duke@435: /*
mikael@6198:  * Copyright (c) 1999, 2013, Oracle and/or its affiliates. All rights reserved.
duke@435:  * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
duke@435:  *
duke@435:  * This code is free software; you can redistribute it and/or modify it
duke@435:  * under the terms of the GNU General Public License version 2 only, as
duke@435:  * published by the Free Software Foundation.
duke@435:  *
duke@435:  * This code is distributed in the hope that it will be useful, but WITHOUT
duke@435:  * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
duke@435:  * FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License
duke@435:  * version 2 for more details (a copy is included in the LICENSE file that
duke@435:  * accompanied this code).
duke@435:  *
duke@435:  * You should have received a copy of the GNU General Public License version
duke@435:  * 2 along with this work; if not, write to the Free Software Foundation,
duke@435:  * Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA.
duke@435:  *
trims@1907:  * Please contact Oracle, 500 Oracle Parkway, Redwood Shores, CA 94065 USA
trims@1907:  * or visit www.oracle.com if you need additional information or have any
trims@1907:  * questions.
duke@435:  *
duke@435:  */
duke@435: 
stefank@2314: #include "precompiled.hpp"
stefank@2314: #include "c1/c1_Compilation.hpp"
stefank@2314: #include "c1/c1_FrameMap.hpp"
stefank@2314: #include "c1/c1_GraphBuilder.hpp"
stefank@2314: #include "c1/c1_IR.hpp"
stefank@2314: #include "c1/c1_InstructionPrinter.hpp"
stefank@2314: #include "c1/c1_Optimizer.hpp"
stefank@2314: #include "utilities/bitMap.inline.hpp"
duke@435: 
duke@435: 
duke@435: // Implementation of XHandlers
duke@435: //
duke@435: // Note: This code could eventually go away if we are
duke@435: //       just using the ciExceptionHandlerStream.
duke@435: 
duke@435: XHandlers::XHandlers(ciMethod* method) : _list(method->exception_table_length()) {
duke@435:   ciExceptionHandlerStream s(method);
duke@435:   while (!s.is_done()) {
duke@435:     _list.append(new XHandler(s.handler()));
duke@435:     s.next();
duke@435:   }
duke@435:   assert(s.count() == method->exception_table_length(), "exception table lengths inconsistent");
duke@435: }
duke@435: 
duke@435: // deep copy of all XHandler contained in list
duke@435: XHandlers::XHandlers(XHandlers* other) :
duke@435:   _list(other->length())
duke@435: {
duke@435:   for (int i = 0; i < other->length(); i++) {
duke@435:     _list.append(new XHandler(other->handler_at(i)));
duke@435:   }
duke@435: }
duke@435: 
duke@435: // Returns whether a particular exception type can be caught.  Also
duke@435: // returns true if klass is unloaded or any exception handler
duke@435: // classes are unloaded.  type_is_exact indicates whether the throw
duke@435: // is known to be exactly that class or it might throw a subtype.
duke@435: bool XHandlers::could_catch(ciInstanceKlass* klass, bool type_is_exact) const {
duke@435:   // the type is unknown so be conservative
duke@435:   if (!klass->is_loaded()) {
duke@435:     return true;
duke@435:   }
duke@435: 
duke@435:   for (int i = 0; i < length(); i++) {
duke@435:     XHandler* handler = handler_at(i);
duke@435:     if (handler->is_catch_all()) {
duke@435:       // catch of ANY
duke@435:       return true;
duke@435:     }
duke@435:     ciInstanceKlass* handler_klass = handler->catch_klass();
duke@435:     // if it's unknown it might be catchable
duke@435:     if (!handler_klass->is_loaded()) {
duke@435:       return true;
duke@435:     }
duke@435:     // if the throw type is definitely a subtype of the catch type
duke@435:     // then it can be caught.
duke@435:     if (klass->is_subtype_of(handler_klass)) {
duke@435:       return true;
duke@435:     }
duke@435:     if (!type_is_exact) {
duke@435:       // If the type isn't exactly known then it can also be caught by
duke@435:       // catch statements where the inexact type is a subtype of the
duke@435:       // catch type.
duke@435:       // given: foo extends bar extends Exception
duke@435:       // throw bar can be caught by catch foo, catch bar, and catch
duke@435:       // Exception, however it can't be caught by any handlers without
duke@435:       // bar in its type hierarchy.
duke@435:       if (handler_klass->is_subtype_of(klass)) {
duke@435:         return true;
duke@435:       }
duke@435:     }
duke@435:   }
duke@435: 
duke@435:   return false;
duke@435: }
duke@435: 
duke@435: 
duke@435: bool XHandlers::equals(XHandlers* others) const {
duke@435:   if (others == NULL) return false;
duke@435:   if (length() != others->length()) return false;
duke@435: 
duke@435:   for (int i = 0; i < length(); i++) {
duke@435:     if (!handler_at(i)->equals(others->handler_at(i))) return false;
duke@435:   }
duke@435:   return true;
duke@435: }
duke@435: 
duke@435: bool XHandler::equals(XHandler* other) const {
duke@435:   assert(entry_pco() != -1 && other->entry_pco() != -1, "must have entry_pco");
duke@435: 
duke@435:   if (entry_pco() != other->entry_pco()) return false;
duke@435:   if (scope_count() != other->scope_count()) return false;
duke@435:   if (_desc != other->_desc) return false;
duke@435: 
duke@435:   assert(entry_block() == other->entry_block(), "entry_block must be equal when entry_pco is equal");
duke@435:   return true;
duke@435: }
duke@435: 
duke@435: 
duke@435: // Implementation of IRScope
duke@435: BlockBegin* IRScope::build_graph(Compilation* compilation, int osr_bci) {
duke@435:   GraphBuilder gm(compilation, this);
duke@435:   NOT_PRODUCT(if (PrintValueNumbering && Verbose) gm.print_stats());
duke@435:   if (compilation->bailed_out()) return NULL;
duke@435:   return gm.start();
duke@435: }
duke@435: 
duke@435: 
duke@435: IRScope::IRScope(Compilation* compilation, IRScope* caller, int caller_bci, ciMethod* method, int osr_bci, bool create_graph)
duke@435: : _callees(2)
duke@435: , _compilation(compilation)
duke@435: , _requires_phi_function(method->max_locals())
duke@435: {
duke@435:   _caller             = caller;
duke@435:   _level              = caller == NULL ?  0 : caller->level() + 1;
duke@435:   _method             = method;
duke@435:   _xhandlers          = new XHandlers(method);
duke@435:   _number_of_locks    = 0;
duke@435:   _monitor_pairing_ok = method->has_balanced_monitors();
jiangli@3592:   _wrote_final        = false;
duke@435:   _start              = NULL;
duke@435: 
duke@435:   if (osr_bci == -1) {
duke@435:     _requires_phi_function.clear();
duke@435:   } else {
duke@435:         // selective creation of phi functions is not possibel in osr-methods
duke@435:     _requires_phi_function.set_range(0, method->max_locals());
duke@435:   }
duke@435: 
duke@435:   assert(method->holder()->is_loaded() , "method holder must be loaded");
duke@435: 
duke@435:   // build graph if monitor pairing is ok
duke@435:   if (create_graph && monitor_pairing_ok()) _start = build_graph(compilation, osr_bci);
duke@435: }
duke@435: 
duke@435: 
duke@435: int IRScope::max_stack() const {
duke@435:   int my_max = method()->max_stack();
duke@435:   int callee_max = 0;
duke@435:   for (int i = 0; i < number_of_callees(); i++) {
duke@435:     callee_max = MAX2(callee_max, callee_no(i)->max_stack());
duke@435:   }
duke@435:   return my_max + callee_max;
duke@435: }
duke@435: 
duke@435: 
cfang@1335: bool IRScopeDebugInfo::should_reexecute() {
cfang@1335:   ciMethod* cur_method = scope()->method();
cfang@1335:   int       cur_bci    = bci();
cfang@1335:   if (cur_method != NULL && cur_bci != SynchronizationEntryBCI) {
cfang@1335:     Bytecodes::Code code = cur_method->java_code_at_bci(cur_bci);
cfang@1335:     return Interpreter::bytecode_should_reexecute(code);
cfang@1335:   } else
cfang@1335:     return false;
cfang@1335: }
duke@435: 
duke@435: 
duke@435: // Implementation of CodeEmitInfo
duke@435: 
duke@435: // Stack must be NON-null
roland@4860: CodeEmitInfo::CodeEmitInfo(ValueStack* stack, XHandlers* exception_handlers, bool deoptimize_on_exception)
duke@435:   : _scope(stack->scope())
duke@435:   , _scope_debug_info(NULL)
duke@435:   , _oop_map(NULL)
duke@435:   , _stack(stack)
duke@435:   , _exception_handlers(exception_handlers)
roland@4860:   , _is_method_handle_invoke(false)
roland@4860:   , _deoptimize_on_exception(deoptimize_on_exception) {
duke@435:   assert(_stack != NULL, "must be non null");
duke@435: }
duke@435: 
duke@435: 
roland@2174: CodeEmitInfo::CodeEmitInfo(CodeEmitInfo* info, ValueStack* stack)
duke@435:   : _scope(info->_scope)
duke@435:   , _exception_handlers(NULL)
duke@435:   , _scope_debug_info(NULL)
twisti@1919:   , _oop_map(NULL)
roland@2174:   , _stack(stack == NULL ? info->_stack : stack)
roland@4860:   , _is_method_handle_invoke(info->_is_method_handle_invoke)
roland@4860:   , _deoptimize_on_exception(info->_deoptimize_on_exception) {
duke@435: 
duke@435:   // deep copy of exception handlers
duke@435:   if (info->_exception_handlers != NULL) {
duke@435:     _exception_handlers = new XHandlers(info->_exception_handlers);
duke@435:   }
duke@435: }
duke@435: 
duke@435: 
twisti@1919: void CodeEmitInfo::record_debug_info(DebugInformationRecorder* recorder, int pc_offset) {
duke@435:   // record the safepoint before recording the debug info for enclosing scopes
duke@435:   recorder->add_safepoint(pc_offset, _oop_map->deep_copy());
twisti@1919:   _scope_debug_info->record_debug_info(recorder, pc_offset, true/*topmost*/, _is_method_handle_invoke);
duke@435:   recorder->end_safepoint(pc_offset);
duke@435: }
duke@435: 
duke@435: 
duke@435: void CodeEmitInfo::add_register_oop(LIR_Opr opr) {
duke@435:   assert(_oop_map != NULL, "oop map must already exist");
duke@435:   assert(opr->is_single_cpu(), "should not call otherwise");
duke@435: 
duke@435:   VMReg name = frame_map()->regname(opr);
duke@435:   _oop_map->set_oop(name);
duke@435: }
duke@435: 
roland@6723: // Mirror the stack size calculation in the deopt code
roland@6723: // How much stack space would we need at this point in the program in
roland@6723: // case of deoptimization?
roland@6723: int CodeEmitInfo::interpreter_frame_size() const {
roland@6723:   ValueStack* state = _stack;
roland@6723:   int size = 0;
roland@6723:   int callee_parameters = 0;
roland@6723:   int callee_locals = 0;
roland@6723:   int extra_args = state->scope()->method()->max_stack() - state->stack_size();
duke@435: 
roland@6723:   while (state != NULL) {
roland@6723:     int locks = state->locks_size();
roland@6723:     int temps = state->stack_size();
roland@6723:     bool is_top_frame = (state == _stack);
roland@6723:     ciMethod* method = state->scope()->method();
duke@435: 
roland@6723:     int frame_size = BytesPerWord * Interpreter::size_activation(method->max_stack(),
roland@6723:                                                                  temps + callee_parameters,
roland@6723:                                                                  extra_args,
roland@6723:                                                                  locks,
roland@6723:                                                                  callee_parameters,
roland@6723:                                                                  callee_locals,
roland@6723:                                                                  is_top_frame);
roland@6723:     size += frame_size;
roland@6723: 
roland@6723:     callee_parameters = method->size_of_parameters();
roland@6723:     callee_locals = method->max_locals();
roland@6723:     extra_args = 0;
roland@6723:     state = state->caller_state();
roland@6723:   }
roland@6723:   return size + Deoptimization::last_frame_adjust(0, callee_locals) * BytesPerWord;
roland@6723: }
duke@435: 
duke@435: // Implementation of IR
duke@435: 
duke@435: IR::IR(Compilation* compilation, ciMethod* method, int osr_bci) :
duke@435:     _locals_size(in_WordSize(-1))
duke@435:   , _num_loops(0) {
duke@435:   // setup IR fields
duke@435:   _compilation = compilation;
duke@435:   _top_scope   = new IRScope(compilation, NULL, -1, method, osr_bci, true);
duke@435:   _code        = NULL;
duke@435: }
duke@435: 
duke@435: 
roland@4860: void IR::optimize_blocks() {
duke@435:   Optimizer opt(this);
iveresov@2138:   if (!compilation()->profile_branches()) {
iveresov@2138:     if (DoCEE) {
iveresov@2138:       opt.eliminate_conditional_expressions();
duke@435: #ifndef PRODUCT
iveresov@2138:       if (PrintCFG || PrintCFG1) { tty->print_cr("CFG after CEE"); print(true); }
iveresov@2138:       if (PrintIR  || PrintIR1 ) { tty->print_cr("IR after CEE"); print(false); }
duke@435: #endif
iveresov@2138:     }
iveresov@2138:     if (EliminateBlocks) {
iveresov@2138:       opt.eliminate_blocks();
duke@435: #ifndef PRODUCT
iveresov@2138:       if (PrintCFG || PrintCFG1) { tty->print_cr("CFG after block elimination"); print(true); }
iveresov@2138:       if (PrintIR  || PrintIR1 ) { tty->print_cr("IR after block elimination"); print(false); }
duke@435: #endif
iveresov@2138:     }
duke@435:   }
roland@4860: }
roland@4860: 
roland@4860: void IR::eliminate_null_checks() {
roland@4860:   Optimizer opt(this);
duke@435:   if (EliminateNullChecks) {
duke@435:     opt.eliminate_null_checks();
duke@435: #ifndef PRODUCT
duke@435:     if (PrintCFG || PrintCFG1) { tty->print_cr("CFG after null check elimination"); print(true); }
duke@435:     if (PrintIR  || PrintIR1 ) { tty->print_cr("IR after null check elimination"); print(false); }
duke@435: #endif
duke@435:   }
duke@435: }
duke@435: 
duke@435: 
duke@435: static int sort_pairs(BlockPair** a, BlockPair** b) {
duke@435:   if ((*a)->from() == (*b)->from()) {
duke@435:     return (*a)->to()->block_id() - (*b)->to()->block_id();
duke@435:   } else {
duke@435:     return (*a)->from()->block_id() - (*b)->from()->block_id();
duke@435:   }
duke@435: }
duke@435: 
duke@435: 
duke@435: class CriticalEdgeFinder: public BlockClosure {
duke@435:   BlockPairList blocks;
duke@435:   IR*       _ir;
duke@435: 
duke@435:  public:
duke@435:   CriticalEdgeFinder(IR* ir): _ir(ir) {}
duke@435:   void block_do(BlockBegin* bb) {
duke@435:     BlockEnd* be = bb->end();
duke@435:     int nos = be->number_of_sux();
duke@435:     if (nos >= 2) {
duke@435:       for (int i = 0; i < nos; i++) {
duke@435:         BlockBegin* sux = be->sux_at(i);
duke@435:         if (sux->number_of_preds() >= 2) {
duke@435:           blocks.append(new BlockPair(bb, sux));
duke@435:         }
duke@435:       }
duke@435:     }
duke@435:   }
duke@435: 
duke@435:   void split_edges() {
duke@435:     BlockPair* last_pair = NULL;
duke@435:     blocks.sort(sort_pairs);
duke@435:     for (int i = 0; i < blocks.length(); i++) {
duke@435:       BlockPair* pair = blocks.at(i);
duke@435:       if (last_pair != NULL && pair->is_same(last_pair)) continue;
duke@435:       BlockBegin* from = pair->from();
duke@435:       BlockBegin* to = pair->to();
duke@435:       BlockBegin* split = from->insert_block_between(to);
duke@435: #ifndef PRODUCT
duke@435:       if ((PrintIR || PrintIR1) && Verbose) {
duke@435:         tty->print_cr("Split critical edge B%d -> B%d (new block B%d)",
duke@435:                       from->block_id(), to->block_id(), split->block_id());
duke@435:       }
duke@435: #endif
duke@435:       last_pair = pair;
duke@435:     }
duke@435:   }
duke@435: };
duke@435: 
duke@435: void IR::split_critical_edges() {
duke@435:   CriticalEdgeFinder cef(this);
duke@435: 
duke@435:   iterate_preorder(&cef);
duke@435:   cef.split_edges();
duke@435: }
duke@435: 
duke@435: 
iveresov@1939: class UseCountComputer: public ValueVisitor, BlockClosure {
duke@435:  private:
iveresov@1939:   void visit(Value* n) {
duke@435:     // Local instructions and Phis for expression stack values at the
duke@435:     // start of basic blocks are not added to the instruction list
roland@2254:     if (!(*n)->is_linked() && (*n)->can_be_linked()) {
duke@435:       assert(false, "a node was not appended to the graph");
iveresov@1939:       Compilation::current()->bailout("a node was not appended to the graph");
duke@435:     }
duke@435:     // use n's input if not visited before
duke@435:     if (!(*n)->is_pinned() && !(*n)->has_uses()) {
duke@435:       // note: a) if the instruction is pinned, it will be handled by compute_use_count
duke@435:       //       b) if the instruction has uses, it was touched before
duke@435:       //       => in both cases we don't need to update n's values
duke@435:       uses_do(n);
duke@435:     }
duke@435:     // use n
duke@435:     (*n)->_use_count++;
duke@435:   }
duke@435: 
iveresov@1939:   Values* worklist;
iveresov@1939:   int depth;
duke@435:   enum {
duke@435:     max_recurse_depth = 20
duke@435:   };
duke@435: 
iveresov@1939:   void uses_do(Value* n) {
duke@435:     depth++;
duke@435:     if (depth > max_recurse_depth) {
duke@435:       // don't allow the traversal to recurse too deeply
duke@435:       worklist->push(*n);
duke@435:     } else {
iveresov@1939:       (*n)->input_values_do(this);
duke@435:       // special handling for some instructions
duke@435:       if ((*n)->as_BlockEnd() != NULL) {
duke@435:         // note on BlockEnd:
duke@435:         //   must 'use' the stack only if the method doesn't
duke@435:         //   terminate, however, in those cases stack is empty
iveresov@1939:         (*n)->state_values_do(this);
duke@435:       }
duke@435:     }
duke@435:     depth--;
duke@435:   }
duke@435: 
iveresov@1939:   void block_do(BlockBegin* b) {
duke@435:     depth = 0;
duke@435:     // process all pinned nodes as the roots of expression trees
duke@435:     for (Instruction* n = b; n != NULL; n = n->next()) {
duke@435:       if (n->is_pinned()) uses_do(&n);
duke@435:     }
duke@435:     assert(depth == 0, "should have counted back down");
duke@435: 
duke@435:     // now process any unpinned nodes which recursed too deeply
duke@435:     while (worklist->length() > 0) {
duke@435:       Value t = worklist->pop();
duke@435:       if (!t->is_pinned()) {
duke@435:         // compute the use count
duke@435:         uses_do(&t);
duke@435: 
duke@435:         // pin the instruction so that LIRGenerator doesn't recurse
duke@435:         // too deeply during it's evaluation.
duke@435:         t->pin();
duke@435:       }
duke@435:     }
duke@435:     assert(depth == 0, "should have counted back down");
duke@435:   }
duke@435: 
iveresov@1939:   UseCountComputer() {
iveresov@1939:     worklist = new Values();
iveresov@1939:     depth = 0;
iveresov@1939:   }
iveresov@1939: 
duke@435:  public:
duke@435:   static void compute(BlockList* blocks) {
iveresov@1939:     UseCountComputer ucc;
iveresov@1939:     blocks->iterate_backward(&ucc);
duke@435:   }
duke@435: };
duke@435: 
duke@435: 
duke@435: // helper macro for short definition of trace-output inside code
duke@435: #ifndef PRODUCT
duke@435:   #define TRACE_LINEAR_SCAN(level, code)       \
duke@435:     if (TraceLinearScanLevel >= level) {       \
duke@435:       code;                                    \
duke@435:     }
duke@435: #else
duke@435:   #define TRACE_LINEAR_SCAN(level, code)
duke@435: #endif
duke@435: 
duke@435: class ComputeLinearScanOrder : public StackObj {
duke@435:  private:
duke@435:   int        _max_block_id;        // the highest block_id of a block
duke@435:   int        _num_blocks;          // total number of blocks (smaller than _max_block_id)
duke@435:   int        _num_loops;           // total number of loops
duke@435:   bool       _iterative_dominators;// method requires iterative computation of dominatiors
duke@435: 
duke@435:   BlockList* _linear_scan_order;   // the resulting list of blocks in correct order
duke@435: 
duke@435:   BitMap     _visited_blocks;      // used for recursive processing of blocks
duke@435:   BitMap     _active_blocks;       // used for recursive processing of blocks
duke@435:   BitMap     _dominator_blocks;    // temproary BitMap used for computation of dominator
duke@435:   intArray   _forward_branches;    // number of incoming forward branches for each block
duke@435:   BlockList  _loop_end_blocks;     // list of all loop end blocks collected during count_edges
duke@435:   BitMap2D   _loop_map;            // two-dimensional bit set: a bit is set if a block is contained in a loop
duke@435:   BlockList  _work_list;           // temporary list (used in mark_loops and compute_order)
roland@4860:   BlockList  _loop_headers;
duke@435: 
iveresov@2138:   Compilation* _compilation;
iveresov@2138: 
duke@435:   // accessors for _visited_blocks and _active_blocks
duke@435:   void init_visited()                     { _active_blocks.clear(); _visited_blocks.clear(); }
duke@435:   bool is_visited(BlockBegin* b) const    { return _visited_blocks.at(b->block_id()); }
duke@435:   bool is_active(BlockBegin* b) const     { return _active_blocks.at(b->block_id()); }
duke@435:   void set_visited(BlockBegin* b)         { assert(!is_visited(b), "already set"); _visited_blocks.set_bit(b->block_id()); }
duke@435:   void set_active(BlockBegin* b)          { assert(!is_active(b), "already set");  _active_blocks.set_bit(b->block_id()); }
duke@435:   void clear_active(BlockBegin* b)        { assert(is_active(b), "not already");   _active_blocks.clear_bit(b->block_id()); }
duke@435: 
duke@435:   // accessors for _forward_branches
duke@435:   void inc_forward_branches(BlockBegin* b) { _forward_branches.at_put(b->block_id(), _forward_branches.at(b->block_id()) + 1); }
duke@435:   int  dec_forward_branches(BlockBegin* b) { _forward_branches.at_put(b->block_id(), _forward_branches.at(b->block_id()) - 1); return _forward_branches.at(b->block_id()); }
duke@435: 
duke@435:   // accessors for _loop_map
duke@435:   bool is_block_in_loop   (int loop_idx, BlockBegin* b) const { return _loop_map.at(loop_idx, b->block_id()); }
duke@435:   void set_block_in_loop  (int loop_idx, BlockBegin* b)       { _loop_map.set_bit(loop_idx, b->block_id()); }
duke@435:   void clear_block_in_loop(int loop_idx, int block_id)        { _loop_map.clear_bit(loop_idx, block_id); }
duke@435: 
duke@435:   // count edges between blocks
duke@435:   void count_edges(BlockBegin* cur, BlockBegin* parent);
duke@435: 
duke@435:   // loop detection
duke@435:   void mark_loops();
duke@435:   void clear_non_natural_loops(BlockBegin* start_block);
duke@435:   void assign_loop_depth(BlockBegin* start_block);
duke@435: 
duke@435:   // computation of final block order
duke@435:   BlockBegin* common_dominator(BlockBegin* a, BlockBegin* b);
duke@435:   void compute_dominator(BlockBegin* cur, BlockBegin* parent);
duke@435:   int  compute_weight(BlockBegin* cur);
duke@435:   bool ready_for_processing(BlockBegin* cur);
duke@435:   void sort_into_work_list(BlockBegin* b);
duke@435:   void append_block(BlockBegin* cur);
duke@435:   void compute_order(BlockBegin* start_block);
duke@435: 
duke@435:   // fixup of dominators for non-natural loops
duke@435:   bool compute_dominators_iter();
duke@435:   void compute_dominators();
duke@435: 
duke@435:   // debug functions
duke@435:   NOT_PRODUCT(void print_blocks();)
duke@435:   DEBUG_ONLY(void verify();)
duke@435: 
iveresov@2138:   Compilation* compilation() const { return _compilation; }
duke@435:  public:
iveresov@2138:   ComputeLinearScanOrder(Compilation* c, BlockBegin* start_block);
duke@435: 
duke@435:   // accessors for final result
duke@435:   BlockList* linear_scan_order() const    { return _linear_scan_order; }
duke@435:   int        num_loops() const            { return _num_loops; }
duke@435: };
duke@435: 
duke@435: 
iveresov@2138: ComputeLinearScanOrder::ComputeLinearScanOrder(Compilation* c, BlockBegin* start_block) :
duke@435:   _max_block_id(BlockBegin::number_of_blocks()),
duke@435:   _num_blocks(0),
duke@435:   _num_loops(0),
duke@435:   _iterative_dominators(false),
duke@435:   _visited_blocks(_max_block_id),
duke@435:   _active_blocks(_max_block_id),
duke@435:   _dominator_blocks(_max_block_id),
duke@435:   _forward_branches(_max_block_id, 0),
duke@435:   _loop_end_blocks(8),
duke@435:   _work_list(8),
duke@435:   _linear_scan_order(NULL), // initialized later with correct size
iveresov@2138:   _loop_map(0, 0),          // initialized later with correct size
iveresov@2138:   _compilation(c)
duke@435: {
ccheung@5259:   TRACE_LINEAR_SCAN(2, tty->print_cr("***** computing linear-scan block order"));
duke@435: 
duke@435:   init_visited();
duke@435:   count_edges(start_block, NULL);
duke@435: 
iveresov@2138:   if (compilation()->is_profiling()) {
iveresov@2349:     ciMethod *method = compilation()->method();
iveresov@2349:     if (!method->is_accessor()) {
iveresov@2349:       ciMethodData* md = method->method_data_or_null();
iveresov@2349:       assert(md != NULL, "Sanity");
iveresov@2349:       md->set_compilation_stats(_num_loops, _num_blocks);
iveresov@2349:     }
iveresov@2138:   }
iveresov@2138: 
duke@435:   if (_num_loops > 0) {
duke@435:     mark_loops();
duke@435:     clear_non_natural_loops(start_block);
duke@435:     assign_loop_depth(start_block);
duke@435:   }
duke@435: 
duke@435:   compute_order(start_block);
duke@435:   compute_dominators();
duke@435: 
duke@435:   NOT_PRODUCT(print_blocks());
duke@435:   DEBUG_ONLY(verify());
duke@435: }
duke@435: 
duke@435: 
duke@435: // Traverse the CFG:
duke@435: // * count total number of blocks
duke@435: // * count all incoming edges and backward incoming edges
duke@435: // * number loop header blocks
duke@435: // * create a list with all loop end blocks
duke@435: void ComputeLinearScanOrder::count_edges(BlockBegin* cur, BlockBegin* parent) {
duke@435:   TRACE_LINEAR_SCAN(3, tty->print_cr("Enter count_edges for block B%d coming from B%d", cur->block_id(), parent != NULL ? parent->block_id() : -1));
duke@435:   assert(cur->dominator() == NULL, "dominator already initialized");
duke@435: 
duke@435:   if (is_active(cur)) {
duke@435:     TRACE_LINEAR_SCAN(3, tty->print_cr("backward branch"));
duke@435:     assert(is_visited(cur), "block must be visisted when block is active");
duke@435:     assert(parent != NULL, "must have parent");
duke@435: 
duke@435:     cur->set(BlockBegin::linear_scan_loop_header_flag);
duke@435:     cur->set(BlockBegin::backward_branch_target_flag);
duke@435: 
duke@435:     parent->set(BlockBegin::linear_scan_loop_end_flag);
never@863: 
never@863:     // When a loop header is also the start of an exception handler, then the backward branch is
never@863:     // an exception edge. Because such edges are usually critical edges which cannot be split, the
never@863:     // loop must be excluded here from processing.
never@863:     if (cur->is_set(BlockBegin::exception_entry_flag)) {
never@863:       // Make sure that dominators are correct in this weird situation
never@863:       _iterative_dominators = true;
never@863:       return;
never@863:     }
never@863:     assert(parent->number_of_sux() == 1 && parent->sux_at(0) == cur,
never@863:            "loop end blocks must have one successor (critical edges are split)");
never@863: 
duke@435:     _loop_end_blocks.append(parent);
duke@435:     return;
duke@435:   }
duke@435: 
duke@435:   // increment number of incoming forward branches
duke@435:   inc_forward_branches(cur);
duke@435: 
duke@435:   if (is_visited(cur)) {
duke@435:     TRACE_LINEAR_SCAN(3, tty->print_cr("block already visited"));
duke@435:     return;
duke@435:   }
duke@435: 
duke@435:   _num_blocks++;
duke@435:   set_visited(cur);
duke@435:   set_active(cur);
duke@435: 
duke@435:   // recursive call for all successors
duke@435:   int i;
duke@435:   for (i = cur->number_of_sux() - 1; i >= 0; i--) {
duke@435:     count_edges(cur->sux_at(i), cur);
duke@435:   }
duke@435:   for (i = cur->number_of_exception_handlers() - 1; i >= 0; i--) {
duke@435:     count_edges(cur->exception_handler_at(i), cur);
duke@435:   }
duke@435: 
duke@435:   clear_active(cur);
duke@435: 
duke@435:   // Each loop has a unique number.
duke@435:   // When multiple loops are nested, assign_loop_depth assumes that the
duke@435:   // innermost loop has the lowest number. This is guaranteed by setting
duke@435:   // the loop number after the recursive calls for the successors above
duke@435:   // have returned.
duke@435:   if (cur->is_set(BlockBegin::linear_scan_loop_header_flag)) {
duke@435:     assert(cur->loop_index() == -1, "cannot set loop-index twice");
duke@435:     TRACE_LINEAR_SCAN(3, tty->print_cr("Block B%d is loop header of loop %d", cur->block_id(), _num_loops));
duke@435: 
duke@435:     cur->set_loop_index(_num_loops);
roland@4860:     _loop_headers.append(cur);
duke@435:     _num_loops++;
duke@435:   }
duke@435: 
duke@435:   TRACE_LINEAR_SCAN(3, tty->print_cr("Finished count_edges for block B%d", cur->block_id()));
duke@435: }
duke@435: 
duke@435: 
duke@435: void ComputeLinearScanOrder::mark_loops() {
duke@435:   TRACE_LINEAR_SCAN(3, tty->print_cr("----- marking loops"));
duke@435: 
duke@435:   _loop_map = BitMap2D(_num_loops, _max_block_id);
duke@435:   _loop_map.clear();
duke@435: 
duke@435:   for (int i = _loop_end_blocks.length() - 1; i >= 0; i--) {
duke@435:     BlockBegin* loop_end   = _loop_end_blocks.at(i);
duke@435:     BlockBegin* loop_start = loop_end->sux_at(0);
duke@435:     int         loop_idx   = loop_start->loop_index();
duke@435: 
duke@435:     TRACE_LINEAR_SCAN(3, tty->print_cr("Processing loop from B%d to B%d (loop %d):", loop_start->block_id(), loop_end->block_id(), loop_idx));
duke@435:     assert(loop_end->is_set(BlockBegin::linear_scan_loop_end_flag), "loop end flag must be set");
duke@435:     assert(loop_end->number_of_sux() == 1, "incorrect number of successors");
duke@435:     assert(loop_start->is_set(BlockBegin::linear_scan_loop_header_flag), "loop header flag must be set");
duke@435:     assert(loop_idx >= 0 && loop_idx < _num_loops, "loop index not set");
duke@435:     assert(_work_list.is_empty(), "work list must be empty before processing");
duke@435: 
duke@435:     // add the end-block of the loop to the working list
duke@435:     _work_list.push(loop_end);
duke@435:     set_block_in_loop(loop_idx, loop_end);
duke@435:     do {
duke@435:       BlockBegin* cur = _work_list.pop();
duke@435: 
duke@435:       TRACE_LINEAR_SCAN(3, tty->print_cr("    processing B%d", cur->block_id()));
duke@435:       assert(is_block_in_loop(loop_idx, cur), "bit in loop map must be set when block is in work list");
duke@435: 
duke@435:       // recursive processing of all predecessors ends when start block of loop is reached
duke@435:       if (cur != loop_start && !cur->is_set(BlockBegin::osr_entry_flag)) {
duke@435:         for (int j = cur->number_of_preds() - 1; j >= 0; j--) {
duke@435:           BlockBegin* pred = cur->pred_at(j);
duke@435: 
duke@435:           if (!is_block_in_loop(loop_idx, pred) /*&& !pred->is_set(BlockBeginosr_entry_flag)*/) {
duke@435:             // this predecessor has not been processed yet, so add it to work list
duke@435:             TRACE_LINEAR_SCAN(3, tty->print_cr("    pushing B%d", pred->block_id()));
duke@435:             _work_list.push(pred);
duke@435:             set_block_in_loop(loop_idx, pred);
duke@435:           }
duke@435:         }
duke@435:       }
duke@435:     } while (!_work_list.is_empty());
duke@435:   }
duke@435: }
duke@435: 
duke@435: 
duke@435: // check for non-natural loops (loops where the loop header does not dominate
duke@435: // all other loop blocks = loops with mulitple entries).
duke@435: // such loops are ignored
duke@435: void ComputeLinearScanOrder::clear_non_natural_loops(BlockBegin* start_block) {
duke@435:   for (int i = _num_loops - 1; i >= 0; i--) {
duke@435:     if (is_block_in_loop(i, start_block)) {
duke@435:       // loop i contains the entry block of the method
duke@435:       // -> this is not a natural loop, so ignore it
duke@435:       TRACE_LINEAR_SCAN(2, tty->print_cr("Loop %d is non-natural, so it is ignored", i));
duke@435: 
roland@4860:       BlockBegin *loop_header = _loop_headers.at(i);
roland@4860:       assert(loop_header->is_set(BlockBegin::linear_scan_loop_header_flag), "Must be loop header");
roland@4860: 
roland@4860:       for (int j = 0; j < loop_header->number_of_preds(); j++) {
roland@4860:         BlockBegin *pred = loop_header->pred_at(j);
roland@4860:         pred->clear(BlockBegin::linear_scan_loop_end_flag);
roland@4860:       }
roland@4860: 
roland@4860:       loop_header->clear(BlockBegin::linear_scan_loop_header_flag);
roland@4860: 
duke@435:       for (int block_id = _max_block_id - 1; block_id >= 0; block_id--) {
duke@435:         clear_block_in_loop(i, block_id);
duke@435:       }
duke@435:       _iterative_dominators = true;
duke@435:     }
duke@435:   }
duke@435: }
duke@435: 
duke@435: void ComputeLinearScanOrder::assign_loop_depth(BlockBegin* start_block) {
ccheung@5259:   TRACE_LINEAR_SCAN(3, tty->print_cr("----- computing loop-depth and weight"));
duke@435:   init_visited();
duke@435: 
duke@435:   assert(_work_list.is_empty(), "work list must be empty before processing");
duke@435:   _work_list.append(start_block);
duke@435: 
duke@435:   do {
duke@435:     BlockBegin* cur = _work_list.pop();
duke@435: 
duke@435:     if (!is_visited(cur)) {
duke@435:       set_visited(cur);
duke@435:       TRACE_LINEAR_SCAN(4, tty->print_cr("Computing loop depth for block B%d", cur->block_id()));
duke@435: 
duke@435:       // compute loop-depth and loop-index for the block
duke@435:       assert(cur->loop_depth() == 0, "cannot set loop-depth twice");
duke@435:       int i;
duke@435:       int loop_depth = 0;
duke@435:       int min_loop_idx = -1;
duke@435:       for (i = _num_loops - 1; i >= 0; i--) {
duke@435:         if (is_block_in_loop(i, cur)) {
duke@435:           loop_depth++;
duke@435:           min_loop_idx = i;
duke@435:         }
duke@435:       }
duke@435:       cur->set_loop_depth(loop_depth);
duke@435:       cur->set_loop_index(min_loop_idx);
duke@435: 
duke@435:       // append all unvisited successors to work list
duke@435:       for (i = cur->number_of_sux() - 1; i >= 0; i--) {
duke@435:         _work_list.append(cur->sux_at(i));
duke@435:       }
duke@435:       for (i = cur->number_of_exception_handlers() - 1; i >= 0; i--) {
duke@435:         _work_list.append(cur->exception_handler_at(i));
duke@435:       }
duke@435:     }
duke@435:   } while (!_work_list.is_empty());
duke@435: }
duke@435: 
duke@435: 
duke@435: BlockBegin* ComputeLinearScanOrder::common_dominator(BlockBegin* a, BlockBegin* b) {
duke@435:   assert(a != NULL && b != NULL, "must have input blocks");
duke@435: 
duke@435:   _dominator_blocks.clear();
duke@435:   while (a != NULL) {
duke@435:     _dominator_blocks.set_bit(a->block_id());
duke@435:     assert(a->dominator() != NULL || a == _linear_scan_order->at(0), "dominator must be initialized");
duke@435:     a = a->dominator();
duke@435:   }
duke@435:   while (b != NULL && !_dominator_blocks.at(b->block_id())) {
duke@435:     assert(b->dominator() != NULL || b == _linear_scan_order->at(0), "dominator must be initialized");
duke@435:     b = b->dominator();
duke@435:   }
duke@435: 
duke@435:   assert(b != NULL, "could not find dominator");
duke@435:   return b;
duke@435: }
duke@435: 
duke@435: void ComputeLinearScanOrder::compute_dominator(BlockBegin* cur, BlockBegin* parent) {
duke@435:   if (cur->dominator() == NULL) {
duke@435:     TRACE_LINEAR_SCAN(4, tty->print_cr("DOM: initializing dominator of B%d to B%d", cur->block_id(), parent->block_id()));
duke@435:     cur->set_dominator(parent);
duke@435: 
duke@435:   } else if (!(cur->is_set(BlockBegin::linear_scan_loop_header_flag) && parent->is_set(BlockBegin::linear_scan_loop_end_flag))) {
duke@435:     TRACE_LINEAR_SCAN(4, tty->print_cr("DOM: computing dominator of B%d: common dominator of B%d and B%d is B%d", cur->block_id(), parent->block_id(), cur->dominator()->block_id(), common_dominator(cur->dominator(), parent)->block_id()));
roland@4860:     // Does not hold for exception blocks
roland@4860:     assert(cur->number_of_preds() > 1 || cur->is_set(BlockBegin::exception_entry_flag), "");
duke@435:     cur->set_dominator(common_dominator(cur->dominator(), parent));
duke@435:   }
roland@4860: 
roland@4860:   // Additional edge to xhandler of all our successors
roland@4860:   // range check elimination needs that the state at the end of a
roland@4860:   // block be valid in every block it dominates so cur must dominate
roland@4860:   // the exception handlers of its successors.
roland@4860:   int num_cur_xhandler = cur->number_of_exception_handlers();
roland@4860:   for (int j = 0; j < num_cur_xhandler; j++) {
roland@4860:     BlockBegin* xhandler = cur->exception_handler_at(j);
roland@4860:     compute_dominator(xhandler, parent);
roland@4860:   }
duke@435: }
duke@435: 
duke@435: 
duke@435: int ComputeLinearScanOrder::compute_weight(BlockBegin* cur) {
duke@435:   BlockBegin* single_sux = NULL;
duke@435:   if (cur->number_of_sux() == 1) {
duke@435:     single_sux = cur->sux_at(0);
duke@435:   }
duke@435: 
duke@435:   // limit loop-depth to 15 bit (only for security reason, it will never be so big)
duke@435:   int weight = (cur->loop_depth() & 0x7FFF) << 16;
duke@435: 
duke@435:   // general macro for short definition of weight flags
duke@435:   // the first instance of INC_WEIGHT_IF has the highest priority
duke@435:   int cur_bit = 15;
duke@435:   #define INC_WEIGHT_IF(condition) if ((condition)) { weight |= (1 << cur_bit); } cur_bit--;
duke@435: 
duke@435:   // this is necessery for the (very rare) case that two successing blocks have
duke@435:   // the same loop depth, but a different loop index (can happen for endless loops
duke@435:   // with exception handlers)
duke@435:   INC_WEIGHT_IF(!cur->is_set(BlockBegin::linear_scan_loop_header_flag));
duke@435: 
duke@435:   // loop end blocks (blocks that end with a backward branch) are added
duke@435:   // after all other blocks of the loop.
duke@435:   INC_WEIGHT_IF(!cur->is_set(BlockBegin::linear_scan_loop_end_flag));
duke@435: 
duke@435:   // critical edge split blocks are prefered because than they have a bigger
duke@435:   // proability to be completely empty
duke@435:   INC_WEIGHT_IF(cur->is_set(BlockBegin::critical_edge_split_flag));
duke@435: 
duke@435:   // exceptions should not be thrown in normal control flow, so these blocks
duke@435:   // are added as late as possible
duke@435:   INC_WEIGHT_IF(cur->end()->as_Throw() == NULL  && (single_sux == NULL || single_sux->end()->as_Throw()  == NULL));
duke@435:   INC_WEIGHT_IF(cur->end()->as_Return() == NULL && (single_sux == NULL || single_sux->end()->as_Return() == NULL));
duke@435: 
duke@435:   // exceptions handlers are added as late as possible
duke@435:   INC_WEIGHT_IF(!cur->is_set(BlockBegin::exception_entry_flag));
duke@435: 
duke@435:   // guarantee that weight is > 0
duke@435:   weight |= 1;
duke@435: 
duke@435:   #undef INC_WEIGHT_IF
duke@435:   assert(cur_bit >= 0, "too many flags");
duke@435:   assert(weight > 0, "weight cannot become negative");
duke@435: 
duke@435:   return weight;
duke@435: }
duke@435: 
duke@435: bool ComputeLinearScanOrder::ready_for_processing(BlockBegin* cur) {
duke@435:   // Discount the edge just traveled.
duke@435:   // When the number drops to zero, all forward branches were processed
duke@435:   if (dec_forward_branches(cur) != 0) {
duke@435:     return false;
duke@435:   }
duke@435: 
duke@435:   assert(_linear_scan_order->index_of(cur) == -1, "block already processed (block can be ready only once)");
duke@435:   assert(_work_list.index_of(cur) == -1, "block already in work-list (block can be ready only once)");
duke@435:   return true;
duke@435: }
duke@435: 
duke@435: void ComputeLinearScanOrder::sort_into_work_list(BlockBegin* cur) {
duke@435:   assert(_work_list.index_of(cur) == -1, "block already in work list");
duke@435: 
duke@435:   int cur_weight = compute_weight(cur);
duke@435: 
duke@435:   // the linear_scan_number is used to cache the weight of a block
duke@435:   cur->set_linear_scan_number(cur_weight);
duke@435: 
duke@435: #ifndef PRODUCT
duke@435:   if (StressLinearScan) {
duke@435:     _work_list.insert_before(0, cur);
duke@435:     return;
duke@435:   }
duke@435: #endif
duke@435: 
duke@435:   _work_list.append(NULL); // provide space for new element
duke@435: 
duke@435:   int insert_idx = _work_list.length() - 1;
duke@435:   while (insert_idx > 0 && _work_list.at(insert_idx - 1)->linear_scan_number() > cur_weight) {
duke@435:     _work_list.at_put(insert_idx, _work_list.at(insert_idx - 1));
duke@435:     insert_idx--;
duke@435:   }
duke@435:   _work_list.at_put(insert_idx, cur);
duke@435: 
duke@435:   TRACE_LINEAR_SCAN(3, tty->print_cr("Sorted B%d into worklist. new worklist:", cur->block_id()));
duke@435:   TRACE_LINEAR_SCAN(3, for (int i = 0; i < _work_list.length(); i++) tty->print_cr("%8d B%2d  weight:%6x", i, _work_list.at(i)->block_id(), _work_list.at(i)->linear_scan_number()));
duke@435: 
duke@435: #ifdef ASSERT
duke@435:   for (int i = 0; i < _work_list.length(); i++) {
duke@435:     assert(_work_list.at(i)->linear_scan_number() > 0, "weight not set");
duke@435:     assert(i == 0 || _work_list.at(i - 1)->linear_scan_number() <= _work_list.at(i)->linear_scan_number(), "incorrect order in worklist");
duke@435:   }
duke@435: #endif
duke@435: }
duke@435: 
duke@435: void ComputeLinearScanOrder::append_block(BlockBegin* cur) {
duke@435:   TRACE_LINEAR_SCAN(3, tty->print_cr("appending block B%d (weight 0x%6x) to linear-scan order", cur->block_id(), cur->linear_scan_number()));
duke@435:   assert(_linear_scan_order->index_of(cur) == -1, "cannot add the same block twice");
duke@435: 
duke@435:   // currently, the linear scan order and code emit order are equal.
duke@435:   // therefore the linear_scan_number and the weight of a block must also
duke@435:   // be equal.
duke@435:   cur->set_linear_scan_number(_linear_scan_order->length());
duke@435:   _linear_scan_order->append(cur);
duke@435: }
duke@435: 
duke@435: void ComputeLinearScanOrder::compute_order(BlockBegin* start_block) {
ccheung@5259:   TRACE_LINEAR_SCAN(3, tty->print_cr("----- computing final block order"));
duke@435: 
duke@435:   // the start block is always the first block in the linear scan order
duke@435:   _linear_scan_order = new BlockList(_num_blocks);
duke@435:   append_block(start_block);
duke@435: 
duke@435:   assert(start_block->end()->as_Base() != NULL, "start block must end with Base-instruction");
duke@435:   BlockBegin* std_entry = ((Base*)start_block->end())->std_entry();
duke@435:   BlockBegin* osr_entry = ((Base*)start_block->end())->osr_entry();
duke@435: 
duke@435:   BlockBegin* sux_of_osr_entry = NULL;
duke@435:   if (osr_entry != NULL) {
duke@435:     // special handling for osr entry:
duke@435:     // ignore the edge between the osr entry and its successor for processing
duke@435:     // the osr entry block is added manually below
duke@435:     assert(osr_entry->number_of_sux() == 1, "osr entry must have exactly one successor");
duke@435:     assert(osr_entry->sux_at(0)->number_of_preds() >= 2, "sucessor of osr entry must have two predecessors (otherwise it is not present in normal control flow");
duke@435: 
duke@435:     sux_of_osr_entry = osr_entry->sux_at(0);
duke@435:     dec_forward_branches(sux_of_osr_entry);
duke@435: 
duke@435:     compute_dominator(osr_entry, start_block);
duke@435:     _iterative_dominators = true;
duke@435:   }
duke@435:   compute_dominator(std_entry, start_block);
duke@435: 
duke@435:   // start processing with standard entry block
duke@435:   assert(_work_list.is_empty(), "list must be empty before processing");
duke@435: 
duke@435:   if (ready_for_processing(std_entry)) {
duke@435:     sort_into_work_list(std_entry);
duke@435:   } else {
duke@435:     assert(false, "the std_entry must be ready for processing (otherwise, the method has no start block)");
duke@435:   }
duke@435: 
duke@435:   do {
duke@435:     BlockBegin* cur = _work_list.pop();
duke@435: 
duke@435:     if (cur == sux_of_osr_entry) {
duke@435:       // the osr entry block is ignored in normal processing, it is never added to the
duke@435:       // work list. Instead, it is added as late as possible manually here.
duke@435:       append_block(osr_entry);
duke@435:       compute_dominator(cur, osr_entry);
duke@435:     }
duke@435:     append_block(cur);
duke@435: 
duke@435:     int i;
duke@435:     int num_sux = cur->number_of_sux();
duke@435:     // changed loop order to get "intuitive" order of if- and else-blocks
duke@435:     for (i = 0; i < num_sux; i++) {
duke@435:       BlockBegin* sux = cur->sux_at(i);
duke@435:       compute_dominator(sux, cur);
duke@435:       if (ready_for_processing(sux)) {
duke@435:         sort_into_work_list(sux);
duke@435:       }
duke@435:     }
duke@435:     num_sux = cur->number_of_exception_handlers();
duke@435:     for (i = 0; i < num_sux; i++) {
duke@435:       BlockBegin* sux = cur->exception_handler_at(i);
duke@435:       if (ready_for_processing(sux)) {
duke@435:         sort_into_work_list(sux);
duke@435:       }
duke@435:     }
duke@435:   } while (_work_list.length() > 0);
duke@435: }
duke@435: 
duke@435: 
duke@435: bool ComputeLinearScanOrder::compute_dominators_iter() {
duke@435:   bool changed = false;
duke@435:   int num_blocks = _linear_scan_order->length();
duke@435: 
duke@435:   assert(_linear_scan_order->at(0)->dominator() == NULL, "must not have dominator");
duke@435:   assert(_linear_scan_order->at(0)->number_of_preds() == 0, "must not have predecessors");
duke@435:   for (int i = 1; i < num_blocks; i++) {
duke@435:     BlockBegin* block = _linear_scan_order->at(i);
duke@435: 
duke@435:     BlockBegin* dominator = block->pred_at(0);
duke@435:     int num_preds = block->number_of_preds();
roland@4860: 
roland@4860:     TRACE_LINEAR_SCAN(4, tty->print_cr("DOM: Processing B%d", block->block_id()));
roland@4860: 
roland@4860:     for (int j = 0; j < num_preds; j++) {
roland@4860: 
roland@4860:       BlockBegin *pred = block->pred_at(j);
roland@4860:       TRACE_LINEAR_SCAN(4, tty->print_cr("   DOM: Subrocessing B%d", pred->block_id()));
roland@4860: 
roland@4860:       if (block->is_set(BlockBegin::exception_entry_flag)) {
roland@4860:         dominator = common_dominator(dominator, pred);
roland@4860:         int num_pred_preds = pred->number_of_preds();
roland@4860:         for (int k = 0; k < num_pred_preds; k++) {
roland@4860:           dominator = common_dominator(dominator, pred->pred_at(k));
roland@4860:         }
roland@4860:       } else {
roland@4860:         dominator = common_dominator(dominator, pred);
roland@4860:       }
duke@435:     }
duke@435: 
duke@435:     if (dominator != block->dominator()) {
duke@435:       TRACE_LINEAR_SCAN(4, tty->print_cr("DOM: updating dominator of B%d from B%d to B%d", block->block_id(), block->dominator()->block_id(), dominator->block_id()));
duke@435: 
duke@435:       block->set_dominator(dominator);
duke@435:       changed = true;
duke@435:     }
duke@435:   }
duke@435:   return changed;
duke@435: }
duke@435: 
duke@435: void ComputeLinearScanOrder::compute_dominators() {
duke@435:   TRACE_LINEAR_SCAN(3, tty->print_cr("----- computing dominators (iterative computation reqired: %d)", _iterative_dominators));
duke@435: 
duke@435:   // iterative computation of dominators is only required for methods with non-natural loops
duke@435:   // and OSR-methods. For all other methods, the dominators computed when generating the
duke@435:   // linear scan block order are correct.
duke@435:   if (_iterative_dominators) {
duke@435:     do {
duke@435:       TRACE_LINEAR_SCAN(1, tty->print_cr("DOM: next iteration of fix-point calculation"));
duke@435:     } while (compute_dominators_iter());
duke@435:   }
duke@435: 
duke@435:   // check that dominators are correct
duke@435:   assert(!compute_dominators_iter(), "fix point not reached");
roland@4860: 
roland@4860:   // Add Blocks to dominates-Array
roland@4860:   int num_blocks = _linear_scan_order->length();
roland@4860:   for (int i = 0; i < num_blocks; i++) {
roland@4860:     BlockBegin* block = _linear_scan_order->at(i);
roland@4860: 
roland@4860:     BlockBegin *dom = block->dominator();
roland@4860:     if (dom) {
roland@4860:       assert(dom->dominator_depth() != -1, "Dominator must have been visited before");
roland@4860:       dom->dominates()->append(block);
roland@4860:       block->set_dominator_depth(dom->dominator_depth() + 1);
roland@4860:     } else {
roland@4860:       block->set_dominator_depth(0);
roland@4860:     }
roland@4860:   }
duke@435: }
duke@435: 
duke@435: 
duke@435: #ifndef PRODUCT
duke@435: void ComputeLinearScanOrder::print_blocks() {
duke@435:   if (TraceLinearScanLevel >= 2) {
duke@435:     tty->print_cr("----- loop information:");
duke@435:     for (int block_idx = 0; block_idx < _linear_scan_order->length(); block_idx++) {
duke@435:       BlockBegin* cur = _linear_scan_order->at(block_idx);
duke@435: 
duke@435:       tty->print("%4d: B%2d: ", cur->linear_scan_number(), cur->block_id());
duke@435:       for (int loop_idx = 0; loop_idx < _num_loops; loop_idx++) {
duke@435:         tty->print ("%d ", is_block_in_loop(loop_idx, cur));
duke@435:       }
duke@435:       tty->print_cr(" -> loop_index: %2d, loop_depth: %2d", cur->loop_index(), cur->loop_depth());
duke@435:     }
duke@435:   }
duke@435: 
duke@435:   if (TraceLinearScanLevel >= 1) {
duke@435:     tty->print_cr("----- linear-scan block order:");
duke@435:     for (int block_idx = 0; block_idx < _linear_scan_order->length(); block_idx++) {
duke@435:       BlockBegin* cur = _linear_scan_order->at(block_idx);
duke@435:       tty->print("%4d: B%2d    loop: %2d  depth: %2d", cur->linear_scan_number(), cur->block_id(), cur->loop_index(), cur->loop_depth());
duke@435: 
duke@435:       tty->print(cur->is_set(BlockBegin::exception_entry_flag)         ? " ex" : "   ");
duke@435:       tty->print(cur->is_set(BlockBegin::critical_edge_split_flag)     ? " ce" : "   ");
duke@435:       tty->print(cur->is_set(BlockBegin::linear_scan_loop_header_flag) ? " lh" : "   ");
duke@435:       tty->print(cur->is_set(BlockBegin::linear_scan_loop_end_flag)    ? " le" : "   ");
duke@435: 
duke@435:       if (cur->dominator() != NULL) {
duke@435:         tty->print("    dom: B%d ", cur->dominator()->block_id());
duke@435:       } else {
duke@435:         tty->print("    dom: NULL ");
duke@435:       }
duke@435: 
duke@435:       if (cur->number_of_preds() > 0) {
duke@435:         tty->print("    preds: ");
duke@435:         for (int j = 0; j < cur->number_of_preds(); j++) {
duke@435:           BlockBegin* pred = cur->pred_at(j);
duke@435:           tty->print("B%d ", pred->block_id());
duke@435:         }
duke@435:       }
duke@435:       if (cur->number_of_sux() > 0) {
duke@435:         tty->print("    sux: ");
duke@435:         for (int j = 0; j < cur->number_of_sux(); j++) {
duke@435:           BlockBegin* sux = cur->sux_at(j);
duke@435:           tty->print("B%d ", sux->block_id());
duke@435:         }
duke@435:       }
duke@435:       if (cur->number_of_exception_handlers() > 0) {
duke@435:         tty->print("    ex: ");
duke@435:         for (int j = 0; j < cur->number_of_exception_handlers(); j++) {
duke@435:           BlockBegin* ex = cur->exception_handler_at(j);
duke@435:           tty->print("B%d ", ex->block_id());
duke@435:         }
duke@435:       }
duke@435:       tty->cr();
duke@435:     }
duke@435:   }
duke@435: }
duke@435: #endif
duke@435: 
duke@435: #ifdef ASSERT
duke@435: void ComputeLinearScanOrder::verify() {
duke@435:   assert(_linear_scan_order->length() == _num_blocks, "wrong number of blocks in list");
duke@435: 
duke@435:   if (StressLinearScan) {
duke@435:     // blocks are scrambled when StressLinearScan is used
duke@435:     return;
duke@435:   }
duke@435: 
duke@435:   // check that all successors of a block have a higher linear-scan-number
duke@435:   // and that all predecessors of a block have a lower linear-scan-number
duke@435:   // (only backward branches of loops are ignored)
duke@435:   int i;
duke@435:   for (i = 0; i < _linear_scan_order->length(); i++) {
duke@435:     BlockBegin* cur = _linear_scan_order->at(i);
duke@435: 
duke@435:     assert(cur->linear_scan_number() == i, "incorrect linear_scan_number");
duke@435:     assert(cur->linear_scan_number() >= 0 && cur->linear_scan_number() == _linear_scan_order->index_of(cur), "incorrect linear_scan_number");
duke@435: 
duke@435:     int j;
duke@435:     for (j = cur->number_of_sux() - 1; j >= 0; j--) {
duke@435:       BlockBegin* sux = cur->sux_at(j);
duke@435: 
duke@435:       assert(sux->linear_scan_number() >= 0 && sux->linear_scan_number() == _linear_scan_order->index_of(sux), "incorrect linear_scan_number");
roland@4860:       if (!sux->is_set(BlockBegin::backward_branch_target_flag)) {
duke@435:         assert(cur->linear_scan_number() < sux->linear_scan_number(), "invalid order");
duke@435:       }
duke@435:       if (cur->loop_depth() == sux->loop_depth()) {
duke@435:         assert(cur->loop_index() == sux->loop_index() || sux->is_set(BlockBegin::linear_scan_loop_header_flag), "successing blocks with same loop depth must have same loop index");
duke@435:       }
duke@435:     }
duke@435: 
duke@435:     for (j = cur->number_of_preds() - 1; j >= 0; j--) {
duke@435:       BlockBegin* pred = cur->pred_at(j);
duke@435: 
duke@435:       assert(pred->linear_scan_number() >= 0 && pred->linear_scan_number() == _linear_scan_order->index_of(pred), "incorrect linear_scan_number");
roland@4860:       if (!cur->is_set(BlockBegin::backward_branch_target_flag)) {
duke@435:         assert(cur->linear_scan_number() > pred->linear_scan_number(), "invalid order");
duke@435:       }
duke@435:       if (cur->loop_depth() == pred->loop_depth()) {
duke@435:         assert(cur->loop_index() == pred->loop_index() || cur->is_set(BlockBegin::linear_scan_loop_header_flag), "successing blocks with same loop depth must have same loop index");
duke@435:       }
duke@435: 
duke@435:       assert(cur->dominator()->linear_scan_number() <= cur->pred_at(j)->linear_scan_number(), "dominator must be before predecessors");
duke@435:     }
duke@435: 
duke@435:     // check dominator
duke@435:     if (i == 0) {
duke@435:       assert(cur->dominator() == NULL, "first block has no dominator");
duke@435:     } else {
duke@435:       assert(cur->dominator() != NULL, "all but first block must have dominator");
duke@435:     }
roland@4860:     // Assertion does not hold for exception handlers
roland@4860:     assert(cur->number_of_preds() != 1 || cur->dominator() == cur->pred_at(0) || cur->is_set(BlockBegin::exception_entry_flag), "Single predecessor must also be dominator");
duke@435:   }
duke@435: 
duke@435:   // check that all loops are continuous
duke@435:   for (int loop_idx = 0; loop_idx < _num_loops; loop_idx++) {
duke@435:     int block_idx = 0;
duke@435:     assert(!is_block_in_loop(loop_idx, _linear_scan_order->at(block_idx)), "the first block must not be present in any loop");
duke@435: 
duke@435:     // skip blocks before the loop
duke@435:     while (block_idx < _num_blocks && !is_block_in_loop(loop_idx, _linear_scan_order->at(block_idx))) {
duke@435:       block_idx++;
duke@435:     }
duke@435:     // skip blocks of loop
duke@435:     while (block_idx < _num_blocks && is_block_in_loop(loop_idx, _linear_scan_order->at(block_idx))) {
duke@435:       block_idx++;
duke@435:     }
duke@435:     // after the first non-loop block, there must not be another loop-block
duke@435:     while (block_idx < _num_blocks) {
duke@435:       assert(!is_block_in_loop(loop_idx, _linear_scan_order->at(block_idx)), "loop not continuous in linear-scan order");
duke@435:       block_idx++;
duke@435:     }
duke@435:   }
duke@435: }
duke@435: #endif
duke@435: 
duke@435: 
duke@435: void IR::compute_code() {
duke@435:   assert(is_valid(), "IR must be valid");
duke@435: 
iveresov@2138:   ComputeLinearScanOrder compute_order(compilation(), start());
duke@435:   _num_loops = compute_order.num_loops();
duke@435:   _code = compute_order.linear_scan_order();
duke@435: }
duke@435: 
duke@435: 
duke@435: void IR::compute_use_counts() {
duke@435:   // make sure all values coming out of this block get evaluated.
duke@435:   int num_blocks = _code->length();
duke@435:   for (int i = 0; i < num_blocks; i++) {
duke@435:     _code->at(i)->end()->state()->pin_stack_for_linear_scan();
duke@435:   }
duke@435: 
duke@435:   // compute use counts
duke@435:   UseCountComputer::compute(_code);
duke@435: }
duke@435: 
duke@435: 
duke@435: void IR::iterate_preorder(BlockClosure* closure) {
duke@435:   assert(is_valid(), "IR must be valid");
duke@435:   start()->iterate_preorder(closure);
duke@435: }
duke@435: 
duke@435: 
duke@435: void IR::iterate_postorder(BlockClosure* closure) {
duke@435:   assert(is_valid(), "IR must be valid");
duke@435:   start()->iterate_postorder(closure);
duke@435: }
duke@435: 
duke@435: void IR::iterate_linear_scan_order(BlockClosure* closure) {
duke@435:   linear_scan_order()->iterate_forward(closure);
duke@435: }
duke@435: 
duke@435: 
duke@435: #ifndef PRODUCT
duke@435: class BlockPrinter: public BlockClosure {
duke@435:  private:
duke@435:   InstructionPrinter* _ip;
duke@435:   bool                _cfg_only;
duke@435:   bool                _live_only;
duke@435: 
duke@435:  public:
duke@435:   BlockPrinter(InstructionPrinter* ip, bool cfg_only, bool live_only = false) {
duke@435:     _ip       = ip;
duke@435:     _cfg_only = cfg_only;
duke@435:     _live_only = live_only;
duke@435:   }
duke@435: 
duke@435:   virtual void block_do(BlockBegin* block) {
duke@435:     if (_cfg_only) {
duke@435:       _ip->print_instr(block); tty->cr();
duke@435:     } else {
duke@435:       block->print_block(*_ip, _live_only);
duke@435:     }
duke@435:   }
duke@435: };
duke@435: 
duke@435: 
duke@435: void IR::print(BlockBegin* start, bool cfg_only, bool live_only) {
duke@435:   ttyLocker ttyl;
duke@435:   InstructionPrinter ip(!cfg_only);
duke@435:   BlockPrinter bp(&ip, cfg_only, live_only);
duke@435:   start->iterate_preorder(&bp);
duke@435:   tty->cr();
duke@435: }
duke@435: 
duke@435: void IR::print(bool cfg_only, bool live_only) {
duke@435:   if (is_valid()) {
duke@435:     print(start(), cfg_only, live_only);
duke@435:   } else {
duke@435:     tty->print_cr("invalid IR");
duke@435:   }
duke@435: }
duke@435: 
duke@435: 
duke@435: define_array(BlockListArray, BlockList*)
duke@435: define_stack(BlockListList, BlockListArray)
duke@435: 
duke@435: class PredecessorValidator : public BlockClosure {
duke@435:  private:
duke@435:   BlockListList* _predecessors;
duke@435:   BlockList*     _blocks;
duke@435: 
duke@435:   static int cmp(BlockBegin** a, BlockBegin** b) {
duke@435:     return (*a)->block_id() - (*b)->block_id();
duke@435:   }
duke@435: 
duke@435:  public:
duke@435:   PredecessorValidator(IR* hir) {
duke@435:     ResourceMark rm;
duke@435:     _predecessors = new BlockListList(BlockBegin::number_of_blocks(), NULL);
duke@435:     _blocks = new BlockList();
duke@435: 
duke@435:     int i;
duke@435:     hir->start()->iterate_preorder(this);
duke@435:     if (hir->code() != NULL) {
duke@435:       assert(hir->code()->length() == _blocks->length(), "must match");
duke@435:       for (i = 0; i < _blocks->length(); i++) {
duke@435:         assert(hir->code()->contains(_blocks->at(i)), "should be in both lists");
duke@435:       }
duke@435:     }
duke@435: 
duke@435:     for (i = 0; i < _blocks->length(); i++) {
duke@435:       BlockBegin* block = _blocks->at(i);
duke@435:       BlockList* preds = _predecessors->at(block->block_id());
duke@435:       if (preds == NULL) {
duke@435:         assert(block->number_of_preds() == 0, "should be the same");
duke@435:         continue;
duke@435:       }
duke@435: 
duke@435:       // clone the pred list so we can mutate it
duke@435:       BlockList* pred_copy = new BlockList();
duke@435:       int j;
duke@435:       for (j = 0; j < block->number_of_preds(); j++) {
duke@435:         pred_copy->append(block->pred_at(j));
duke@435:       }
duke@435:       // sort them in the same order
duke@435:       preds->sort(cmp);
duke@435:       pred_copy->sort(cmp);
duke@435:       int length = MIN2(preds->length(), block->number_of_preds());
duke@435:       for (j = 0; j < block->number_of_preds(); j++) {
duke@435:         assert(preds->at(j) == pred_copy->at(j), "must match");
duke@435:       }
duke@435: 
duke@435:       assert(preds->length() == block->number_of_preds(), "should be the same");
duke@435:     }
duke@435:   }
duke@435: 
duke@435:   virtual void block_do(BlockBegin* block) {
duke@435:     _blocks->append(block);
duke@435:     BlockEnd* be = block->end();
duke@435:     int n = be->number_of_sux();
duke@435:     int i;
duke@435:     for (i = 0; i < n; i++) {
duke@435:       BlockBegin* sux = be->sux_at(i);
duke@435:       assert(!sux->is_set(BlockBegin::exception_entry_flag), "must not be xhandler");
duke@435: 
duke@435:       BlockList* preds = _predecessors->at_grow(sux->block_id(), NULL);
duke@435:       if (preds == NULL) {
duke@435:         preds = new BlockList();
duke@435:         _predecessors->at_put(sux->block_id(), preds);
duke@435:       }
duke@435:       preds->append(block);
duke@435:     }
duke@435: 
duke@435:     n = block->number_of_exception_handlers();
duke@435:     for (i = 0; i < n; i++) {
duke@435:       BlockBegin* sux = block->exception_handler_at(i);
duke@435:       assert(sux->is_set(BlockBegin::exception_entry_flag), "must be xhandler");
duke@435: 
duke@435:       BlockList* preds = _predecessors->at_grow(sux->block_id(), NULL);
duke@435:       if (preds == NULL) {
duke@435:         preds = new BlockList();
duke@435:         _predecessors->at_put(sux->block_id(), preds);
duke@435:       }
duke@435:       preds->append(block);
duke@435:     }
duke@435:   }
duke@435: };
duke@435: 
roland@4860: class VerifyBlockBeginField : public BlockClosure {
roland@4860: 
roland@4860: public:
roland@4860: 
roland@4860:   virtual void block_do(BlockBegin *block) {
roland@4860:     for ( Instruction *cur = block; cur != NULL; cur = cur->next()) {
roland@4860:       assert(cur->block() == block, "Block begin is not correct");
roland@4860:     }
roland@4860:   }
roland@4860: };
roland@4860: 
duke@435: void IR::verify() {
duke@435: #ifdef ASSERT
duke@435:   PredecessorValidator pv(this);
roland@4860:   VerifyBlockBeginField verifier;
roland@4860:   this->iterate_postorder(&verifier);
duke@435: #endif
duke@435: }
duke@435: 
duke@435: #endif // PRODUCT
duke@435: 
iveresov@1939: void SubstitutionResolver::visit(Value* v) {
duke@435:   Value v0 = *v;
duke@435:   if (v0) {
duke@435:     Value vs = v0->subst();
duke@435:     if (vs != v0) {
duke@435:       *v = v0->subst();
duke@435:     }
duke@435:   }
duke@435: }
duke@435: 
duke@435: #ifdef ASSERT
iveresov@1939: class SubstitutionChecker: public ValueVisitor {
iveresov@1939:   void visit(Value* v) {
iveresov@1939:     Value v0 = *v;
iveresov@1939:     if (v0) {
iveresov@1939:       Value vs = v0->subst();
iveresov@1939:       assert(vs == v0, "missed substitution");
iveresov@1939:     }
duke@435:   }
iveresov@1939: };
duke@435: #endif
duke@435: 
duke@435: 
duke@435: void SubstitutionResolver::block_do(BlockBegin* block) {
duke@435:   Instruction* last = NULL;
duke@435:   for (Instruction* n = block; n != NULL;) {
iveresov@1939:     n->values_do(this);
duke@435:     // need to remove this instruction from the instruction stream
duke@435:     if (n->subst() != n) {
duke@435:       assert(last != NULL, "must have last");
roland@2174:       last->set_next(n->next());
duke@435:     } else {
duke@435:       last = n;
duke@435:     }
duke@435:     n = last->next();
duke@435:   }
duke@435: 
duke@435: #ifdef ASSERT
iveresov@1939:   SubstitutionChecker check_substitute;
iveresov@1939:   if (block->state()) block->state()->values_do(&check_substitute);
iveresov@1939:   block->block_values_do(&check_substitute);
iveresov@1939:   if (block->end() && block->end()->state()) block->end()->state()->values_do(&check_substitute);
duke@435: #endif
duke@435: }