compile.cpp

/*
 * Copyright (c) 1997, 2013, 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 "asm/macroAssembler.hpp"
#include "asm/macroAssembler.inline.hpp"
#include "classfile/systemDictionary.hpp"
#include "code/exceptionHandlerTable.hpp"
#include "code/nmethod.hpp"
#include "compiler/compileLog.hpp"
#include "compiler/disassembler.hpp"
#include "compiler/oopMap.hpp"
#include "opto/addnode.hpp"
#include "opto/block.hpp"
#include "opto/c2compiler.hpp"
#include "opto/callGenerator.hpp"
#include "opto/callnode.hpp"
#include "opto/cfgnode.hpp"
#include "opto/chaitin.hpp"
#include "opto/compile.hpp"
#include "opto/connode.hpp"
#include "opto/divnode.hpp"
#include "opto/escape.hpp"
#include "opto/idealGraphPrinter.hpp"
#include "opto/loopnode.hpp"
#include "opto/machnode.hpp"
#include "opto/macro.hpp"
#include "opto/matcher.hpp"
#include "opto/mathexactnode.hpp"
#include "opto/memnode.hpp"
#include "opto/mulnode.hpp"
#include "opto/node.hpp"
#include "opto/opcodes.hpp"
#include "opto/output.hpp"
#include "opto/parse.hpp"
#include "opto/phaseX.hpp"
#include "opto/rootnode.hpp"
#include "opto/runtime.hpp"
#include "opto/stringopts.hpp"
#include "opto/type.hpp"
#include "opto/vectornode.hpp"
#include "runtime/arguments.hpp"
#include "runtime/signature.hpp"
#include "runtime/stubRoutines.hpp"
#include "runtime/timer.hpp"
#include "trace/tracing.hpp"
#include "utilities/copy.hpp"
#ifdef TARGET_ARCH_MODEL_x86_32
# include "adfiles/ad_x86_32.hpp"
#endif
#ifdef TARGET_ARCH_MODEL_x86_64
# include "adfiles/ad_x86_64.hpp"
#endif
#ifdef TARGET_ARCH_MODEL_sparc
# include "adfiles/ad_sparc.hpp"
#endif
#ifdef TARGET_ARCH_MODEL_zero
# include "adfiles/ad_zero.hpp"
#endif
#ifdef TARGET_ARCH_MODEL_arm
# include "adfiles/ad_arm.hpp"
#endif
#ifdef TARGET_ARCH_MODEL_ppc_32
# include "adfiles/ad_ppc_32.hpp"
#endif
#ifdef TARGET_ARCH_MODEL_ppc_64
# include "adfiles/ad_ppc_64.hpp"
#endif


// -------------------- Compile::mach_constant_base_node -----------------------
// Constant table base node singleton.
MachConstantBaseNode* Compile::mach_constant_base_node() {
  if (_mach_constant_base_node == NULL) {
    _mach_constant_base_node = new (C) MachConstantBaseNode();
    _mach_constant_base_node->add_req(C->root());
  }
  return _mach_constant_base_node;
}


/// Support for intrinsics.

// Return the index at which m must be inserted (or already exists).
// The sort order is by the address of the ciMethod, with is_virtual as minor key.
int Compile::intrinsic_insertion_index(ciMethod* m, bool is_virtual) {
#ifdef ASSERT
  for (int i = 1; i < _intrinsics->length(); i++) {
    CallGenerator* cg1 = _intrinsics->at(i-1);
    CallGenerator* cg2 = _intrinsics->at(i);
    assert(cg1->method() != cg2->method()
           ? cg1->method()     < cg2->method()
           : cg1->is_virtual() < cg2->is_virtual(),
           "compiler intrinsics list must stay sorted");
  }
#endif
  // Binary search sorted list, in decreasing intervals [lo, hi].
  int lo = 0, hi = _intrinsics->length()-1;
  while (lo <= hi) {
    int mid = (uint)(hi + lo) / 2;
    ciMethod* mid_m = _intrinsics->at(mid)->method();
    if (m < mid_m) {
      hi = mid-1;
    } else if (m > mid_m) {
      lo = mid+1;
    } else {
      // look at minor sort key
      bool mid_virt = _intrinsics->at(mid)->is_virtual();
      if (is_virtual < mid_virt) {
        hi = mid-1;
      } else if (is_virtual > mid_virt) {
        lo = mid+1;
      } else {
        return mid;  // exact match
      }
    }
  }
  return lo;  // inexact match
}

void Compile::register_intrinsic(CallGenerator* cg) {
  if (_intrinsics == NULL) {
    _intrinsics = new (comp_arena())GrowableArray<CallGenerator*>(comp_arena(), 60, 0, NULL);
  }
  // This code is stolen from ciObjectFactory::insert.
  // Really, GrowableArray should have methods for
  // insert_at, remove_at, and binary_search.
  int len = _intrinsics->length();
  int index = intrinsic_insertion_index(cg->method(), cg->is_virtual());
  if (index == len) {
    _intrinsics->append(cg);
  } else {
#ifdef ASSERT
    CallGenerator* oldcg = _intrinsics->at(index);
    assert(oldcg->method() != cg->method() || oldcg->is_virtual() != cg->is_virtual(), "don't register twice");
#endif
    _intrinsics->append(_intrinsics->at(len-1));
    int pos;
    for (pos = len-2; pos >= index; pos--) {
      _intrinsics->at_put(pos+1,_intrinsics->at(pos));
    }
    _intrinsics->at_put(index, cg);
  }
  assert(find_intrinsic(cg->method(), cg->is_virtual()) == cg, "registration worked");
}

CallGenerator* Compile::find_intrinsic(ciMethod* m, bool is_virtual) {
  assert(m->is_loaded(), "don't try this on unloaded methods");
  if (_intrinsics != NULL) {
    int index = intrinsic_insertion_index(m, is_virtual);
    if (index < _intrinsics->length()
        && _intrinsics->at(index)->method() == m
        && _intrinsics->at(index)->is_virtual() == is_virtual) {
      return _intrinsics->at(index);
    }
  }
  // Lazily create intrinsics for intrinsic IDs well-known in the runtime.
  if (m->intrinsic_id() != vmIntrinsics::_none &&
      m->intrinsic_id() <= vmIntrinsics::LAST_COMPILER_INLINE) {
    CallGenerator* cg = make_vm_intrinsic(m, is_virtual);
    if (cg != NULL) {
      // Save it for next time:
      register_intrinsic(cg);
      return cg;
    } else {
      gather_intrinsic_statistics(m->intrinsic_id(), is_virtual, _intrinsic_disabled);
    }
  }
  return NULL;
}

// Compile:: register_library_intrinsics and make_vm_intrinsic are defined
// in library_call.cpp.


#ifndef PRODUCT
// statistics gathering...

juint  Compile::_intrinsic_hist_count[vmIntrinsics::ID_LIMIT] = {0};
jubyte Compile::_intrinsic_hist_flags[vmIntrinsics::ID_LIMIT] = {0};

bool Compile::gather_intrinsic_statistics(vmIntrinsics::ID id, bool is_virtual, int flags) {
  assert(id > vmIntrinsics::_none && id < vmIntrinsics::ID_LIMIT, "oob");
  int oflags = _intrinsic_hist_flags[id];
  assert(flags != 0, "what happened?");
  if (is_virtual) {
    flags |= _intrinsic_virtual;
  }
  bool changed = (flags != oflags);
  if ((flags & _intrinsic_worked) != 0) {
    juint count = (_intrinsic_hist_count[id] += 1);
    if (count == 1) {
      changed = true;           // first time
    }
    // increment the overall count also:
    _intrinsic_hist_count[vmIntrinsics::_none] += 1;
  }
  if (changed) {
    if (((oflags ^ flags) & _intrinsic_virtual) != 0) {
      // Something changed about the intrinsic's virtuality.
      if ((flags & _intrinsic_virtual) != 0) {
        // This is the first use of this intrinsic as a virtual call.
        if (oflags != 0) {
          // We already saw it as a non-virtual, so note both cases.
          flags |= _intrinsic_both;
        }
      } else if ((oflags & _intrinsic_both) == 0) {
        // This is the first use of this intrinsic as a non-virtual
        flags |= _intrinsic_both;
      }
    }
    _intrinsic_hist_flags[id] = (jubyte) (oflags | flags);
  }
  // update the overall flags also:
  _intrinsic_hist_flags[vmIntrinsics::_none] |= (jubyte) flags;
  return changed;
}

static char* format_flags(int flags, char* buf) {
  buf[0] = 0;
  if ((flags & Compile::_intrinsic_worked) != 0)    strcat(buf, ",worked");
  if ((flags & Compile::_intrinsic_failed) != 0)    strcat(buf, ",failed");
  if ((flags & Compile::_intrinsic_disabled) != 0)  strcat(buf, ",disabled");
  if ((flags & Compile::_intrinsic_virtual) != 0)   strcat(buf, ",virtual");
  if ((flags & Compile::_intrinsic_both) != 0)      strcat(buf, ",nonvirtual");
  if (buf[0] == 0)  strcat(buf, ",");
  assert(buf[0] == ',', "must be");
  return &buf[1];
}

void Compile::print_intrinsic_statistics() {
  char flagsbuf[100];
  ttyLocker ttyl;
  if (xtty != NULL)  xtty->head("statistics type='intrinsic'");
  tty->print_cr("Compiler intrinsic usage:");
  juint total = _intrinsic_hist_count[vmIntrinsics::_none];
  if (total == 0)  total = 1;  // avoid div0 in case of no successes
  #define PRINT_STAT_LINE(name, c, f) \
    tty->print_cr("  %4d (%4.1f%%) %s (%s)", (int)(c), ((c) * 100.0) / total, name, f);
  for (int index = 1 + (int)vmIntrinsics::_none; index < (int)vmIntrinsics::ID_LIMIT; index++) {
    vmIntrinsics::ID id = (vmIntrinsics::ID) index;
    int   flags = _intrinsic_hist_flags[id];
    juint count = _intrinsic_hist_count[id];
    if ((flags | count) != 0) {
      PRINT_STAT_LINE(vmIntrinsics::name_at(id), count, format_flags(flags, flagsbuf));
    }
  }
  PRINT_STAT_LINE("total", total, format_flags(_intrinsic_hist_flags[vmIntrinsics::_none], flagsbuf));
  if (xtty != NULL)  xtty->tail("statistics");
}

void Compile::print_statistics() {
  { ttyLocker ttyl;
    if (xtty != NULL)  xtty->head("statistics type='opto'");
    Parse::print_statistics();
    PhaseCCP::print_statistics();
    PhaseRegAlloc::print_statistics();
    Scheduling::print_statistics();
    PhasePeephole::print_statistics();
    PhaseIdealLoop::print_statistics();
    if (xtty != NULL)  xtty->tail("statistics");
  }
  if (_intrinsic_hist_flags[vmIntrinsics::_none] != 0) {
    // put this under its own <statistics> element.
    print_intrinsic_statistics();
  }
}
#endif //PRODUCT

// Support for bundling info
Bundle* Compile::node_bundling(const Node *n) {
  assert(valid_bundle_info(n), "oob");
  return &_node_bundling_base[n->_idx];
}

bool Compile::valid_bundle_info(const Node *n) {
  return (_node_bundling_limit > n->_idx);
}


void Compile::gvn_replace_by(Node* n, Node* nn) {
  for (DUIterator_Last imin, i = n->last_outs(imin); i >= imin; ) {
    Node* use = n->last_out(i);
    bool is_in_table = initial_gvn()->hash_delete(use);
    uint uses_found = 0;
    for (uint j = 0; j < use->len(); j++) {
      if (use->in(j) == n) {
        if (j < use->req())
          use->set_req(j, nn);
        else
          use->set_prec(j, nn);
        uses_found++;
      }
    }
    if (is_in_table) {
      // reinsert into table
      initial_gvn()->hash_find_insert(use);
    }
    record_for_igvn(use);
    i -= uses_found;    // we deleted 1 or more copies of this edge
  }
}


static inline bool not_a_node(const Node* n) {
  if (n == NULL)                   return true;
  if (((intptr_t)n & 1) != 0)      return true;  // uninitialized, etc.
  if (*(address*)n == badAddress)  return true;  // kill by Node::destruct
  return false;
}

// Identify all nodes that are reachable from below, useful.
// Use breadth-first pass that records state in a Unique_Node_List,
// recursive traversal is slower.
void Compile::identify_useful_nodes(Unique_Node_List &useful) {
  int estimated_worklist_size = unique();
  useful.map( estimated_worklist_size, NULL );  // preallocate space

  // Initialize worklist
  if (root() != NULL)     { useful.push(root()); }
  // If 'top' is cached, declare it useful to preserve cached node
  if( cached_top_node() ) { useful.push(cached_top_node()); }

  // Push all useful nodes onto the list, breadthfirst
  for( uint next = 0; next < useful.size(); ++next ) {
    assert( next < unique(), "Unique useful nodes < total nodes");
    Node *n  = useful.at(next);
    uint max = n->len();
    for( uint i = 0; i < max; ++i ) {
      Node *m = n->in(i);
      if (not_a_node(m))  continue;
      useful.push(m);
    }
  }
}

// Update dead_node_list with any missing dead nodes using useful
// list. Consider all non-useful nodes to be useless i.e., dead nodes.
void Compile::update_dead_node_list(Unique_Node_List &useful) {
  uint max_idx = unique();
  VectorSet& useful_node_set = useful.member_set();

  for (uint node_idx = 0; node_idx < max_idx; node_idx++) {
    // If node with index node_idx is not in useful set,
    // mark it as dead in dead node list.
    if (! useful_node_set.test(node_idx) ) {
      record_dead_node(node_idx);
    }
  }
}

void Compile::remove_useless_late_inlines(GrowableArray<CallGenerator*>* inlines, Unique_Node_List &useful) {
  int shift = 0;
  for (int i = 0; i < inlines->length(); i++) {
    CallGenerator* cg = inlines->at(i);
    CallNode* call = cg->call_node();
    if (shift > 0) {
      inlines->at_put(i-shift, cg);
    }
    if (!useful.member(call)) {
      shift++;
    }
  }
  inlines->trunc_to(inlines->length()-shift);
}

// Disconnect all useless nodes by disconnecting those at the boundary.
void Compile::remove_useless_nodes(Unique_Node_List &useful) {
  uint next = 0;
  while (next < useful.size()) {
    Node *n = useful.at(next++);
    // Use raw traversal of out edges since this code removes out edges
    int max = n->outcnt();
    for (int j = 0; j < max; ++j) {
      Node* child = n->raw_out(j);
      if (! useful.member(child)) {
        assert(!child->is_top() || child != top(),
               "If top is cached in Compile object it is in useful list");
        // Only need to remove this out-edge to the useless node
        n->raw_del_out(j);
        --j;
        --max;
      }
    }
    if (n->outcnt() == 1 && n->has_special_unique_user()) {
      record_for_igvn(n->unique_out());
    }
  }
  // Remove useless macro and predicate opaq nodes
  for (int i = C->macro_count()-1; i >= 0; i--) {
    Node* n = C->macro_node(i);
    if (!useful.member(n)) {
      remove_macro_node(n);
    }
  }
  // Remove useless expensive node
  for (int i = C->expensive_count()-1; i >= 0; i--) {
    Node* n = C->expensive_node(i);
    if (!useful.member(n)) {
      remove_expensive_node(n);
    }
  }
  // clean up the late inline lists
  remove_useless_late_inlines(&_string_late_inlines, useful);
  remove_useless_late_inlines(&_boxing_late_inlines, useful);
  remove_useless_late_inlines(&_late_inlines, useful);
  debug_only(verify_graph_edges(true/*check for no_dead_code*/);)
}

//------------------------------frame_size_in_words-----------------------------
// frame_slots in units of words
int Compile::frame_size_in_words() const {
  // shift is 0 in LP32 and 1 in LP64
  const int shift = (LogBytesPerWord - LogBytesPerInt);
  int words = _frame_slots >> shift;
  assert( words << shift == _frame_slots, "frame size must be properly aligned in LP64" );
  return words;
}

// ============================================================================
//------------------------------CompileWrapper---------------------------------
class CompileWrapper : public StackObj {
  Compile *const _compile;
 public:
  CompileWrapper(Compile* compile);

  ~CompileWrapper();
};

CompileWrapper::CompileWrapper(Compile* compile) : _compile(compile) {
  // the Compile* pointer is stored in the current ciEnv:
  ciEnv* env = compile->env();
  assert(env == ciEnv::current(), "must already be a ciEnv active");
  assert(env->compiler_data() == NULL, "compile already active?");
  env->set_compiler_data(compile);
  assert(compile == Compile::current(), "sanity");

  compile->set_type_dict(NULL);
  compile->set_type_hwm(NULL);
  compile->set_type_last_size(0);
  compile->set_last_tf(NULL, NULL);
  compile->set_indexSet_arena(NULL);
  compile->set_indexSet_free_block_list(NULL);
  compile->init_type_arena();
  Type::Initialize(compile);
  _compile->set_scratch_buffer_blob(NULL);
  _compile->begin_method();
}
CompileWrapper::~CompileWrapper() {
  _compile->end_method();
  if (_compile->scratch_buffer_blob() != NULL)
    BufferBlob::free(_compile->scratch_buffer_blob());
  _compile->env()->set_compiler_data(NULL);
}


//----------------------------print_compile_messages---------------------------
void Compile::print_compile_messages() {
#ifndef PRODUCT
  // Check if recompiling
  if (_subsume_loads == false && PrintOpto) {
    // Recompiling without allowing machine instructions to subsume loads
    tty->print_cr("*********************************************************");
    tty->print_cr("** Bailout: Recompile without subsuming loads          **");
    tty->print_cr("*********************************************************");
  }
  if (_do_escape_analysis != DoEscapeAnalysis && PrintOpto) {
    // Recompiling without escape analysis
    tty->print_cr("*********************************************************");
    tty->print_cr("** Bailout: Recompile without escape analysis          **");
    tty->print_cr("*********************************************************");
  }
  if (_eliminate_boxing != EliminateAutoBox && PrintOpto) {
    // Recompiling without boxing elimination
    tty->print_cr("*********************************************************");
    tty->print_cr("** Bailout: Recompile without boxing elimination       **");
    tty->print_cr("*********************************************************");
  }
  if (env()->break_at_compile()) {
    // Open the debugger when compiling this method.
    tty->print("### Breaking when compiling: ");
    method()->print_short_name();
    tty->cr();
    BREAKPOINT;
  }

  if( PrintOpto ) {
    if (is_osr_compilation()) {
      tty->print("[OSR]%3d", _compile_id);
    } else {
      tty->print("%3d", _compile_id);
    }
  }
#endif
}


//-----------------------init_scratch_buffer_blob------------------------------
// Construct a temporary BufferBlob and cache it for this compile.
void Compile::init_scratch_buffer_blob(int const_size) {
  // If there is already a scratch buffer blob allocated and the
  // constant section is big enough, use it.  Otherwise free the
  // current and allocate a new one.
  BufferBlob* blob = scratch_buffer_blob();
  if ((blob != NULL) && (const_size <= _scratch_const_size)) {
    // Use the current blob.
  } else {
    if (blob != NULL) {
      BufferBlob::free(blob);
    }

    ResourceMark rm;
    _scratch_const_size = const_size;
    int size = (MAX_inst_size + MAX_stubs_size + _scratch_const_size);
    blob = BufferBlob::create("Compile::scratch_buffer", size);
    // Record the buffer blob for next time.
    set_scratch_buffer_blob(blob);
    // Have we run out of code space?
    if (scratch_buffer_blob() == NULL) {
      // Let CompilerBroker disable further compilations.
      record_failure("Not enough space for scratch buffer in CodeCache");
      return;
    }
  }

  // Initialize the relocation buffers
  relocInfo* locs_buf = (relocInfo*) blob->content_end() - MAX_locs_size;
  set_scratch_locs_memory(locs_buf);
}


//-----------------------scratch_emit_size-------------------------------------
// Helper function that computes size by emitting code
uint Compile::scratch_emit_size(const Node* n) {
  // Start scratch_emit_size section.
  set_in_scratch_emit_size(true);

  // Emit into a trash buffer and count bytes emitted.
  // This is a pretty expensive way to compute a size,
  // but it works well enough if seldom used.
  // All common fixed-size instructions are given a size
  // method by the AD file.
  // Note that the scratch buffer blob and locs memory are
  // allocated at the beginning of the compile task, and
  // may be shared by several calls to scratch_emit_size.
  // The allocation of the scratch buffer blob is particularly
  // expensive, since it has to grab the code cache lock.
  BufferBlob* blob = this->scratch_buffer_blob();
  assert(blob != NULL, "Initialize BufferBlob at start");
  assert(blob->size() > MAX_inst_size, "sanity");
  relocInfo* locs_buf = scratch_locs_memory();
  address blob_begin = blob->content_begin();
  address blob_end   = (address)locs_buf;
  assert(blob->content_contains(blob_end), "sanity");
  CodeBuffer buf(blob_begin, blob_end - blob_begin);
  buf.initialize_consts_size(_scratch_const_size);
  buf.initialize_stubs_size(MAX_stubs_size);
  assert(locs_buf != NULL, "sanity");
  int lsize = MAX_locs_size / 3;
  buf.consts()->initialize_shared_locs(&locs_buf[lsize * 0], lsize);
  buf.insts()->initialize_shared_locs( &locs_buf[lsize * 1], lsize);
  buf.stubs()->initialize_shared_locs( &locs_buf[lsize * 2], lsize);

  // Do the emission.

  Label fakeL; // Fake label for branch instructions.
  Label*   saveL = NULL;
  uint save_bnum = 0;
  bool is_branch = n->is_MachBranch();
  if (is_branch) {
    MacroAssembler masm(&buf);
    masm.bind(fakeL);
    n->as_MachBranch()->save_label(&saveL, &save_bnum);
    n->as_MachBranch()->label_set(&fakeL, 0);
  }
  n->emit(buf, this->regalloc());
  if (is_branch) // Restore label.
    n->as_MachBranch()->label_set(saveL, save_bnum);

  // End scratch_emit_size section.
  set_in_scratch_emit_size(false);

  return buf.insts_size();
}


// ============================================================================
//------------------------------Compile standard-------------------------------
debug_only( int Compile::_debug_idx = 100000; )

// Compile a method.  entry_bci is -1 for normal compilations and indicates
// the continuation bci for on stack replacement.


Compile::Compile( ciEnv* ci_env, C2Compiler* compiler, ciMethod* target, int osr_bci,
                  bool subsume_loads, bool do_escape_analysis, bool eliminate_boxing )
                : Phase(Compiler),
                  _env(ci_env),
                  _log(ci_env->log()),
                  _compile_id(ci_env->compile_id()),
                  _save_argument_registers(false),
                  _stub_name(NULL),
                  _stub_function(NULL),
                  _stub_entry_point(NULL),
                  _method(target),
                  _entry_bci(osr_bci),
                  _initial_gvn(NULL),
                  _for_igvn(NULL),
                  _warm_calls(NULL),
                  _subsume_loads(subsume_loads),
                  _do_escape_analysis(do_escape_analysis),
                  _eliminate_boxing(eliminate_boxing),
                  _failure_reason(NULL),
                  _code_buffer("Compile::Fill_buffer"),
                  _orig_pc_slot(0),
                  _orig_pc_slot_offset_in_bytes(0),
                  _has_method_handle_invokes(false),
                  _mach_constant_base_node(NULL),
                  _node_bundling_limit(0),
                  _node_bundling_base(NULL),
                  _java_calls(0),
                  _inner_loops(0),
                  _scratch_const_size(-1),
                  _in_scratch_emit_size(false),
                  _dead_node_list(comp_arena()),
                  _dead_node_count(0),
#ifndef PRODUCT
                  _trace_opto_output(TraceOptoOutput || method()->has_option("TraceOptoOutput")),
                  _in_dump_cnt(0),
                  _printer(IdealGraphPrinter::printer()),
#endif
                  _congraph(NULL),
                  _late_inlines(comp_arena(), 2, 0, NULL),
                  _string_late_inlines(comp_arena(), 2, 0, NULL),
                  _boxing_late_inlines(comp_arena(), 2, 0, NULL),
                  _late_inlines_pos(0),
                  _number_of_mh_late_inlines(0),
                  _inlining_progress(false),
                  _inlining_incrementally(false),
                  _print_inlining_list(NULL),
                  _print_inlining_idx(0),
                  _preserve_jvm_state(0) {
  C = this;

  CompileWrapper cw(this);
#ifndef PRODUCT
  if (TimeCompiler2) {
    tty->print(" ");
    target->holder()->name()->print();
    tty->print(".");
    target->print_short_name();
    tty->print("  ");
  }
  TraceTime t1("Total compilation time", &_t_totalCompilation, TimeCompiler, TimeCompiler2);
  TraceTime t2(NULL, &_t_methodCompilation, TimeCompiler, false);
  bool print_opto_assembly = PrintOptoAssembly || _method->has_option("PrintOptoAssembly");
  if (!print_opto_assembly) {
    bool print_assembly = (PrintAssembly || _method->should_print_assembly());
    if (print_assembly && !Disassembler::can_decode()) {
      tty->print_cr("PrintAssembly request changed to PrintOptoAssembly");
      print_opto_assembly = true;
    }
  }
  set_print_assembly(print_opto_assembly);
  set_parsed_irreducible_loop(false);
#endif
  set_print_inlining(PrintInlining || method()->has_option("PrintInlining") NOT_PRODUCT( || PrintOptoInlining));
  set_print_intrinsics(PrintIntrinsics || method()->has_option("PrintIntrinsics"));

  if (ProfileTraps) {
    // Make sure the method being compiled gets its own MDO,
    // so we can at least track the decompile_count().
    method()->ensure_method_data();
  }

  Init(::AliasLevel);


  print_compile_messages();

  if (UseOldInlining || PrintCompilation NOT_PRODUCT( || PrintOpto) )
    _ilt = InlineTree::build_inline_tree_root();
  else
    _ilt = NULL;

  // Even if NO memory addresses are used, MergeMem nodes must have at least 1 slice
  assert(num_alias_types() >= AliasIdxRaw, "");

#define MINIMUM_NODE_HASH  1023
  // Node list that Iterative GVN will start with
  Unique_Node_List for_igvn(comp_arena());
  set_for_igvn(&for_igvn);

  // GVN that will be run immediately on new nodes
  uint estimated_size = method()->code_size()*4+64;
  estimated_size = (estimated_size < MINIMUM_NODE_HASH ? MINIMUM_NODE_HASH : estimated_size);
  PhaseGVN gvn(node_arena(), estimated_size);
  set_initial_gvn(&gvn);

  if (print_inlining() || print_intrinsics()) {
    _print_inlining_list = new (comp_arena())GrowableArray<PrintInliningBuffer>(comp_arena(), 1, 1, PrintInliningBuffer());
  }
  { // Scope for timing the parser
    TracePhase t3("parse", &_t_parser, true);

    // Put top into the hash table ASAP.
    initial_gvn()->transform_no_reclaim(top());

    // Set up tf(), start(), and find a CallGenerator.
    CallGenerator* cg = NULL;
    if (is_osr_compilation()) {
      const TypeTuple *domain = StartOSRNode::osr_domain();
      const TypeTuple *range = TypeTuple::make_range(method()->signature());
      init_tf(TypeFunc::make(domain, range));
      StartNode* s = new (this) StartOSRNode(root(), domain);
      initial_gvn()->set_type_bottom(s);
      init_start(s);
      cg = CallGenerator::for_osr(method(), entry_bci());
    } else {
      // Normal case.
      init_tf(TypeFunc::make(method()));
      StartNode* s = new (this) StartNode(root(), tf()->domain());
      initial_gvn()->set_type_bottom(s);
      init_start(s);
      if (method()->intrinsic_id() == vmIntrinsics::_Reference_get && UseG1GC) {
        // With java.lang.ref.reference.get() we must go through the
        // intrinsic when G1 is enabled - even when get() is the root
        // method of the compile - so that, if necessary, the value in
        // the referent field of the reference object gets recorded by
        // the pre-barrier code.
        // Specifically, if G1 is enabled, the value in the referent
        // field is recorded by the G1 SATB pre barrier. This will
        // result in the referent being marked live and the reference
        // object removed from the list of discovered references during
        // reference processing.
        cg = find_intrinsic(method(), false);
      }
      if (cg == NULL) {
        float past_uses = method()->interpreter_invocation_count();
        float expected_uses = past_uses;
        cg = CallGenerator::for_inline(method(), expected_uses);
      }
    }
    if (failing())  return;
    if (cg == NULL) {
      record_method_not_compilable_all_tiers("cannot parse method");
      return;
    }
    JVMState* jvms = build_start_state(start(), tf());
    if ((jvms = cg->generate(jvms, NULL)) == NULL) {
      record_method_not_compilable("method parse failed");
      return;
    }
    GraphKit kit(jvms);

    if (!kit.stopped()) {
      // Accept return values, and transfer control we know not where.
      // This is done by a special, unique ReturnNode bound to root.
      return_values(kit.jvms());
    }

    if (kit.has_exceptions()) {
      // Any exceptions that escape from this call must be rethrown
      // to whatever caller is dynamically above us on the stack.
      // This is done by a special, unique RethrowNode bound to root.
      rethrow_exceptions(kit.transfer_exceptions_into_jvms());
    }

    assert(IncrementalInline || (_late_inlines.length() == 0 && !has_mh_late_inlines()), "incremental inlining is off");

    if (_late_inlines.length() == 0 && !has_mh_late_inlines() && !failing() && has_stringbuilder()) {
      inline_string_calls(true);
    }

    if (failing())  return;

    print_method(PHASE_BEFORE_REMOVEUSELESS, 3);

    // Remove clutter produced by parsing.
    if (!failing()) {
      ResourceMark rm;
      PhaseRemoveUseless pru(initial_gvn(), &for_igvn);
    }
  }

  // Note:  Large methods are capped off in do_one_bytecode().
  if (failing())  return;

  // After parsing, node notes are no longer automagic.
  // They must be propagated by register_new_node_with_optimizer(),
  // clone(), or the like.
  set_default_node_notes(NULL);

  for (;;) {
    int successes = Inline_Warm();
    if (failing())  return;
    if (successes == 0)  break;
  }

  // Drain the list.
  Finish_Warm();
#ifndef PRODUCT
  if (_printer) {
    _printer->print_inlining(this);
  }
#endif

  if (failing())  return;
  NOT_PRODUCT( verify_graph_edges(); )

  // Now optimize
  Optimize();
  if (failing())  return;
  NOT_PRODUCT( verify_graph_edges(); )

#ifndef PRODUCT
  if (PrintIdeal) {
    ttyLocker ttyl;  // keep the following output all in one block
    // This output goes directly to the tty, not the compiler log.
    // To enable tools to match it up with the compilation activity,
    // be sure to tag this tty output with the compile ID.
    if (xtty != NULL) {
      xtty->head("ideal compile_id='%d'%s", compile_id(),
                 is_osr_compilation()    ? " compile_kind='osr'" :
                 "");
    }
    root()->dump(9999);
    if (xtty != NULL) {
      xtty->tail("ideal");
    }
  }
#endif

  NOT_PRODUCT( verify_barriers(); )
  // Now that we know the size of all the monitors we can add a fixed slot
  // for the original deopt pc.

  _orig_pc_slot =  fixed_slots();
  int next_slot = _orig_pc_slot + (sizeof(address) / VMRegImpl::stack_slot_size);
  set_fixed_slots(next_slot);

  // Compute when to use implicit null checks. Used by matching trap based
  // nodes and NullCheck optimization.
  set_allowed_deopt_reasons();

  // Now generate code
  Code_Gen();
  if (failing())  return;

  // Check if we want to skip execution of all compiled code.
  {
#ifndef PRODUCT
    if (OptoNoExecute) {
      record_method_not_compilable("+OptoNoExecute");  // Flag as failed
      return;
    }
    TracePhase t2("install_code", &_t_registerMethod, TimeCompiler);
#endif

    if (is_osr_compilation()) {
      _code_offsets.set_value(CodeOffsets::Verified_Entry, 0);
      _code_offsets.set_value(CodeOffsets::OSR_Entry, _first_block_size);
    } else {
      _code_offsets.set_value(CodeOffsets::Verified_Entry, _first_block_size);
      _code_offsets.set_value(CodeOffsets::OSR_Entry, 0);
    }

    env()->register_method(_method, _entry_bci,
                           &_code_offsets,
                           _orig_pc_slot_offset_in_bytes,
                           code_buffer(),
                           frame_size_in_words(), _oop_map_set,
                           &_handler_table, &_inc_table,
                           compiler,
                           env()->comp_level(),
                           has_unsafe_access(),
                           SharedRuntime::is_wide_vector(max_vector_size())
                           );

    if (log() != NULL) // Print code cache state into compiler log
      log()->code_cache_state();
  }
}

//------------------------------Compile----------------------------------------
// Compile a runtime stub
Compile::Compile( ciEnv* ci_env,
                  TypeFunc_generator generator,
                  address stub_function,
                  const char *stub_name,
                  int is_fancy_jump,
                  bool pass_tls,
                  bool save_arg_registers,
                  bool return_pc )
  : Phase(Compiler),
    _env(ci_env),
    _log(ci_env->log()),
    _compile_id(0),
    _save_argument_registers(save_arg_registers),
    _method(NULL),
    _stub_name(stub_name),
    _stub_function(stub_function),
    _stub_entry_point(NULL),
    _entry_bci(InvocationEntryBci),
    _initial_gvn(NULL),
    _for_igvn(NULL),
    _warm_calls(NULL),
    _orig_pc_slot(0),
    _orig_pc_slot_offset_in_bytes(0),
    _subsume_loads(true),
    _do_escape_analysis(false),
    _eliminate_boxing(false),
    _failure_reason(NULL),
    _code_buffer("Compile::Fill_buffer"),
    _has_method_handle_invokes(false),
    _mach_constant_base_node(NULL),
    _node_bundling_limit(0),
    _node_bundling_base(NULL),
    _java_calls(0),
    _inner_loops(0),
#ifndef PRODUCT
    _trace_opto_output(TraceOptoOutput),
    _in_dump_cnt(0),
    _printer(NULL),
#endif
    _dead_node_list(comp_arena()),
    _dead_node_count(0),
    _congraph(NULL),
    _number_of_mh_late_inlines(0),
    _inlining_progress(false),
    _inlining_incrementally(false),
    _print_inlining_list(NULL),
    _print_inlining_idx(0),
    _preserve_jvm_state(0),
    _allowed_reasons(0) {
  C = this;

#ifndef PRODUCT
  TraceTime t1(NULL, &_t_totalCompilation, TimeCompiler, false);
  TraceTime t2(NULL, &_t_stubCompilation, TimeCompiler, false);
  set_print_assembly(PrintFrameConverterAssembly);
  set_parsed_irreducible_loop(false);
#endif
  CompileWrapper cw(this);
  Init(/*AliasLevel=*/ 0);
  init_tf((*generator)());

  {
    // The following is a dummy for the sake of GraphKit::gen_stub
    Unique_Node_List for_igvn(comp_arena());
    set_for_igvn(&for_igvn);  // not used, but some GraphKit guys push on this
    PhaseGVN gvn(Thread::current()->resource_area(),255);
    set_initial_gvn(&gvn);    // not significant, but GraphKit guys use it pervasively
    gvn.transform_no_reclaim(top());

    GraphKit kit;
    kit.gen_stub(stub_function, stub_name, is_fancy_jump, pass_tls, return_pc);
  }

  NOT_PRODUCT( verify_graph_edges(); )
  Code_Gen();
  if (failing())  return;


  // Entry point will be accessed using compile->stub_entry_point();
  if (code_buffer() == NULL) {
    Matcher::soft_match_failure();
  } else {
    if (PrintAssembly && (WizardMode || Verbose))
      tty->print_cr("### Stub::%s", stub_name);

    if (!failing()) {
      assert(_fixed_slots == 0, "no fixed slots used for runtime stubs");

      // Make the NMethod
      // For now we mark the frame as never safe for profile stackwalking
      RuntimeStub *rs = RuntimeStub::new_runtime_stub(stub_name,
                                                      code_buffer(),
                                                      CodeOffsets::frame_never_safe,
                                                      // _code_offsets.value(CodeOffsets::Frame_Complete),
                                                      frame_size_in_words(),
                                                      _oop_map_set,
                                                      save_arg_registers);
      assert(rs != NULL && rs->is_runtime_stub(), "sanity check");

      _stub_entry_point = rs->entry_point();
    }
  }
}

//------------------------------Init-------------------------------------------
// Prepare for a single compilation
void Compile::Init(int aliaslevel) {
  _unique  = 0;
  _regalloc = NULL;

  _tf      = NULL;  // filled in later
  _top     = NULL;  // cached later
  _matcher = NULL;  // filled in later
  _cfg     = NULL;  // filled in later

  set_24_bit_selection_and_mode(Use24BitFP, false);

  _node_note_array = NULL;
  _default_node_notes = NULL;

  _immutable_memory = NULL; // filled in at first inquiry

  // Globally visible Nodes
  // First set TOP to NULL to give safe behavior during creation of RootNode
  set_cached_top_node(NULL);
  set_root(new (this) RootNode());
  // Now that you have a Root to point to, create the real TOP
  set_cached_top_node( new (this) ConNode(Type::TOP) );
  set_recent_alloc(NULL, NULL);

  // Create Debug Information Recorder to record scopes, oopmaps, etc.
  env()->set_oop_recorder(new OopRecorder(env()->arena()));
  env()->set_debug_info(new DebugInformationRecorder(env()->oop_recorder()));
  env()->set_dependencies(new Dependencies(env()));

  _fixed_slots = 0;
  set_has_split_ifs(false);
  set_has_loops(has_method() && method()->has_loops()); // first approximation
  set_has_stringbuilder(false);
  set_has_boxed_value(false);
  _trap_can_recompile = false;  // no traps emitted yet
  _major_progress = true; // start out assuming good things will happen
  set_has_unsafe_access(false);
  set_max_vector_size(0);
  Copy::zero_to_bytes(_trap_hist, sizeof(_trap_hist));
  set_decompile_count(0);

  set_do_freq_based_layout(BlockLayoutByFrequency || method_has_option("BlockLayoutByFrequency"));
  set_num_loop_opts(LoopOptsCount);
  set_do_inlining(Inline);
  set_max_inline_size(MaxInlineSize);
  set_freq_inline_size(FreqInlineSize);
  set_do_scheduling(OptoScheduling);
  set_do_count_invocations(false);
  set_do_method_data_update(false);

  if (debug_info()->recording_non_safepoints()) {
    set_node_note_array(new(comp_arena()) GrowableArray<Node_Notes*>
                        (comp_arena(), 8, 0, NULL));
    set_default_node_notes(Node_Notes::make(this));
  }

  // // -- Initialize types before each compile --
  // // Update cached type information
  // if( _method && _method->constants() )
  //   Type::update_loaded_types(_method, _method->constants());

  // Init alias_type map.
  if (!_do_escape_analysis && aliaslevel == 3)
    aliaslevel = 2;  // No unique types without escape analysis
  _AliasLevel = aliaslevel;
  const int grow_ats = 16;
  _max_alias_types = grow_ats;
  _alias_types   = NEW_ARENA_ARRAY(comp_arena(), AliasType*, grow_ats);
  AliasType* ats = NEW_ARENA_ARRAY(comp_arena(), AliasType,  grow_ats);
  Copy::zero_to_bytes(ats, sizeof(AliasType)*grow_ats);
  {
    for (int i = 0; i < grow_ats; i++)  _alias_types[i] = &ats[i];
  }
  // Initialize the first few types.
  _alias_types[AliasIdxTop]->Init(AliasIdxTop, NULL);
  _alias_types[AliasIdxBot]->Init(AliasIdxBot, TypePtr::BOTTOM);
  _alias_types[AliasIdxRaw]->Init(AliasIdxRaw, TypeRawPtr::BOTTOM);
  _num_alias_types = AliasIdxRaw+1;
  // Zero out the alias type cache.
  Copy::zero_to_bytes(_alias_cache, sizeof(_alias_cache));
  // A NULL adr_type hits in the cache right away.  Preload the right answer.
  probe_alias_cache(NULL)->_index = AliasIdxTop;

  _intrinsics = NULL;
  _macro_nodes = new(comp_arena()) GrowableArray<Node*>(comp_arena(), 8,  0, NULL);
  _predicate_opaqs = new(comp_arena()) GrowableArray<Node*>(comp_arena(), 8,  0, NULL);
  _expensive_nodes = new(comp_arena()) GrowableArray<Node*>(comp_arena(), 8,  0, NULL);
  register_library_intrinsics();
}

//---------------------------init_start----------------------------------------
// Install the StartNode on this compile object.
void Compile::init_start(StartNode* s) {
  if (failing())
    return; // already failing
  assert(s == start(), "");
}

StartNode* Compile::start() const {
  assert(!failing(), "");
  for (DUIterator_Fast imax, i = root()->fast_outs(imax); i < imax; i++) {
    Node* start = root()->fast_out(i);
    if( start->is_Start() )
      return start->as_Start();
  }
  ShouldNotReachHere();
  return NULL;
}

//-------------------------------immutable_memory-------------------------------------
// Access immutable memory
Node* Compile::immutable_memory() {
  if (_immutable_memory != NULL) {
    return _immutable_memory;
  }
  StartNode* s = start();
  for (DUIterator_Fast imax, i = s->fast_outs(imax); true; i++) {
    Node *p = s->fast_out(i);
    if (p != s && p->as_Proj()->_con == TypeFunc::Memory) {
      _immutable_memory = p;
      return _immutable_memory;
    }
  }
  ShouldNotReachHere();
  return NULL;
}

//----------------------set_cached_top_node------------------------------------
// Install the cached top node, and make sure Node::is_top works correctly.
void Compile::set_cached_top_node(Node* tn) {
  if (tn != NULL)  verify_top(tn);
  Node* old_top = _top;
  _top = tn;
  // Calling Node::setup_is_top allows the nodes the chance to adjust
  // their _out arrays.
  if (_top != NULL)     _top->setup_is_top();
  if (old_top != NULL)  old_top->setup_is_top();
  assert(_top == NULL || top()->is_top(), "");
}

#ifdef ASSERT
uint Compile::count_live_nodes_by_graph_walk() {
  Unique_Node_List useful(comp_arena());
  // Get useful node list by walking the graph.
  identify_useful_nodes(useful);
  return useful.size();
}

void Compile::print_missing_nodes() {

  // Return if CompileLog is NULL and PrintIdealNodeCount is false.
  if ((_log == NULL) && (! PrintIdealNodeCount)) {
    return;
  }

  // This is an expensive function. It is executed only when the user
  // specifies VerifyIdealNodeCount option or otherwise knows the
  // additional work that needs to be done to identify reachable nodes
  // by walking the flow graph and find the missing ones using
  // _dead_node_list.

  Unique_Node_List useful(comp_arena());
  // Get useful node list by walking the graph.
  identify_useful_nodes(useful);

  uint l_nodes = C->live_nodes();
  uint l_nodes_by_walk = useful.size();

  if (l_nodes != l_nodes_by_walk) {
    if (_log != NULL) {
      _log->begin_head("mismatched_nodes count='%d'", abs((int) (l_nodes - l_nodes_by_walk)));
      _log->stamp();
      _log->end_head();
    }
    VectorSet& useful_member_set = useful.member_set();
    int last_idx = l_nodes_by_walk;
    for (int i = 0; i < last_idx; i++) {
      if (useful_member_set.test(i)) {
        if (_dead_node_list.test(i)) {
          if (_log != NULL) {
            _log->elem("mismatched_node_info node_idx='%d' type='both live and dead'", i);
          }
          if (PrintIdealNodeCount) {
            // Print the log message to tty
              tty->print_cr("mismatched_node idx='%d' both live and dead'", i);
              useful.at(i)->dump();
          }
        }
      }
      else if (! _dead_node_list.test(i)) {
        if (_log != NULL) {
          _log->elem("mismatched_node_info node_idx='%d' type='neither live nor dead'", i);
        }
        if (PrintIdealNodeCount) {
          // Print the log message to tty
          tty->print_cr("mismatched_node idx='%d' type='neither live nor dead'", i);
        }
      }
    }
    if (_log != NULL) {
      _log->tail("mismatched_nodes");
    }
  }
}
#endif

#ifndef PRODUCT
void Compile::verify_top(Node* tn) const {
  if (tn != NULL) {
    assert(tn->is_Con(), "top node must be a constant");
    assert(((ConNode*)tn)->type() == Type::TOP, "top node must have correct type");
    assert(tn->in(0) != NULL, "must have live top node");
  }
}
#endif


///-------------------Managing Per-Node Debug & Profile Info-------------------

void Compile::grow_node_notes(GrowableArray<Node_Notes*>* arr, int grow_by) {
  guarantee(arr != NULL, "");
  int num_blocks = arr->length();
  if (grow_by < num_blocks)  grow_by = num_blocks;
  int num_notes = grow_by * _node_notes_block_size;
  Node_Notes* notes = NEW_ARENA_ARRAY(node_arena(), Node_Notes, num_notes);
  Copy::zero_to_bytes(notes, num_notes * sizeof(Node_Notes));
  while (num_notes > 0) {
    arr->append(notes);
    notes     += _node_notes_block_size;
    num_notes -= _node_notes_block_size;
  }
  assert(num_notes == 0, "exact multiple, please");
}

bool Compile::copy_node_notes_to(Node* dest, Node* source) {
  if (source == NULL || dest == NULL)  return false;

  if (dest->is_Con())
    return false;               // Do not push debug info onto constants.

#ifdef ASSERT
  // Leave a bread crumb trail pointing to the original node:
  if (dest != NULL && dest != source && dest->debug_orig() == NULL) {
    dest->set_debug_orig(source);
  }
#endif

  if (node_note_array() == NULL)
    return false;               // Not collecting any notes now.

  // This is a copy onto a pre-existing node, which may already have notes.
  // If both nodes have notes, do not overwrite any pre-existing notes.
  Node_Notes* source_notes = node_notes_at(source->_idx);
  if (source_notes == NULL || source_notes->is_clear())  return false;
  Node_Notes* dest_notes   = node_notes_at(dest->_idx);
  if (dest_notes == NULL || dest_notes->is_clear()) {
    return set_node_notes_at(dest->_idx, source_notes);
  }

  Node_Notes merged_notes = (*source_notes);
  // The order of operations here ensures that dest notes will win...
  merged_notes.update_from(dest_notes);
  return set_node_notes_at(dest->_idx, &merged_notes);
}


//--------------------------allow_range_check_smearing-------------------------
// Gating condition for coalescing similar range checks.
// Sometimes we try 'speculatively' replacing a series of a range checks by a
// single covering check that is at least as strong as any of them.
// If the optimization succeeds, the simplified (strengthened) range check
// will always succeed.  If it fails, we will deopt, and then give up
// on the optimization.
bool Compile::allow_range_check_smearing() const {
  // If this method has already thrown a range-check,
  // assume it was because we already tried range smearing
  // and it failed.
  uint already_trapped = trap_count(Deoptimization::Reason_range_check);
  return !already_trapped;
}


//------------------------------flatten_alias_type-----------------------------
const TypePtr *Compile::flatten_alias_type( const TypePtr *tj ) const {
  int offset = tj->offset();
  TypePtr::PTR ptr = tj->ptr();

  // Known instance (scalarizable allocation) alias only with itself.
  bool is_known_inst = tj->isa_oopptr() != NULL &&
                       tj->is_oopptr()->is_known_instance();

  // Process weird unsafe references.
  if (offset == Type::OffsetBot && (tj->isa_instptr() /*|| tj->isa_klassptr()*/)) {
    assert(InlineUnsafeOps, "indeterminate pointers come only from unsafe ops");
    assert(!is_known_inst, "scalarizable allocation should not have unsafe references");
    tj = TypeOopPtr::BOTTOM;
    ptr = tj->ptr();
    offset = tj->offset();
  }

  // Array pointers need some flattening
  const TypeAryPtr *ta = tj->isa_aryptr();
  if (ta && ta->is_stable()) {
    // Erase stability property for alias analysis.
    tj = ta = ta->cast_to_stable(false);
  }
  if( ta && is_known_inst ) {
    if ( offset != Type::OffsetBot &&
         offset > arrayOopDesc::length_offset_in_bytes() ) {
      offset = Type::OffsetBot; // Flatten constant access into array body only
      tj = ta = TypeAryPtr::make(ptr, ta->ary(), ta->klass(), true, offset, ta->instance_id());
    }
  } else if( ta && _AliasLevel >= 2 ) {
    // For arrays indexed by constant indices, we flatten the alias
    // space to include all of the array body.  Only the header, klass
    // and array length can be accessed un-aliased.
    if( offset != Type::OffsetBot ) {
      if( ta->const_oop() ) { // MethodData* or Method*
        offset = Type::OffsetBot;   // Flatten constant access into array body
        tj = ta = TypeAryPtr::make(ptr,ta->const_oop(),ta->ary(),ta->klass(),false,offset);
      } else if( offset == arrayOopDesc::length_offset_in_bytes() ) {
        // range is OK as-is.
        tj = ta = TypeAryPtr::RANGE;
      } else if( offset == oopDesc::klass_offset_in_bytes() ) {
        tj = TypeInstPtr::KLASS; // all klass loads look alike
        ta = TypeAryPtr::RANGE; // generic ignored junk
        ptr = TypePtr::BotPTR;
      } else if( offset == oopDesc::mark_offset_in_bytes() ) {
        tj = TypeInstPtr::MARK;
        ta = TypeAryPtr::RANGE; // generic ignored junk
        ptr = TypePtr::BotPTR;
      } else {                  // Random constant offset into array body
        offset = Type::OffsetBot;   // Flatten constant access into array body
        tj = ta = TypeAryPtr::make(ptr,ta->ary(),ta->klass(),false,offset);
      }
    }
    // Arrays of fixed size alias with arrays of unknown size.
    if (ta->size() != TypeInt::POS) {
      const TypeAry *tary = TypeAry::make(ta->elem(), TypeInt::POS);
      tj = ta = TypeAryPtr::make(ptr,ta->const_oop(),tary,ta->klass(),false,offset);
    }
    // Arrays of known objects become arrays of unknown objects.
    if (ta->elem()->isa_narrowoop() && ta->elem() != TypeNarrowOop::BOTTOM) {
      const TypeAry *tary = TypeAry::make(TypeNarrowOop::BOTTOM, ta->size());
      tj = ta = TypeAryPtr::make(ptr,ta->const_oop(),tary,NULL,false,offset);
    }
    if (ta->elem()->isa_oopptr() && ta->elem() != TypeInstPtr::BOTTOM) {
      const TypeAry *tary = TypeAry::make(TypeInstPtr::BOTTOM, ta->size());
      tj = ta = TypeAryPtr::make(ptr,ta->const_oop(),tary,NULL,false,offset);
    }
    // Arrays of bytes and of booleans both use 'bastore' and 'baload' so
    // cannot be distinguished by bytecode alone.
    if (ta->elem() == TypeInt::BOOL) {
      const TypeAry *tary = TypeAry::make(TypeInt::BYTE, ta->size());
      ciKlass* aklass = ciTypeArrayKlass::make(T_BYTE);
      tj = ta = TypeAryPtr::make(ptr,ta->const_oop(),tary,aklass,false,offset);
    }
    // During the 2nd round of IterGVN, NotNull castings are removed.
    // Make sure the Bottom and NotNull variants alias the same.
    // Also, make sure exact and non-exact variants alias the same.
    if (ptr == TypePtr::NotNull || ta->klass_is_exact() || ta->speculative() != NULL) {
      tj = ta = TypeAryPtr::make(TypePtr::BotPTR,ta->ary(),ta->klass(),false,offset);
    }
  }

  // Oop pointers need some flattening
  const TypeInstPtr *to = tj->isa_instptr();
  if( to && _AliasLevel >= 2 && to != TypeOopPtr::BOTTOM ) {
    ciInstanceKlass *k = to->klass()->as_instance_klass();
    if( ptr == TypePtr::Constant ) {
      if (to->klass() != ciEnv::current()->Class_klass() ||
          offset < k->size_helper() * wordSize) {
        // No constant oop pointers (such as Strings); they alias with
        // unknown strings.
        assert(!is_known_inst, "not scalarizable allocation");
        tj = to = TypeInstPtr::make(TypePtr::BotPTR,to->klass(),false,0,offset);
      }
    } else if( is_known_inst ) {
      tj = to; // Keep NotNull and klass_is_exact for instance type
    } else if( ptr == TypePtr::NotNull || to->klass_is_exact() ) {
      // During the 2nd round of IterGVN, NotNull castings are removed.
      // Make sure the Bottom and NotNull variants alias the same.
      // Also, make sure exact and non-exact variants alias the same.
      tj = to = TypeInstPtr::make(TypePtr::BotPTR,to->klass(),false,0,offset);
    }
    if (to->speculative() != NULL) {
      tj = to = TypeInstPtr::make(to->ptr(),to->klass(),to->klass_is_exact(),to->const_oop(),to->offset(), to->instance_id());
    }
    // Canonicalize the holder of this field
    if (offset >= 0 && offset < instanceOopDesc::base_offset_in_bytes()) {
      // First handle header references such as a LoadKlassNode, even if the
      // object's klass is unloaded at compile time (4965979).
      if (!is_known_inst) { // Do it only for non-instance types
        tj = to = TypeInstPtr::make(TypePtr::BotPTR, env()->Object_klass(), false, NULL, offset);
      }
    } else if (offset < 0 || offset >= k->size_helper() * wordSize) {
      // Static fields are in the space above the normal instance
      // fields in the java.lang.Class instance.
      if (to->klass() != ciEnv::current()->Class_klass()) {
        to = NULL;
        tj = TypeOopPtr::BOTTOM;
        offset = tj->offset();
      }
    } else {
      ciInstanceKlass *canonical_holder = k->get_canonical_holder(offset);
      if (!k->equals(canonical_holder) || tj->offset() != offset) {
        if( is_known_inst ) {
          tj = to = TypeInstPtr::make(to->ptr(), canonical_holder, true, NULL, offset, to->instance_id());
        } else {
          tj = to = TypeInstPtr::make(to->ptr(), canonical_holder, false, NULL, offset);
        }
      }
    }
  }

  // Klass pointers to object array klasses need some flattening
  const TypeKlassPtr *tk = tj->isa_klassptr();
  if( tk ) {
    // If we are referencing a field within a Klass, we need
    // to assume the worst case of an Object.  Both exact and
    // inexact types must flatten to the same alias class so
    // use NotNull as the PTR.
    if ( offset == Type::OffsetBot || (offset >= 0 && (size_t)offset < sizeof(Klass)) ) {

      tj = tk = TypeKlassPtr::make(TypePtr::NotNull,
                                   TypeKlassPtr::OBJECT->klass(),
                                   offset);
    }

    ciKlass* klass = tk->klass();
    if( klass->is_obj_array_klass() ) {
      ciKlass* k = TypeAryPtr::OOPS->klass();
      if( !k || !k->is_loaded() )                  // Only fails for some -Xcomp runs
        k = TypeInstPtr::BOTTOM->klass();
      tj = tk = TypeKlassPtr::make( TypePtr::NotNull, k, offset );
    }

    // Check for precise loads from the primary supertype array and force them
    // to the supertype cache alias index.  Check for generic array loads from
    // the primary supertype array and also force them to the supertype cache
    // alias index.  Since the same load can reach both, we need to merge
    // these 2 disparate memories into the same alias class.  Since the
    // primary supertype array is read-only, there's no chance of confusion
    // where we bypass an array load and an array store.
    int primary_supers_offset = in_bytes(Klass::primary_supers_offset());
    if (offset == Type::OffsetBot ||
        (offset >= primary_supers_offset &&
         offset < (int)(primary_supers_offset + Klass::primary_super_limit() * wordSize)) ||
        offset == (int)in_bytes(Klass::secondary_super_cache_offset())) {
      offset = in_bytes(Klass::secondary_super_cache_offset());
      tj = tk = TypeKlassPtr::make( TypePtr::NotNull, tk->klass(), offset );
    }
  }

  // Flatten all Raw pointers together.
  if (tj->base() == Type::RawPtr)
    tj = TypeRawPtr::BOTTOM;

  if (tj->base() == Type::AnyPtr)
    tj = TypePtr::BOTTOM;      // An error, which the caller must check for.

  // Flatten all to bottom for now
  switch( _AliasLevel ) {
  case 0:
    tj = TypePtr::BOTTOM;
    break;
  case 1:                       // Flatten to: oop, static, field or array
    switch (tj->base()) {
    //case Type::AryPtr: tj = TypeAryPtr::RANGE;    break;
    case Type::RawPtr:   tj = TypeRawPtr::BOTTOM;   break;
    case Type::AryPtr:   // do not distinguish arrays at all
    case Type::InstPtr:  tj = TypeInstPtr::BOTTOM;  break;
    case Type::KlassPtr: tj = TypeKlassPtr::OBJECT; break;
    case Type::AnyPtr:   tj = TypePtr::BOTTOM;      break;  // caller checks it
    default: ShouldNotReachHere();
    }
    break;
  case 2:                       // No collapsing at level 2; keep all splits
  case 3:                       // No collapsing at level 3; keep all splits
    break;
  default:
    Unimplemented();
  }

  offset = tj->offset();
  assert( offset != Type::OffsetTop, "Offset has fallen from constant" );

  assert( (offset != Type::OffsetBot && tj->base() != Type::AryPtr) ||
          (offset == Type::OffsetBot && tj->base() == Type::AryPtr) ||
          (offset == Type::OffsetBot && tj == TypeOopPtr::BOTTOM) ||
          (offset == Type::OffsetBot && tj == TypePtr::BOTTOM) ||
          (offset == oopDesc::mark_offset_in_bytes() && tj->base() == Type::AryPtr) ||
          (offset == oopDesc::klass_offset_in_bytes() && tj->base() == Type::AryPtr) ||
          (offset == arrayOopDesc::length_offset_in_bytes() && tj->base() == Type::AryPtr)  ,
          "For oops, klasses, raw offset must be constant; for arrays the offset is never known" );
  assert( tj->ptr() != TypePtr::TopPTR &&
          tj->ptr() != TypePtr::AnyNull &&
          tj->ptr() != TypePtr::Null, "No imprecise addresses" );
//    assert( tj->ptr() != TypePtr::Constant ||
//            tj->base() == Type::RawPtr ||
//            tj->base() == Type::KlassPtr, "No constant oop addresses" );

  return tj;
}

void Compile::AliasType::Init(int i, const TypePtr* at) {
  _index = i;
  _adr_type = at;
  _field = NULL;
  _element = NULL;
  _is_rewritable = true; // default
  const TypeOopPtr *atoop = (at != NULL) ? at->isa_oopptr() : NULL;
  if (atoop != NULL && atoop->is_known_instance()) {
    const TypeOopPtr *gt = atoop->cast_to_instance_id(TypeOopPtr::InstanceBot);
    _general_index = Compile::current()->get_alias_index(gt);
  } else {
    _general_index = 0;
  }
}

//---------------------------------print_on------------------------------------
#ifndef PRODUCT
void Compile::AliasType::print_on(outputStream* st) {
  if (index() < 10)
        st->print("@ <%d> ", index());
  else  st->print("@ <%d>",  index());
  st->print(is_rewritable() ? "   " : " RO");
  int offset = adr_type()->offset();
  if (offset == Type::OffsetBot)
        st->print(" +any");
  else  st->print(" +%-3d", offset);
  st->print(" in ");
  adr_type()->dump_on(st);
  const TypeOopPtr* tjp = adr_type()->isa_oopptr();
  if (field() != NULL && tjp) {
    if (tjp->klass()  != field()->holder() ||
        tjp->offset() != field()->offset_in_bytes()) {
      st->print(" != ");
      field()->print();
      st->print(" ***");
    }
  }
}

void print_alias_types() {
  Compile* C = Compile::current();
  tty->print_cr("--- Alias types, AliasIdxBot .. %d", C->num_alias_types()-1);
  for (int idx = Compile::AliasIdxBot; idx < C->num_alias_types(); idx++) {
    C->alias_type(idx)->print_on(tty);
    tty->cr();
  }
}
#endif


//----------------------------probe_alias_cache--------------------------------
Compile::AliasCacheEntry* Compile::probe_alias_cache(const TypePtr* adr_type) {
  intptr_t key = (intptr_t) adr_type;
  key ^= key >> logAliasCacheSize;
  return &_alias_cache[key & right_n_bits(logAliasCacheSize)];
}


//-----------------------------grow_alias_types--------------------------------
void Compile::grow_alias_types() {
  const int old_ats  = _max_alias_types; // how many before?
  const int new_ats  = old_ats;          // how many more?
  const int grow_ats = old_ats+new_ats;  // how many now?
  _max_alias_types = grow_ats;
  _alias_types =  REALLOC_ARENA_ARRAY(comp_arena(), AliasType*, _alias_types, old_ats, grow_ats);
  AliasType* ats =    NEW_ARENA_ARRAY(comp_arena(), AliasType, new_ats);
  Copy::zero_to_bytes(ats, sizeof(AliasType)*new_ats);
  for (int i = 0; i < new_ats; i++)  _alias_types[old_ats+i] = &ats[i];
}


//--------------------------------find_alias_type------------------------------
Compile::AliasType* Compile::find_alias_type(const TypePtr* adr_type, bool no_create, ciField* original_field) {
  if (_AliasLevel == 0)
    return alias_type(AliasIdxBot);

  AliasCacheEntry* ace = probe_alias_cache(adr_type);
  if (ace->_adr_type == adr_type) {
    return alias_type(ace->_index);
  }

  // Handle special cases.
  if (adr_type == NULL)             return alias_type(AliasIdxTop);
  if (adr_type == TypePtr::BOTTOM)  return alias_type(AliasIdxBot);

  // Do it the slow way.
  const TypePtr* flat = flatten_alias_type(adr_type);

#ifdef ASSERT
  assert(flat == flatten_alias_type(flat), "idempotent");
  assert(flat != TypePtr::BOTTOM,     "cannot alias-analyze an untyped ptr");
  if (flat->isa_oopptr() && !flat->isa_klassptr()) {
    const TypeOopPtr* foop = flat->is_oopptr();
    // Scalarizable allocations have exact klass always.
    bool exact = !foop->klass_is_exact() || foop->is_known_instance();
    const TypePtr* xoop = foop->cast_to_exactness(exact)->is_ptr();
    assert(foop == flatten_alias_type(xoop), "exactness must not affect alias type");
  }
  assert(flat == flatten_alias_type(flat), "exact bit doesn't matter");
#endif

  int idx = AliasIdxTop;
  for (int i = 0; i < num_alias_types(); i++) {
    if (alias_type(i)->adr_type() == flat) {
      idx = i;
      break;
    }
  }

  if (idx == AliasIdxTop) {
    if (no_create)  return NULL;
    // Grow the array if necessary.
    if (_num_alias_types == _max_alias_types)  grow_alias_types();
    // Add a new alias type.
    idx = _num_alias_types++;
    _alias_types[idx]->Init(idx, flat);
    if (flat == TypeInstPtr::KLASS)  alias_type(idx)->set_rewritable(false);
    if (flat == TypeAryPtr::RANGE)   alias_type(idx)->set_rewritable(false);
    if (flat->isa_instptr()) {
      if (flat->offset() == java_lang_Class::klass_offset_in_bytes()
          && flat->is_instptr()->klass() == env()->Class_klass())
        alias_type(idx)->set_rewritable(false);
    }
    if (flat->isa_aryptr()) {
#ifdef ASSERT
      const int header_size_min  = arrayOopDesc::base_offset_in_bytes(T_BYTE);
      // (T_BYTE has the weakest alignment and size restrictions...)
      assert(flat->offset() < header_size_min, "array body reference must be OffsetBot");
#endif
      if (flat->offset() == TypePtr::OffsetBot) {
        alias_type(idx)->set_element(flat->is_aryptr()->elem());
      }
    }
    if (flat->isa_klassptr()) {
      if (flat->offset() == in_bytes(Klass::super_check_offset_offset()))
        alias_type(idx)->set_rewritable(false);
      if (flat->offset() == in_bytes(Klass::modifier_flags_offset()))
        alias_type(idx)->set_rewritable(false);
      if (flat->offset() == in_bytes(Klass::access_flags_offset()))
        alias_type(idx)->set_rewritable(false);
      if (flat->offset() == in_bytes(Klass::java_mirror_offset()))
        alias_type(idx)->set_rewritable(false);
    }
    // %%% (We would like to finalize JavaThread::threadObj_offset(),
    // but the base pointer type is not distinctive enough to identify
    // references into JavaThread.)

    // Check for final fields.
    const TypeInstPtr* tinst = flat->isa_instptr();
    if (tinst && tinst->offset() >= instanceOopDesc::base_offset_in_bytes()) {
      ciField* field;
      if (tinst->const_oop() != NULL &&
          tinst->klass() == ciEnv::current()->Class_klass() &&
          tinst->offset() >= (tinst->klass()->as_instance_klass()->size_helper() * wordSize)) {
        // static field
        ciInstanceKlass* k = tinst->const_oop()->as_instance()->java_lang_Class_klass()->as_instance_klass();
        field = k->get_field_by_offset(tinst->offset(), true);
      } else {
        ciInstanceKlass *k = tinst->klass()->as_instance_klass();
        field = k->get_field_by_offset(tinst->offset(), false);
      }
      assert(field == NULL ||
             original_field == NULL ||
             (field->holder() == original_field->holder() &&
              field->offset() == original_field->offset() &&
              field->is_static() == original_field->is_static()), "wrong field?");
      // Set field() and is_rewritable() attributes.
      if (field != NULL)  alias_type(idx)->set_field(field);
    }
  }

  // Fill the cache for next time.
  ace->_adr_type = adr_type;
  ace->_index    = idx;
  assert(alias_type(adr_type) == alias_type(idx),  "type must be installed");

  // Might as well try to fill the cache for the flattened version, too.
  AliasCacheEntry* face = probe_alias_cache(flat);
  if (face->_adr_type == NULL) {
    face->_adr_type = flat;
    face->_index    = idx;
    assert(alias_type(flat) == alias_type(idx), "flat type must work too");
  }

  return alias_type(idx);
}


Compile::AliasType* Compile::alias_type(ciField* field) {
  const TypeOopPtr* t;
  if (field->is_static())
    t = TypeInstPtr::make(field->holder()->java_mirror());
  else
    t = TypeOopPtr::make_from_klass_raw(field->holder());
  AliasType* atp = alias_type(t->add_offset(field->offset_in_bytes()), field);
  assert((field->is_final() || field->is_stable()) == !atp->is_rewritable(), "must get the rewritable bits correct");
  return atp;
}


//------------------------------have_alias_type--------------------------------
bool Compile::have_alias_type(const TypePtr* adr_type) {
  AliasCacheEntry* ace = probe_alias_cache(adr_type);
  if (ace->_adr_type == adr_type) {
    return true;
  }

  // Handle special cases.
  if (adr_type == NULL)             return true;
  if (adr_type == TypePtr::BOTTOM)  return true;

  return find_alias_type(adr_type, true, NULL) != NULL;
}

//-----------------------------must_alias--------------------------------------
// True if all values of the given address type are in the given alias category.
bool Compile::must_alias(const TypePtr* adr_type, int alias_idx) {
  if (alias_idx == AliasIdxBot)         return true;  // the universal category
  if (adr_type == NULL)                 return true;  // NULL serves as TypePtr::TOP
  if (alias_idx == AliasIdxTop)         return false; // the empty category
  if (adr_type->base() == Type::AnyPtr) return false; // TypePtr::BOTTOM or its twins

  // the only remaining possible overlap is identity
  int adr_idx = get_alias_index(adr_type);
  assert(adr_idx != AliasIdxBot && adr_idx != AliasIdxTop, "");
  assert(adr_idx == alias_idx ||
         (alias_type(alias_idx)->adr_type() != TypeOopPtr::BOTTOM
          && adr_type                       != TypeOopPtr::BOTTOM),
         "should not be testing for overlap with an unsafe pointer");
  return adr_idx == alias_idx;
}

//------------------------------can_alias--------------------------------------
// True if any values of the given address type are in the given alias category.
bool Compile::can_alias(const TypePtr* adr_type, int alias_idx) {
  if (alias_idx == AliasIdxTop)         return false; // the empty category
  if (adr_type == NULL)                 return false; // NULL serves as TypePtr::TOP
  if (alias_idx == AliasIdxBot)         return true;  // the universal category
  if (adr_type->base() == Type::AnyPtr) return true;  // TypePtr::BOTTOM or its twins

  // the only remaining possible overlap is identity
  int adr_idx = get_alias_index(adr_type);
  assert(adr_idx != AliasIdxBot && adr_idx != AliasIdxTop, "");
  return adr_idx == alias_idx;
}



//---------------------------pop_warm_call-------------------------------------
WarmCallInfo* Compile::pop_warm_call() {
  WarmCallInfo* wci = _warm_calls;
  if (wci != NULL)  _warm_calls = wci->remove_from(wci);
  return wci;
}

//----------------------------Inline_Warm--------------------------------------
int Compile::Inline_Warm() {
  // If there is room, try to inline some more warm call sites.
  // %%% Do a graph index compaction pass when we think we're out of space?
  if (!InlineWarmCalls)  return 0;

  int calls_made_hot = 0;
  int room_to_grow   = NodeCountInliningCutoff - unique();
  int amount_to_grow = MIN2(room_to_grow, (int)NodeCountInliningStep);
  int amount_grown   = 0;
  WarmCallInfo* call;
  while (amount_to_grow > 0 && (call = pop_warm_call()) != NULL) {
    int est_size = (int)call->size();
    if (est_size > (room_to_grow - amount_grown)) {
      // This one won't fit anyway.  Get rid of it.
      call->make_cold();
      continue;
    }
    call->make_hot();
    calls_made_hot++;
    amount_grown   += est_size;
    amount_to_grow -= est_size;
  }

  if (calls_made_hot > 0)  set_major_progress();
  return calls_made_hot;
}


//----------------------------Finish_Warm--------------------------------------
void Compile::Finish_Warm() {
  if (!InlineWarmCalls)  return;
  if (failing())  return;
  if (warm_calls() == NULL)  return;

  // Clean up loose ends, if we are out of space for inlining.
  WarmCallInfo* call;
  while ((call = pop_warm_call()) != NULL) {
    call->make_cold();
  }
}

//---------------------cleanup_loop_predicates-----------------------
// Remove the opaque nodes that protect the predicates so that all unused
// checks and uncommon_traps will be eliminated from the ideal graph
void Compile::cleanup_loop_predicates(PhaseIterGVN &igvn) {
  if (predicate_count()==0) return;
  for (int i = predicate_count(); i > 0; i--) {
    Node * n = predicate_opaque1_node(i-1);
    assert(n->Opcode() == Op_Opaque1, "must be");
    igvn.replace_node(n, n->in(1));
  }
  assert(predicate_count()==0, "should be clean!");
}

// StringOpts and late inlining of string methods
void Compile::inline_string_calls(bool parse_time) {
  {
    // remove useless nodes to make the usage analysis simpler
    ResourceMark rm;
    PhaseRemoveUseless pru(initial_gvn(), for_igvn());
  }

  {
    ResourceMark rm;
    print_method(PHASE_BEFORE_STRINGOPTS, 3);
    PhaseStringOpts pso(initial_gvn(), for_igvn());
    print_method(PHASE_AFTER_STRINGOPTS, 3);
  }

  // now inline anything that we skipped the first time around
  if (!parse_time) {
    _late_inlines_pos = _late_inlines.length();
  }

  while (_string_late_inlines.length() > 0) {
    CallGenerator* cg = _string_late_inlines.pop();
    cg->do_late_inline();
    if (failing())  return;
  }
  _string_late_inlines.trunc_to(0);
}

// Late inlining of boxing methods
void Compile::inline_boxing_calls(PhaseIterGVN& igvn) {
  if (_boxing_late_inlines.length() > 0) {
    assert(has_boxed_value(), "inconsistent");

    PhaseGVN* gvn = initial_gvn();
    set_inlining_incrementally(true);

    assert( igvn._worklist.size() == 0, "should be done with igvn" );
    for_igvn()->clear();
    gvn->replace_with(&igvn);

    while (_boxing_late_inlines.length() > 0) {
      CallGenerator* cg = _boxing_late_inlines.pop();
      cg->do_late_inline();
      if (failing())  return;
    }
    _boxing_late_inlines.trunc_to(0);

    {
      ResourceMark rm;
      PhaseRemoveUseless pru(gvn, for_igvn());
    }

    igvn = PhaseIterGVN(gvn);
    igvn.optimize();

    set_inlining_progress(false);
    set_inlining_incrementally(false);
  }
}

void Compile::inline_incrementally_one(PhaseIterGVN& igvn) {
  assert(IncrementalInline, "incremental inlining should be on");
  PhaseGVN* gvn = initial_gvn();

  set_inlining_progress(false);
  for_igvn()->clear();
  gvn->replace_with(&igvn);

  int i = 0;

  for (; i <_late_inlines.length() && !inlining_progress(); i++) {
    CallGenerator* cg = _late_inlines.at(i);
    _late_inlines_pos = i+1;
    cg->do_late_inline();
    if (failing())  return;
  }
  int j = 0;
  for (; i < _late_inlines.length(); i++, j++) {
    _late_inlines.at_put(j, _late_inlines.at(i));
  }
  _late_inlines.trunc_to(j);

  {
    ResourceMark rm;
    PhaseRemoveUseless pru(gvn, for_igvn());
  }

  igvn = PhaseIterGVN(gvn);
}

// Perform incremental inlining until bound on number of live nodes is reached
void Compile::inline_incrementally(PhaseIterGVN& igvn) {
  PhaseGVN* gvn = initial_gvn();

  set_inlining_incrementally(true);
  set_inlining_progress(true);
  uint low_live_nodes = 0;

  while(inlining_progress() && _late_inlines.length() > 0) {

    if (live_nodes() > (uint)LiveNodeCountInliningCutoff) {
      if (low_live_nodes < (uint)LiveNodeCountInliningCutoff * 8 / 10) {
        // PhaseIdealLoop is expensive so we only try it once we are
        // out of loop and we only try it again if the previous helped
        // got the number of nodes down significantly
        PhaseIdealLoop ideal_loop( igvn, false, true );
        if (failing())  return;
        low_live_nodes = live_nodes();
        _major_progress = true;
      }

      if (live_nodes() > (uint)LiveNodeCountInliningCutoff) {
        break;
      }
    }

    inline_incrementally_one(igvn);

    if (failing())  return;

    igvn.optimize();

    if (failing())  return;
  }

  assert( igvn._worklist.size() == 0, "should be done with igvn" );

  if (_string_late_inlines.length() > 0) {
    assert(has_stringbuilder(), "inconsistent");
    for_igvn()->clear();
    initial_gvn()->replace_with(&igvn);

    inline_string_calls(false);

    if (failing())  return;

    {
      ResourceMark rm;
      PhaseRemoveUseless pru(initial_gvn(), for_igvn());
    }

    igvn = PhaseIterGVN(gvn);

    igvn.optimize();
  }

  set_inlining_incrementally(false);
}


//------------------------------Optimize---------------------------------------
// Given a graph, optimize it.
void Compile::Optimize() {
  TracePhase t1("optimizer", &_t_optimizer, true);

#ifndef PRODUCT
  if (env()->break_at_compile()) {
    BREAKPOINT;
  }

#endif

  ResourceMark rm;
  int          loop_opts_cnt;

  NOT_PRODUCT( verify_graph_edges(); )

  print_method(PHASE_AFTER_PARSING);

 {
  // Iterative Global Value Numbering, including ideal transforms
  // Initialize IterGVN with types and values from parse-time GVN
  PhaseIterGVN igvn(initial_gvn());
  {
    NOT_PRODUCT( TracePhase t2("iterGVN", &_t_iterGVN, TimeCompiler); )
    igvn.optimize();
  }

  print_method(PHASE_ITER_GVN1, 2);

  if (failing())  return;

  {
    NOT_PRODUCT( TracePhase t2("incrementalInline", &_t_incrInline, TimeCompiler); )
    inline_incrementally(igvn);
  }

  print_method(PHASE_INCREMENTAL_INLINE, 2);

  if (failing())  return;

  if (eliminate_boxing()) {
    NOT_PRODUCT( TracePhase t2("incrementalInline", &_t_incrInline, TimeCompiler); )
    // Inline valueOf() methods now.
    inline_boxing_calls(igvn);

    print_method(PHASE_INCREMENTAL_BOXING_INLINE, 2);

    if (failing())  return;
  }

  // Remove the speculative part of types and clean up the graph from
  // the extra CastPP nodes whose only purpose is to carry them. Do
  // that early so that optimizations are not disrupted by the extra
  // CastPP nodes.
  remove_speculative_types(igvn);

  // No more new expensive nodes will be added to the list from here
  // so keep only the actual candidates for optimizations.
  cleanup_expensive_nodes(igvn);

  // Perform escape analysis
  if (_do_escape_analysis && ConnectionGraph::has_candidates(this)) {
    if (has_loops()) {
      // Cleanup graph (remove dead nodes).
      TracePhase t2("idealLoop", &_t_idealLoop, true);
      PhaseIdealLoop ideal_loop( igvn, false, true );
      if (major_progress()) print_method(PHASE_PHASEIDEAL_BEFORE_EA, 2);
      if (failing())  return;
    }
    ConnectionGraph::do_analysis(this, &igvn);

    if (failing())  return;

    // Optimize out fields loads from scalar replaceable allocations.
    igvn.optimize();
    print_method(PHASE_ITER_GVN_AFTER_EA, 2);

    if (failing())  return;

    if (congraph() != NULL && macro_count() > 0) {
      NOT_PRODUCT( TracePhase t2("macroEliminate", &_t_macroEliminate, TimeCompiler); )
      PhaseMacroExpand mexp(igvn);
      mexp.eliminate_macro_nodes();
      igvn.set_delay_transform(false);

      igvn.optimize();
      print_method(PHASE_ITER_GVN_AFTER_ELIMINATION, 2);

      if (failing())  return;
    }
  }

  // Loop transforms on the ideal graph.  Range Check Elimination,
  // peeling, unrolling, etc.

  // Set loop opts counter
  loop_opts_cnt = num_loop_opts();
  if((loop_opts_cnt > 0) && (has_loops() || has_split_ifs())) {
    {
      TracePhase t2("idealLoop", &_t_idealLoop, true);
      PhaseIdealLoop ideal_loop( igvn, true );
      loop_opts_cnt--;
      if (major_progress()) print_method(PHASE_PHASEIDEALLOOP1, 2);
      if (failing())  return;
    }
    // Loop opts pass if partial peeling occurred in previous pass
    if(PartialPeelLoop && major_progress() && (loop_opts_cnt > 0)) {
      TracePhase t3("idealLoop", &_t_idealLoop, true);
      PhaseIdealLoop ideal_loop( igvn, false );
      loop_opts_cnt--;
      if (major_progress()) print_method(PHASE_PHASEIDEALLOOP2, 2);
      if (failing())  return;
    }
    // Loop opts pass for loop-unrolling before CCP
    if(major_progress() && (loop_opts_cnt > 0)) {
      TracePhase t4("idealLoop", &_t_idealLoop, true);
      PhaseIdealLoop ideal_loop( igvn, false );
      loop_opts_cnt--;
      if (major_progress()) print_method(PHASE_PHASEIDEALLOOP3, 2);
    }
    if (!failing()) {
      // Verify that last round of loop opts produced a valid graph
      NOT_PRODUCT( TracePhase t2("idealLoopVerify", &_t_idealLoopVerify, TimeCompiler); )
      PhaseIdealLoop::verify(igvn);
    }
  }
  if (failing())  return;

  // Conditional Constant Propagation;
  PhaseCCP ccp( &igvn );
  assert( true, "Break here to ccp.dump_nodes_and_types(_root,999,1)");
  {
    TracePhase t2("ccp", &_t_ccp, true);
    ccp.do_transform();
  }
  print_method(PHASE_CPP1, 2);

  assert( true, "Break here to ccp.dump_old2new_map()");

  // Iterative Global Value Numbering, including ideal transforms
  {
    NOT_PRODUCT( TracePhase t2("iterGVN2", &_t_iterGVN2, TimeCompiler); )
    igvn = ccp;
    igvn.optimize();
  }

  print_method(PHASE_ITER_GVN2, 2);

  if (failing())  return;

  // Loop transforms on the ideal graph.  Range Check Elimination,
  // peeling, unrolling, etc.
  if(loop_opts_cnt > 0) {
    debug_only( int cnt = 0; );
    while(major_progress() && (loop_opts_cnt > 0)) {
      TracePhase t2("idealLoop", &_t_idealLoop, true);
      assert( cnt++ < 40, "infinite cycle in loop optimization" );
      PhaseIdealLoop ideal_loop( igvn, true);
      loop_opts_cnt--;
      if (major_progress()) print_method(PHASE_PHASEIDEALLOOP_ITERATIONS, 2);
      if (failing())  return;
    }
  }

  {
    // Verify that all previous optimizations produced a valid graph
    // at least to this point, even if no loop optimizations were done.
    NOT_PRODUCT( TracePhase t2("idealLoopVerify", &_t_idealLoopVerify, TimeCompiler); )
    PhaseIdealLoop::verify(igvn);
  }

  {
    NOT_PRODUCT( TracePhase t2("macroExpand", &_t_macroExpand, TimeCompiler); )
    PhaseMacroExpand  mex(igvn);
    if (mex.expand_macro_nodes()) {
      assert(failing(), "must bail out w/ explicit message");
      return;
    }
  }

 } // (End scope of igvn; run destructor if necessary for asserts.)

  dump_inlining();
  // A method with only infinite loops has no edges entering loops from root
  {
    NOT_PRODUCT( TracePhase t2("graphReshape", &_t_graphReshaping, TimeCompiler); )
    if (final_graph_reshaping()) {
      assert(failing(), "must bail out w/ explicit message");
      return;
    }
  }

  print_method(PHASE_OPTIMIZE_FINISHED, 2);
}


//------------------------------Code_Gen---------------------------------------
// Given a graph, generate code for it
void Compile::Code_Gen() {
  if (failing()) {
    return;
  }

  // Perform instruction selection.  You might think we could reclaim Matcher
  // memory PDQ, but actually the Matcher is used in generating spill code.
  // Internals of the Matcher (including some VectorSets) must remain live
  // for awhile - thus I cannot reclaim Matcher memory lest a VectorSet usage
  // set a bit in reclaimed memory.

  // In debug mode can dump m._nodes.dump() for mapping of ideal to machine
  // nodes.  Mapping is only valid at the root of each matched subtree.
  NOT_PRODUCT( verify_graph_edges(); )

  Matcher matcher;
  _matcher = &matcher;
  {
    TracePhase t2("matcher", &_t_matcher, true);
    matcher.match();
  }
  // In debug mode can dump m._nodes.dump() for mapping of ideal to machine
  // nodes.  Mapping is only valid at the root of each matched subtree.
  NOT_PRODUCT( verify_graph_edges(); )

  // If you have too many nodes, or if matching has failed, bail out
  check_node_count(0, "out of nodes matching instructions");
  if (failing()) {
    return;
  }

  // Build a proper-looking CFG
  PhaseCFG cfg(node_arena(), root(), matcher);
  _cfg = &cfg;
  {
    NOT_PRODUCT( TracePhase t2("scheduler", &_t_scheduler, TimeCompiler); )
    bool success = cfg.do_global_code_motion();
    if (!success) {
      return;
    }

    print_method(PHASE_GLOBAL_CODE_MOTION, 2);
    NOT_PRODUCT( verify_graph_edges(); )
    debug_only( cfg.verify(); )
  }

  PhaseChaitin regalloc(unique(), cfg, matcher);
  _regalloc = &regalloc;
  {
    TracePhase t2("regalloc", &_t_registerAllocation, true);
    // Perform register allocation.  After Chaitin, use-def chains are
    // no longer accurate (at spill code) and so must be ignored.
    // Node->LRG->reg mappings are still accurate.
    _regalloc->Register_Allocate();

    // Bail out if the allocator builds too many nodes
    if (failing()) {
      return;
    }
  }

  // Prior to register allocation we kept empty basic blocks in case the
  // the allocator needed a place to spill.  After register allocation we
  // are not adding any new instructions.  If any basic block is empty, we
  // can now safely remove it.
  {
    NOT_PRODUCT( TracePhase t2("blockOrdering", &_t_blockOrdering, TimeCompiler); )
    cfg.remove_empty_blocks();
    if (do_freq_based_layout()) {
      PhaseBlockLayout layout(cfg);
    } else {
      cfg.set_loop_alignment();
    }
    cfg.fixup_flow();
  }

  // Apply peephole optimizations
  if( OptoPeephole ) {
    NOT_PRODUCT( TracePhase t2("peephole", &_t_peephole, TimeCompiler); )
    PhasePeephole peep( _regalloc, cfg);
    peep.do_transform();
  }

  // Do late expand if CPU requires this.
  if (Matcher::require_postalloc_expand) {
    NOT_PRODUCT(TracePhase t2c("postalloc_expand", &_t_postalloc_expand, true));
    cfg.postalloc_expand(_regalloc);
  }

  // Convert Nodes to instruction bits in a buffer
  {
    // %%%% workspace merge brought two timers together for one job
    TracePhase t2a("output", &_t_output, true);
    NOT_PRODUCT( TraceTime t2b(NULL, &_t_codeGeneration, TimeCompiler, false); )
    Output();
  }

  print_method(PHASE_FINAL_CODE);

  // He's dead, Jim.
  _cfg     = (PhaseCFG*)0xdeadbeef;
  _regalloc = (PhaseChaitin*)0xdeadbeef;
}


//------------------------------dump_asm---------------------------------------
// Dump formatted assembly
#ifndef PRODUCT
void Compile::dump_asm(int *pcs, uint pc_limit) {
  bool cut_short = false;
  tty->print_cr("#");
  tty->print("#  ");  _tf->dump();  tty->cr();
  tty->print_cr("#");

  // For all blocks
  int pc = 0x0;                 // Program counter
  char starts_bundle = ' ';
  _regalloc->dump_frame();

  Node *n = NULL;
  for (uint i = 0; i < _cfg->number_of_blocks(); i++) {
    if (VMThread::should_terminate()) {
      cut_short = true;
      break;
    }
    Block* block = _cfg->get_block(i);
    if (block->is_connector() && !Verbose) {
      continue;
    }
    n = block->head();
    if (pcs && n->_idx < pc_limit) {
      tty->print("%3.3x   ", pcs[n->_idx]);
    } else {
      tty->print("      ");
    }
    block->dump_head(_cfg);
    if (block->is_connector()) {
      tty->print_cr("        # Empty connector block");
    } else if (block->num_preds() == 2 && block->pred(1)->is_CatchProj() && block->pred(1)->as_CatchProj()->_con == CatchProjNode::fall_through_index) {
      tty->print_cr("        # Block is sole successor of call");
    }

    // For all instructions
    Node *delay = NULL;
    for (uint j = 0; j < block->number_of_nodes(); j++) {
      if (VMThread::should_terminate()) {
        cut_short = true;
        break;
      }
      n = block->get_node(j);
      if (valid_bundle_info(n)) {
        Bundle* bundle = node_bundling(n);
        if (bundle->used_in_unconditional_delay()) {
          delay = n;
          continue;
        }
        if (bundle->starts_bundle()) {
          starts_bundle = '+';
        }
      }

      if (WizardMode) {
        n->dump();
      }

      if( !n->is_Region() &&    // Dont print in the Assembly
          !n->is_Phi() &&       // a few noisely useless nodes
          !n->is_Proj() &&
          !n->is_MachTemp() &&
          !n->is_SafePointScalarObject() &&
          !n->is_Catch() &&     // Would be nice to print exception table targets
          !n->is_MergeMem() &&  // Not very interesting
          !n->is_top() &&       // Debug info table constants
          !(n->is_Con() && !n->is_Mach())// Debug info table constants
          ) {
        if (pcs && n->_idx < pc_limit)
          tty->print("%3.3x", pcs[n->_idx]);
        else
          tty->print("   ");
        tty->print(" %c ", starts_bundle);
        starts_bundle = ' ';
        tty->print("\t");
        n->format(_regalloc, tty);
        tty->cr();
      }

      // If we have an instruction with a delay slot, and have seen a delay,
      // then back up and print it
      if (valid_bundle_info(n) && node_bundling(n)->use_unconditional_delay()) {
        assert(delay != NULL, "no unconditional delay instruction");
        if (WizardMode) delay->dump();

        if (node_bundling(delay)->starts_bundle())
          starts_bundle = '+';
        if (pcs && n->_idx < pc_limit)
          tty->print("%3.3x", pcs[n->_idx]);
        else
          tty->print("   ");
        tty->print(" %c ", starts_bundle);
        starts_bundle = ' ';
        tty->print("\t");
        delay->format(_regalloc, tty);
        tty->print_cr("");
        delay = NULL;
      }

      // Dump the exception table as well
      if( n->is_Catch() && (Verbose || WizardMode) ) {
        // Print the exception table for this offset
        _handler_table.print_subtable_for(pc);
      }
    }

    if (pcs && n->_idx < pc_limit)
      tty->print_cr("%3.3x", pcs[n->_idx]);
    else
      tty->print_cr("");

    assert(cut_short || delay == NULL, "no unconditional delay branch");

  } // End of per-block dump
  tty->print_cr("");

  if (cut_short)  tty->print_cr("*** disassembly is cut short ***");
}
#endif

//------------------------------Final_Reshape_Counts---------------------------
// This class defines counters to help identify when a method
// may/must be executed using hardware with only 24-bit precision.
struct Final_Reshape_Counts : public StackObj {
  int  _call_count;             // count non-inlined 'common' calls
  int  _float_count;            // count float ops requiring 24-bit precision
  int  _double_count;           // count double ops requiring more precision
  int  _java_call_count;        // count non-inlined 'java' calls
  int  _inner_loop_count;       // count loops which need alignment
  VectorSet _visited;           // Visitation flags
  Node_List _tests;             // Set of IfNodes & PCTableNodes

  Final_Reshape_Counts() :
    _call_count(0), _float_count(0), _double_count(0),
    _java_call_count(0), _inner_loop_count(0),
    _visited( Thread::current()->resource_area() ) { }

  void inc_call_count  () { _call_count  ++; }
  void inc_float_count () { _float_count ++; }
  void inc_double_count() { _double_count++; }
  void inc_java_call_count() { _java_call_count++; }
  void inc_inner_loop_count() { _inner_loop_count++; }

  int  get_call_count  () const { return _call_count  ; }
  int  get_float_count () const { return _float_count ; }
  int  get_double_count() const { return _double_count; }
  int  get_java_call_count() const { return _java_call_count; }
  int  get_inner_loop_count() const { return _inner_loop_count; }
};

#ifdef ASSERT
static bool oop_offset_is_sane(const TypeInstPtr* tp) {
  ciInstanceKlass *k = tp->klass()->as_instance_klass();
  // Make sure the offset goes inside the instance layout.
  return k->contains_field_offset(tp->offset());
  // Note that OffsetBot and OffsetTop are very negative.
}
#endif

// Eliminate trivially redundant StoreCMs and accumulate their
// precedence edges.
void Compile::eliminate_redundant_card_marks(Node* n) {
  assert(n->Opcode() == Op_StoreCM, "expected StoreCM");
  if (n->in(MemNode::Address)->outcnt() > 1) {
    // There are multiple users of the same address so it might be
    // possible to eliminate some of the StoreCMs
    Node* mem = n->in(MemNode::Memory);
    Node* adr = n->in(MemNode::Address);
    Node* val = n->in(MemNode::ValueIn);
    Node* prev = n;
    bool done = false;
    // Walk the chain of StoreCMs eliminating ones that match.  As
    // long as it's a chain of single users then the optimization is
    // safe.  Eliminating partially redundant StoreCMs would require
    // cloning copies down the other paths.
    while (mem->Opcode() == Op_StoreCM && mem->outcnt() == 1 && !done) {
      if (adr == mem->in(MemNode::Address) &&
          val == mem->in(MemNode::ValueIn)) {
        // redundant StoreCM
        if (mem->req() > MemNode::OopStore) {
          // Hasn't been processed by this code yet.
          n->add_prec(mem->in(MemNode::OopStore));
        } else {
          // Already converted to precedence edge
          for (uint i = mem->req(); i < mem->len(); i++) {
            // Accumulate any precedence edges
            if (mem->in(i) != NULL) {
              n->add_prec(mem->in(i));
            }
          }
          // Everything above this point has been processed.
          done = true;
        }
        // Eliminate the previous StoreCM
        prev->set_req(MemNode::Memory, mem->in(MemNode::Memory));
        assert(mem->outcnt() == 0, "should be dead");
        mem->disconnect_inputs(NULL, this);
      } else {
        prev = mem;
      }
      mem = prev->in(MemNode::Memory);
    }
  }
}

//------------------------------final_graph_reshaping_impl----------------------
// Implement items 1-5 from final_graph_reshaping below.
void Compile::final_graph_reshaping_impl( Node *n, Final_Reshape_Counts &frc) {

  if ( n->outcnt() == 0 ) return; // dead node
  uint nop = n->Opcode();

  // Check for 2-input instruction with "last use" on right input.
  // Swap to left input.  Implements item (2).
  if( n->req() == 3 &&          // two-input instruction
      n->in(1)->outcnt() > 1 && // left use is NOT a last use
      (!n->in(1)->is_Phi() || n->in(1)->in(2) != n) && // it is not data loop
      n->in(2)->outcnt() == 1 &&// right use IS a last use
      !n->in(2)->is_Con() ) {   // right use is not a constant
    // Check for commutative opcode
    switch( nop ) {
    case Op_AddI:  case Op_AddF:  case Op_AddD:  case Op_AddL:
    case Op_MaxI:  case Op_MinI:
    case Op_MulI:  case Op_MulF:  case Op_MulD:  case Op_MulL:
    case Op_AndL:  case Op_XorL:  case Op_OrL:
    case Op_AndI:  case Op_XorI:  case Op_OrI: {
      // Move "last use" input to left by swapping inputs
      n->swap_edges(1, 2);
      break;
    }
    default:
      break;
    }
  }

#ifdef ASSERT
  if( n->is_Mem() ) {
    int alias_idx = get_alias_index(n->as_Mem()->adr_type());
    assert( n->in(0) != NULL || alias_idx != Compile::AliasIdxRaw ||
            // oop will be recorded in oop map if load crosses safepoint
            n->is_Load() && (n->as_Load()->bottom_type()->isa_oopptr() ||
                             LoadNode::is_immutable_value(n->in(MemNode::Address))),
            "raw memory operations should have control edge");
  }
#endif
  // Count FPU ops and common calls, implements item (3)
  switch( nop ) {
  // Count all float operations that may use FPU
  case Op_AddF:
  case Op_SubF:
  case Op_MulF:
  case Op_DivF:
  case Op_NegF:
  case Op_ModF:
  case Op_ConvI2F:
  case Op_ConF:
  case Op_CmpF:
  case Op_CmpF3:
  // case Op_ConvL2F: // longs are split into 32-bit halves
    frc.inc_float_count();
    break;

  case Op_ConvF2D:
  case Op_ConvD2F:
    frc.inc_float_count();
    frc.inc_double_count();
    break;

  // Count all double operations that may use FPU
  case Op_AddD:
  case Op_SubD:
  case Op_MulD:
  case Op_DivD:
  case Op_NegD:
  case Op_ModD:
  case Op_ConvI2D:
  case Op_ConvD2I:
  // case Op_ConvL2D: // handled by leaf call
  // case Op_ConvD2L: // handled by leaf call
  case Op_ConD:
  case Op_CmpD:
  case Op_CmpD3:
    frc.inc_double_count();
    break;
  case Op_Opaque1:              // Remove Opaque Nodes before matching
  case Op_Opaque2:              // Remove Opaque Nodes before matching
    n->subsume_by(n->in(1), this);
    break;
  case Op_CallStaticJava:
  case Op_CallJava:
  case Op_CallDynamicJava:
    frc.inc_java_call_count(); // Count java call site;
  case Op_CallRuntime:
  case Op_CallLeaf:
  case Op_CallLeafNoFP: {
    assert( n->is_Call(), "" );
    CallNode *call = n->as_Call();
    // Count call sites where the FP mode bit would have to be flipped.
    // Do not count uncommon runtime calls:
    // uncommon_trap, _complete_monitor_locking, _complete_monitor_unlocking,
    // _new_Java, _new_typeArray, _new_objArray, _rethrow_Java, ...
    if( !call->is_CallStaticJava() || !call->as_CallStaticJava()->_name ) {
      frc.inc_call_count();   // Count the call site
    } else {                  // See if uncommon argument is shared
      Node *n = call->in(TypeFunc::Parms);
      int nop = n->Opcode();
      // Clone shared simple arguments to uncommon calls, item (1).
      if( n->outcnt() > 1 &&
          !n->is_Proj() &&
          nop != Op_CreateEx &&
          nop != Op_CheckCastPP &&
          nop != Op_DecodeN &&
          nop != Op_DecodeNKlass &&
          !n->is_Mem() ) {
        Node *x = n->clone();
        call->set_req( TypeFunc::Parms, x );
      }
    }
    break;
  }

  case Op_StoreD:
  case Op_LoadD:
  case Op_LoadD_unaligned:
    frc.inc_double_count();
    goto handle_mem;
  case Op_StoreF:
  case Op_LoadF:
    frc.inc_float_count();
    goto handle_mem;

  case Op_StoreCM:
    {
      // Convert OopStore dependence into precedence edge
      Node* prec = n->in(MemNode::OopStore);
      n->del_req(MemNode::OopStore);
      n->add_prec(prec);
      eliminate_redundant_card_marks(n);
    }

    // fall through

  case Op_StoreB:
  case Op_StoreC:
  case Op_StorePConditional:
  case Op_StoreI:
  case Op_StoreL:
  case Op_StoreIConditional:
  case Op_StoreLConditional:
  case Op_CompareAndSwapI:
  case Op_CompareAndSwapL:
  case Op_CompareAndSwapP:
  case Op_CompareAndSwapN:
  case Op_GetAndAddI:
  case Op_GetAndAddL:
  case Op_GetAndSetI:
  case Op_GetAndSetL:
  case Op_GetAndSetP:
  case Op_GetAndSetN:
  case Op_StoreP:
  case Op_StoreN:
  case Op_StoreNKlass:
  case Op_LoadB:
  case Op_LoadUB:
  case Op_LoadUS:
  case Op_LoadI:
  case Op_LoadKlass:
  case Op_LoadNKlass:
  case Op_LoadL:
  case Op_LoadL_unaligned:
  case Op_LoadPLocked:
  case Op_LoadP:
  case Op_LoadN:
  case Op_LoadRange:
  case Op_LoadS: {
  handle_mem:
#ifdef ASSERT
    if( VerifyOptoOopOffsets ) {
      assert( n->is_Mem(), "" );
      MemNode *mem  = (MemNode*)n;
      // Check to see if address types have grounded out somehow.
      const TypeInstPtr *tp = mem->in(MemNode::Address)->bottom_type()->isa_instptr();
      assert( !tp || oop_offset_is_sane(tp), "" );
    }
#endif
    break;
  }

  case Op_AddP: {               // Assert sane base pointers
    Node *addp = n->in(AddPNode::Address);
    assert( !addp->is_AddP() ||
            addp->in(AddPNode::Base)->is_top() || // Top OK for allocation
            addp->in(AddPNode::Base) == n->in(AddPNode::Base),
            "Base pointers must match" );
#ifdef _LP64
    if ((UseCompressedOops || UseCompressedClassPointers) &&
        addp->Opcode() == Op_ConP &&
        addp == n->in(AddPNode::Base) &&
        n->in(AddPNode::Offset)->is_Con()) {
      // Use addressing with narrow klass to load with offset on x86.
      // On sparc loading 32-bits constant and decoding it have less
      // instructions (4) then load 64-bits constant (7).
      // Do this transformation here since IGVN will convert ConN back to ConP.
      const Type* t = addp->bottom_type();
      if (t->isa_oopptr() || t->isa_klassptr()) {
        Node* nn = NULL;

        int op = t->isa_oopptr() ? Op_ConN : Op_ConNKlass;

        // Look for existing ConN node of the same exact type.
        Node* r  = root();
        uint cnt = r->outcnt();
        for (uint i = 0; i < cnt; i++) {
          Node* m = r->raw_out(i);
          if (m!= NULL && m->Opcode() == op &&
              m->bottom_type()->make_ptr() == t) {
            nn = m;
            break;
          }
        }
        if (nn != NULL) {
          // Decode a narrow oop to match address
          // [R12 + narrow_oop_reg<<3 + offset]
          if (t->isa_oopptr()) {
            nn = new (this) DecodeNNode(nn, t);
          } else {
            nn = new (this) DecodeNKlassNode(nn, t);
          }
          n->set_req(AddPNode::Base, nn);
          n->set_req(AddPNode::Address, nn);
          if (addp->outcnt() == 0) {
            addp->disconnect_inputs(NULL, this);
          }
        }
      }
    }
#endif
    break;
  }

#ifdef _LP64
  case Op_CastPP:
    if (n->in(1)->is_DecodeN() && Matcher::gen_narrow_oop_implicit_null_checks()) {
      Node* in1 = n->in(1);
      const Type* t = n->bottom_type();
      Node* new_in1 = in1->clone();
      new_in1->as_DecodeN()->set_type(t);

      if (!Matcher::narrow_oop_use_complex_address()) {
        //
        // x86, ARM and friends can handle 2 adds in addressing mode
        // and Matcher can fold a DecodeN node into address by using
        // a narrow oop directly and do implicit NULL check in address:
        //
        // [R12 + narrow_oop_reg<<3 + offset]
        // NullCheck narrow_oop_reg
        //
        // On other platforms (Sparc) we have to keep new DecodeN node and
        // use it to do implicit NULL check in address:
        //
        // decode_not_null narrow_oop_reg, base_reg
        // [base_reg + offset]
        // NullCheck base_reg
        //
        // Pin the new DecodeN node to non-null path on these platform (Sparc)
        // to keep the information to which NULL check the new DecodeN node
        // corresponds to use it as value in implicit_null_check().
        //
        new_in1->set_req(0, n->in(0));
      }

      n->subsume_by(new_in1, this);
      if (in1->outcnt() == 0) {
        in1->disconnect_inputs(NULL, this);
      }
    }
    break;

  case Op_CmpP:
    // Do this transformation here to preserve CmpPNode::sub() and
    // other TypePtr related Ideal optimizations (for example, ptr nullness).
    if (n->in(1)->is_DecodeNarrowPtr() || n->in(2)->is_DecodeNarrowPtr()) {
      Node* in1 = n->in(1);
      Node* in2 = n->in(2);
      if (!in1->is_DecodeNarrowPtr()) {
        in2 = in1;
        in1 = n->in(2);
      }
      assert(in1->is_DecodeNarrowPtr(), "sanity");

      Node* new_in2 = NULL;
      if (in2->is_DecodeNarrowPtr()) {
        assert(in2->Opcode() == in1->Opcode(), "must be same node type");
        new_in2 = in2->in(1);
      } else if (in2->Opcode() == Op_ConP) {
        const Type* t = in2->bottom_type();
        if (t == TypePtr::NULL_PTR) {
          assert(in1->is_DecodeN(), "compare klass to null?");
          // Don't convert CmpP null check into CmpN if compressed
          // oops implicit null check is not generated.
          // This will allow to generate normal oop implicit null check.
          if (Matcher::gen_narrow_oop_implicit_null_checks())
            new_in2 = ConNode::make(this, TypeNarrowOop::NULL_PTR);
          //
          // This transformation together with CastPP transformation above
          // will generated code for implicit NULL checks for compressed oops.
          //
          // The original code after Optimize()
          //
          //    LoadN memory, narrow_oop_reg
          //    decode narrow_oop_reg, base_reg
          //    CmpP base_reg, NULL
          //    CastPP base_reg // NotNull
          //    Load [base_reg + offset], val_reg
          //
          // after these transformations will be
          //
          //    LoadN memory, narrow_oop_reg
          //    CmpN narrow_oop_reg, NULL
          //    decode_not_null narrow_oop_reg, base_reg
          //    Load [base_reg + offset], val_reg
          //
          // and the uncommon path (== NULL) will use narrow_oop_reg directly
          // since narrow oops can be used in debug info now (see the code in
          // final_graph_reshaping_walk()).
          //
          // At the end the code will be matched to
          // on x86:
          //
          //    Load_narrow_oop memory, narrow_oop_reg
          //    Load [R12 + narrow_oop_reg<<3 + offset], val_reg
          //    NullCheck narrow_oop_reg
          //
          // and on sparc:
          //
          //    Load_narrow_oop memory, narrow_oop_reg
          //    decode_not_null narrow_oop_reg, base_reg
          //    Load [base_reg + offset], val_reg
          //    NullCheck base_reg
          //
        } else if (t->isa_oopptr()) {
          new_in2 = ConNode::make(this, t->make_narrowoop());
        } else if (t->isa_klassptr()) {
          new_in2 = ConNode::make(this, t->make_narrowklass());
        }
      }
      if (new_in2 != NULL) {
        Node* cmpN = new (this) CmpNNode(in1->in(1), new_in2);
        n->subsume_by(cmpN, this);
        if (in1->outcnt() == 0) {
          in1->disconnect_inputs(NULL, this);
        }
        if (in2->outcnt() == 0) {
          in2->disconnect_inputs(NULL, this);
        }
      }
    }
    break;

  case Op_DecodeN:
  case Op_DecodeNKlass:
    assert(!n->in(1)->is_EncodeNarrowPtr(), "should be optimized out");
    // DecodeN could be pinned when it can't be fold into
    // an address expression, see the code for Op_CastPP above.
    assert(n->in(0) == NULL || (UseCompressedOops && !Matcher::narrow_oop_use_complex_address()), "no control");
    break;

  case Op_EncodeP:
  case Op_EncodePKlass: {
    Node* in1 = n->in(1);
    if (in1->is_DecodeNarrowPtr()) {
      n->subsume_by(in1->in(1), this);
    } else if (in1->Opcode() == Op_ConP) {
      const Type* t = in1->bottom_type();
      if (t == TypePtr::NULL_PTR) {
        assert(t->isa_oopptr(), "null klass?");
        n->subsume_by(ConNode::make(this, TypeNarrowOop::NULL_PTR), this);
      } else if (t->isa_oopptr()) {
        n->subsume_by(ConNode::make(this, t->make_narrowoop()), this);
      } else if (t->isa_klassptr()) {
        n->subsume_by(ConNode::make(this, t->make_narrowklass()), this);
      }
    }
    if (in1->outcnt() == 0) {
      in1->disconnect_inputs(NULL, this);
    }
    break;
  }

  case Op_Proj: {
    if (OptimizeStringConcat) {
      ProjNode* p = n->as_Proj();
      if (p->_is_io_use) {
        // Separate projections were used for the exception path which
        // are normally removed by a late inline.  If it wasn't inlined
        // then they will hang around and should just be replaced with
        // the original one.
        Node* proj = NULL;
        // Replace with just one
        for (SimpleDUIterator i(p->in(0)); i.has_next(); i.next()) {
          Node *use = i.get();
          if (use->is_Proj() && p != use && use->as_Proj()->_con == p->_con) {
            proj = use;
            break;
          }
        }
        assert(proj != NULL, "must be found");
        p->subsume_by(proj, this);
      }
    }
    break;
  }

  case Op_Phi:
    if (n->as_Phi()->bottom_type()->isa_narrowoop() || n->as_Phi()->bottom_type()->isa_narrowklass()) {
      // The EncodeP optimization may create Phi with the same edges
      // for all paths. It is not handled well by Register Allocator.
      Node* unique_in = n->in(1);
      assert(unique_in != NULL, "");
      uint cnt = n->req();
      for (uint i = 2; i < cnt; i++) {
        Node* m = n->in(i);
        assert(m != NULL, "");
        if (unique_in != m)
          unique_in = NULL;
      }
      if (unique_in != NULL) {
        n->subsume_by(unique_in, this);
      }
    }
    break;

#endif

  case Op_ModI:
    if (UseDivMod) {
      // Check if a%b and a/b both exist
      Node* d = n->find_similar(Op_DivI);
      if (d) {
        // Replace them with a fused divmod if supported
        if (Matcher::has_match_rule(Op_DivModI)) {
          DivModINode* divmod = DivModINode::make(this, n);
          d->subsume_by(divmod->div_proj(), this);
          n->subsume_by(divmod->mod_proj(), this);
        } else {
          // replace a%b with a-((a/b)*b)
          Node* mult = new (this) MulINode(d, d->in(2));
          Node* sub  = new (this) SubINode(d->in(1), mult);
          n->subsume_by(sub, this);
        }
      }
    }
    break;

  case Op_ModL:
    if (UseDivMod) {
      // Check if a%b and a/b both exist
      Node* d = n->find_similar(Op_DivL);
      if (d) {
        // Replace them with a fused divmod if supported
        if (Matcher::has_match_rule(Op_DivModL)) {
          DivModLNode* divmod = DivModLNode::make(this, n);
          d->subsume_by(divmod->div_proj(), this);
          n->subsume_by(divmod->mod_proj(), this);
        } else {
          // replace a%b with a-((a/b)*b)
          Node* mult = new (this) MulLNode(d, d->in(2));
          Node* sub  = new (this) SubLNode(d->in(1), mult);
          n->subsume_by(sub, this);
        }
      }
    }
    break;

  case Op_LoadVector:
  case Op_StoreVector:
    break;

  case Op_PackB:
  case Op_PackS:
  case Op_PackI:
  case Op_PackF:
  case Op_PackL:
  case Op_PackD:
    if (n->req()-1 > 2) {
      // Replace many operand PackNodes with a binary tree for matching
      PackNode* p = (PackNode*) n;
      Node* btp = p->binary_tree_pack(this, 1, n->req());
      n->subsume_by(btp, this);
    }
    break;
  case Op_Loop:
  case Op_CountedLoop:
    if (n->as_Loop()->is_inner_loop()) {
      frc.inc_inner_loop_count();
    }
    break;
  case Op_LShiftI:
  case Op_RShiftI:
  case Op_URShiftI:
  case Op_LShiftL:
  case Op_RShiftL:
  case Op_URShiftL:
    if (Matcher::need_masked_shift_count) {
      // The cpu's shift instructions don't restrict the count to the
      // lower 5/6 bits. We need to do the masking ourselves.
      Node* in2 = n->in(2);
      juint mask = (n->bottom_type() == TypeInt::INT) ? (BitsPerInt - 1) : (BitsPerLong - 1);
      const TypeInt* t = in2->find_int_type();
      if (t != NULL && t->is_con()) {
        juint shift = t->get_con();
        if (shift > mask) { // Unsigned cmp
          n->set_req(2, ConNode::make(this, TypeInt::make(shift & mask)));
        }
      } else {
        if (t == NULL || t->_lo < 0 || t->_hi > (int)mask) {
          Node* shift = new (this) AndINode(in2, ConNode::make(this, TypeInt::make(mask)));
          n->set_req(2, shift);
        }
      }
      if (in2->outcnt() == 0) { // Remove dead node
        in2->disconnect_inputs(NULL, this);
      }
    }
    break;
  case Op_MemBarStoreStore:
  case Op_MemBarRelease:
    // Break the link with AllocateNode: it is no longer useful and
    // confuses register allocation.
    if (n->req() > MemBarNode::Precedent) {
      n->set_req(MemBarNode::Precedent, top());
    }
    break;
    // Must set a control edge on all nodes that produce a FlagsProj
    // so they can't escape the block that consumes the flags.
    // Must also set the non throwing branch as the control
    // for all nodes that depends on the result. Unless the node
    // already have a control that isn't the control of the
    // flag producer
  case Op_FlagsProj:
    {
      MathExactNode* math = (MathExactNode*)  n->in(0);
      Node* ctrl = math->control_node();
      Node* non_throwing = math->non_throwing_branch();
      math->set_req(0, ctrl);

      Node* result = math->result_node();
      if (result != NULL) {
        for (DUIterator_Fast jmax, j = result->fast_outs(jmax); j < jmax; j++) {
          Node* out = result->fast_out(j);
          // Phi nodes shouldn't be moved. They would only match below if they
          // had the same control as the MathExactNode. The only time that
          // would happen is if the Phi is also an input to the MathExact
          //
          // Cmp nodes shouldn't have control set at all.
          if (out->is_Phi() ||
              out->is_Cmp()) {
            continue;
          }

          if (out->in(0) == NULL) {
            out->set_req(0, non_throwing);
          } else if (out->in(0) == ctrl) {
            out->set_req(0, non_throwing);
          }
        }
      }
    }
    break;
  default:
    assert( !n->is_Call(), "" );
    assert( !n->is_Mem(), "" );
    break;
  }

  // Collect CFG split points
  if (n->is_MultiBranch())
    frc._tests.push(n);
}

//------------------------------final_graph_reshaping_walk---------------------
// Replacing Opaque nodes with their input in final_graph_reshaping_impl(),
// requires that the walk visits a node's inputs before visiting the node.
void Compile::final_graph_reshaping_walk( Node_Stack &nstack, Node *root, Final_Reshape_Counts &frc ) {
  ResourceArea *area = Thread::current()->resource_area();
  Unique_Node_List sfpt(area);

  frc._visited.set(root->_idx); // first, mark node as visited
  uint cnt = root->req();
  Node *n = root;
  uint  i = 0;
  while (true) {
    if (i < cnt) {
      // Place all non-visited non-null inputs onto stack
      Node* m = n->in(i);
      ++i;
      if (m != NULL && !frc._visited.test_set(m->_idx)) {
        if (m->is_SafePoint() && m->as_SafePoint()->jvms() != NULL)
          sfpt.push(m);
        cnt = m->req();
        nstack.push(n, i); // put on stack parent and next input's index
        n = m;
        i = 0;
      }
    } else {
      // Now do post-visit work
      final_graph_reshaping_impl( n, frc );
      if (nstack.is_empty())
        break;             // finished
      n = nstack.node();   // Get node from stack
      cnt = n->req();
      i = nstack.index();
      nstack.pop();        // Shift to the next node on stack
    }
  }

  // Skip next transformation if compressed oops are not used.
  if ((UseCompressedOops && !Matcher::gen_narrow_oop_implicit_null_checks()) ||
      (!UseCompressedOops && !UseCompressedClassPointers))
    return;

  // Go over safepoints nodes to skip DecodeN/DecodeNKlass nodes for debug edges.
  // It could be done for an uncommon traps or any safepoints/calls
  // if the DecodeN/DecodeNKlass node is referenced only in a debug info.
  while (sfpt.size() > 0) {
    n = sfpt.pop();
    JVMState *jvms = n->as_SafePoint()->jvms();
    assert(jvms != NULL, "sanity");
    int start = jvms->debug_start();
    int end   = n->req();
    bool is_uncommon = (n->is_CallStaticJava() &&
                        n->as_CallStaticJava()->uncommon_trap_request() != 0);
    for (int j = start; j < end; j++) {
      Node* in = n->in(j);
      if (in->is_DecodeNarrowPtr()) {
        bool safe_to_skip = true;
        if (!is_uncommon ) {
          // Is it safe to skip?
          for (uint i = 0; i < in->outcnt(); i++) {
            Node* u = in->raw_out(i);
            if (!u->is_SafePoint() ||
                 u->is_Call() && u->as_Call()->has_non_debug_use(n)) {
              safe_to_skip = false;
            }
          }
        }
        if (safe_to_skip) {
          n->set_req(j, in->in(1));
        }
        if (in->outcnt() == 0) {
          in->disconnect_inputs(NULL, this);
        }
      }
    }
  }
}

//------------------------------final_graph_reshaping--------------------------
// Final Graph Reshaping.
//
// (1) Clone simple inputs to uncommon calls, so they can be scheduled late
//     and not commoned up and forced early.  Must come after regular
//     optimizations to avoid GVN undoing the cloning.  Clone constant
//     inputs to Loop Phis; these will be split by the allocator anyways.
//     Remove Opaque nodes.
// (2) Move last-uses by commutative operations to the left input to encourage
//     Intel update-in-place two-address operations and better register usage
//     on RISCs.  Must come after regular optimizations to avoid GVN Ideal
//     calls canonicalizing them back.
// (3) Count the number of double-precision FP ops, single-precision FP ops
//     and call sites.  On Intel, we can get correct rounding either by
//     forcing singles to memory (requires extra stores and loads after each
//     FP bytecode) or we can set a rounding mode bit (requires setting and
//     clearing the mode bit around call sites).  The mode bit is only used
//     if the relative frequency of single FP ops to calls is low enough.
//     This is a key transform for SPEC mpeg_audio.
// (4) Detect infinite loops; blobs of code reachable from above but not
//     below.  Several of the Code_Gen algorithms fail on such code shapes,
//     so we simply bail out.  Happens a lot in ZKM.jar, but also happens
//     from time to time in other codes (such as -Xcomp finalizer loops, etc).
//     Detection is by looking for IfNodes where only 1 projection is
//     reachable from below or CatchNodes missing some targets.
// (5) Assert for insane oop offsets in debug mode.

bool Compile::final_graph_reshaping() {
  // an infinite loop may have been eliminated by the optimizer,
  // in which case the graph will be empty.
  if (root()->req() == 1) {
    record_method_not_compilable("trivial infinite loop");
    return true;
  }

  // Expensive nodes have their control input set to prevent the GVN
  // from freely commoning them. There's no GVN beyond this point so
  // no need to keep the control input. We want the expensive nodes to
  // be freely moved to the least frequent code path by gcm.
  assert(OptimizeExpensiveOps || expensive_count() == 0, "optimization off but list non empty?");
  for (int i = 0; i < expensive_count(); i++) {
    _expensive_nodes->at(i)->set_req(0, NULL);
  }

  Final_Reshape_Counts frc;

  // Visit everybody reachable!
  // Allocate stack of size C->unique()/2 to avoid frequent realloc
  Node_Stack nstack(unique() >> 1);
  final_graph_reshaping_walk(nstack, root(), frc);

  // Check for unreachable (from below) code (i.e., infinite loops).
  for( uint i = 0; i < frc._tests.size(); i++ ) {
    MultiBranchNode *n = frc._tests[i]->as_MultiBranch();
    // Get number of CFG targets.
    // Note that PCTables include exception targets after calls.
    uint required_outcnt = n->required_outcnt();
    if (n->outcnt() != required_outcnt) {
      // Check for a few special cases.  Rethrow Nodes never take the
      // 'fall-thru' path, so expected kids is 1 less.
      if (n->is_PCTable() && n->in(0) && n->in(0)->in(0)) {
        if (n->in(0)->in(0)->is_Call()) {
          CallNode *call = n->in(0)->in(0)->as_Call();
          if (call->entry_point() == OptoRuntime::rethrow_stub()) {
            required_outcnt--;      // Rethrow always has 1 less kid
          } else if (call->req() > TypeFunc::Parms &&
                     call->is_CallDynamicJava()) {
            // Check for null receiver. In such case, the optimizer has
            // detected that the virtual call will always result in a null
            // pointer exception. The fall-through projection of this CatchNode
            // will not be populated.
            Node *arg0 = call->in(TypeFunc::Parms);
            if (arg0->is_Type() &&
                arg0->as_Type()->type()->higher_equal(TypePtr::NULL_PTR)) {
              required_outcnt--;
            }
          } else if (call->entry_point() == OptoRuntime::new_array_Java() &&
                     call->req() > TypeFunc::Parms+1 &&
                     call->is_CallStaticJava()) {
            // Check for negative array length. In such case, the optimizer has
            // detected that the allocation attempt will always result in an
            // exception. There is no fall-through projection of this CatchNode .
            Node *arg1 = call->in(TypeFunc::Parms+1);
            if (arg1->is_Type() &&
                arg1->as_Type()->type()->join(TypeInt::POS)->empty()) {
              required_outcnt--;
            }
          }
        }
      }
      // Recheck with a better notion of 'required_outcnt'
      if (n->outcnt() != required_outcnt) {
        record_method_not_compilable("malformed control flow");
        return true;            // Not all targets reachable!
      }
    }
    // Check that I actually visited all kids.  Unreached kids
    // must be infinite loops.
    for (DUIterator_Fast jmax, j = n->fast_outs(jmax); j < jmax; j++)
      if (!frc._visited.test(n->fast_out(j)->_idx)) {
        record_method_not_compilable("infinite loop");
        return true;            // Found unvisited kid; must be unreach
      }
  }

  // If original bytecodes contained a mixture of floats and doubles
  // check if the optimizer has made it homogenous, item (3).
  if( Use24BitFPMode && Use24BitFP && UseSSE == 0 &&
      frc.get_float_count() > 32 &&
      frc.get_double_count() == 0 &&
      (10 * frc.get_call_count() < frc.get_float_count()) ) {
    set_24_bit_selection_and_mode( false,  true );
  }

  set_java_calls(frc.get_java_call_count());
  set_inner_loops(frc.get_inner_loop_count());

  // No infinite loops, no reason to bail out.
  return false;
}

//-----------------------------too_many_traps----------------------------------
// Report if there are too many traps at the current method and bci.
// Return true if there was a trap, and/or PerMethodTrapLimit is exceeded.
bool Compile::too_many_traps(ciMethod* method,
                             int bci,
                             Deoptimization::DeoptReason reason) {
  ciMethodData* md = method->method_data();
  if (md->is_empty()) {
    // Assume the trap has not occurred, or that it occurred only
    // because of a transient condition during start-up in the interpreter.
    return false;
  }
  if (md->has_trap_at(bci, reason) != 0) {
    // Assume PerBytecodeTrapLimit==0, for a more conservative heuristic.
    // Also, if there are multiple reasons, or if there is no per-BCI record,
    // assume the worst.
    if (log())
      log()->elem("observe trap='%s' count='%d'",
                  Deoptimization::trap_reason_name(reason),
                  md->trap_count(reason));
    return true;
  } else {
    // Ignore method/bci and see if there have been too many globally.
    return too_many_traps(reason, md);
  }
}

// Less-accurate variant which does not require a method and bci.
bool Compile::too_many_traps(Deoptimization::DeoptReason reason,
                             ciMethodData* logmd) {
 if (trap_count(reason) >= (uint)PerMethodTrapLimit) {
    // Too many traps globally.
    // Note that we use cumulative trap_count, not just md->trap_count.
    if (log()) {
      int mcount = (logmd == NULL)? -1: (int)logmd->trap_count(reason);
      log()->elem("observe trap='%s' count='0' mcount='%d' ccount='%d'",
                  Deoptimization::trap_reason_name(reason),
                  mcount, trap_count(reason));
    }
    return true;
  } else {
    // The coast is clear.
    return false;
  }
}

//--------------------------too_many_recompiles--------------------------------
// Report if there are too many recompiles at the current method and bci.
// Consults PerBytecodeRecompilationCutoff and PerMethodRecompilationCutoff.
// Is not eager to return true, since this will cause the compiler to use
// Action_none for a trap point, to avoid too many recompilations.
bool Compile::too_many_recompiles(ciMethod* method,
                                  int bci,
                                  Deoptimization::DeoptReason reason) {
  ciMethodData* md = method->method_data();
  if (md->is_empty()) {
    // Assume the trap has not occurred, or that it occurred only
    // because of a transient condition during start-up in the interpreter.
    return false;
  }
  // Pick a cutoff point well within PerBytecodeRecompilationCutoff.
  uint bc_cutoff = (uint) PerBytecodeRecompilationCutoff / 8;
  uint m_cutoff  = (uint) PerMethodRecompilationCutoff / 2 + 1;  // not zero
  Deoptimization::DeoptReason per_bc_reason
    = Deoptimization::reason_recorded_per_bytecode_if_any(reason);
  if ((per_bc_reason == Deoptimization::Reason_none
       || md->has_trap_at(bci, reason) != 0)
      // The trap frequency measure we care about is the recompile count:
      && md->trap_recompiled_at(bci)
      && md->overflow_recompile_count() >= bc_cutoff) {
    // Do not emit a trap here if it has already caused recompilations.
    // Also, if there are multiple reasons, or if there is no per-BCI record,
    // assume the worst.
    if (log())
      log()->elem("observe trap='%s recompiled' count='%d' recompiles2='%d'",
                  Deoptimization::trap_reason_name(reason),
                  md->trap_count(reason),
                  md->overflow_recompile_count());
    return true;
  } else if (trap_count(reason) != 0
             && decompile_count() >= m_cutoff) {
    // Too many recompiles globally, and we have seen this sort of trap.
    // Use cumulative decompile_count, not just md->decompile_count.
    if (log())
      log()->elem("observe trap='%s' count='%d' mcount='%d' decompiles='%d' mdecompiles='%d'",
                  Deoptimization::trap_reason_name(reason),
                  md->trap_count(reason), trap_count(reason),
                  md->decompile_count(), decompile_count());
    return true;
  } else {
    // The coast is clear.
    return false;
  }
}

// Compute when not to trap. Used by matching trap based nodes and
// NullCheck optimization.
void Compile::set_allowed_deopt_reasons() {
  _allowed_reasons = 0;
  if (is_method_compilation()) {
    for (int rs = (int)Deoptimization::Reason_none+1; rs < Compile::trapHistLength; rs++) {
      assert(rs < BitsPerInt, "recode bit map");
      if (!too_many_traps((Deoptimization::DeoptReason) rs)) {
        _allowed_reasons |= nth_bit(rs);
      }
    }
  }
}

#ifndef PRODUCT
//------------------------------verify_graph_edges---------------------------
// Walk the Graph and verify that there is a one-to-one correspondence
// between Use-Def edges and Def-Use edges in the graph.
void Compile::verify_graph_edges(bool no_dead_code) {
  if (VerifyGraphEdges) {
    ResourceArea *area = Thread::current()->resource_area();
    Unique_Node_List visited(area);
    // Call recursive graph walk to check edges
    _root->verify_edges(visited);
    if (no_dead_code) {
      // Now make sure that no visited node is used by an unvisited node.
      bool dead_nodes = 0;
      Unique_Node_List checked(area);
      while (visited.size() > 0) {
        Node* n = visited.pop();
        checked.push(n);
        for (uint i = 0; i < n->outcnt(); i++) {
          Node* use = n->raw_out(i);
          if (checked.member(use))  continue;  // already checked
          if (visited.member(use))  continue;  // already in the graph
          if (use->is_Con())        continue;  // a dead ConNode is OK
          // At this point, we have found a dead node which is DU-reachable.
          if (dead_nodes++ == 0)
            tty->print_cr("*** Dead nodes reachable via DU edges:");
          use->dump(2);
          tty->print_cr("---");
          checked.push(use);  // No repeats; pretend it is now checked.
        }
      }
      assert(dead_nodes == 0, "using nodes must be reachable from root");
    }
  }
}

// Verify GC barriers consistency
// Currently supported:
// - G1 pre-barriers (see GraphKit::g1_write_barrier_pre())
void Compile::verify_barriers() {
  if (UseG1GC) {
    // Verify G1 pre-barriers
    const int marking_offset = in_bytes(JavaThread::satb_mark_queue_offset() + PtrQueue::byte_offset_of_active());

    ResourceArea *area = Thread::current()->resource_area();
    Unique_Node_List visited(area);
    Node_List worklist(area);
    // We're going to walk control flow backwards starting from the Root
    worklist.push(_root);
    while (worklist.size() > 0) {
      Node* x = worklist.pop();
      if (x == NULL || x == top()) continue;
      if (visited.member(x)) {
        continue;
      } else {
        visited.push(x);
      }

      if (x->is_Region()) {
        for (uint i = 1; i < x->req(); i++) {
          worklist.push(x->in(i));
        }
      } else {
        worklist.push(x->in(0));
        // We are looking for the pattern:
        //                            /->ThreadLocal
        // If->Bool->CmpI->LoadB->AddP->ConL(marking_offset)
        //              \->ConI(0)
        // We want to verify that the If and the LoadB have the same control
        // See GraphKit::g1_write_barrier_pre()
        if (x->is_If()) {
          IfNode *iff = x->as_If();
          if (iff->in(1)->is_Bool() && iff->in(1)->in(1)->is_Cmp()) {
            CmpNode *cmp = iff->in(1)->in(1)->as_Cmp();
            if (cmp->Opcode() == Op_CmpI && cmp->in(2)->is_Con() && cmp->in(2)->bottom_type()->is_int()->get_con() == 0
                && cmp->in(1)->is_Load()) {
              LoadNode* load = cmp->in(1)->as_Load();
              if (load->Opcode() == Op_LoadB && load->in(2)->is_AddP() && load->in(2)->in(2)->Opcode() == Op_ThreadLocal
                  && load->in(2)->in(3)->is_Con()
                  && load->in(2)->in(3)->bottom_type()->is_intptr_t()->get_con() == marking_offset) {

                Node* if_ctrl = iff->in(0);
                Node* load_ctrl = load->in(0);

                if (if_ctrl != load_ctrl) {
                  // Skip possible CProj->NeverBranch in infinite loops
                  if ((if_ctrl->is_Proj() && if_ctrl->Opcode() == Op_CProj)
                      && (if_ctrl->in(0)->is_MultiBranch() && if_ctrl->in(0)->Opcode() == Op_NeverBranch)) {
                    if_ctrl = if_ctrl->in(0)->in(0);
                  }
                }
                assert(load_ctrl != NULL && if_ctrl == load_ctrl, "controls must match");
              }
            }
          }
        }
      }
    }
  }
}

#endif

// The Compile object keeps track of failure reasons separately from the ciEnv.
// This is required because there is not quite a 1-1 relation between the
// ciEnv and its compilation task and the Compile object.  Note that one
// ciEnv might use two Compile objects, if C2Compiler::compile_method decides
// to backtrack and retry without subsuming loads.  Other than this backtracking
// behavior, the Compile's failure reason is quietly copied up to the ciEnv
// by the logic in C2Compiler.
void Compile::record_failure(const char* reason) {
  if (log() != NULL) {
    log()->elem("failure reason='%s' phase='compile'", reason);
  }
  if (_failure_reason == NULL) {
    // Record the first failure reason.
    _failure_reason = reason;
  }

  EventCompilerFailure event;
  if (event.should_commit()) {
    event.set_compileID(Compile::compile_id());
    event.set_failure(reason);
    event.commit();
  }

  if (!C->failure_reason_is(C2Compiler::retry_no_subsuming_loads())) {
    C->print_method(PHASE_FAILURE);
  }
  _root = NULL;  // flush the graph, too
}

Compile::TracePhase::TracePhase(const char* name, elapsedTimer* accumulator, bool dolog)
  : TraceTime(NULL, accumulator, false NOT_PRODUCT( || TimeCompiler ), false),
    _phase_name(name), _dolog(dolog)
{
  if (dolog) {
    C = Compile::current();
    _log = C->log();
  } else {
    C = NULL;
    _log = NULL;
  }
  if (_log != NULL) {
    _log->begin_head("phase name='%s' nodes='%d' live='%d'", _phase_name, C->unique(), C->live_nodes());
    _log->stamp();
    _log->end_head();
  }
}

Compile::TracePhase::~TracePhase() {

  C = Compile::current();
  if (_dolog) {
    _log = C->log();
  } else {
    _log = NULL;
  }

#ifdef ASSERT
  if (PrintIdealNodeCount) {
    tty->print_cr("phase name='%s' nodes='%d' live='%d' live_graph_walk='%d'",
                  _phase_name, C->unique(), C->live_nodes(), C->count_live_nodes_by_graph_walk());
  }

  if (VerifyIdealNodeCount) {
    Compile::current()->print_missing_nodes();
  }
#endif

  if (_log != NULL) {
    _log->done("phase name='%s' nodes='%d' live='%d'", _phase_name, C->unique(), C->live_nodes());
  }
}

//=============================================================================
// Two Constant's are equal when the type and the value are equal.
bool Compile::Constant::operator==(const Constant& other) {
  if (type()          != other.type()         )  return false;
  if (can_be_reused() != other.can_be_reused())  return false;
  // For floating point values we compare the bit pattern.
  switch (type()) {
  case T_FLOAT:   return (_v._value.i == other._v._value.i);
  case T_LONG:
  case T_DOUBLE:  return (_v._value.j == other._v._value.j);
  case T_OBJECT:
  case T_ADDRESS: return (_v._value.l == other._v._value.l);
  case T_VOID:    return (_v._value.l == other._v._value.l);  // jump-table entries
  case T_METADATA: return (_v._metadata == other._v._metadata);
  default: ShouldNotReachHere();
  }
  return false;
}

static int type_to_size_in_bytes(BasicType t) {
  switch (t) {
  case T_LONG:    return sizeof(jlong  );
  case T_FLOAT:   return sizeof(jfloat );
  case T_DOUBLE:  return sizeof(jdouble);
  case T_METADATA: return sizeof(Metadata*);
    // We use T_VOID as marker for jump-table entries (labels) which
    // need an internal word relocation.
  case T_VOID:
  case T_ADDRESS:
  case T_OBJECT:  return sizeof(jobject);
  }

  ShouldNotReachHere();
  return -1;
}

int Compile::ConstantTable::qsort_comparator(Constant* a, Constant* b) {
  // sort descending
  if (a->freq() > b->freq())  return -1;
  if (a->freq() < b->freq())  return  1;
  return 0;
}

void Compile::ConstantTable::calculate_offsets_and_size() {
  // First, sort the array by frequencies.
  _constants.sort(qsort_comparator);

#ifdef ASSERT
  // Make sure all jump-table entries were sorted to the end of the
  // array (they have a negative frequency).
  bool found_void = false;
  for (int i = 0; i < _constants.length(); i++) {
    Constant con = _constants.at(i);
    if (con.type() == T_VOID)
      found_void = true;  // jump-tables
    else
      assert(!found_void, "wrong sorting");
  }
#endif

  int offset = 0;
  for (int i = 0; i < _constants.length(); i++) {
    Constant* con = _constants.adr_at(i);

    // Align offset for type.
    int typesize = type_to_size_in_bytes(con->type());
    offset = align_size_up(offset, typesize);
    con->set_offset(offset);   // set constant's offset

    if (con->type() == T_VOID) {
      MachConstantNode* n = (MachConstantNode*) con->get_jobject();
      offset = offset + typesize * n->outcnt();  // expand jump-table
    } else {
      offset = offset + typesize;
    }
  }

  // Align size up to the next section start (which is insts; see
  // CodeBuffer::align_at_start).
  assert(_size == -1, "already set?");
  _size = align_size_up(offset, CodeEntryAlignment);
}

void Compile::ConstantTable::emit(CodeBuffer& cb) {
  MacroAssembler _masm(&cb);
  for (int i = 0; i < _constants.length(); i++) {
    Constant con = _constants.at(i);
    address constant_addr;
    switch (con.type()) {
    case T_LONG:   constant_addr = _masm.long_constant(  con.get_jlong()  ); break;
    case T_FLOAT:  constant_addr = _masm.float_constant( con.get_jfloat() ); break;
    case T_DOUBLE: constant_addr = _masm.double_constant(con.get_jdouble()); break;
    case T_OBJECT: {
      jobject obj = con.get_jobject();
      int oop_index = _masm.oop_recorder()->find_index(obj);
      constant_addr = _masm.address_constant((address) obj, oop_Relocation::spec(oop_index));
      break;
    }
    case T_ADDRESS: {
      address addr = (address) con.get_jobject();
      constant_addr = _masm.address_constant(addr);
      break;
    }
    // We use T_VOID as marker for jump-table entries (labels) which
    // need an internal word relocation.
    case T_VOID: {
      MachConstantNode* n = (MachConstantNode*) con.get_jobject();
      // Fill the jump-table with a dummy word.  The real value is
      // filled in later in fill_jump_table.
      address dummy = (address) n;
      constant_addr = _masm.address_constant(dummy);
      // Expand jump-table
      for (uint i = 1; i < n->outcnt(); i++) {
        address temp_addr = _masm.address_constant(dummy + i);
        assert(temp_addr, "consts section too small");
      }
      break;
    }
    case T_METADATA: {
      Metadata* obj = con.get_metadata();
      int metadata_index = _masm.oop_recorder()->find_index(obj);
      constant_addr = _masm.address_constant((address) obj, metadata_Relocation::spec(metadata_index));
      break;
    }
    default: ShouldNotReachHere();
    }
    assert(constant_addr, "consts section too small");
    assert((constant_addr - _masm.code()->consts()->start()) == con.offset(), err_msg_res("must be: %d == %d", constant_addr - _masm.code()->consts()->start(), con.offset()));
  }
}

int Compile::ConstantTable::find_offset(Constant& con) const {
  int idx = _constants.find(con);
  assert(idx != -1, "constant must be in constant table");
  int offset = _constants.at(idx).offset();
  assert(offset != -1, "constant table not emitted yet?");
  return offset;
}

void Compile::ConstantTable::add(Constant& con) {
  if (con.can_be_reused()) {
    int idx = _constants.find(con);
    if (idx != -1 && _constants.at(idx).can_be_reused()) {
      _constants.adr_at(idx)->inc_freq(con.freq());  // increase the frequency by the current value
      return;
    }
  }
  (void) _constants.append(con);
}

Compile::Constant Compile::ConstantTable::add(MachConstantNode* n, BasicType type, jvalue value) {
  Block* b = Compile::current()->cfg()->get_block_for_node(n);
  Constant con(type, value, b->_freq);
  add(con);
  return con;
}

Compile::Constant Compile::ConstantTable::add(Metadata* metadata) {
  Constant con(metadata);
  add(con);
  return con;
}

Compile::Constant Compile::ConstantTable::add(MachConstantNode* n, MachOper* oper) {
  jvalue value;
  BasicType type = oper->type()->basic_type();
  switch (type) {
  case T_LONG:    value.j = oper->constantL(); break;
  case T_FLOAT:   value.f = oper->constantF(); break;
  case T_DOUBLE:  value.d = oper->constantD(); break;
  case T_OBJECT:
  case T_ADDRESS: value.l = (jobject) oper->constant(); break;
  case T_METADATA: return add((Metadata*)oper->constant()); break;
  default: guarantee(false, err_msg_res("unhandled type: %s", type2name(type)));
  }
  return add(n, type, value);
}

Compile::Constant Compile::ConstantTable::add_jump_table(MachConstantNode* n) {
  jvalue value;
  // We can use the node pointer here to identify the right jump-table
  // as this method is called from Compile::Fill_buffer right before
  // the MachNodes are emitted and the jump-table is filled (means the
  // MachNode pointers do not change anymore).
  value.l = (jobject) n;
  Constant con(T_VOID, value, next_jump_table_freq(), false);  // Labels of a jump-table cannot be reused.
  add(con);
  return con;
}

void Compile::ConstantTable::fill_jump_table(CodeBuffer& cb, MachConstantNode* n, GrowableArray<Label*> labels) const {
  // If called from Compile::scratch_emit_size do nothing.
  if (Compile::current()->in_scratch_emit_size())  return;

  assert(labels.is_nonempty(), "must be");
  assert((uint) labels.length() == n->outcnt(), err_msg_res("must be equal: %d == %d", labels.length(), n->outcnt()));

  // Since MachConstantNode::constant_offset() also contains
  // table_base_offset() we need to subtract the table_base_offset()
  // to get the plain offset into the constant table.
  int offset = n->constant_offset() - table_base_offset();

  MacroAssembler _masm(&cb);
  address* jump_table_base = (address*) (_masm.code()->consts()->start() + offset);

  for (uint i = 0; i < n->outcnt(); i++) {
    address* constant_addr = &jump_table_base[i];
    assert(*constant_addr == (((address) n) + i), err_msg_res("all jump-table entries must contain adjusted node pointer: " INTPTR_FORMAT " == " INTPTR_FORMAT, *constant_addr, (((address) n) + i)));
    *constant_addr = cb.consts()->target(*labels.at(i), (address) constant_addr);
    cb.consts()->relocate((address) constant_addr, relocInfo::internal_word_type);
  }
}

void Compile::dump_inlining() {
  if (print_inlining() || print_intrinsics()) {
    // Print inlining message for candidates that we couldn't inline
    // for lack of space or non constant receiver
    for (int i = 0; i < _late_inlines.length(); i++) {
      CallGenerator* cg = _late_inlines.at(i);
      cg->print_inlining_late("live nodes > LiveNodeCountInliningCutoff");
    }
    Unique_Node_List useful;
    useful.push(root());
    for (uint next = 0; next < useful.size(); ++next) {
      Node* n  = useful.at(next);
      if (n->is_Call() && n->as_Call()->generator() != NULL && n->as_Call()->generator()->call_node() == n) {
        CallNode* call = n->as_Call();
        CallGenerator* cg = call->generator();
        cg->print_inlining_late("receiver not constant");
      }
      uint max = n->len();
      for ( uint i = 0; i < max; ++i ) {
        Node *m = n->in(i);
        if ( m == NULL ) continue;
        useful.push(m);
      }
    }
    for (int i = 0; i < _print_inlining_list->length(); i++) {
      tty->print(_print_inlining_list->adr_at(i)->ss()->as_string());
    }
  }
}

int Compile::cmp_expensive_nodes(Node* n1, Node* n2) {
  if (n1->Opcode() < n2->Opcode())      return -1;
  else if (n1->Opcode() > n2->Opcode()) return 1;

  assert(n1->req() == n2->req(), err_msg_res("can't compare %s nodes: n1->req() = %d, n2->req() = %d", NodeClassNames[n1->Opcode()], n1->req(), n2->req()));
  for (uint i = 1; i < n1->req(); i++) {
    if (n1->in(i) < n2->in(i))      return -1;
    else if (n1->in(i) > n2->in(i)) return 1;
  }

  return 0;
}

int Compile::cmp_expensive_nodes(Node** n1p, Node** n2p) {
  Node* n1 = *n1p;
  Node* n2 = *n2p;

  return cmp_expensive_nodes(n1, n2);
}

void Compile::sort_expensive_nodes() {
  if (!expensive_nodes_sorted()) {
    _expensive_nodes->sort(cmp_expensive_nodes);
  }
}

bool Compile::expensive_nodes_sorted() const {
  for (int i = 1; i < _expensive_nodes->length(); i++) {
    if (cmp_expensive_nodes(_expensive_nodes->adr_at(i), _expensive_nodes->adr_at(i-1)) < 0) {
      return false;
    }
  }
  return true;
}

bool Compile::should_optimize_expensive_nodes(PhaseIterGVN &igvn) {
  if (_expensive_nodes->length() == 0) {
    return false;
  }

  assert(OptimizeExpensiveOps, "optimization off?");

  // Take this opportunity to remove dead nodes from the list
  int j = 0;
  for (int i = 0; i < _expensive_nodes->length(); i++) {
    Node* n = _expensive_nodes->at(i);
    if (!n->is_unreachable(igvn)) {
      assert(n->is_expensive(), "should be expensive");
      _expensive_nodes->at_put(j, n);
      j++;
    }
  }
  _expensive_nodes->trunc_to(j);

  // Then sort the list so that similar nodes are next to each other
  // and check for at least two nodes of identical kind with same data
  // inputs.
  sort_expensive_nodes();

  for (int i = 0; i < _expensive_nodes->length()-1; i++) {
    if (cmp_expensive_nodes(_expensive_nodes->adr_at(i), _expensive_nodes->adr_at(i+1)) == 0) {
      return true;
    }
  }

  return false;
}

void Compile::cleanup_expensive_nodes(PhaseIterGVN &igvn) {
  if (_expensive_nodes->length() == 0) {
    return;
  }

  assert(OptimizeExpensiveOps, "optimization off?");

  // Sort to bring similar nodes next to each other and clear the
  // control input of nodes for which there's only a single copy.
  sort_expensive_nodes();

  int j = 0;
  int identical = 0;
  int i = 0;
  for (; i < _expensive_nodes->length()-1; i++) {
    assert(j <= i, "can't write beyond current index");
    if (_expensive_nodes->at(i)->Opcode() == _expensive_nodes->at(i+1)->Opcode()) {
      identical++;
      _expensive_nodes->at_put(j++, _expensive_nodes->at(i));
      continue;
    }
    if (identical > 0) {
      _expensive_nodes->at_put(j++, _expensive_nodes->at(i));
      identical = 0;
    } else {
      Node* n = _expensive_nodes->at(i);
      igvn.hash_delete(n);
      n->set_req(0, NULL);
      igvn.hash_insert(n);
    }
  }
  if (identical > 0) {
    _expensive_nodes->at_put(j++, _expensive_nodes->at(i));
  } else if (_expensive_nodes->length() >= 1) {
    Node* n = _expensive_nodes->at(i);
    igvn.hash_delete(n);
    n->set_req(0, NULL);
    igvn.hash_insert(n);
  }
  _expensive_nodes->trunc_to(j);
}

void Compile::add_expensive_node(Node * n) {
  assert(!_expensive_nodes->contains(n), "duplicate entry in expensive list");
  assert(n->is_expensive(), "expensive nodes with non-null control here only");
  assert(!n->is_CFG() && !n->is_Mem(), "no cfg or memory nodes here");
  if (OptimizeExpensiveOps) {
    _expensive_nodes->append(n);
  } else {
    // Clear control input and let IGVN optimize expensive nodes if
    // OptimizeExpensiveOps is off.
    n->set_req(0, NULL);
  }
}

/**
 * Remove the speculative part of types and clean up the graph
 */
void Compile::remove_speculative_types(PhaseIterGVN &igvn) {
  if (UseTypeSpeculation) {
    Unique_Node_List worklist;
    worklist.push(root());
    int modified = 0;
    // Go over all type nodes that carry a speculative type, drop the
    // speculative part of the type and enqueue the node for an igvn
    // which may optimize it out.
    for (uint next = 0; next < worklist.size(); ++next) {
      Node *n  = worklist.at(next);
      if (n->is_Type() && n->as_Type()->type()->isa_oopptr() != NULL &&
          n->as_Type()->type()->is_oopptr()->speculative() != NULL) {
        TypeNode* tn = n->as_Type();
        const TypeOopPtr* t = tn->type()->is_oopptr();
        bool in_hash = igvn.hash_delete(n);
        assert(in_hash, "node should be in igvn hash table");
        tn->set_type(t->remove_speculative());
        igvn.hash_insert(n);
        igvn._worklist.push(n); // give it a chance to go away
        modified++;
      }
      uint max = n->len();
      for( uint i = 0; i < max; ++i ) {
        Node *m = n->in(i);
        if (not_a_node(m))  continue;
        worklist.push(m);
      }
    }
    // Drop the speculative part of all types in the igvn's type table
    igvn.remove_speculative_types();
    if (modified > 0) {
      igvn.optimize();
    }
  }
}

// Auxiliary method to support randomized stressing/fuzzing.
//
// This method can be called the arbitrary number of times, with current count
// as the argument. The logic allows selecting a single candidate from the
// running list of candidates as follows:
//    int count = 0;
//    Cand* selected = null;
//    while(cand = cand->next()) {
//      if (randomized_select(++count)) {
//        selected = cand;
//      }
//    }
//
// Including count equalizes the chances any candidate is "selected".
// This is useful when we don't have the complete list of candidates to choose
// from uniformly. In this case, we need to adjust the randomicity of the
// selection, or else we will end up biasing the selection towards the latter
// candidates.
//
// Quick back-envelope calculation shows that for the list of n candidates
// the equal probability for the candidate to persist as "best" can be
// achieved by replacing it with "next" k-th candidate with the probability
// of 1/k. It can be easily shown that by the end of the run, the
// probability for any candidate is converged to 1/n, thus giving the
// uniform distribution among all the candidates.
//
// We don't care about the domain size as long as (RANDOMIZED_DOMAIN / count) is large.
#define RANDOMIZED_DOMAIN_POW 29
#define RANDOMIZED_DOMAIN (1 << RANDOMIZED_DOMAIN_POW)
#define RANDOMIZED_DOMAIN_MASK ((1 << (RANDOMIZED_DOMAIN_POW + 1)) - 1)
bool Compile::randomized_select(int count) {
  assert(count > 0, "only positive");
  return (os::random() & RANDOMIZED_DOMAIN_MASK) < (RANDOMIZED_DOMAIN / count);
}