1 /*
   2  * Copyright (c) 1998, 2011, Oracle and/or its affiliates. All rights reserved.
   3  * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
   4  *
   5  * This code is free software; you can redistribute it and/or modify it
   6  * under the terms of the GNU General Public License version 2 only, as
   7  * published by the Free Software Foundation.
   8  *
   9  * This code is distributed in the hope that it will be useful, but WITHOUT
  10  * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
  11  * FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License
  12  * version 2 for more details (a copy is included in the LICENSE file that
  13  * accompanied this code).
  14  *
  15  * You should have received a copy of the GNU General Public License version
  16  * 2 along with this work; if not, write to the Free Software Foundation,
  17  * Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA.
  18  *
  19  * Please contact Oracle, 500 Oracle Parkway, Redwood Shores, CA 94065 USA
  20  * or visit www.oracle.com if you need additional information or have any
  21  * questions.
  22  *
  23  */
  24 
  25 #ifndef SHARE_VM_OPTO_LOOPNODE_HPP
  26 #define SHARE_VM_OPTO_LOOPNODE_HPP
  27 
  28 #include "opto/cfgnode.hpp"
  29 #include "opto/multnode.hpp"
  30 #include "opto/phaseX.hpp"
  31 #include "opto/subnode.hpp"
  32 #include "opto/type.hpp"
  33 
  34 class CmpNode;
  35 class CountedLoopEndNode;
  36 class CountedLoopNode;
  37 class IdealLoopTree;
  38 class LoopNode;
  39 class Node;
  40 class PhaseIdealLoop;
  41 class VectorSet;
  42 class Invariance;
  43 struct small_cache;
  44 
  45 //
  46 //                  I D E A L I Z E D   L O O P S
  47 //
  48 // Idealized loops are the set of loops I perform more interesting
  49 // transformations on, beyond simple hoisting.
  50 
  51 //------------------------------LoopNode---------------------------------------
  52 // Simple loop header.  Fall in path on left, loop-back path on right.
  53 class LoopNode : public RegionNode {
  54   // Size is bigger to hold the flags.  However, the flags do not change
  55   // the semantics so it does not appear in the hash & cmp functions.
  56   virtual uint size_of() const { return sizeof(*this); }
  57 protected:
  58   short _loop_flags;
  59   // Names for flag bitfields
  60   enum { Normal=0, Pre=1, Main=2, Post=3, PreMainPostFlagsMask=3,
  61          MainHasNoPreLoop=4,
  62          HasExactTripCount=8,
  63          InnerLoop=16,
  64          PartialPeelLoop=32,
  65          PartialPeelFailed=64 };
  66   char _unswitch_count;
  67   enum { _unswitch_max=3 };
  68 
  69 public:
  70   // Names for edge indices
  71   enum { Self=0, EntryControl, LoopBackControl };
  72 
  73   int is_inner_loop() const { return _loop_flags & InnerLoop; }
  74   void set_inner_loop() { _loop_flags |= InnerLoop; }
  75 
  76   int is_partial_peel_loop() const { return _loop_flags & PartialPeelLoop; }
  77   void set_partial_peel_loop() { _loop_flags |= PartialPeelLoop; }
  78   int partial_peel_has_failed() const { return _loop_flags & PartialPeelFailed; }
  79   void mark_partial_peel_failed() { _loop_flags |= PartialPeelFailed; }
  80 
  81   int unswitch_max() { return _unswitch_max; }
  82   int unswitch_count() { return _unswitch_count; }
  83   void set_unswitch_count(int val) {
  84     assert (val <= unswitch_max(), "too many unswitches");
  85     _unswitch_count = val;
  86   }
  87 
  88   LoopNode( Node *entry, Node *backedge ) : RegionNode(3), _loop_flags(0), _unswitch_count(0) {
  89     init_class_id(Class_Loop);
  90     init_req(EntryControl, entry);
  91     init_req(LoopBackControl, backedge);
  92   }
  93 
  94   virtual Node *Ideal(PhaseGVN *phase, bool can_reshape);
  95   virtual int Opcode() const;
  96   bool can_be_counted_loop(PhaseTransform* phase) const {
  97     return req() == 3 && in(0) != NULL &&
  98       in(1) != NULL && phase->type(in(1)) != Type::TOP &&
  99       in(2) != NULL && phase->type(in(2)) != Type::TOP;
 100   }
 101   bool is_valid_counted_loop() const;
 102 #ifndef PRODUCT
 103   virtual void dump_spec(outputStream *st) const;
 104 #endif
 105 };
 106 
 107 //------------------------------Counted Loops----------------------------------
 108 // Counted loops are all trip-counted loops, with exactly 1 trip-counter exit
 109 // path (and maybe some other exit paths).  The trip-counter exit is always
 110 // last in the loop.  The trip-counter have to stride by a constant;
 111 // the exit value is also loop invariant.
 112 
 113 // CountedLoopNodes and CountedLoopEndNodes come in matched pairs.  The
 114 // CountedLoopNode has the incoming loop control and the loop-back-control
 115 // which is always the IfTrue before the matching CountedLoopEndNode.  The
 116 // CountedLoopEndNode has an incoming control (possibly not the
 117 // CountedLoopNode if there is control flow in the loop), the post-increment
 118 // trip-counter value, and the limit.  The trip-counter value is always of
 119 // the form (Op old-trip-counter stride).  The old-trip-counter is produced
 120 // by a Phi connected to the CountedLoopNode.  The stride is constant.
 121 // The Op is any commutable opcode, including Add, Mul, Xor.  The
 122 // CountedLoopEndNode also takes in the loop-invariant limit value.
 123 
 124 // From a CountedLoopNode I can reach the matching CountedLoopEndNode via the
 125 // loop-back control.  From CountedLoopEndNodes I can reach CountedLoopNodes
 126 // via the old-trip-counter from the Op node.
 127 
 128 //------------------------------CountedLoopNode--------------------------------
 129 // CountedLoopNodes head simple counted loops.  CountedLoopNodes have as
 130 // inputs the incoming loop-start control and the loop-back control, so they
 131 // act like RegionNodes.  They also take in the initial trip counter, the
 132 // loop-invariant stride and the loop-invariant limit value.  CountedLoopNodes
 133 // produce a loop-body control and the trip counter value.  Since
 134 // CountedLoopNodes behave like RegionNodes I still have a standard CFG model.
 135 
 136 class CountedLoopNode : public LoopNode {
 137   // Size is bigger to hold _main_idx.  However, _main_idx does not change
 138   // the semantics so it does not appear in the hash & cmp functions.
 139   virtual uint size_of() const { return sizeof(*this); }
 140 
 141   // For Pre- and Post-loops during debugging ONLY, this holds the index of
 142   // the Main CountedLoop.  Used to assert that we understand the graph shape.
 143   node_idx_t _main_idx;
 144 
 145   // Known trip count calculated by compute_exact_trip_count()
 146   uint  _trip_count;
 147 
 148   // Expected trip count from profile data
 149   float _profile_trip_cnt;
 150 
 151   // Log2 of original loop bodies in unrolled loop
 152   int _unrolled_count_log2;
 153 
 154   // Node count prior to last unrolling - used to decide if
 155   // unroll,optimize,unroll,optimize,... is making progress
 156   int _node_count_before_unroll;
 157 
 158 public:
 159   CountedLoopNode( Node *entry, Node *backedge )
 160     : LoopNode(entry, backedge), _main_idx(0), _trip_count(max_juint),
 161       _profile_trip_cnt(COUNT_UNKNOWN), _unrolled_count_log2(0),
 162       _node_count_before_unroll(0) {
 163     init_class_id(Class_CountedLoop);
 164     // Initialize _trip_count to the largest possible value.
 165     // Will be reset (lower) if the loop's trip count is known.
 166   }
 167 
 168   virtual int Opcode() const;
 169   virtual Node *Ideal(PhaseGVN *phase, bool can_reshape);
 170 
 171   Node *init_control() const { return in(EntryControl); }
 172   Node *back_control() const { return in(LoopBackControl); }
 173   CountedLoopEndNode *loopexit() const;
 174   Node *init_trip() const;
 175   Node *stride() const;
 176   int   stride_con() const;
 177   bool  stride_is_con() const;
 178   Node *limit() const;
 179   Node *incr() const;
 180   Node *phi() const;
 181 
 182   // Match increment with optional truncation
 183   static Node* match_incr_with_optional_truncation(Node* expr, Node** trunc1, Node** trunc2, const TypeInt** trunc_type);
 184 
 185   // A 'main' loop has a pre-loop and a post-loop.  The 'main' loop
 186   // can run short a few iterations and may start a few iterations in.
 187   // It will be RCE'd and unrolled and aligned.
 188 
 189   // A following 'post' loop will run any remaining iterations.  Used
 190   // during Range Check Elimination, the 'post' loop will do any final
 191   // iterations with full checks.  Also used by Loop Unrolling, where
 192   // the 'post' loop will do any epilog iterations needed.  Basically,
 193   // a 'post' loop can not profitably be further unrolled or RCE'd.
 194 
 195   // A preceding 'pre' loop will run at least 1 iteration (to do peeling),
 196   // it may do under-flow checks for RCE and may do alignment iterations
 197   // so the following main loop 'knows' that it is striding down cache
 198   // lines.
 199 
 200   // A 'main' loop that is ONLY unrolled or peeled, never RCE'd or
 201   // Aligned, may be missing it's pre-loop.
 202   int is_normal_loop() const { return (_loop_flags&PreMainPostFlagsMask) == Normal; }
 203   int is_pre_loop   () const { return (_loop_flags&PreMainPostFlagsMask) == Pre;    }
 204   int is_main_loop  () const { return (_loop_flags&PreMainPostFlagsMask) == Main;   }
 205   int is_post_loop  () const { return (_loop_flags&PreMainPostFlagsMask) == Post;   }
 206   int is_main_no_pre_loop() const { return _loop_flags & MainHasNoPreLoop; }
 207   void set_main_no_pre_loop() { _loop_flags |= MainHasNoPreLoop; }
 208 
 209   int main_idx() const { return _main_idx; }
 210 
 211 
 212   void set_pre_loop  (CountedLoopNode *main) { assert(is_normal_loop(),""); _loop_flags |= Pre ; _main_idx = main->_idx; }
 213   void set_main_loop (                     ) { assert(is_normal_loop(),""); _loop_flags |= Main;                         }
 214   void set_post_loop (CountedLoopNode *main) { assert(is_normal_loop(),""); _loop_flags |= Post; _main_idx = main->_idx; }
 215   void set_normal_loop(                    ) { _loop_flags &= ~PreMainPostFlagsMask; }
 216 
 217   void set_trip_count(uint tc) { _trip_count = tc; }
 218   uint trip_count()            { return _trip_count; }
 219 
 220   bool has_exact_trip_count() const { return (_loop_flags & HasExactTripCount) != 0; }
 221   void set_exact_trip_count(uint tc) {
 222     _trip_count = tc;
 223     _loop_flags |= HasExactTripCount;
 224   }
 225   void set_nonexact_trip_count() {
 226     _loop_flags &= ~HasExactTripCount;
 227   }
 228 
 229   void set_profile_trip_cnt(float ptc) { _profile_trip_cnt = ptc; }
 230   float profile_trip_cnt()             { return _profile_trip_cnt; }
 231 
 232   void double_unrolled_count() { _unrolled_count_log2++; }
 233   int  unrolled_count()        { return 1 << MIN2(_unrolled_count_log2, BitsPerInt-3); }
 234 
 235   void set_node_count_before_unroll(int ct) { _node_count_before_unroll = ct; }
 236   int  node_count_before_unroll()           { return _node_count_before_unroll; }
 237 
 238 #ifndef PRODUCT
 239   virtual void dump_spec(outputStream *st) const;
 240 #endif
 241 };
 242 
 243 //------------------------------CountedLoopEndNode-----------------------------
 244 // CountedLoopEndNodes end simple trip counted loops.  They act much like
 245 // IfNodes.
 246 class CountedLoopEndNode : public IfNode {
 247 public:
 248   enum { TestControl, TestValue };
 249 
 250   CountedLoopEndNode( Node *control, Node *test, float prob, float cnt )
 251     : IfNode( control, test, prob, cnt) {
 252     init_class_id(Class_CountedLoopEnd);
 253   }
 254   virtual int Opcode() const;
 255 
 256   Node *cmp_node() const            { return (in(TestValue)->req() >=2) ? in(TestValue)->in(1) : NULL; }
 257   Node *incr() const                { Node *tmp = cmp_node(); return (tmp && tmp->req()==3) ? tmp->in(1) : NULL; }
 258   Node *limit() const               { Node *tmp = cmp_node(); return (tmp && tmp->req()==3) ? tmp->in(2) : NULL; }
 259   Node *stride() const              { Node *tmp = incr    (); return (tmp && tmp->req()==3) ? tmp->in(2) : NULL; }
 260   Node *phi() const                 { Node *tmp = incr    (); return (tmp && tmp->req()==3) ? tmp->in(1) : NULL; }
 261   Node *init_trip() const           { Node *tmp = phi     (); return (tmp && tmp->req()==3) ? tmp->in(1) : NULL; }
 262   int stride_con() const;
 263   bool stride_is_con() const        { Node *tmp = stride  (); return (tmp != NULL && tmp->is_Con()); }
 264   BoolTest::mask test_trip() const  { return in(TestValue)->as_Bool()->_test._test; }
 265   CountedLoopNode *loopnode() const {
 266     Node *ln = phi()->in(0);
 267     assert( ln->Opcode() == Op_CountedLoop, "malformed loop" );
 268     return (CountedLoopNode*)ln; }
 269 
 270 #ifndef PRODUCT
 271   virtual void dump_spec(outputStream *st) const;
 272 #endif
 273 };
 274 
 275 
 276 inline CountedLoopEndNode *CountedLoopNode::loopexit() const {
 277   Node *bc = back_control();
 278   if( bc == NULL ) return NULL;
 279   Node *le = bc->in(0);
 280   if( le->Opcode() != Op_CountedLoopEnd )
 281     return NULL;
 282   return (CountedLoopEndNode*)le;
 283 }
 284 inline Node *CountedLoopNode::init_trip() const { return loopexit() ? loopexit()->init_trip() : NULL; }
 285 inline Node *CountedLoopNode::stride() const { return loopexit() ? loopexit()->stride() : NULL; }
 286 inline int CountedLoopNode::stride_con() const { return loopexit() ? loopexit()->stride_con() : 0; }
 287 inline bool CountedLoopNode::stride_is_con() const { return loopexit() && loopexit()->stride_is_con(); }
 288 inline Node *CountedLoopNode::limit() const { return loopexit() ? loopexit()->limit() : NULL; }
 289 inline Node *CountedLoopNode::incr() const { return loopexit() ? loopexit()->incr() : NULL; }
 290 inline Node *CountedLoopNode::phi() const { return loopexit() ? loopexit()->phi() : NULL; }
 291 
 292 //------------------------------LoopLimitNode-----------------------------
 293 // Counted Loop limit node which represents exact final iterator value:
 294 // trip_count = (limit - init_trip + stride - 1)/stride
 295 // final_value= trip_count * stride + init_trip.
 296 // Use HW instructions to calculate it when it can overflow in integer.
 297 // Note, final_value should fit into integer since counted loop has
 298 // limit check: limit <= max_int-stride.
 299 class LoopLimitNode : public Node {
 300   enum { Init=1, Limit=2, Stride=3 };
 301  public:
 302   LoopLimitNode( Compile* C, Node *init, Node *limit, Node *stride ) : Node(0,init,limit,stride) {
 303     // Put it on the Macro nodes list to optimize during macro nodes expansion.
 304     init_flags(Flag_is_macro);
 305     C->add_macro_node(this);
 306   }
 307   virtual int Opcode() const;
 308   virtual const Type *bottom_type() const { return TypeInt::INT; }
 309   virtual uint ideal_reg() const { return Op_RegI; }
 310   virtual const Type *Value( PhaseTransform *phase ) const;
 311   virtual Node *Ideal(PhaseGVN *phase, bool can_reshape);
 312   virtual Node *Identity( PhaseTransform *phase );
 313 };
 314 
 315 // -----------------------------IdealLoopTree----------------------------------
 316 class IdealLoopTree : public ResourceObj {
 317 public:
 318   IdealLoopTree *_parent;       // Parent in loop tree
 319   IdealLoopTree *_next;         // Next sibling in loop tree
 320   IdealLoopTree *_child;        // First child in loop tree
 321 
 322   // The head-tail backedge defines the loop.
 323   // If tail is NULL then this loop has multiple backedges as part of the
 324   // same loop.  During cleanup I'll peel off the multiple backedges; merge
 325   // them at the loop bottom and flow 1 real backedge into the loop.
 326   Node *_head;                  // Head of loop
 327   Node *_tail;                  // Tail of loop
 328   inline Node *tail();          // Handle lazy update of _tail field
 329   PhaseIdealLoop* _phase;
 330 
 331   Node_List _body;              // Loop body for inner loops
 332 
 333   uint8 _nest;                  // Nesting depth
 334   uint8 _irreducible:1,         // True if irreducible
 335         _has_call:1,            // True if has call safepoint
 336         _has_sfpt:1,            // True if has non-call safepoint
 337         _rce_candidate:1;       // True if candidate for range check elimination
 338 
 339   Node_List* _required_safept;  // A inner loop cannot delete these safepts;
 340   bool  _allow_optimizations;   // Allow loop optimizations
 341 
 342   IdealLoopTree( PhaseIdealLoop* phase, Node *head, Node *tail )
 343     : _parent(0), _next(0), _child(0),
 344       _head(head), _tail(tail),
 345       _phase(phase),
 346       _required_safept(NULL),
 347       _allow_optimizations(true),
 348       _nest(0), _irreducible(0), _has_call(0), _has_sfpt(0), _rce_candidate(0)
 349   { }
 350 
 351   // Is 'l' a member of 'this'?
 352   int is_member( const IdealLoopTree *l ) const; // Test for nested membership
 353 
 354   // Set loop nesting depth.  Accumulate has_call bits.
 355   int set_nest( uint depth );
 356 
 357   // Split out multiple fall-in edges from the loop header.  Move them to a
 358   // private RegionNode before the loop.  This becomes the loop landing pad.
 359   void split_fall_in( PhaseIdealLoop *phase, int fall_in_cnt );
 360 
 361   // Split out the outermost loop from this shared header.
 362   void split_outer_loop( PhaseIdealLoop *phase );
 363 
 364   // Merge all the backedges from the shared header into a private Region.
 365   // Feed that region as the one backedge to this loop.
 366   void merge_many_backedges( PhaseIdealLoop *phase );
 367 
 368   // Split shared headers and insert loop landing pads.
 369   // Insert a LoopNode to replace the RegionNode.
 370   // Returns TRUE if loop tree is structurally changed.
 371   bool beautify_loops( PhaseIdealLoop *phase );
 372 
 373   // Perform optimization to use the loop predicates for null checks and range checks.
 374   // Applies to any loop level (not just the innermost one)
 375   bool loop_predication( PhaseIdealLoop *phase);
 376 
 377   // Perform iteration-splitting on inner loops.  Split iterations to
 378   // avoid range checks or one-shot null checks.  Returns false if the
 379   // current round of loop opts should stop.
 380   bool iteration_split( PhaseIdealLoop *phase, Node_List &old_new );
 381 
 382   // Driver for various flavors of iteration splitting.  Returns false
 383   // if the current round of loop opts should stop.
 384   bool iteration_split_impl( PhaseIdealLoop *phase, Node_List &old_new );
 385 
 386   // Given dominators, try to find loops with calls that must always be
 387   // executed (call dominates loop tail).  These loops do not need non-call
 388   // safepoints (ncsfpt).
 389   void check_safepts(VectorSet &visited, Node_List &stack);
 390 
 391   // Allpaths backwards scan from loop tail, terminating each path at first safepoint
 392   // encountered.
 393   void allpaths_check_safepts(VectorSet &visited, Node_List &stack);
 394 
 395   // Convert to counted loops where possible
 396   void counted_loop( PhaseIdealLoop *phase );
 397 
 398   // Check for Node being a loop-breaking test
 399   Node *is_loop_exit(Node *iff) const;
 400 
 401   // Returns true if ctrl is executed on every complete iteration
 402   bool dominates_backedge(Node* ctrl);
 403 
 404   // Remove simplistic dead code from loop body
 405   void DCE_loop_body();
 406 
 407   // Look for loop-exit tests with my 50/50 guesses from the Parsing stage.
 408   // Replace with a 1-in-10 exit guess.
 409   void adjust_loop_exit_prob( PhaseIdealLoop *phase );
 410 
 411   // Return TRUE or FALSE if the loop should never be RCE'd or aligned.
 412   // Useful for unrolling loops with NO array accesses.
 413   bool policy_peel_only( PhaseIdealLoop *phase ) const;
 414 
 415   // Return TRUE or FALSE if the loop should be unswitched -- clone
 416   // loop with an invariant test
 417   bool policy_unswitching( PhaseIdealLoop *phase ) const;
 418 
 419   // Micro-benchmark spamming.  Remove empty loops.
 420   bool policy_do_remove_empty_loop( PhaseIdealLoop *phase );
 421 
 422   // Convert one iteration loop into normal code.
 423   bool policy_do_one_iteration_loop( PhaseIdealLoop *phase );
 424 
 425   // Return TRUE or FALSE if the loop should be peeled or not.  Peel if we can
 426   // make some loop-invariant test (usually a null-check) happen before the
 427   // loop.
 428   bool policy_peeling( PhaseIdealLoop *phase ) const;
 429 
 430   // Return TRUE or FALSE if the loop should be maximally unrolled. Stash any
 431   // known trip count in the counted loop node.
 432   bool policy_maximally_unroll( PhaseIdealLoop *phase ) const;
 433 
 434   // Return TRUE or FALSE if the loop should be unrolled or not.  Unroll if
 435   // the loop is a CountedLoop and the body is small enough.
 436   bool policy_unroll( PhaseIdealLoop *phase ) const;
 437 
 438   // Return TRUE or FALSE if the loop should be range-check-eliminated.
 439   // Gather a list of IF tests that are dominated by iteration splitting;
 440   // also gather the end of the first split and the start of the 2nd split.
 441   bool policy_range_check( PhaseIdealLoop *phase ) const;
 442 
 443   // Return TRUE or FALSE if the loop should be cache-line aligned.
 444   // Gather the expression that does the alignment.  Note that only
 445   // one array base can be aligned in a loop (unless the VM guarantees
 446   // mutual alignment).  Note that if we vectorize short memory ops
 447   // into longer memory ops, we may want to increase alignment.
 448   bool policy_align( PhaseIdealLoop *phase ) const;
 449 
 450   // Return TRUE if "iff" is a range check.
 451   bool is_range_check_if(IfNode *iff, PhaseIdealLoop *phase, Invariance& invar) const;
 452 
 453   // Compute loop exact trip count if possible
 454   void compute_exact_trip_count( PhaseIdealLoop *phase );
 455 
 456   // Compute loop trip count from profile data
 457   void compute_profile_trip_cnt( PhaseIdealLoop *phase );
 458 
 459   // Reassociate invariant expressions.
 460   void reassociate_invariants(PhaseIdealLoop *phase);
 461   // Reassociate invariant add and subtract expressions.
 462   Node* reassociate_add_sub(Node* n1, PhaseIdealLoop *phase);
 463   // Return nonzero index of invariant operand if invariant and variant
 464   // are combined with an Add or Sub. Helper for reassociate_invariants.
 465   int is_invariant_addition(Node* n, PhaseIdealLoop *phase);
 466 
 467   // Return true if n is invariant
 468   bool is_invariant(Node* n) const;
 469 
 470   // Put loop body on igvn work list
 471   void record_for_igvn();
 472 
 473   bool is_loop()    { return !_irreducible && _tail && !_tail->is_top(); }
 474   bool is_inner()   { return is_loop() && _child == NULL; }
 475   bool is_counted() { return is_loop() && _head != NULL && _head->is_CountedLoop(); }
 476 
 477 #ifndef PRODUCT
 478   void dump_head( ) const;      // Dump loop head only
 479   void dump() const;            // Dump this loop recursively
 480   void verify_tree(IdealLoopTree *loop, const IdealLoopTree *parent) const;
 481 #endif
 482 
 483 };
 484 
 485 // -----------------------------PhaseIdealLoop---------------------------------
 486 // Computes the mapping from Nodes to IdealLoopTrees.  Organizes IdealLoopTrees into a
 487 // loop tree.  Drives the loop-based transformations on the ideal graph.
 488 class PhaseIdealLoop : public PhaseTransform {
 489   friend class IdealLoopTree;
 490   friend class SuperWord;
 491   // Pre-computed def-use info
 492   PhaseIterGVN &_igvn;
 493 
 494   // Head of loop tree
 495   IdealLoopTree *_ltree_root;
 496 
 497   // Array of pre-order numbers, plus post-visited bit.
 498   // ZERO for not pre-visited.  EVEN for pre-visited but not post-visited.
 499   // ODD for post-visited.  Other bits are the pre-order number.
 500   uint *_preorders;
 501   uint _max_preorder;
 502 
 503   const PhaseIdealLoop* _verify_me;
 504   bool _verify_only;
 505 
 506   // Allocate _preorders[] array
 507   void allocate_preorders() {
 508     _max_preorder = C->unique()+8;
 509     _preorders = NEW_RESOURCE_ARRAY(uint, _max_preorder);
 510     memset(_preorders, 0, sizeof(uint) * _max_preorder);
 511   }
 512 
 513   // Allocate _preorders[] array
 514   void reallocate_preorders() {
 515     if ( _max_preorder < C->unique() ) {
 516       _preorders = REALLOC_RESOURCE_ARRAY(uint, _preorders, _max_preorder, C->unique());
 517       _max_preorder = C->unique();
 518     }
 519     memset(_preorders, 0, sizeof(uint) * _max_preorder);
 520   }
 521 
 522   // Check to grow _preorders[] array for the case when build_loop_tree_impl()
 523   // adds new nodes.
 524   void check_grow_preorders( ) {
 525     if ( _max_preorder < C->unique() ) {
 526       uint newsize = _max_preorder<<1;  // double size of array
 527       _preorders = REALLOC_RESOURCE_ARRAY(uint, _preorders, _max_preorder, newsize);
 528       memset(&_preorders[_max_preorder],0,sizeof(uint)*(newsize-_max_preorder));
 529       _max_preorder = newsize;
 530     }
 531   }
 532   // Check for pre-visited.  Zero for NOT visited; non-zero for visited.
 533   int is_visited( Node *n ) const { return _preorders[n->_idx]; }
 534   // Pre-order numbers are written to the Nodes array as low-bit-set values.
 535   void set_preorder_visited( Node *n, int pre_order ) {
 536     assert( !is_visited( n ), "already set" );
 537     _preorders[n->_idx] = (pre_order<<1);
 538   };
 539   // Return pre-order number.
 540   int get_preorder( Node *n ) const { assert( is_visited(n), "" ); return _preorders[n->_idx]>>1; }
 541 
 542   // Check for being post-visited.
 543   // Should be previsited already (checked with assert(is_visited(n))).
 544   int is_postvisited( Node *n ) const { assert( is_visited(n), "" ); return _preorders[n->_idx]&1; }
 545 
 546   // Mark as post visited
 547   void set_postvisited( Node *n ) { assert( !is_postvisited( n ), "" ); _preorders[n->_idx] |= 1; }
 548 
 549   // Set/get control node out.  Set lower bit to distinguish from IdealLoopTree
 550   // Returns true if "n" is a data node, false if it's a control node.
 551   bool has_ctrl( Node *n ) const { return ((intptr_t)_nodes[n->_idx]) & 1; }
 552 
 553   // clear out dead code after build_loop_late
 554   Node_List _deadlist;
 555 
 556   // Support for faster execution of get_late_ctrl()/dom_lca()
 557   // when a node has many uses and dominator depth is deep.
 558   Node_Array _dom_lca_tags;
 559   void   init_dom_lca_tags();
 560   void   clear_dom_lca_tags();
 561 
 562   // Helper for debugging bad dominance relationships
 563   bool verify_dominance(Node* n, Node* use, Node* LCA, Node* early);
 564 
 565   Node* compute_lca_of_uses(Node* n, Node* early, bool verify = false);
 566 
 567   // Inline wrapper for frequent cases:
 568   // 1) only one use
 569   // 2) a use is the same as the current LCA passed as 'n1'
 570   Node *dom_lca_for_get_late_ctrl( Node *lca, Node *n, Node *tag ) {
 571     assert( n->is_CFG(), "" );
 572     // Fast-path NULL lca
 573     if( lca != NULL && lca != n ) {
 574       assert( lca->is_CFG(), "" );
 575       // find LCA of all uses
 576       n = dom_lca_for_get_late_ctrl_internal( lca, n, tag );
 577     }
 578     return find_non_split_ctrl(n);
 579   }
 580   Node *dom_lca_for_get_late_ctrl_internal( Node *lca, Node *n, Node *tag );
 581 
 582   // Helper function for directing control inputs away from CFG split
 583   // points.
 584   Node *find_non_split_ctrl( Node *ctrl ) const {
 585     if (ctrl != NULL) {
 586       if (ctrl->is_MultiBranch()) {
 587         ctrl = ctrl->in(0);
 588       }
 589       assert(ctrl->is_CFG(), "CFG");
 590     }
 591     return ctrl;
 592   }
 593 
 594 public:
 595   bool has_node( Node* n ) const { return _nodes[n->_idx] != NULL; }
 596   // check if transform created new nodes that need _ctrl recorded
 597   Node *get_late_ctrl( Node *n, Node *early );
 598   Node *get_early_ctrl( Node *n );
 599   void set_early_ctrl( Node *n );
 600   void set_subtree_ctrl( Node *root );
 601   void set_ctrl( Node *n, Node *ctrl ) {
 602     assert( !has_node(n) || has_ctrl(n), "" );
 603     assert( ctrl->in(0), "cannot set dead control node" );
 604     assert( ctrl == find_non_split_ctrl(ctrl), "must set legal crtl" );
 605     _nodes.map( n->_idx, (Node*)((intptr_t)ctrl + 1) );
 606   }
 607   // Set control and update loop membership
 608   void set_ctrl_and_loop(Node* n, Node* ctrl) {
 609     IdealLoopTree* old_loop = get_loop(get_ctrl(n));
 610     IdealLoopTree* new_loop = get_loop(ctrl);
 611     if (old_loop != new_loop) {
 612       if (old_loop->_child == NULL) old_loop->_body.yank(n);
 613       if (new_loop->_child == NULL) new_loop->_body.push(n);
 614     }
 615     set_ctrl(n, ctrl);
 616   }
 617   // Control nodes can be replaced or subsumed.  During this pass they
 618   // get their replacement Node in slot 1.  Instead of updating the block
 619   // location of all Nodes in the subsumed block, we lazily do it.  As we
 620   // pull such a subsumed block out of the array, we write back the final
 621   // correct block.
 622   Node *get_ctrl( Node *i ) {
 623     assert(has_node(i), "");
 624     Node *n = get_ctrl_no_update(i);
 625     _nodes.map( i->_idx, (Node*)((intptr_t)n + 1) );
 626     assert(has_node(i) && has_ctrl(i), "");
 627     assert(n == find_non_split_ctrl(n), "must return legal ctrl" );
 628     return n;
 629   }
 630   // true if CFG node d dominates CFG node n
 631   bool is_dominator(Node *d, Node *n);
 632   // return get_ctrl for a data node and self(n) for a CFG node
 633   Node* ctrl_or_self(Node* n) {
 634     if (has_ctrl(n))
 635       return get_ctrl(n);
 636     else {
 637       assert (n->is_CFG(), "must be a CFG node");
 638       return n;
 639     }
 640   }
 641 
 642 private:
 643   Node *get_ctrl_no_update( Node *i ) const {
 644     assert( has_ctrl(i), "" );
 645     Node *n = (Node*)(((intptr_t)_nodes[i->_idx]) & ~1);
 646     if (!n->in(0)) {
 647       // Skip dead CFG nodes
 648       do {
 649         n = (Node*)(((intptr_t)_nodes[n->_idx]) & ~1);
 650       } while (!n->in(0));
 651       n = find_non_split_ctrl(n);
 652     }
 653     return n;
 654   }
 655 
 656   // Check for loop being set
 657   // "n" must be a control node. Returns true if "n" is known to be in a loop.
 658   bool has_loop( Node *n ) const {
 659     assert(!has_node(n) || !has_ctrl(n), "");
 660     return has_node(n);
 661   }
 662   // Set loop
 663   void set_loop( Node *n, IdealLoopTree *loop ) {
 664     _nodes.map(n->_idx, (Node*)loop);
 665   }
 666   // Lazy-dazy update of 'get_ctrl' and 'idom_at' mechanisms.  Replace
 667   // the 'old_node' with 'new_node'.  Kill old-node.  Add a reference
 668   // from old_node to new_node to support the lazy update.  Reference
 669   // replaces loop reference, since that is not needed for dead node.
 670 public:
 671   void lazy_update( Node *old_node, Node *new_node ) {
 672     assert( old_node != new_node, "no cycles please" );
 673     //old_node->set_req( 1, new_node /*NO DU INFO*/ );
 674     // Nodes always have DU info now, so re-use the side array slot
 675     // for this node to provide the forwarding pointer.
 676     _nodes.map( old_node->_idx, (Node*)((intptr_t)new_node + 1) );
 677   }
 678   void lazy_replace( Node *old_node, Node *new_node ) {
 679     _igvn.replace_node( old_node, new_node );
 680     lazy_update( old_node, new_node );
 681   }
 682   void lazy_replace_proj( Node *old_node, Node *new_node ) {
 683     assert( old_node->req() == 1, "use this for Projs" );
 684     _igvn.hash_delete(old_node); // Must hash-delete before hacking edges
 685     old_node->add_req( NULL );
 686     lazy_replace( old_node, new_node );
 687   }
 688 
 689 private:
 690 
 691   // Place 'n' in some loop nest, where 'n' is a CFG node
 692   void build_loop_tree();
 693   int build_loop_tree_impl( Node *n, int pre_order );
 694   // Insert loop into the existing loop tree.  'innermost' is a leaf of the
 695   // loop tree, not the root.
 696   IdealLoopTree *sort( IdealLoopTree *loop, IdealLoopTree *innermost );
 697 
 698   // Place Data nodes in some loop nest
 699   void build_loop_early( VectorSet &visited, Node_List &worklist, Node_Stack &nstack );
 700   void build_loop_late ( VectorSet &visited, Node_List &worklist, Node_Stack &nstack );
 701   void build_loop_late_post ( Node* n );
 702 
 703   // Array of immediate dominance info for each CFG node indexed by node idx
 704 private:
 705   uint _idom_size;
 706   Node **_idom;                 // Array of immediate dominators
 707   uint *_dom_depth;           // Used for fast LCA test
 708   GrowableArray<uint>* _dom_stk; // For recomputation of dom depth
 709 
 710   Node* idom_no_update(Node* d) const {
 711     assert(d->_idx < _idom_size, "oob");
 712     Node* n = _idom[d->_idx];
 713     assert(n != NULL,"Bad immediate dominator info.");
 714     while (n->in(0) == NULL) {  // Skip dead CFG nodes
 715       //n = n->in(1);
 716       n = (Node*)(((intptr_t)_nodes[n->_idx]) & ~1);
 717       assert(n != NULL,"Bad immediate dominator info.");
 718     }
 719     return n;
 720   }
 721   Node *idom(Node* d) const {
 722     uint didx = d->_idx;
 723     Node *n = idom_no_update(d);
 724     _idom[didx] = n;            // Lazily remove dead CFG nodes from table.
 725     return n;
 726   }
 727   uint dom_depth(Node* d) const {
 728     assert(d->_idx < _idom_size, "");
 729     return _dom_depth[d->_idx];
 730   }
 731   void set_idom(Node* d, Node* n, uint dom_depth);
 732   // Locally compute IDOM using dom_lca call
 733   Node *compute_idom( Node *region ) const;
 734   // Recompute dom_depth
 735   void recompute_dom_depth();
 736 
 737   // Is safept not required by an outer loop?
 738   bool is_deleteable_safept(Node* sfpt);
 739 
 740   // Replace parallel induction variable (parallel to trip counter)
 741   void replace_parallel_iv(IdealLoopTree *loop);
 742 
 743   // Perform verification that the graph is valid.
 744   PhaseIdealLoop( PhaseIterGVN &igvn) :
 745     PhaseTransform(Ideal_Loop),
 746     _igvn(igvn),
 747     _dom_lca_tags(arena()), // Thread::resource_area
 748     _verify_me(NULL),
 749     _verify_only(true) {
 750     build_and_optimize(false);
 751   }
 752 
 753   // build the loop tree and perform any requested optimizations
 754   void build_and_optimize(bool do_split_if);
 755 
 756 public:
 757   // Dominators for the sea of nodes
 758   void Dominators();
 759   Node *dom_lca( Node *n1, Node *n2 ) const {
 760     return find_non_split_ctrl(dom_lca_internal(n1, n2));
 761   }
 762   Node *dom_lca_internal( Node *n1, Node *n2 ) const;
 763 
 764   // Compute the Ideal Node to Loop mapping
 765   PhaseIdealLoop( PhaseIterGVN &igvn, bool do_split_ifs) :
 766     PhaseTransform(Ideal_Loop),
 767     _igvn(igvn),
 768     _dom_lca_tags(arena()), // Thread::resource_area
 769     _verify_me(NULL),
 770     _verify_only(false) {
 771     build_and_optimize(do_split_ifs);
 772   }
 773 
 774   // Verify that verify_me made the same decisions as a fresh run.
 775   PhaseIdealLoop( PhaseIterGVN &igvn, const PhaseIdealLoop *verify_me) :
 776     PhaseTransform(Ideal_Loop),
 777     _igvn(igvn),
 778     _dom_lca_tags(arena()), // Thread::resource_area
 779     _verify_me(verify_me),
 780     _verify_only(false) {
 781     build_and_optimize(false);
 782   }
 783 
 784   // Build and verify the loop tree without modifying the graph.  This
 785   // is useful to verify that all inputs properly dominate their uses.
 786   static void verify(PhaseIterGVN& igvn) {
 787 #ifdef ASSERT
 788     PhaseIdealLoop v(igvn);
 789 #endif
 790   }
 791 
 792   // True if the method has at least 1 irreducible loop
 793   bool _has_irreducible_loops;
 794 
 795   // Per-Node transform
 796   virtual Node *transform( Node *a_node ) { return 0; }
 797 
 798   bool is_counted_loop( Node *x, IdealLoopTree *loop );
 799 
 800   Node* exact_limit( IdealLoopTree *loop );
 801 
 802   // Return a post-walked LoopNode
 803   IdealLoopTree *get_loop( Node *n ) const {
 804     // Dead nodes have no loop, so return the top level loop instead
 805     if (!has_node(n))  return _ltree_root;
 806     assert(!has_ctrl(n), "");
 807     return (IdealLoopTree*)_nodes[n->_idx];
 808   }
 809 
 810   // Is 'n' a (nested) member of 'loop'?
 811   int is_member( const IdealLoopTree *loop, Node *n ) const {
 812     return loop->is_member(get_loop(n)); }
 813 
 814   // This is the basic building block of the loop optimizations.  It clones an
 815   // entire loop body.  It makes an old_new loop body mapping; with this
 816   // mapping you can find the new-loop equivalent to an old-loop node.  All
 817   // new-loop nodes are exactly equal to their old-loop counterparts, all
 818   // edges are the same.  All exits from the old-loop now have a RegionNode
 819   // that merges the equivalent new-loop path.  This is true even for the
 820   // normal "loop-exit" condition.  All uses of loop-invariant old-loop values
 821   // now come from (one or more) Phis that merge their new-loop equivalents.
 822   // Parameter side_by_side_idom:
 823   //   When side_by_size_idom is NULL, the dominator tree is constructed for
 824   //      the clone loop to dominate the original.  Used in construction of
 825   //      pre-main-post loop sequence.
 826   //   When nonnull, the clone and original are side-by-side, both are
 827   //      dominated by the passed in side_by_side_idom node.  Used in
 828   //      construction of unswitched loops.
 829   void clone_loop( IdealLoopTree *loop, Node_List &old_new, int dom_depth,
 830                    Node* side_by_side_idom = NULL);
 831 
 832   // If we got the effect of peeling, either by actually peeling or by
 833   // making a pre-loop which must execute at least once, we can remove
 834   // all loop-invariant dominated tests in the main body.
 835   void peeled_dom_test_elim( IdealLoopTree *loop, Node_List &old_new );
 836 
 837   // Generate code to do a loop peel for the given loop (and body).
 838   // old_new is a temp array.
 839   void do_peeling( IdealLoopTree *loop, Node_List &old_new );
 840 
 841   // Add pre and post loops around the given loop.  These loops are used
 842   // during RCE, unrolling and aligning loops.
 843   void insert_pre_post_loops( IdealLoopTree *loop, Node_List &old_new, bool peel_only );
 844   // If Node n lives in the back_ctrl block, we clone a private version of n
 845   // in preheader_ctrl block and return that, otherwise return n.
 846   Node *clone_up_backedge_goo( Node *back_ctrl, Node *preheader_ctrl, Node *n, VectorSet &visited, Node_Stack &clones );
 847 
 848   // Take steps to maximally unroll the loop.  Peel any odd iterations, then
 849   // unroll to do double iterations.  The next round of major loop transforms
 850   // will repeat till the doubled loop body does all remaining iterations in 1
 851   // pass.
 852   void do_maximally_unroll( IdealLoopTree *loop, Node_List &old_new );
 853 
 854   // Unroll the loop body one step - make each trip do 2 iterations.
 855   void do_unroll( IdealLoopTree *loop, Node_List &old_new, bool adjust_min_trip );
 856 
 857   // Return true if exp is a constant times an induction var
 858   bool is_scaled_iv(Node* exp, Node* iv, int* p_scale);
 859 
 860   // Return true if exp is a scaled induction var plus (or minus) constant
 861   bool is_scaled_iv_plus_offset(Node* exp, Node* iv, int* p_scale, Node** p_offset, int depth = 0);
 862 
 863   // Return true if proj is for "proj->[region->..]call_uct"
 864   static bool is_uncommon_trap_proj(ProjNode* proj, Deoptimization::DeoptReason reason);
 865   // Return true for    "if(test)-> proj -> ...
 866   //                          |
 867   //                          V
 868   //                      other_proj->[region->..]call_uct"
 869   static bool is_uncommon_trap_if_pattern(ProjNode* proj, Deoptimization::DeoptReason reason);
 870   // Create a new if above the uncommon_trap_if_pattern for the predicate to be promoted
 871   ProjNode* create_new_if_for_predicate(ProjNode* cont_proj, Node* new_entry,
 872                                         Deoptimization::DeoptReason reason);
 873   void register_control(Node* n, IdealLoopTree *loop, Node* pred);
 874 
 875   // Clone loop predicates to cloned loops (peeled, unswitched)
 876   static ProjNode* clone_predicate(ProjNode* predicate_proj, Node* new_entry,
 877                                    Deoptimization::DeoptReason reason,
 878                                    PhaseIdealLoop* loop_phase,
 879                                    PhaseIterGVN* igvn);
 880 
 881   static Node* clone_loop_predicates(Node* old_entry, Node* new_entry,
 882                                          bool clone_limit_check,
 883                                          PhaseIdealLoop* loop_phase,
 884                                          PhaseIterGVN* igvn);
 885   Node* clone_loop_predicates(Node* old_entry, Node* new_entry, bool clone_limit_check);
 886 
 887   static Node* skip_loop_predicates(Node* entry);
 888 
 889   // Find a good location to insert a predicate
 890   static ProjNode* find_predicate_insertion_point(Node* start_c, Deoptimization::DeoptReason reason);
 891   // Find a predicate
 892   static Node* find_predicate(Node* entry);
 893   // Construct a range check for a predicate if
 894   BoolNode* rc_predicate(IdealLoopTree *loop, Node* ctrl,
 895                          int scale, Node* offset,
 896                          Node* init, Node* limit, Node* stride,
 897                          Node* range, bool upper);
 898 
 899   // Implementation of the loop predication to promote checks outside the loop
 900   bool loop_predication_impl(IdealLoopTree *loop);
 901 
 902   // Helper function to collect predicate for eliminating the useless ones
 903   void collect_potentially_useful_predicates(IdealLoopTree *loop, Unique_Node_List &predicate_opaque1);
 904   void eliminate_useless_predicates();
 905 
 906   // Eliminate range-checks and other trip-counter vs loop-invariant tests.
 907   void do_range_check( IdealLoopTree *loop, Node_List &old_new );
 908 
 909   // Create a slow version of the loop by cloning the loop
 910   // and inserting an if to select fast-slow versions.
 911   ProjNode* create_slow_version_of_loop(IdealLoopTree *loop,
 912                                         Node_List &old_new);
 913 
 914   // Clone loop with an invariant test (that does not exit) and
 915   // insert a clone of the test that selects which version to
 916   // execute.
 917   void do_unswitching (IdealLoopTree *loop, Node_List &old_new);
 918 
 919   // Find candidate "if" for unswitching
 920   IfNode* find_unswitching_candidate(const IdealLoopTree *loop) const;
 921 
 922   // Range Check Elimination uses this function!
 923   // Constrain the main loop iterations so the affine function:
 924   //    low_limit <= scale_con * I + offset  <  upper_limit
 925   // always holds true.  That is, either increase the number of iterations in
 926   // the pre-loop or the post-loop until the condition holds true in the main
 927   // loop.  Scale_con, offset and limit are all loop invariant.
 928   void add_constraint( int stride_con, int scale_con, Node *offset, Node *low_limit, Node *upper_limit, Node *pre_ctrl, Node **pre_limit, Node **main_limit );
 929   // Helper function for add_constraint().
 930   Node* adjust_limit( int stride_con, Node * scale, Node *offset, Node *rc_limit, Node *loop_limit, Node *pre_ctrl );
 931 
 932   // Partially peel loop up through last_peel node.
 933   bool partial_peel( IdealLoopTree *loop, Node_List &old_new );
 934 
 935   // Create a scheduled list of nodes control dependent on ctrl set.
 936   void scheduled_nodelist( IdealLoopTree *loop, VectorSet& ctrl, Node_List &sched );
 937   // Has a use in the vector set
 938   bool has_use_in_set( Node* n, VectorSet& vset );
 939   // Has use internal to the vector set (ie. not in a phi at the loop head)
 940   bool has_use_internal_to_set( Node* n, VectorSet& vset, IdealLoopTree *loop );
 941   // clone "n" for uses that are outside of loop
 942   void clone_for_use_outside_loop( IdealLoopTree *loop, Node* n, Node_List& worklist );
 943   // clone "n" for special uses that are in the not_peeled region
 944   void clone_for_special_use_inside_loop( IdealLoopTree *loop, Node* n,
 945                                           VectorSet& not_peel, Node_List& sink_list, Node_List& worklist );
 946   // Insert phi(lp_entry_val, back_edge_val) at use->in(idx) for loop lp if phi does not already exist
 947   void insert_phi_for_loop( Node* use, uint idx, Node* lp_entry_val, Node* back_edge_val, LoopNode* lp );
 948 #ifdef ASSERT
 949   // Validate the loop partition sets: peel and not_peel
 950   bool is_valid_loop_partition( IdealLoopTree *loop, VectorSet& peel, Node_List& peel_list, VectorSet& not_peel );
 951   // Ensure that uses outside of loop are of the right form
 952   bool is_valid_clone_loop_form( IdealLoopTree *loop, Node_List& peel_list,
 953                                  uint orig_exit_idx, uint clone_exit_idx);
 954   bool is_valid_clone_loop_exit_use( IdealLoopTree *loop, Node* use, uint exit_idx);
 955 #endif
 956 
 957   // Returns nonzero constant stride if-node is a possible iv test (otherwise returns zero.)
 958   int stride_of_possible_iv( Node* iff );
 959   bool is_possible_iv_test( Node* iff ) { return stride_of_possible_iv(iff) != 0; }
 960   // Return the (unique) control output node that's in the loop (if it exists.)
 961   Node* stay_in_loop( Node* n, IdealLoopTree *loop);
 962   // Insert a signed compare loop exit cloned from an unsigned compare.
 963   IfNode* insert_cmpi_loop_exit(IfNode* if_cmpu, IdealLoopTree *loop);
 964   void remove_cmpi_loop_exit(IfNode* if_cmp, IdealLoopTree *loop);
 965   // Utility to register node "n" with PhaseIdealLoop
 966   void register_node(Node* n, IdealLoopTree *loop, Node* pred, int ddepth);
 967   // Utility to create an if-projection
 968   ProjNode* proj_clone(ProjNode* p, IfNode* iff);
 969   // Force the iff control output to be the live_proj
 970   Node* short_circuit_if(IfNode* iff, ProjNode* live_proj);
 971   // Insert a region before an if projection
 972   RegionNode* insert_region_before_proj(ProjNode* proj);
 973   // Insert a new if before an if projection
 974   ProjNode* insert_if_before_proj(Node* left, bool Signed, BoolTest::mask relop, Node* right, ProjNode* proj);
 975 
 976   // Passed in a Phi merging (recursively) some nearly equivalent Bool/Cmps.
 977   // "Nearly" because all Nodes have been cloned from the original in the loop,
 978   // but the fall-in edges to the Cmp are different.  Clone bool/Cmp pairs
 979   // through the Phi recursively, and return a Bool.
 980   BoolNode *clone_iff( PhiNode *phi, IdealLoopTree *loop );
 981   CmpNode *clone_bool( PhiNode *phi, IdealLoopTree *loop );
 982 
 983 
 984   // Rework addressing expressions to get the most loop-invariant stuff
 985   // moved out.  We'd like to do all associative operators, but it's especially
 986   // important (common) to do address expressions.
 987   Node *remix_address_expressions( Node *n );
 988 
 989   // Attempt to use a conditional move instead of a phi/branch
 990   Node *conditional_move( Node *n );
 991 
 992   // Reorganize offset computations to lower register pressure.
 993   // Mostly prevent loop-fallout uses of the pre-incremented trip counter
 994   // (which are then alive with the post-incremented trip counter
 995   // forcing an extra register move)
 996   void reorg_offsets( IdealLoopTree *loop );
 997 
 998   // Check for aggressive application of 'split-if' optimization,
 999   // using basic block level info.
1000   void  split_if_with_blocks     ( VectorSet &visited, Node_Stack &nstack );
1001   Node *split_if_with_blocks_pre ( Node *n );
1002   void  split_if_with_blocks_post( Node *n );
1003   Node *has_local_phi_input( Node *n );
1004   // Mark an IfNode as being dominated by a prior test,
1005   // without actually altering the CFG (and hence IDOM info).
1006   void dominated_by( Node *prevdom, Node *iff, bool flip = false );
1007 
1008   // Split Node 'n' through merge point
1009   Node *split_thru_region( Node *n, Node *region );
1010   // Split Node 'n' through merge point if there is enough win.
1011   Node *split_thru_phi( Node *n, Node *region, int policy );
1012   // Found an If getting its condition-code input from a Phi in the
1013   // same block.  Split thru the Region.
1014   void do_split_if( Node *iff );
1015 
1016   // Conversion of fill/copy patterns into intrisic versions
1017   bool do_intrinsify_fill();
1018   bool intrinsify_fill(IdealLoopTree* lpt);
1019   bool match_fill_loop(IdealLoopTree* lpt, Node*& store, Node*& store_value,
1020                        Node*& shift, Node*& offset);
1021 
1022 private:
1023   // Return a type based on condition control flow
1024   const TypeInt* filtered_type( Node *n, Node* n_ctrl);
1025   const TypeInt* filtered_type( Node *n ) { return filtered_type(n, NULL); }
1026  // Helpers for filtered type
1027   const TypeInt* filtered_type_from_dominators( Node* val, Node *val_ctrl);
1028 
1029   // Helper functions
1030   Node *spinup( Node *iff, Node *new_false, Node *new_true, Node *region, Node *phi, small_cache *cache );
1031   Node *find_use_block( Node *use, Node *def, Node *old_false, Node *new_false, Node *old_true, Node *new_true );
1032   void handle_use( Node *use, Node *def, small_cache *cache, Node *region_dom, Node *new_false, Node *new_true, Node *old_false, Node *old_true );
1033   bool split_up( Node *n, Node *blk1, Node *blk2 );
1034   void sink_use( Node *use, Node *post_loop );
1035   Node *place_near_use( Node *useblock ) const;
1036 
1037   bool _created_loop_node;
1038 public:
1039   void set_created_loop_node() { _created_loop_node = true; }
1040   bool created_loop_node()     { return _created_loop_node; }
1041   void register_new_node( Node *n, Node *blk );
1042 
1043 #ifndef PRODUCT
1044   void dump( ) const;
1045   void dump( IdealLoopTree *loop, uint rpo_idx, Node_List &rpo_list ) const;
1046   void rpo( Node *start, Node_Stack &stk, VectorSet &visited, Node_List &rpo_list ) const;
1047   void verify() const;          // Major slow  :-)
1048   void verify_compare( Node *n, const PhaseIdealLoop *loop_verify, VectorSet &visited ) const;
1049   IdealLoopTree *get_loop_idx(Node* n) const {
1050     // Dead nodes have no loop, so return the top level loop instead
1051     return _nodes[n->_idx] ? (IdealLoopTree*)_nodes[n->_idx] : _ltree_root;
1052   }
1053   // Print some stats
1054   static void print_statistics();
1055   static int _loop_invokes;     // Count of PhaseIdealLoop invokes
1056   static int _loop_work;        // Sum of PhaseIdealLoop x _unique
1057 #endif
1058 };
1059 
1060 inline Node* IdealLoopTree::tail() {
1061 // Handle lazy update of _tail field
1062   Node *n = _tail;
1063   //while( !n->in(0) )  // Skip dead CFG nodes
1064     //n = n->in(1);
1065   if (n->in(0) == NULL)
1066     n = _phase->get_ctrl(n);
1067   _tail = n;
1068   return n;
1069 }
1070 
1071 
1072 // Iterate over the loop tree using a preorder, left-to-right traversal.
1073 //
1074 // Example that visits all counted loops from within PhaseIdealLoop
1075 //
1076 //  for (LoopTreeIterator iter(_ltree_root); !iter.done(); iter.next()) {
1077 //   IdealLoopTree* lpt = iter.current();
1078 //   if (!lpt->is_counted()) continue;
1079 //   ...
1080 class LoopTreeIterator : public StackObj {
1081 private:
1082   IdealLoopTree* _root;
1083   IdealLoopTree* _curnt;
1084 
1085 public:
1086   LoopTreeIterator(IdealLoopTree* root) : _root(root), _curnt(root) {}
1087 
1088   bool done() { return _curnt == NULL; }       // Finished iterating?
1089 
1090   void next();                                 // Advance to next loop tree
1091 
1092   IdealLoopTree* current() { return _curnt; }  // Return current value of iterator.
1093 };
1094 
1095 #endif // SHARE_VM_OPTO_LOOPNODE_HPP