![Compiler data structures
$ ~/jdk1.7.0-b147/fastdebug/bin/java -XX:+PrintCompilation -XX:+PrintIdeal
-XX:CICompilerCount=1 sum
214 1 sum::doit (22 bytes)
VM option '+PrintIdeal' ...
21" ConI" === 0 [[ 180 ]] #int:0
180" Phi" === 184 21 70 [[ 179 ]] #int !orig=[159],[139],[66] !
jvms: sum::doit @ bci:10
179" AddI" === _ 180 181 [[ 178 ]] !orig=[154],[137],70,[145] !jvms:
sum::doit @ bci:12
178" AddI" === _ 179 181 [[ 177 ]] !orig=[153],[146],[135],86,[71] !
jvms: sum::doit @ bci:14
177" AddI" === _ 178 181 [[ 176 ]] !orig=[165],[152],86,[71] !jvms:
sum::doit @ bci:14
148" ConI" === 0 [[ 87 ]] #int:97
176" AddI" === _ 177 181 [[ 190 ]] !orig=[168],[152],86,[71] !jvms:
sum::doit @ bci:14
// <idx> <node type> === <in[]> [[out[]]] <additional desc>
// jvms = JVMState, root()->dump(9999) would dump IR as above;
Real example: https://gist.github.com/1369656
9](https://image.slidesharecdn.com/jvm-code-reading-c2-111118121955-phpapp01/85/JVM-code-reading-C2-9-320.jpg)
![Compiler data structures
Options to generate data for Ideal
Graph Visualizer
-XX:PrintIdealGraphLevel=0 [0:None, 4: most verbose]
-XX:PrintIdealGraphPort=4444
-XX:PrintIdealGraphAddress=”127.0.0.1”
-XX:PrintIdealGraphFile=<path to IR xml file>
IdealGraphVisualizer listens to port 4444 by default.
14](https://image.slidesharecdn.com/jvm-code-reading-c2-111118121955-phpapp01/85/JVM-code-reading-C2-14-320.jpg)
![Compiler data structures
// Insert a new required input at the end
void Node::ins_req( uint idx, Node *n ) {
assert( is_not_dead(n), "can not use dead node");
add_req(NULL); // Make space
...
_in[idx] = n; // Stuff over old required edge
if (n != NULL) n->add_out((Node *)this); // Add
reciprocal def-use edge
}
void add_out( Node *n ) {
if (is_top()) return;
if( _outcnt == _outmax ) out_grow(_outcnt);
_out[_outcnt++] = n;
}
17](https://image.slidesharecdn.com/jvm-code-reading-c2-111118121955-phpapp01/85/JVM-code-reading-C2-17-320.jpg)
![Compiler data structures
class PhaseTransform : public Phase {
protected:
Arena* _arena;
Node_Array _nodes;
Type_Array _types;
ConINode* _icons[...];
ConLNode* _lcons[...];
ConNode* _zcons[...];
:
}
45](https://image.slidesharecdn.com/jvm-code-reading-c2-111118121955-phpapp01/85/JVM-code-reading-C2-45-320.jpg)
![Parse do_one_bytecode
getClass はインライン展開され、
LoadKlass -> メモリアクセスに。
hashCode は static に
public class Call {
public static void main(String[] args) {
Call c = new Call();
for (int i = 0; i < 100000; i++) {
c.doit();
}
}
int doit() {
return getClass().hashCode();
}
}
72](https://image.slidesharecdn.com/jvm-code-reading-c2-111118121955-phpapp01/85/JVM-code-reading-C2-72-320.jpg)
JVM code reading -- C2
- 1. C2 The Server Compiler 第5回JVMソースコードリーディングの会 @ytoshima 1
- 2. Compilation trigger invoke* と goto bytecode (to negative offset) の呼び出し数を interpreter でカウ ントし，しきい値を越えると CompilerBroker にコンパイルのリクエストを出す。キューにリク エストが入り、キューを見ている CompilerThread がコンパイルを開始する。GC 動作とも並列して動作できる様になっている。 2
- 3. Compilation trigger // Method invocation SimpleCompPolicy::method_invocation_e vent CompileBroker::compile_method(...) // OSR SimpleCompPolicy::method_back_branch_ event CompileBroker::compile_method(...) 3
- 4. Compilation trigger CompileBroker::compile_method CompileBroker::compile_method_base : CompileQueue* queue = compile_queue(comp_level); task = create_compile_task(queue, compile_id, method, osr_bci, comp_level, hot_method, hot_count, comment, blocking); 4
- 5. Compile Compile::Compile : if ((jvms = cg->generate(jvms)) == NULL) // Parse : Optimize(); : Code_Gen(); 5
- 7. Compiler data structures Compile::build_start_state + aload_0 (this) 7
- 8. Compiler data structures Graph Text Representation 8
- 9. Compiler data structures $ ~/jdk1.7.0-b147/fastdebug/bin/java -XX:+PrintCompilation -XX:+PrintIdeal -XX:CICompilerCount=1 sum 214 1 sum::doit (22 bytes) VM option '+PrintIdeal' ... 21" ConI" === 0 [[ 180 ]] #int:0 180" Phi" === 184 21 70 [[ 179 ]] #int !orig=[159],[139],[66] ! jvms: sum::doit @ bci:10 179" AddI" === _ 180 181 [[ 178 ]] !orig=[154],[137],70,[145] !jvms: sum::doit @ bci:12 178" AddI" === _ 179 181 [[ 177 ]] !orig=[153],[146],[135],86,[71] ! jvms: sum::doit @ bci:14 177" AddI" === _ 178 181 [[ 176 ]] !orig=[165],[152],86,[71] !jvms: sum::doit @ bci:14 148" ConI" === 0 [[ 87 ]] #int:97 176" AddI" === _ 177 181 [[ 190 ]] !orig=[168],[152],86,[71] !jvms: sum::doit @ bci:14 // <idx> <node type> === <in[]> [[out[]]] <additional desc> // jvms = JVMState, root()->dump(9999) would dump IR as above; Real example: https://gist.github.com/1369656 9
- 10. Ideal Graph Visualizer グラフ表示オプション level 4 ではパースの各段階と最適 化の各段階の IR が表示可能 10
- 11. Ideal Graph Visualizer public int getValue() { return value; } enum { Control, I_O, Memory, FramePtr, ReturnAdr, Parms }; 11
- 12. Ideal Graph Visualizer Final Code: Mach* node や Epilog, Prolog, Ret などマシン依存のノードに なっている 12
- 13. Compiler data structures Ideal Graph Visualizer http://ssw.jku.at/General/Staff/TW/igv.html etc/idealgraphvisualizer.conf: default_options="-J-Xmx400m --branding idealgraphvisualizer" You need a fastdebug or a debug build In OpenJDK build dir $ make fastdebug_build OR $ make debug_build 13
- 14. Compiler data structures Options to generate data for Ideal Graph Visualizer -XX:PrintIdealGraphLevel=0 [0:None, 4: most verbose] -XX:PrintIdealGraphPort=4444 -XX:PrintIdealGraphAddress=”127.0.0.1” -XX:PrintIdealGraphFile=<path to IR xml file> IdealGraphVisualizer listens to port 4444 by default. 14
- 15. Compiler data structures // -XX:PrintIdealGraphFile=<path> , IdealGraphViewer can display this <graphDocument> <group> <properties> <p name="name"> virtual jint Call.doit()</p> </properties> <graph name="Bytecode 0: aload_0"> <nodes> <node id="159337448"> <properties> <p name="name"> Root</p> <p name="type"> bottom</p> <p name="idx"> 0</p> ... </properties> </node> <node id="159414516"> ... </node> <node id="159337448"> ... </node> ... </nodes> <edges> <edge index="0" to="159337448" from="159337448"></edge> <edge index="0" to="159414516" from="159414516"></edge> Example: https://gist.github.com/1369620 15
- 16. Compiler data structures Node # _in : Node** // use-def # _out: Node** // def-use # _cnt: node_idx_t // # of required inputs # _max: node_idx_t // actual input array length # _outcnt: node_idx_t # _outmax: node_idx_t - _class_id: jushort - _flags: jushort // _in は ordered, 位置も重要 // サブクラスの合計 340 個 <-> C1 62 個 // _class_id は 16 bit 値、ideal, mach で node の型を判断 // _flags は enum NodeFlags: Flag_is_Copy, Flag_is_Call, // Flag_is_macro, Flag_is_con... 16
- 17. Compiler data structures // Insert a new required input at the end void Node::ins_req( uint idx, Node *n ) { assert( is_not_dead(n), "can not use dead node"); add_req(NULL); // Make space ... _in[idx] = n; // Stuff over old required edge if (n != NULL) n->add_out((Node *)this); // Add reciprocal def-use edge } void add_out( Node *n ) { if (is_top()) return; if( _outcnt == _outmax ) out_grow(_outcnt); _out[_outcnt++] = n; } 17
- 18. Compiler data structures RegionNode Control の merge PhiNode Control の merge に伴うデータのマージ. 対応する RegionNode を指す。 18
- 19. Compiler data structures Node // Optimize functions // more ideal node, canonicalize virtual Node *Ideal(PhaseGVN *phase, bool can_reshape); // set of values this node can take virtual const Type *Value ( PhaseTransform *phase ) const; // existing node which computes same virtual Node *Identity ( PhaseTransform *phase ); 19
- 20. Compiler data structures Node Ideal defined in Add*Node, MinINode, StartNode, ReturnNode, RethrowNode, SafePointNode, AllocateArrayNode, LockNode, UnlockNode, RegionNode, PhiNode, PCTableNode, NeverBranchNode, CMove*Node, ConstraintCastNode, CheckCastPPNode, Conv?2?Node, Div?Node, Mod?Node, IfNode, LoopNode, CountedLoopNode, LoopLimitNode, Load*Node, Store*Node, ClearArrayNode, StrIntrinsicNode, MemBarNode, MergeMemNode, Mul*Node, And*Node, LShift*Node, URShift*Node, RootNode, HaltNode, Sub*Node, Cmp*Node, BoolNode Ideal, Value, Identity は多くのサブクラスが目的に応じた物を定義 している。 20
- 21. Compiler data structures AddNode Ideal Convert "(x+1)+2" into "x+(1+2)" Convert "(x+1)+y" into "(x+y)+1" Convert "x+(y+1)" into "(x+y)+1" x Con1 Con2 x Con1 Con2 Add Add Add Add 21
- 22. Compiler data structures AddINode Ideal Node* in1 = in(1); Node* in2 = in(2); int op1 = in1->Opcode(); int op2 = in2->Opcode(); // Fold (con1-x)+con2 into (con1+con2)-x if ( op1 == Op_AddI && op2 == Op_SubI ) { // Swap edges to try optimizations below in1 = in2; in2 = in(1); op1 = op2; op2 = in2->Opcode(); } if( op1 == Op_SubI ) { "(a-b)+(c-d)" into "(a+c)-(b+d)" "(a-b)+(b+c)" into "(a+c)" "(a-b)+(c+b)" into "(a+c)" 22
- 23. Compiler data structures const Type *AddNode::Value(...) // Either input is TOP ==> the result is TOP // Either input is BOTTOM ==> the result is the local BOTTOM // Check for an addition involving the additive identity 23
- 24. Compiler data structures Node *AddNode::Identity(...) // If either input is a constant 0, return the other input. const Type *zero = add_id(); // The additive identity if( phase->type( in(1) )->higher_equal( zero ) ) return in(2); if( phase->type( in(2) )->higher_equal( zero ) ) return in(1); return this; 24
- 25. Compiler data structures Node *AddINode::Identity(...) // Fold (x-y)+y OR y+(x-y) into x if( in(1)->Opcode() == Op_SubI && phase->eqv(in(1)->in (2),in(2)) ) { return in(1)->in(1); } else if( in(2)->Opcode() == Op_SubI && phase->eqv(in (2)->in(2),in(1)) ) { return in(2)->in(1); } return AddNode::Identity(phase); 25
- 26. Compiler data structures Node *PhaseGVN::transform_no_reclaim // Return a node which computes the same function // as this node, but in a faster or cheaper fashion. while( 1 ) { Node *i = k->Ideal(this, /*can_reshape=*/false); if( !i ) break;... } const Type *t = k->Value(this); // Get runtime Value set k->raise_bottom_type(t); Node *i = k->Identity(this); if (i != k) return i; i = hash_find_insert(k); if( i && (i != k)) return i; Parse で Node を作ると transform する。Parse しながら GVN 26
- 27. Compiler data structures Node // raise_bottom_type related TypeNode // Type* _type ConNode // ConINode, ConPNode .. PhiNode // TypePtr* _adr_type // int _inst_id, // inst_index, _inst_offset ConvI2LNode MemNode // TypePtr* _adr_type LoadNode // Type* _type LoadPNode // load obj or arr LoadINode https://gist.github.com/1369608 27
- 28. Compiler data structures Node ^ RegionNode basic blocks にマップできる。入力は Control sources. PhiNode は RegionNode を指す 入力を持つ。PhiNode へのマージされるデータの入力は RegionNode の 入力と一対一の 対応を持つ。 PhiNode の 0 の入力は RegionNode で RegionNode の入力 0 は自身。 PhiNode* has_phi() const ^ LoopNode // Simple Loop Header short _loop_flags ^ RootNode 28
- 31. Compiler data structures Node ^ ProjNode // project a single elem out of a tuple or signature type ^ ParmNode // incoming Parameters const uint _con; // The field in the tuple we are projecting const bool _is_io_use; // Used to distinguish between the projections // used on the control and io paths from a macro node 31
- 32. Compiler data structures Node ^ MergeMem // (See comment in memnode.cpp near MergeMemNode::MergeMemNode for semantics.) in(AliasIdxTop) = in(1) is always the top node in(0) is NULL in(AliasIdxBot) is a "wide" memory state. For in(AliasIdxRaw) = in(3) and above, mem state for alias type <N> or top base_memory() // wide state memory_at(N) // for alias type <N> Identity: base が empty なら base を返す，さも無ければ this Ideal: Simplify stacked MergeMem 32
- 33. Compiler data structures TypeNode ^ PhiNode 異なるコントロールパスからの値をマージする。Slot 0 は control す る RegionNode 33
- 34. Compiler data structures class ConINode : public ConNode { public: ConINode( const TypeInt *t ) : ConNode(t) {} virtual int Opcode() const; // Factory method: static ConINode* make( Compile* C, int con ) { return new (C, 1) ConINode( TypeInt::make(con) ); } class ConNode : public TypeNode { public: ConNode( const Type *t ) : TypeNode(t,1) { init_req(0, (Node*)Compile::current()->root()); init_flags(Flag_is_Con); } class TypeNode : public Node { const Type* const _type; TypeNode( const Type *t, uint required ) : Node 34
- 35. Compiler data structures // Add pointer plus integer to get pointer. NOT commutative, really. // So not really an AddNode. Lives here, because people associate it with // an add. class AddPNode : public Node { public: enum { Control, // When is it safe to do this add? Base, // Base oop, for GC purposes Address, // Actually address, derived from base Offset } ; // Offset added to address AddPNode( Node *base, Node *ptr, Node *off ) : Node(0,base,ptr,off) { init_class_id(Class_AddP); } Identity: if one input is 0, return in(Address), otherwise this Ideal: 左が定数の加算であれば， expression tree を平坦化 raw pointer で NULL なら CastX2PNode(offset) 右が constant の加算なら (ptr + (offset+cn)) を (ptr + offset) + con に変更 35
- 36. Compiler data structures // Return from subroutine node class ReturnNode : public Node { public: ReturnNode( uint edges, Node *cntrl, Node *i_o, Node *memory, Node *retadr, Node *frameptr ); virtual int Opcode() const; virtual bool is_CFG() const { return true; } 36
- 37. Compiler data structures JVMState JVMState* _caller // for scope chains uint _depth, _locoff, _stkoff, _monoff, uint _scloff // offset of scalar objs uint _endoff uint _sp int _bci ReexecuteState _reexecute ciMethod* _method SafePointNode* _map 37
- 38. Compiler data structures class Type { public: enum TYPES { Bad = 0, Control, Top, Int, Long, Half, NarrowOop, Tuple, Array, AnyPtr, RawPtr, OopPtr, InstPtr, AryPtr, KlassPtr, Function, Abio, Return_Address, Memory, FloatTop, FloatCon, FloatBot, DoubleTop, DoubleCon, DoubleBot, Bottom, lasttype }; private: const Type __dual; protected: const TYPES _base; 38
- 39. Compiler data structures class Type { : public: TYPES base(); static const Type *make(enum TYPES); static int cmp(Type*, Type*); int higher_equal( Type *t) const Type *meet(Type *t); virtual const Type *widen(Type *old, Type* limit) virtual const Type *narrow(Type *old) 39
- 40. Compiler data structures class Dict; class Type; class TypeD; class TypeF; class TypeInt; class TypeLong; class TypeNarrowOop; class TypeAry; class TypeTuple; class TypePtr; class TypeRawPtr; class TypeOopPtr; class TypeInstPtr; class TypeAryPtr; class TypeKlassPtr; 40
- 41. Compiler data structures Phase PhaseTransform Compile PhaseIdealLoop GraphKit Matcher PhaseCFG PhaseValues PhaseBlockLayout PhaseGVN PhaseCoalesce PhaseIterGVN PhaseAggressiveCoalesce PhaseCCP PhaseConservativeCoalesce PhasePeephole PhaseIFG PhaseStringOpts PhaseLive PhaseMacroExpand PhaseRegAlloc PhaseChaitin PhaseRemoveUseless 41
- 42. Compiler data structures class Phase : public StackObj { public: enum PhaseNumber { Compiler, Parser,Remove_Useless, ...} protected: enum PhaseNumber _pnum; public: Compile * C; } 42
- 43. Compiler data structures class Compile : public Phase { const int _compile_id; ciMethod* _method; int _entry_bci; const TypeFunc* _tf; InlineTree* _ilt; Arena _comp_arena; ConnectionGraph* _congraph; uint _unique; Arena _node_arena; RootNode* _root; Node* _top; : } 43
- 44. Compiler data structures class Compile : public Phase { : PhaseGVN* _initial_gvn; Unique_Node_List _for_igvn; WarmCallInfo* _warm_calls; PhaseCFG* _cfg; Matcher* _matcher; PhaseRegAlloc* _regalloc; OopMapSet* _oop_map_set; : } 44
- 45. Compiler data structures class PhaseTransform : public Phase { protected: Arena* _arena; Node_Array _nodes; Type_Array _types; ConINode* _icons[...]; ConLNode* _lcons[...]; ConNode* _zcons[...]; : } 45
- 46. Compiler data structures class PhaseTransform : public Phase { public: const Type* type(const Node* n) const; const Type* type_or_null(const Node* n) const; void set_type(const Node* n, const Type* t); void set_type_bottom(const Node* n); void ensure_type_or_null(const Node* n); ConNode* makecon(const Type* t); ConINode* intcon(jint i); ConLNode* longcon(jlong l); virtual Node *transform(Node *) = 0; : } 46
- 47. Compiler data structures 値をテーブルで管理する機能 class PhaseValues : public PhaseTransform { protected: NodeHash _table; // for value-numbering public: bool hash_delete(Node *n); bool hash_insert(Node *n); Node *hash_find_insert(Node* n); Node *hash_find(Node* n); : } 47
- 48. Compiler data structures ローカルの悲観的な GVN-style の最適化 class PhaseGVN : public PhaseValues { public: Node *transform(Node *n); Node *transform_no_reclaim(Node *n); : } 48
- 49. Compiler data structures 繰り返しのローカル、悲観的 GVN-style 最適化と ideal の変形 class PhaseIterGVN : public PhaseGVN { private: bool _delay_transform; virtual Node *transform_old(Node *a_node); void subsume_node(Node *old, Node *nn); protected: virtual Node *transform(Node *a_node); void init_worklist(Node *a_root); virtual const Type* saturate(Type*, Type*, Type*) public: Unique_Node_List _worklist; void optimize(); : } 49
- 50. Parse 最初のパスでブロックを認識、２番目のパスで各 ブロックを訪れ、そのなかのバイトコードを処理 して、Node のサブクラスのオブジェクトを作っ たり、JVMState を作ったり、更新したり、最適 化したり。値の伝播がうまくいく様にブロックに 入ってくるブロックが極力 Parse されている様 にする。 50
- 51. Parse #0 Parse::do_one_bytecode() #1 Parse::do_one_block() #2 Parse::do_all_blocks() #3 Parse::Parse(JVMState*, ciMethod*, float) () #4 ParseGenerator::generate (JVMState*) #5 Compile::Compile(ciEnv*, C2Compiler*, ciMethod*, int, bool, bool) 51
- 52. Parse do_one_bytecode switch (bc()) { case Bytecodes::_nop: // do nothing break; case Bytecodes::_lconst_0: push_pair(longcon(0)); break; : case Bytecodes::_iconst_5: push(intcon( 5)); break; case Bytecodes::_bipush: push(intcon(iter ().get_constant_u1())); break; case Bytecodes::_sipush: push(intcon(iter ().get_constant_u2())); break; makecon, ingcon など定数を表すノードを返す static 関数もある。 52
- 53. Parse do_one_bytecode case Bytecodes::_ldc: case Bytecodes::_ldc_w: case Bytecodes::_ldc2_w: // If the constant is unresolved, run this BC once in the interpreter. { ciConstant constant = iter().get_constant(); if (constant.basic_type() == T_OBJECT && !constant.as_object()->is_loaded()) { int index = iter().get_constant_pool_index(); 53
- 54. Parse do_one_bytecode case Bytecodes::_aload_0: push( local(0) ); break; : case Bytecodes::_aload: push( local(iter().get_index()) ); break; push, local は結果的に JVMState, SafePointNode の状態を変 更。 iter() を使って bytecode の引き数を取って来る事ができる。 54
- 55. Parse do_one_bytecode case Bytecodes::_fstore_0: case Bytecodes::_istore_0: case Bytecodes::_astore_0: set_local( 0, pop() ); break; : case Bytecodes::_fstore: case Bytecodes::_istore: case Bytecodes::_astore: set_local( iter().get_index(), pop() ); break; 55
- 56. Parse do_one_bytecode case Bytecodes::_pop: _sp -= 1; break; case Bytecodes::_pop2: _sp -= 2; break; case Bytecodes::_swap: a = pop(); b = pop(); push(a); push(b); break; case Bytecodes::_dup: a = pop(); push(a); push(a); break; 56
- 57. Parse do_one_bytecode case Bytecodes::_baload: array_load(T_BYTE); break; case Bytecodes::_caload: array_load(T_CHAR); break; case Bytecodes::_iaload: array_load(T_INT); break; case Bytecodes::_saload: array_load(T_SHORT); break; case Bytecodes::_faload: array_load(T_FLOAT); break; case Bytecodes::_aaload: array_load(T_OBJECT); break; case Bytecodes::_laload: { a = array_addressing(T_LONG, 0); if (stopped()) return; // guaranteed null or range check _sp -= 2; // Pop array and index push_pair( make_load(control(), a, TypeLong::LONG, T_LONG, TypeAryPtr::LONGS)); break; } 57
- 58. Parse do_one_bytecode case Bytecodes::_bastore: array_store(T_BYTE); break; case Bytecodes::_castore: array_store(T_CHAR); break; case Bytecodes::_iastore: array_store(T_INT); break; case Bytecodes::_sastore: array_store(T_SHORT); break; case Bytecodes::_fastore: array_store(T_FLOAT); break; case Bytecodes::_aastore: { d = array_addressing(T_OBJECT, 1); if (stopped()) return; // guaranteed null or range check array_store_check(); c = pop(); // Oop to store b = pop(); // index (already used) a = pop(); // the array itself const TypeOopPtr* elemtype = _gvn.type(a)- >is_aryptr()->elem()->make_oopptr(); const TypeAryPtr* adr_type = TypeAryPtr::OOPS; Node* store = store_oop_to_array(control(), a, d, adr_type, c, elemtype, T_OBJECT); 58
- 59. Parse do_one_bytecode case Bytecodes::_getfield: do_getfield(); break; case Bytecodes::_getstatic: do_getstatic(); break; case Bytecodes::_putfield: do_putfield(); break; case Bytecodes::_putstatic: do_putstatic(); break; 59
- 60. Parse do_one_bytecode // implementation of _get* and _put* bytecodes void do_getstatic() { do_field_access(true, false); } void do_getfield () { do_field_access(true, true); } void do_putstatic() { do_field_access(false, false); } void do_putfield () { do_field_access(false, true); } 60
- 61. Parse do_one_bytecode Parse::do_field_access Parse::do_get_xxx(Node* obj, ciField* field, bool is_field) Node *adr = basic_plus_adr(obj, obj, offset); : Node* ld = make_load(NULL, adr, type, bt, adr_type, is_vol); Node* GraphKit::basic_plus_adr(Node* base, Node* ptr, Node* offset) { // short-circuit a common case if (offset == intcon(0)) return ptr; return _gvn.transform( new (C, 4) AddPNode(base, ptr, offset) ); } 61
- 62. Parse do_one_bytecode // factory methods in "int adr_idx" Node* GraphKit::make_load(Node* ctl, Node* adr, const Type* t, BasicType bt,int adr_idx, bool require_atomic_access) { Node* mem = memory(adr_idx); Node* ld; if (require_atomic_access && bt == T_LONG) { ld = LoadLNode::make_atomic(C, ctl, mem, adr, adr_type, t); } else { ld = LoadNode::make(_gvn, ctl, mem, adr, adr_type, t, bt); } return _gvn.transform(ld); } 62
- 63. Parse do_one_bytecode Node* GraphKit::memory(uint alias_idx) { MergeMemNode* mem = merged_memory(); Node* p = mem->memory_at(alias_idx); _gvn.set_type(p, Type::MEMORY); // must be mapped return p; } 63
- 64. Parse do_one_bytecode _iadd b = pop(), a = pop() push(_gvn.transform( new (C, 3) AddINode(a,b))) // GraphKit::pop() Node* pop() { ..; return _map->stack(_map->_jvms,--_sp); } // SefePointNode::stack Node *stack(JVMState* jvms, uint idx) const { return in(jvms->stkoff() + idx); } 64
- 65. Parse do_one_bytecode case Bytecodes::_iinc: // Increment local i = iter().get_index(); // Get local index set_local( i, _gvn.transform( new (C, 3) AddINode( _gvn.intcon(iter().get_iinc_con()), local(i) ) ) ); break; 65
- 66. Parse do_one_bytecode _goto, _goto_w int target_bci = (bc() == Bytecodes::_goto) ? iter().get_dest() : iter().get_far_dest(); // If this is a backwards branch in the bytecodes, add Safepoint maybe_add_safepoint(target_bci); // Update method data profile_taken_branch(target_bci); // Add loop predicate if it goes to a loop if (should_add_predicate(target_bci)){ add_predicate(); } // Merge the current control into the target basic block merge(target_bci); ... 66
- 67. Parse do_one_bytecode _goto, _goto_w ...// See if we can get some profile data and hand it off to the next block Block *target_block = block()->successor_for_bci (target_bci); if (target_block->pred_count() != 1) break; ciMethodData* methodData = method()->method_data(); if (!methodData->is_mature()) break; ciProfileData* data = methodData->bci_to_data(bci ()); assert( data->is_JumpData(), "" ); int taken = ((ciJumpData*)data)->taken(); taken = method()->scale_count(taken); target_block->set_count(taken); break; 67
- 68. Parse do_one_bytecode case _ifnull: btest = BoolTest::eq; goto handle_if_null; case _ifnonnull: btest = BoolTest::ne; goto handle_if_null; handle_if_null: // If this is a backwards branch in the bytecodes, add Safepoint maybe_add_safepoint(iter().get_dest()); a = null(); b = pop(); c = _gvn.transform( new (C, 3) CmpPNode(b, a) ); do_ifnull(btest, c); break; 68
- 69. Parse do_one_bytecode case _if_acmpeq: btest = BoolTest::eq; goto handle_if_acmp; case _if_acmpne: btest = BoolTest::ne; goto handle_if_acmp; handle_if_acmp: // If this is a backwards branch in the bytecodes, add Safepoint maybe_add_safepoint(iter().get_dest()); a = pop(); b = pop(); c = _gvn.transform( new (C, 3) CmpPNode(b, a) ); do_if(btest, c); break; 69
- 70. Parse do_one_bytecode case Bytecodes::_tableswitch: do_tableswitch(); break; case Bytecodes::_lookupswitch: do_lookupswitch(); break; 70
- 71. Parse do_one_bytecode case Bytecodes::_invokestatic: case Bytecodes::_invokedynamic: case Bytecodes::_invokespecial: case Bytecodes::_invokevirtual: case Bytecodes::_invokeinterface: do_call(); break; case Bytecodes::_checkcast: do_checkcast(); break; case Bytecodes::_instanceof: do_instanceof(); break; 71
- 72. Parse do_one_bytecode getClass はインライン展開され、 LoadKlass -> メモリアクセスに。 hashCode は static に public class Call { public static void main(String[] args) { Call c = new Call(); for (int i = 0; i < 100000; i++) { c.doit(); } } int doit() { return getClass().hashCode(); } } 72
- 73. Parse do_one_bytecode case Bytecodes::_anewarray: do_anewarray(); break; case Bytecodes::_newarray: do_newarray((BasicType)iter().get_index()); break; case Bytecodes::_multianewarray: do_multianewarray(); break; case Bytecodes::_new: do_new(); break; 73
- 74. Parse do_one_bytecode case Bytecodes::_jsr: case Bytecodes::_jsr_w: do_jsr(); break; case Bytecodes::_ret: do_ret(); break; 74
- 75. Parse do_one_bytecode case Bytecodes::_monitorenter: do_monitor_enter(); break; case Bytecodes::_monitorexit: do_monitor_exit(); break; 75
- 76. Optimize PhaseIterGVN igvn(initial_gvn) igvn.optimize() ConnectionGraph::do_analysis(this, &igvn) // EscapeAnalysis igvn.optimize() PhaseIdealLoop ideal_loop(igvn, true) PhaseIdealLoop ideal_loop(igvn, ...) PhaseIdealLoop ideal_loop(igvn, ...) PhaseCCP ccp( &igvn ) PhaseMacroExpand mex(igvn) 76
- 77. Optimize PhaseIterGVN igvn(initial_gvn) igvn.optimize() ConnectionGraph::do_analysis(this, &igvn) // EscapeAnalysis igvn.optimize() PhaseIdealLoop ideal_loop(igvn, true) PhaseIdealLoop ideal_loop(igvn, ...) PhaseIdealLoop ideal_loop(igvn, ...) PhaseCCP ccp( &igvn ) PhaseMacroExpand mex(igvn) 77
- 78. Optimize PhaseIterGVN igvn(initial_gvn) igvn.optimize() // worklist から取り出し， node を transform. // node が変わったら，edge 情報を // 更新して， users を worklist に置く while( _worklist.size() ) { Node *n = _worklist.pop(); if (++loop_count >= K * C->unique()) { // 範囲の確認 ...} if (n->outcnt() != 0) { Node *nn = transform_old(n); } else if (!n->is_top()) { remove_dead_node(n); } } 78
- 79. Optimize Node *PhaseIterGVN::transform_old ( Node *n ) Ideal に渡す can_reshape が true である事 Constant に計算される物は subsume_node で user を新しいノード をさす様に変更する事 79
- 80. Optimize PhaseIterGVN igvn(initial_gvn) igvn.optimize() ConnectionGraph::do_analysis(this, &igvn) // EscapeAnalysis igvn.optimize() PhaseIdealLoop ideal_loop(igvn, true) PhaseIdealLoop ideal_loop(igvn, ...) PhaseIdealLoop ideal_loop(igvn, ...) PhaseCCP ccp( &igvn ) PhaseMacroExpand mex(igvn) 80
- 81. Optimize ConnectionGraph::do_analysis(this, &igvn) // EscapeAnalysis if (congraph->compute_escape()) { // There are non escaping objects. C->set_congraph(congraph); } congraph は LockNode, UnlockNode で確認し、これらが non- escape なら処理がなくなる。local なオブジェクトにロック、アンロッ クは無意味。 81
- 82. Optimize ConnectionGraph::compute_escape() java object の allocation がなければ false を返す AddP, MergeMem 等を work list にのせる、それらの out ものせる worklist のノードを細かく調べる GrowableArray<PointsToNode> _nodes に登録して、 GlobalEscape, ArgEscape, NoEscape に分類, 到達可能なノードに 伝播する。 // comment in escape.hpp // flags: PrintEscapeAnalysis PrintEliminateAllocations 82
- 83. Optimize class ConnectionGraph: public ResourceObj // escape state of a node PointsToNode::EscapeState escape_state(Node *n); // other information we have collected bool is_scalar_replaceable(Node *n) { if (_collecting || (n->_idx >= nodes_size())) return false; PointsToNode* ptn = ptnode_adr(n->_idx); return ptn->escape_state() == PointsToNode::NoEscape && ptn->_scalar_replaceable; } 83
- 84. Optimize PhaseIterGVN igvn(initial_gvn) igvn.optimize() ConnectionGraph::do_analysis(this, &igvn) // EscapeAnalysis igvn.optimize() PhaseIdealLoop ideal_loop(igvn, true) PhaseIdealLoop ideal_loop(igvn, ...) PhaseIdealLoop ideal_loop(igvn, ...) PhaseCCP ccp( &igvn ) PhaseMacroExpand mex(igvn) 84
- 85. Optimize PhaseIdealLoop::PhaseIdealLoop (PhaseIterGVN &igvn, bool do_split_ifs) build_and_optimize(do_split_ifs); 85
- 86. Optimize // Convert to counted loops where possible PhaseIdealLoop::is_counted_loop( Node *x, IdealLoopTree *loop ) PhaseIdealLoop::is_counted_loop で CountedLoop への変換 を試みる。再帰的に子のループに関しても counted_loop を呼ぶ void PhaseIdealLoop::do_peeling( IdealLoopTree *loop, Node_List &old_new ) // １回目の実行を切り出す。loopTransform.cpp に図解 void PhaseIdealLoop::do_unroll( IdealLoopTree *loop, Node_List &old_new, bool adjust_min_trip ) void PhaseIdealLoop::do_maximally_unroll( IdealLoopTree *loop, Node_List &old_new ) // Eliminate range-checks and other trip-counter vs loop-invariant tests. void PhaseIdealLoop::do_range_check( IdealLoopTree *loop, Node_List &old_new ) 86
- 87. Optimize -- PhaseIdealLoop After Parsing static int doit() { int sm = 0; for (int i = 0; i < 100; i++) sm += i; return sm; } 87
- 88. Optimize -- PhaseIdealLoop After CountedLoop static int doit() { int sm = 0; for (int i = 0; i < 100; i++) sm += i; return sm; } 88
- 89. Optimize -- PhaseIdealLoop Optimization Finished Unrolling? 89
- 90. Optimize PhaseIterGVN igvn(initial_gvn) igvn.optimize() ConnectionGraph::do_analysis(this, &igvn) // EscapeAnalysis igvn.optimize() PhaseIdealLoop ideal_loop(igvn, true) PhaseIdealLoop ideal_loop(igvn, ...) PhaseIdealLoop ideal_loop(igvn, ...) PhaseCCP ccp( &igvn ) PhaseMacroExpand mex(igvn) 90
- 91. Optimize PhaseCCP ccp( &igvn ) ccp.do_transform C->set_root( transform(C->root())- >as_Root() ); 定数置き換え可能な物を置き換える 91
- 92. Optimize PhaseIterGVN igvn(initial_gvn) igvn.optimize() ConnectionGraph::do_analysis(this, &igvn) // EscapeAnalysis igvn.optimize() PhaseIdealLoop ideal_loop(igvn, true) PhaseIdealLoop ideal_loop(igvn, ...) PhaseIdealLoop ideal_loop(igvn, ...) PhaseCCP ccp( &igvn ) PhaseMacroExpand mex(igvn) 92
- 93. Optimize PhaseMacroExpand mex(igvn) mex.expand_macro_nodes() ... eliminate_allocate_node scalar_replacement Escape Analysis の結果の処理、allocation をスタック操作に変換？ 93
- 94. Code_Gen Matcher m(proj_list) m.match() PhaseCFG cfg(node_arena(), root()) cfg.Dominators() cfg.Estimate_Block_Frequency() cfg.GlobalCodeMotion(m,unique(),proj) PhaseChaitin regalloc(unique, cfg, m) regalloc->Register_Allocate() PaseBlockLayout PhasePeephole Output 94
- 95. Matcher #0 in addI_eRegNode::Expand(State*, Node_List&, Node*) () #1 in Matcher::ReduceInst(State*, int, Node*&) () #2 in Matcher::match_tree(Node const*) () #3 in Matcher::xform(Node*, int) () #4 in Matcher::match() () #5 in Compile::Code_Gen() () #6 in Compile::Compile(ciEnv*, C2Compiler*, ciMethod*, int, bool, bool) () 95
- 96. Matcher // x86_32.ad, an ADLC file instruct addI_eReg(eRegI dst, eRegI src, eFlagsReg cr) %{ match(Set dst (AddI dst src)); effect(KILL cr); size(2); format %{ "ADD $dst,$src" %} opcode(0x03); ... %} 96
- 97. PhaseCFG PhaseCFG::build_cfg() RegionNode, StartNode を元に CFG (Control Flow Graph) を構築。以降のマシ ンよりの操作が行える様にする。 97
- 98. PhaseCFG class PhaseCFG : public Phase + _num_blocks: uint + _blocks: RootNode* + _bbs: Block_Array + _broot: Block* + _rpo_ctr: uint + _root_loop: + _node_latency: GrowableAray<uint>* 98
- 99. PhaseCFG class Block : public CFGElement + _nodes : Node_List + _succs : Block_Array + _num_succs: uint + _pre_order: uint // Pre-order DFS # + _dom_depth: uint + _idom : Block* + _loop : CFGLoop* + _rpo : uint : // reg pressure, etc 99
- 100. PhaseCFG PhaseCFG::Dominators() // Lengauer & Tarjan algorithm // Block の _dom_depth, _idom を設定 // Code Motion の元になるデータ PhaseCFG::Estimate_Block_Frequency() // IfNode の probabilities から block // の frequency を算出, Block の親の // field _freq に設定 100
- 101. PhaseCFG PhaseCFG::GlobalCodeMotion schedule_early schedule_late 101
- 103. Output StartNode を MachPrologNode で置き換え Unverified entry point の設定 MachEpilogNode を各 return の前に配置 ScheduleAndBundle() BuildOopMap() Fill_buffer() CodeBuffer を用意 for (i=0; i < _cfg->numLblocks; i++) for Uj = 0; j < last_inst; j++) … n->emit(*cb, _regalloc) -XX:+PrintOptoAssembly to dump instructions https://gist.github.com/1376858 103
- 104. おまけ Sheet2 bytecode size vs arena use 14000000 12000000 10000000 8000000 comp bytes node res 6000000 4000000 2000000 0 0 1000 2000 3000 4000 5000 6000 7000 bytecode size ページ 1 104
- 105. Sheet2 compiler memory use 14000000 12000000 10000000 8000000 comp_arena bytes node_arena res_area 6000000 4000000 2000000 0 0 5000 10000 15000 20000 25000 30000 unique (number of nodes) ページ 1 105