
    for (int i = 0; i < len; i++) {
      __ cmp(lir_cond_equal, value, x->key_at(i));
      __ branch(lir_cond_equal, T_INT, x->sux_at(i));
    }
    __ jump(x->default_sux());
  }
}


void LIRGenerator::do_Goto(Goto* x) {
  set_no_result(x);

  if (block()->next()->as_OsrEntry()) {
    // need to free up storage used for OSR entry point
    LIR_Opr osrBuffer = block()->next()->operand();
    BasicTypeList signature;
    signature.append(T_INT);
    CallingConvention* cc = frame_map()->c_calling_convention(&signature);
    __ move(osrBuffer, cc->args()->at(0));
    __ call_runtime_leaf(CAST_FROM_FN_PTR(address, SharedRuntime::OSR_migration_end),
                         getThreadTemp(), LIR_OprFact::illegalOpr, cc->args());
  }

  if (x->is_safepoint()) {
    ValueStack* state = x->state_before() ? x->state_before() : x->state();

    // increment backedge counter if needed
    CodeEmitInfo* info = state_for(x, state);
    increment_backedge_counter(info, info->stack()->bci());
    CodeEmitInfo* safepoint_info = state_for(x, state);
    __ safepoint(safepoint_poll_register(), safepoint_info);
  }

  // Gotos can be folded Ifs, handle this case.
  if (x->should_profile()) {
    ciMethod* method = x->profiled_method();
    assert(method != NULL, "method should be set if branch is profiled");
    ciMethodData* md = method->method_data_or_null();
    assert(md != NULL, "Sanity");
    ciProfileData* data = md->bci_to_data(x->profiled_bci());
    assert(data != NULL, "must have profiling data");
    int offset;
    if (x->direction() == Goto::taken) {
      assert(data->is_BranchData(), "need BranchData for two-way branches");
      offset = md->byte_offset_of_slot(data, BranchData::taken_offset());
    } else if (x->direction() == Goto::not_taken) {
      assert(data->is_BranchData(), "need BranchData for two-way branches");
      offset = md->byte_offset_of_slot(data, BranchData::not_taken_offset());
    } else {
      assert(data->is_JumpData(), "need JumpData for branches");
      offset = md->byte_offset_of_slot(data, JumpData::taken_offset());
    }
    LIR_Opr md_reg = new_register(T_OBJECT);
    __ oop2reg(md->constant_encoding(), md_reg);

    increment_counter(new LIR_Address(md_reg, offset,
                                      NOT_LP64(T_INT) LP64_ONLY(T_LONG)), DataLayout::counter_increment);
  }

  // emit phi-instruction move after safepoint since this simplifies
  // describing the state as the safepoint.
  move_to_phi(x->state());

  __ jump(x->default_sux());
}


void LIRGenerator::do_Base(Base* x) {
  __ std_entry(LIR_OprFact::illegalOpr);
  // Emit moves from physical registers / stack slots to virtual registers
  CallingConvention* args = compilation()->frame_map()->incoming_arguments();
  IRScope* irScope = compilation()->hir()->top_scope();
  int java_index = 0;
  for (int i = 0; i < args->length(); i++) {
    LIR_Opr src = args->at(i);
    assert(!src->is_illegal(), "check");
    BasicType t = src->type();

    // Types which are smaller than int are passed as int, so
    // correct the type which passed.
    switch (t) {
    case T_BYTE:
    case T_BOOLEAN:
    case T_SHORT:
    case T_CHAR:
      t = T_INT;
      break;
    }

    LIR_Opr dest = new_register(t);
    __ move(src, dest);

    // Assign new location to Local instruction for this local
    Local* local = x->state()->local_at(java_index)->as_Local();
    assert(local != NULL, "Locals for incoming arguments must have been created");
#ifndef __SOFTFP__
    // The java calling convention passes double as long and float as int.
    assert(as_ValueType(t)->tag() == local->type()->tag(), "check");
#endif // __SOFTFP__
    local->set_operand(dest);
    _instruction_for_operand.at_put_grow(dest->vreg_number(), local, NULL);
    java_index += type2size[t];
  }

  if (compilation()->env()->dtrace_method_probes()) {
    BasicTypeList signature;
    signature.append(LP64_ONLY(T_LONG) NOT_LP64(T_INT));    // thread
    signature.append(T_OBJECT); // methodOop
    LIR_OprList* args = new LIR_OprList();
    args->append(getThreadPointer());
    LIR_Opr meth = new_register(T_OBJECT);
    __ oop2reg(method()->constant_encoding(), meth);
    args->append(meth);
    call_runtime(&signature, args, CAST_FROM_FN_PTR(address, SharedRuntime::dtrace_method_entry), voidType, NULL);
  }

  if (method()->is_synchronized()) {
    LIR_Opr obj;
    if (method()->is_static()) {
      obj = new_register(T_OBJECT);
      __ oop2reg(method()->holder()->java_mirror()->constant_encoding(), obj);
    } else {
      Local* receiver = x->state()->local_at(0)->as_Local();
      assert(receiver != NULL, "must already exist");
      obj = receiver->operand();
    }
    assert(obj->is_valid(), "must be valid");

    if (method()->is_synchronized() && GenerateSynchronizationCode) {
      LIR_Opr lock = new_register(T_INT);
      __ load_stack_address_monitor(0, lock);

      CodeEmitInfo* info = new CodeEmitInfo(scope()->start()->state()->copy(ValueStack::StateBefore, SynchronizationEntryBCI), NULL);
      CodeStub* slow_path = new MonitorEnterStub(obj, lock, info);

      // receiver is guaranteed non-NULL so don't need CodeEmitInfo
      __ lock_object(syncTempOpr(), obj, lock, new_register(T_OBJECT), slow_path, NULL);
    }
  }

  // increment invocation counters if needed
  if (!method()->is_accessor()) { // Accessors do not have MDOs, so no counting.
    CodeEmitInfo* info = new CodeEmitInfo(scope()->start()->state()->copy(ValueStack::StateBefore, SynchronizationEntryBCI), NULL);
    increment_invocation_counter(info);
  }

  // all blocks with a successor must end with an unconditional jump
  // to the successor even if they are consecutive
  __ jump(x->default_sux());
}


void LIRGenerator::do_OsrEntry(OsrEntry* x) {
  // construct our frame and model the production of incoming pointer
  // to the OSR buffer.
  __ osr_entry(LIR_Assembler::osrBufferPointer());
  LIR_Opr result = rlock_result(x);
  __ move(LIR_Assembler::osrBufferPointer(), result);
}


void LIRGenerator::invoke_load_arguments(Invoke* x, LIRItemList* args, const LIR_OprList* arg_list) {
  int i = (x->has_receiver() || x->is_invokedynamic()) ? 1 : 0;
  for (; i < args->length(); i++) {
    LIRItem* param = args->at(i);
    LIR_Opr loc = arg_list->at(i);
    if (loc->is_register()) {
      param->load_item_force(loc);
    } else {
      LIR_Address* addr = loc->as_address_ptr();
      param->load_for_store(addr->type());
      if (addr->type() == T_OBJECT) {
        __ move_wide(param->result(), addr);
      } else
        if (addr->type() == T_LONG || addr->type() == T_DOUBLE) {
          __ unaligned_move(param->result(), addr);
        } else {
          __ move(param->result(), addr);
        }
    }
  }

  if (x->has_receiver()) {
    LIRItem* receiver = args->at(0);
    LIR_Opr loc = arg_list->at(0);
    if (loc->is_register()) {
      receiver->load_item_force(loc);
    } else {
      assert(loc->is_address(), "just checking");
      receiver->load_for_store(T_OBJECT);
      __ move_wide(receiver->result(), loc->as_address_ptr());
    }
  }
}


// Visits all arguments, returns appropriate items without loading them
LIRItemList* LIRGenerator::invoke_visit_arguments(Invoke* x) {
  LIRItemList* argument_items = new LIRItemList();
  if (x->has_receiver()) {
    LIRItem* receiver = new LIRItem(x->receiver(), this);
    argument_items->append(receiver);
  }
  if (x->is_invokedynamic()) {
    // Insert a dummy for the synthetic MethodHandle argument.
    argument_items->append(NULL);
  }
  int idx = x->has_receiver() ? 1 : 0;
  for (int i = 0; i < x->number_of_arguments(); i++) {
    LIRItem* param = new LIRItem(x->argument_at(i), this);
    argument_items->append(param);
    idx += (param->type()->is_double_word() ? 2 : 1);
  }
  return argument_items;
}


// The invoke with receiver has following phases:
//   a) traverse and load/lock receiver;
//   b) traverse all arguments -> item-array (invoke_visit_argument)
//   c) push receiver on stack
//   d) load each of the items and push on stack
//   e) unlock receiver
//   f) move receiver into receiver-register %o0
//   g) lock result registers and emit call operation
//
// Before issuing a call, we must spill-save all values on stack
// that are in caller-save register. "spill-save" moves thos registers
// either in a free callee-save register or spills them if no free
// callee save register is available.
//
// The problem is where to invoke spill-save.
// - if invoked between e) and f), we may lock callee save
//   register in "spill-save" that destroys the receiver register
//   before f) is executed
// - if we rearange the f) to be earlier, by loading %o0, it
//   may destroy a value on the stack that is currently in %o0
//   and is waiting to be spilled
// - if we keep the receiver locked while doing spill-save,
//   we cannot spill it as it is spill-locked
//
void LIRGenerator::do_Invoke(Invoke* x) {
  CallingConvention* cc = frame_map()->java_calling_convention(x->signature(), true);

  LIR_OprList* arg_list = cc->args();
  LIRItemList* args = invoke_visit_arguments(x);
  LIR_Opr receiver = LIR_OprFact::illegalOpr;

  // setup result register
  LIR_Opr result_register = LIR_OprFact::illegalOpr;
  if (x->type() != voidType) {
    result_register = result_register_for(x->type());
  }

  CodeEmitInfo* info = state_for(x, x->state());

  // invokedynamics can deoptimize.
  CodeEmitInfo* deopt_info = x->is_invokedynamic() ? state_for(x, x->state_before()) : NULL;

  invoke_load_arguments(x, args, arg_list);

  if (x->has_receiver()) {
    args->at(0)->load_item_force(LIR_Assembler::receiverOpr());
    receiver = args->at(0)->result();
  }

  // emit invoke code
  bool optimized = x->target_is_loaded() && x->target_is_final();
  assert(receiver->is_illegal() || receiver->is_equal(LIR_Assembler::receiverOpr()), "must match");

  // JSR 292
  // Preserve the SP over MethodHandle call sites.
  ciMethod* target = x->target();
  if (target->is_method_handle_invoke()) {
    info->set_is_method_handle_invoke(true);
    __ move(FrameMap::stack_pointer(), FrameMap::method_handle_invoke_SP_save_opr());
  }

  switch (x->code()) {
    case Bytecodes::_invokestatic:
      __ call_static(target, result_register,
                     SharedRuntime::get_resolve_static_call_stub(),
                     arg_list, info);
      break;
    case Bytecodes::_invokespecial:
    case Bytecodes::_invokevirtual:
    case Bytecodes::_invokeinterface:
      // for final target we still produce an inline cache, in order
      // to be able to call mixed mode
      if (x->code() == Bytecodes::_invokespecial || optimized) {
        __ call_opt_virtual(target, receiver, result_register,
                            SharedRuntime::get_resolve_opt_virtual_call_stub(),
                            arg_list, info);
      } else if (x->vtable_index() < 0) {
        __ call_icvirtual(target, receiver, result_register,
                          SharedRuntime::get_resolve_virtual_call_stub(),
                          arg_list, info);
      } else {
        int entry_offset = instanceKlass::vtable_start_offset() + x->vtable_index() * vtableEntry::size();
        int vtable_offset = entry_offset * wordSize + vtableEntry::method_offset_in_bytes();
        __ call_virtual(target, receiver, result_register, vtable_offset, arg_list, info);
      }
      break;
    case Bytecodes::_invokedynamic: {
      ciBytecodeStream bcs(x->scope()->method());
      bcs.force_bci(x->state()->bci());
      assert(bcs.cur_bc() == Bytecodes::_invokedynamic, "wrong stream");
      ciCPCache* cpcache = bcs.get_cpcache();

      // Get CallSite offset from constant pool cache pointer.
      int index = bcs.get_method_index();
      size_t call_site_offset = cpcache->get_f1_offset(index);

      // If this invokedynamic call site hasn't been executed yet in
      // the interpreter, the CallSite object in the constant pool
      // cache is still null and we need to deoptimize.
      if (cpcache->is_f1_null_at(index)) {
        // Cannot re-use same xhandlers for multiple CodeEmitInfos, so
        // clone all handlers.  This is handled transparently in other
        // places by the CodeEmitInfo cloning logic but is handled
        // specially here because a stub isn't being used.
        x->set_exception_handlers(new XHandlers(x->exception_handlers()));

        DeoptimizeStub* deopt_stub = new DeoptimizeStub(deopt_info);
        __ jump(deopt_stub);
      }

      // Use the receiver register for the synthetic MethodHandle
      // argument.
      receiver = LIR_Assembler::receiverOpr();
      LIR_Opr tmp = new_register(objectType);

      // Load CallSite object from constant pool cache.
      __ oop2reg(cpcache->constant_encoding(), tmp);
      __ load(new LIR_Address(tmp, call_site_offset, T_OBJECT), tmp);

      // Load target MethodHandle from CallSite object.
      __ load(new LIR_Address(tmp, java_lang_invoke_CallSite::target_offset_in_bytes(), T_OBJECT), receiver);

      __ call_dynamic(target, receiver, result_register,
                      SharedRuntime::get_resolve_opt_virtual_call_stub(),
                      arg_list, info);
      break;
    }
    default:
      ShouldNotReachHere();
      break;
  }

  // JSR 292
  // Restore the SP after MethodHandle call sites.
  if (target->is_method_handle_invoke()) {
    __ move(FrameMap::method_handle_invoke_SP_save_opr(), FrameMap::stack_pointer());
  }

  if (x->type()->is_float() || x->type()->is_double()) {
    // Force rounding of results from non-strictfp when in strictfp
    // scope (or when we don't know the strictness of the callee, to
    // be safe.)
    if (method()->is_strict()) {
      if (!x->target_is_loaded() || !x->target_is_strictfp()) {
        result_register = round_item(result_register);
      }
    }
  }

  if (result_register->is_valid()) {
    LIR_Opr result = rlock_result(x);
    __ move(result_register, result);
  }
}


void LIRGenerator::do_FPIntrinsics(Intrinsic* x) {
  assert(x->number_of_arguments() == 1, "wrong type");
  LIRItem value       (x->argument_at(0), this);
  LIR_Opr reg = rlock_result(x);
  value.load_item();
  LIR_Opr tmp = force_to_spill(value.result(), as_BasicType(x->type()));
  __ move(tmp, reg);
}



// Code for  :  x->x() {x->cond()} x->y() ? x->tval() : x->fval()
void LIRGenerator::do_IfOp(IfOp* x) {
#ifdef ASSERT
  {
    ValueTag xtag = x->x()->type()->tag();
    ValueTag ttag = x->tval()->type()->tag();
    assert(xtag == intTag || xtag == objectTag, "cannot handle others");
    assert(ttag == addressTag || ttag == intTag || ttag == objectTag || ttag == longTag, "cannot handle others");
    assert(ttag == x->fval()->type()->tag(), "cannot handle others");
  }
#endif

  LIRItem left(x->x(), this);
  LIRItem right(x->y(), this);
  left.load_item();
  if (can_inline_as_constant(right.value())) {
    right.dont_load_item();
  } else {
    right.load_item();
  }

  LIRItem t_val(x->tval(), this);
  LIRItem f_val(x->fval(), this);
  t_val.dont_load_item();
  f_val.dont_load_item();
  LIR_Opr reg = rlock_result(x);

  __ cmp(lir_cond(x->cond()), left.result(), right.result());
  __ cmove(lir_cond(x->cond()), t_val.result(), f_val.result(), reg, as_BasicType(x->x()->type()));
}


void LIRGenerator::do_Intrinsic(Intrinsic* x) {
  switch (x->id()) {
  case vmIntrinsics::_intBitsToFloat      :
  case vmIntrinsics::_doubleToRawLongBits :
  case vmIntrinsics::_longBitsToDouble    :
  case vmIntrinsics::_floatToRawIntBits   : {
    do_FPIntrinsics(x);
    break;
  }

  case vmIntrinsics::_currentTimeMillis: {
    assert(x->number_of_arguments() == 0, "wrong type");
    LIR_Opr reg = result_register_for(x->type());
    __ call_runtime_leaf(CAST_FROM_FN_PTR(address, os::javaTimeMillis), getThreadTemp(),
                         reg, new LIR_OprList());
    LIR_Opr result = rlock_result(x);
    __ move(reg, result);
    break;
  }

  case vmIntrinsics::_nanoTime: {
    assert(x->number_of_arguments() == 0, "wrong type");
    LIR_Opr reg = result_register_for(x->type());
    __ call_runtime_leaf(CAST_FROM_FN_PTR(address, os::javaTimeNanos), getThreadTemp(),
                         reg, new LIR_OprList());
    LIR_Opr result = rlock_result(x);
    __ move(reg, result);
    break;
  }

  case vmIntrinsics::_Object_init:    do_RegisterFinalizer(x); break;
  case vmIntrinsics::_getClass:       do_getClass(x);      break;
  case vmIntrinsics::_currentThread:  do_currentThread(x); break;

  case vmIntrinsics::_dlog:           // fall through
  case vmIntrinsics::_dlog10:         // fall through
  case vmIntrinsics::_dabs:           // fall through
  case vmIntrinsics::_dsqrt:          // fall through
  case vmIntrinsics::_dtan:           // fall through
  case vmIntrinsics::_dsin :          // fall through
  case vmIntrinsics::_dcos :          do_MathIntrinsic(x); break;
  case vmIntrinsics::_arraycopy:      do_ArrayCopy(x);     break;

  // java.nio.Buffer.checkIndex
  case vmIntrinsics::_checkIndex:     do_NIOCheckIndex(x); break;

  case vmIntrinsics::_compareAndSwapObject:
    do_CompareAndSwap(x, objectType);
    break;
  case vmIntrinsics::_compareAndSwapInt:
    do_CompareAndSwap(x, intType);
    break;
  case vmIntrinsics::_compareAndSwapLong:
    do_CompareAndSwap(x, longType);
    break;

    // sun.misc.AtomicLongCSImpl.attemptUpdate
  case vmIntrinsics::_attemptUpdate:
    do_AttemptUpdate(x);
    break;

  case vmIntrinsics::_Reference_get:
    do_Reference_get(x);
    break;

  default: ShouldNotReachHere(); break;
  }
}

void LIRGenerator::do_ProfileCall(ProfileCall* x) {
  // Need recv in a temporary register so it interferes with the other temporaries
  LIR_Opr recv = LIR_OprFact::illegalOpr;
  LIR_Opr mdo = new_register(T_OBJECT);
  // tmp is used to hold the counters on SPARC
  LIR_Opr tmp = new_pointer_register();
  if (x->recv() != NULL) {
    LIRItem value(x->recv(), this);
    value.load_item();
    recv = new_register(T_OBJECT);
    __ move(value.result(), recv);
  }
  __ profile_call(x->method(), x->bci_of_invoke(), mdo, recv, tmp, x->known_holder());
}

void LIRGenerator::do_ProfileInvoke(ProfileInvoke* x) {
  // We can safely ignore accessors here, since c2 will inline them anyway,
  // accessors are also always mature.
  if (!x->inlinee()->is_accessor()) {
    CodeEmitInfo* info = state_for(x, x->state(), true);
    // Increment invocation counter, don't notify the runtime, because we don't inline loops,
    increment_event_counter_impl(info, x->inlinee(), 0, InvocationEntryBci, false, false);
  }
}

void LIRGenerator::increment_event_counter(CodeEmitInfo* info, int bci, bool backedge) {
  int freq_log;
  int level = compilation()->env()->comp_level();
  if (level == CompLevel_limited_profile) {
    freq_log = (backedge ? Tier2BackedgeNotifyFreqLog : Tier2InvokeNotifyFreqLog);
  } else if (level == CompLevel_full_profile) {
    freq_log = (backedge ? Tier3BackedgeNotifyFreqLog : Tier3InvokeNotifyFreqLog);
  } else {
    ShouldNotReachHere();
  }
  // Increment the appropriate invocation/backedge counter and notify the runtime.
  increment_event_counter_impl(info, info->scope()->method(), (1 << freq_log) - 1, bci, backedge, true);
}

void LIRGenerator::increment_event_counter_impl(CodeEmitInfo* info,
                                                ciMethod *method, int frequency,
                                                int bci, bool backedge, bool notify) {
  assert(frequency == 0 || is_power_of_2(frequency + 1), "Frequency must be x^2 - 1 or 0");
  int level = _compilation->env()->comp_level();
  assert(level > CompLevel_simple, "Shouldn't be here");

  int offset = -1;
  LIR_Opr counter_holder = new_register(T_OBJECT);
  LIR_Opr meth;
  if (level == CompLevel_limited_profile) {
    offset = in_bytes(backedge ? methodOopDesc::backedge_counter_offset() :
                                 methodOopDesc::invocation_counter_offset());
    __ oop2reg(method->constant_encoding(), counter_holder);
    meth = counter_holder;
  } else if (level == CompLevel_full_profile) {
    offset = in_bytes(backedge ? methodDataOopDesc::backedge_counter_offset() :
                                 methodDataOopDesc::invocation_counter_offset());
    ciMethodData* md = method->method_data_or_null();
    assert(md != NULL, "Sanity");
    __ oop2reg(md->constant_encoding(), counter_holder);
    meth = new_register(T_OBJECT);
    __ oop2reg(method->constant_encoding(), meth);
  } else {
    ShouldNotReachHere();
  }
  LIR_Address* counter = new LIR_Address(counter_holder, offset, T_INT);
  LIR_Opr result = new_register(T_INT);
  __ load(counter, result);
  __ add(result, LIR_OprFact::intConst(InvocationCounter::count_increment), result);
  __ store(result, counter);
  if (notify) {
    LIR_Opr mask = load_immediate(frequency << InvocationCounter::count_shift, T_INT);
    __ logical_and(result, mask, result);
    __ cmp(lir_cond_equal, result, LIR_OprFact::intConst(0));
    // The bci for info can point to cmp for if's we want the if bci
    CodeStub* overflow = new CounterOverflowStub(info, bci, meth);
    __ branch(lir_cond_equal, T_INT, overflow);
    __ branch_destination(overflow->continuation());
  }
}

void LIRGenerator::do_RuntimeCall(RuntimeCall* x) {
  LIR_OprList* args = new LIR_OprList(x->number_of_arguments());
  BasicTypeList* signature = new BasicTypeList(x->number_of_arguments());

  if (x->pass_thread()) {
    signature->append(T_ADDRESS);
    args->append(getThreadPointer());
  }

  for (int i = 0; i < x->number_of_arguments(); i++) {
    Value a = x->argument_at(i);
    LIRItem* item = new LIRItem(a, this);
    item->load_item();
    args->append(item->result());
    signature->append(as_BasicType(a->type()));
  }

  LIR_Opr result = call_runtime(signature, args, x->entry(), x->type(), NULL);
  if (x->type() == voidType) {
    set_no_result(x);
  } else {
    __ move(result, rlock_result(x));
  }
}

LIR_Opr LIRGenerator::call_runtime(Value arg1, address entry, ValueType* result_type, CodeEmitInfo* info) {
  LIRItemList args(1);
  LIRItem value(arg1, this);
  args.append(&value);
  BasicTypeList signature;
  signature.append(as_BasicType(arg1->type()));

  return call_runtime(&signature, &args, entry, result_type, info);
}


LIR_Opr LIRGenerator::call_runtime(Value arg1, Value arg2, address entry, ValueType* result_type, CodeEmitInfo* info) {
  LIRItemList args(2);
  LIRItem value1(arg1, this);
  LIRItem value2(arg2, this);
  args.append(&value1);
  args.append(&value2);
  BasicTypeList signature;
  signature.append(as_BasicType(arg1->type()));
  signature.append(as_BasicType(arg2->type()));

  return call_runtime(&signature, &args, entry, result_type, info);
}


LIR_Opr LIRGenerator::call_runtime(BasicTypeArray* signature, LIR_OprList* args,
                                   address entry, ValueType* result_type, CodeEmitInfo* info) {
  // get a result register
  LIR_Opr phys_reg = LIR_OprFact::illegalOpr;
  LIR_Opr result = LIR_OprFact::illegalOpr;
  if (result_type->tag() != voidTag) {
    result = new_register(result_type);
    phys_reg = result_register_for(result_type);
  }

  // move the arguments into the correct location
  CallingConvention* cc = frame_map()->c_calling_convention(signature);
  assert(cc->length() == args->length(), "argument mismatch");
  for (int i = 0; i < args->length(); i++) {
    LIR_Opr arg = args->at(i);
    LIR_Opr loc = cc->at(i);
    if (loc->is_register()) {
      __ move(arg, loc);
    } else {
      LIR_Address* addr = loc->as_address_ptr();
//           if (!can_store_as_constant(arg)) {
//             LIR_Opr tmp = new_register(arg->type());
//             __ move(arg, tmp);
//             arg = tmp;
//           }
      if (addr->type() == T_LONG || addr->type() == T_DOUBLE) {
        __ unaligned_move(arg, addr);
      } else {
        __ move(arg, addr);
      }
    }
  }

  if (info) {
    __ call_runtime(entry, getThreadTemp(), phys_reg, cc->args(), info);
  } else {
    __ call_runtime_leaf(entry, getThreadTemp(), phys_reg, cc->args());
  }
  if (result->is_valid()) {
    __ move(phys_reg, result);
  }
  return result;
}


LIR_Opr LIRGenerator::call_runtime(BasicTypeArray* signature, LIRItemList* args,
                                   address entry, ValueType* result_type, CodeEmitInfo* info) {
  // get a result register
  LIR_Opr phys_reg = LIR_OprFact::illegalOpr;
  LIR_Opr result = LIR_OprFact::illegalOpr;
  if (result_type->tag() != voidTag) {
    result = new_register(result_type);
    phys_reg = result_register_for(result_type);
  }

  // move the arguments into the correct location
  CallingConvention* cc = frame_map()->c_calling_convention(signature);

  assert(cc->length() == args->length(), "argument mismatch");
  for (int i = 0; i < args->length(); i++) {
    LIRItem* arg = args->at(i);
    LIR_Opr loc = cc->at(i);
    if (loc->is_register()) {
      arg->load_item_force(loc);
    } else {
      LIR_Address* addr = loc->as_address_ptr();
      arg->load_for_store(addr->type());
      if (addr->type() == T_LONG || addr->type() == T_DOUBLE) {
        __ unaligned_move(arg->result(), addr);
      } else {
        __ move(arg->result(), addr);
      }
    }
  }

  if (info) {
    __ call_runtime(entry, getThreadTemp(), phys_reg, cc->args(), info);
  } else {
    __ call_runtime_leaf(entry, getThreadTemp(), phys_reg, cc->args());
  }
  if (result->is_valid()) {
    __ move(phys_reg, result);
  }
  return result;
}