• Home
  • Line#
  • Scopes#
  • Navigate#
  • Raw
  • Download
1 /*
2  * Copyright (C) 2013 The Android Open Source Project
3  *
4  * Licensed under the Apache License, Version 2.0 (the "License");
5  * you may not use this file except in compliance with the License.
6  * You may obtain a copy of the License at
7  *
8  *      http://www.apache.org/licenses/LICENSE-2.0
9  *
10  * Unless required by applicable law or agreed to in writing, software
11  * distributed under the License is distributed on an "AS IS" BASIS,
12  * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
13  * See the License for the specific language governing permissions and
14  * limitations under the License.
15  */
16 
17 #ifndef ART_COMPILER_DEX_MIR_GRAPH_H_
18 #define ART_COMPILER_DEX_MIR_GRAPH_H_
19 
20 #include <stdint.h>
21 
22 #include "dex_file.h"
23 #include "dex_instruction.h"
24 #include "compiler_ir.h"
25 #include "invoke_type.h"
26 #include "mir_field_info.h"
27 #include "mir_method_info.h"
28 #include "utils/arena_bit_vector.h"
29 #include "utils/growable_array.h"
30 #include "utils/arena_containers.h"
31 #include "utils/scoped_arena_containers.h"
32 #include "reg_location.h"
33 #include "reg_storage.h"
34 
35 namespace art {
36 
37 class GlobalValueNumbering;
38 
39 enum InstructionAnalysisAttributePos {
40   kUninterestingOp = 0,
41   kArithmeticOp,
42   kFPOp,
43   kSingleOp,
44   kDoubleOp,
45   kIntOp,
46   kLongOp,
47   kBranchOp,
48   kInvokeOp,
49   kArrayOp,
50   kHeavyweightOp,
51   kSimpleConstOp,
52   kMoveOp,
53   kSwitch
54 };
55 
56 #define AN_NONE (1 << kUninterestingOp)
57 #define AN_MATH (1 << kArithmeticOp)
58 #define AN_FP (1 << kFPOp)
59 #define AN_LONG (1 << kLongOp)
60 #define AN_INT (1 << kIntOp)
61 #define AN_SINGLE (1 << kSingleOp)
62 #define AN_DOUBLE (1 << kDoubleOp)
63 #define AN_FLOATMATH (1 << kFPOp)
64 #define AN_BRANCH (1 << kBranchOp)
65 #define AN_INVOKE (1 << kInvokeOp)
66 #define AN_ARRAYOP (1 << kArrayOp)
67 #define AN_HEAVYWEIGHT (1 << kHeavyweightOp)
68 #define AN_SIMPLECONST (1 << kSimpleConstOp)
69 #define AN_MOVE (1 << kMoveOp)
70 #define AN_SWITCH (1 << kSwitch)
71 #define AN_COMPUTATIONAL (AN_MATH | AN_ARRAYOP | AN_MOVE | AN_SIMPLECONST)
72 
73 enum DataFlowAttributePos {
74   kUA = 0,
75   kUB,
76   kUC,
77   kAWide,
78   kBWide,
79   kCWide,
80   kDA,
81   kIsMove,
82   kSetsConst,
83   kFormat35c,
84   kFormat3rc,
85   kFormatExtended,       // Extended format for extended MIRs.
86   kNullCheckSrc0,        // Null check of uses[0].
87   kNullCheckSrc1,        // Null check of uses[1].
88   kNullCheckSrc2,        // Null check of uses[2].
89   kNullCheckOut0,        // Null check out outgoing arg0.
90   kDstNonNull,           // May assume dst is non-null.
91   kRetNonNull,           // May assume retval is non-null.
92   kNullTransferSrc0,     // Object copy src[0] -> dst.
93   kNullTransferSrcN,     // Phi null check state transfer.
94   kRangeCheckSrc1,       // Range check of uses[1].
95   kRangeCheckSrc2,       // Range check of uses[2].
96   kRangeCheckSrc3,       // Range check of uses[3].
97   kFPA,
98   kFPB,
99   kFPC,
100   kCoreA,
101   kCoreB,
102   kCoreC,
103   kRefA,
104   kRefB,
105   kRefC,
106   kUsesMethodStar,       // Implicit use of Method*.
107   kUsesIField,           // Accesses an instance field (IGET/IPUT).
108   kUsesSField,           // Accesses a static field (SGET/SPUT).
109   kDoLVN,                // Worth computing local value numbers.
110 };
111 
112 #define DF_NOP                  UINT64_C(0)
113 #define DF_UA                   (UINT64_C(1) << kUA)
114 #define DF_UB                   (UINT64_C(1) << kUB)
115 #define DF_UC                   (UINT64_C(1) << kUC)
116 #define DF_A_WIDE               (UINT64_C(1) << kAWide)
117 #define DF_B_WIDE               (UINT64_C(1) << kBWide)
118 #define DF_C_WIDE               (UINT64_C(1) << kCWide)
119 #define DF_DA                   (UINT64_C(1) << kDA)
120 #define DF_IS_MOVE              (UINT64_C(1) << kIsMove)
121 #define DF_SETS_CONST           (UINT64_C(1) << kSetsConst)
122 #define DF_FORMAT_35C           (UINT64_C(1) << kFormat35c)
123 #define DF_FORMAT_3RC           (UINT64_C(1) << kFormat3rc)
124 #define DF_FORMAT_EXTENDED      (UINT64_C(1) << kFormatExtended)
125 #define DF_NULL_CHK_0           (UINT64_C(1) << kNullCheckSrc0)
126 #define DF_NULL_CHK_1           (UINT64_C(1) << kNullCheckSrc1)
127 #define DF_NULL_CHK_2           (UINT64_C(1) << kNullCheckSrc2)
128 #define DF_NULL_CHK_OUT0        (UINT64_C(1) << kNullCheckOut0)
129 #define DF_NON_NULL_DST         (UINT64_C(1) << kDstNonNull)
130 #define DF_NON_NULL_RET         (UINT64_C(1) << kRetNonNull)
131 #define DF_NULL_TRANSFER_0      (UINT64_C(1) << kNullTransferSrc0)
132 #define DF_NULL_TRANSFER_N      (UINT64_C(1) << kNullTransferSrcN)
133 #define DF_RANGE_CHK_1          (UINT64_C(1) << kRangeCheckSrc1)
134 #define DF_RANGE_CHK_2          (UINT64_C(1) << kRangeCheckSrc2)
135 #define DF_RANGE_CHK_3          (UINT64_C(1) << kRangeCheckSrc3)
136 #define DF_FP_A                 (UINT64_C(1) << kFPA)
137 #define DF_FP_B                 (UINT64_C(1) << kFPB)
138 #define DF_FP_C                 (UINT64_C(1) << kFPC)
139 #define DF_CORE_A               (UINT64_C(1) << kCoreA)
140 #define DF_CORE_B               (UINT64_C(1) << kCoreB)
141 #define DF_CORE_C               (UINT64_C(1) << kCoreC)
142 #define DF_REF_A                (UINT64_C(1) << kRefA)
143 #define DF_REF_B                (UINT64_C(1) << kRefB)
144 #define DF_REF_C                (UINT64_C(1) << kRefC)
145 #define DF_UMS                  (UINT64_C(1) << kUsesMethodStar)
146 #define DF_IFIELD               (UINT64_C(1) << kUsesIField)
147 #define DF_SFIELD               (UINT64_C(1) << kUsesSField)
148 #define DF_LVN                  (UINT64_C(1) << kDoLVN)
149 
150 #define DF_HAS_USES             (DF_UA | DF_UB | DF_UC)
151 
152 #define DF_HAS_DEFS             (DF_DA)
153 
154 #define DF_HAS_NULL_CHKS        (DF_NULL_CHK_0 | \
155                                  DF_NULL_CHK_1 | \
156                                  DF_NULL_CHK_2 | \
157                                  DF_NULL_CHK_OUT0)
158 
159 #define DF_HAS_RANGE_CHKS       (DF_RANGE_CHK_1 | \
160                                  DF_RANGE_CHK_2 | \
161                                  DF_RANGE_CHK_3)
162 
163 #define DF_HAS_NR_CHKS          (DF_HAS_NULL_CHKS | \
164                                  DF_HAS_RANGE_CHKS)
165 
166 #define DF_A_IS_REG             (DF_UA | DF_DA)
167 #define DF_B_IS_REG             (DF_UB)
168 #define DF_C_IS_REG             (DF_UC)
169 #define DF_IS_GETTER_OR_SETTER  (DF_IS_GETTER | DF_IS_SETTER)
170 #define DF_USES_FP              (DF_FP_A | DF_FP_B | DF_FP_C)
171 #define DF_NULL_TRANSFER        (DF_NULL_TRANSFER_0 | DF_NULL_TRANSFER_N)
172 enum OatMethodAttributes {
173   kIsLeaf,            // Method is leaf.
174   kHasLoop,           // Method contains simple loop.
175 };
176 
177 #define METHOD_IS_LEAF          (1 << kIsLeaf)
178 #define METHOD_HAS_LOOP         (1 << kHasLoop)
179 
180 // Minimum field size to contain Dalvik v_reg number.
181 #define VREG_NUM_WIDTH 16
182 
183 #define INVALID_SREG (-1)
184 #define INVALID_VREG (0xFFFFU)
185 #define INVALID_OFFSET (0xDEADF00FU)
186 
187 #define MIR_IGNORE_NULL_CHECK           (1 << kMIRIgnoreNullCheck)
188 #define MIR_NULL_CHECK_ONLY             (1 << kMIRNullCheckOnly)
189 #define MIR_IGNORE_RANGE_CHECK          (1 << kMIRIgnoreRangeCheck)
190 #define MIR_RANGE_CHECK_ONLY            (1 << kMIRRangeCheckOnly)
191 #define MIR_IGNORE_CLINIT_CHECK         (1 << kMIRIgnoreClInitCheck)
192 #define MIR_INLINED                     (1 << kMIRInlined)
193 #define MIR_INLINED_PRED                (1 << kMIRInlinedPred)
194 #define MIR_CALLEE                      (1 << kMIRCallee)
195 #define MIR_IGNORE_SUSPEND_CHECK        (1 << kMIRIgnoreSuspendCheck)
196 #define MIR_DUP                         (1 << kMIRDup)
197 
198 #define BLOCK_NAME_LEN 80
199 
200 typedef uint16_t BasicBlockId;
201 static const BasicBlockId NullBasicBlockId = 0;
202 static constexpr bool kLeafOptimization = false;
203 
204 /*
205  * In general, vreg/sreg describe Dalvik registers that originated with dx.  However,
206  * it is useful to have compiler-generated temporary registers and have them treated
207  * in the same manner as dx-generated virtual registers.  This struct records the SSA
208  * name of compiler-introduced temporaries.
209  */
210 struct CompilerTemp {
211   int32_t v_reg;      // Virtual register number for temporary.
212   int32_t s_reg_low;  // SSA name for low Dalvik word.
213 };
214 
215 enum CompilerTempType {
216   kCompilerTempVR,                // A virtual register temporary.
217   kCompilerTempSpecialMethodPtr,  // Temporary that keeps track of current method pointer.
218 };
219 
220 // When debug option enabled, records effectiveness of null and range check elimination.
221 struct Checkstats {
222   int32_t null_checks;
223   int32_t null_checks_eliminated;
224   int32_t range_checks;
225   int32_t range_checks_eliminated;
226 };
227 
228 // Dataflow attributes of a basic block.
229 struct BasicBlockDataFlow {
230   ArenaBitVector* use_v;
231   ArenaBitVector* def_v;
232   ArenaBitVector* live_in_v;
233   ArenaBitVector* phi_v;
234   int32_t* vreg_to_ssa_map_exit;
235   ArenaBitVector* ending_check_v;  // For null check and class init check elimination.
236 };
237 
238 /*
239  * Normalized use/def for a MIR operation using SSA names rather than vregs.  Note that
240  * uses/defs retain the Dalvik convention that long operations operate on a pair of 32-bit
241  * vregs.  For example, "ADD_LONG v0, v2, v3" would have 2 defs (v0/v1) and 4 uses (v2/v3, v4/v5).
242  * Following SSA renaming, this is the primary struct used by code generators to locate
243  * operand and result registers.  This is a somewhat confusing and unhelpful convention that
244  * we may want to revisit in the future.
245  *
246  * TODO:
247  *  1. Add accessors for uses/defs and make data private
248  *  2. Change fp_use/fp_def to a bit array (could help memory usage)
249  *  3. Combine array storage into internal array and handled via accessors from 1.
250  */
251 struct SSARepresentation {
252   int32_t* uses;
253   bool* fp_use;
254   int32_t* defs;
255   bool* fp_def;
256   int16_t num_uses_allocated;
257   int16_t num_defs_allocated;
258   int16_t num_uses;
259   int16_t num_defs;
260 
261   static uint32_t GetStartUseIndex(Instruction::Code opcode);
262 };
263 
264 /*
265  * The Midlevel Intermediate Representation node, which may be largely considered a
266  * wrapper around a Dalvik byte code.
267  */
268 struct MIR {
269   /*
270    * TODO: remove embedded DecodedInstruction to save space, keeping only opcode.  Recover
271    * additional fields on as-needed basis.  Question: how to support MIR Pseudo-ops; probably
272    * need to carry aux data pointer.
273    */
274   struct DecodedInstruction {
275     uint32_t vA;
276     uint32_t vB;
277     uint64_t vB_wide;        /* for k51l */
278     uint32_t vC;
279     uint32_t arg[5];         /* vC/D/E/F/G in invoke or filled-new-array */
280     Instruction::Code opcode;
281 
DecodedInstructionMIR::DecodedInstruction282     explicit DecodedInstruction():vA(0), vB(0), vB_wide(0), vC(0), opcode(Instruction::NOP) {
283     }
284 
285     /*
286      * Given a decoded instruction representing a const bytecode, it updates
287      * the out arguments with proper values as dictated by the constant bytecode.
288      */
289     bool GetConstant(int64_t* ptr_value, bool* wide) const;
290 
IsPseudoMirOpMIR::DecodedInstruction291     static bool IsPseudoMirOp(Instruction::Code opcode) {
292       return static_cast<int>(opcode) >= static_cast<int>(kMirOpFirst);
293     }
294 
IsPseudoMirOpMIR::DecodedInstruction295     static bool IsPseudoMirOp(int opcode) {
296       return opcode >= static_cast<int>(kMirOpFirst);
297     }
298 
IsInvokeMIR::DecodedInstruction299     bool IsInvoke() const {
300       return !IsPseudoMirOp(opcode) && ((Instruction::FlagsOf(opcode) & Instruction::kInvoke) == Instruction::kInvoke);
301     }
302 
IsStoreMIR::DecodedInstruction303     bool IsStore() const {
304       return !IsPseudoMirOp(opcode) && ((Instruction::FlagsOf(opcode) & Instruction::kStore) == Instruction::kStore);
305     }
306 
IsLoadMIR::DecodedInstruction307     bool IsLoad() const {
308       return !IsPseudoMirOp(opcode) && ((Instruction::FlagsOf(opcode) & Instruction::kLoad) == Instruction::kLoad);
309     }
310 
IsConditionalBranchMIR::DecodedInstruction311     bool IsConditionalBranch() const {
312       return !IsPseudoMirOp(opcode) && (Instruction::FlagsOf(opcode) == (Instruction::kContinue | Instruction::kBranch));
313     }
314 
315     /**
316      * @brief Is the register C component of the decoded instruction a constant?
317      */
IsCFieldOrConstantMIR::DecodedInstruction318     bool IsCFieldOrConstant() const {
319       return !IsPseudoMirOp(opcode) && ((Instruction::FlagsOf(opcode) & Instruction::kRegCFieldOrConstant) == Instruction::kRegCFieldOrConstant);
320     }
321 
322     /**
323      * @brief Is the register C component of the decoded instruction a constant?
324      */
IsBFieldOrConstantMIR::DecodedInstruction325     bool IsBFieldOrConstant() const {
326       return !IsPseudoMirOp(opcode) && ((Instruction::FlagsOf(opcode) & Instruction::kRegBFieldOrConstant) == Instruction::kRegBFieldOrConstant);
327     }
328 
IsCastMIR::DecodedInstruction329     bool IsCast() const {
330       return !IsPseudoMirOp(opcode) && ((Instruction::FlagsOf(opcode) & Instruction::kCast) == Instruction::kCast);
331     }
332 
333     /**
334      * @brief Does the instruction clobber memory?
335      * @details Clobber means that the instruction changes the memory not in a punctual way.
336      *          Therefore any supposition on memory aliasing or memory contents should be disregarded
337      *            when crossing such an instruction.
338      */
ClobbersMIR::DecodedInstruction339     bool Clobbers() const {
340       return !IsPseudoMirOp(opcode) && ((Instruction::FlagsOf(opcode) & Instruction::kClobber) == Instruction::kClobber);
341     }
342 
IsLinearMIR::DecodedInstruction343     bool IsLinear() const {
344       return !IsPseudoMirOp(opcode) && (Instruction::FlagsOf(opcode) & (Instruction::kAdd | Instruction::kSubtract)) != 0;
345     }
346   } dalvikInsn;
347 
348   NarrowDexOffset offset;         // Offset of the instruction in code units.
349   uint16_t optimization_flags;
350   int16_t m_unit_index;           // From which method was this MIR included
351   BasicBlockId bb;
352   MIR* next;
353   SSARepresentation* ssa_rep;
354   union {
355     // Incoming edges for phi node.
356     BasicBlockId* phi_incoming;
357     // Establish link from check instruction (kMirOpCheck) to the actual throwing instruction.
358     MIR* throw_insn;
359     // Branch condition for fused cmp or select.
360     ConditionCode ccode;
361     // IGET/IPUT lowering info index, points to MIRGraph::ifield_lowering_infos_. Due to limit on
362     // the number of code points (64K) and size of IGET/IPUT insn (2), this will never exceed 32K.
363     uint32_t ifield_lowering_info;
364     // SGET/SPUT lowering info index, points to MIRGraph::sfield_lowering_infos_. Due to limit on
365     // the number of code points (64K) and size of SGET/SPUT insn (2), this will never exceed 32K.
366     uint32_t sfield_lowering_info;
367     // INVOKE data index, points to MIRGraph::method_lowering_infos_.
368     uint32_t method_lowering_info;
369   } meta;
370 
MIRMIR371   explicit MIR():offset(0), optimization_flags(0), m_unit_index(0), bb(NullBasicBlockId),
372                  next(nullptr), ssa_rep(nullptr) {
373     memset(&meta, 0, sizeof(meta));
374   }
375 
GetStartUseIndexMIR376   uint32_t GetStartUseIndex() const {
377     return SSARepresentation::GetStartUseIndex(dalvikInsn.opcode);
378   }
379 
380   MIR* Copy(CompilationUnit *c_unit);
381   MIR* Copy(MIRGraph* mir_Graph);
382 
newMIR383   static void* operator new(size_t size, ArenaAllocator* arena) {
384     return arena->Alloc(sizeof(MIR), kArenaAllocMIR);
385   }
deleteMIR386   static void operator delete(void* p) {}  // Nop.
387 };
388 
389 struct SuccessorBlockInfo;
390 
391 struct BasicBlock {
392   BasicBlockId id;
393   BasicBlockId dfs_id;
394   NarrowDexOffset start_offset;     // Offset in code units.
395   BasicBlockId fall_through;
396   BasicBlockId taken;
397   BasicBlockId i_dom;               // Immediate dominator.
398   uint16_t nesting_depth;
399   BBType block_type:4;
400   BlockListType successor_block_list_type:4;
401   bool visited:1;
402   bool hidden:1;
403   bool catch_entry:1;
404   bool explicit_throw:1;
405   bool conditional_branch:1;
406   bool terminated_by_return:1;  // Block ends with a Dalvik return opcode.
407   bool dominates_return:1;      // Is a member of return extended basic block.
408   bool use_lvn:1;               // Run local value numbering on this block.
409   MIR* first_mir_insn;
410   MIR* last_mir_insn;
411   BasicBlockDataFlow* data_flow_info;
412   ArenaBitVector* dominators;
413   ArenaBitVector* i_dominated;      // Set nodes being immediately dominated.
414   ArenaBitVector* dom_frontier;     // Dominance frontier.
415   GrowableArray<BasicBlockId>* predecessors;
416   GrowableArray<SuccessorBlockInfo*>* successor_blocks;
417 
418   void AppendMIR(MIR* mir);
419   void AppendMIRList(MIR* first_list_mir, MIR* last_list_mir);
420   void AppendMIRList(const std::vector<MIR*>& insns);
421   void PrependMIR(MIR* mir);
422   void PrependMIRList(MIR* first_list_mir, MIR* last_list_mir);
423   void PrependMIRList(const std::vector<MIR*>& to_add);
424   void InsertMIRAfter(MIR* current_mir, MIR* new_mir);
425   void InsertMIRListAfter(MIR* insert_after, MIR* first_list_mir, MIR* last_list_mir);
426   MIR* FindPreviousMIR(MIR* mir);
427   void InsertMIRBefore(MIR* insert_before, MIR* list);
428   void InsertMIRListBefore(MIR* insert_before, MIR* first_list_mir, MIR* last_list_mir);
429   bool RemoveMIR(MIR* mir);
430   bool RemoveMIRList(MIR* first_list_mir, MIR* last_list_mir);
431 
432   BasicBlock* Copy(CompilationUnit* c_unit);
433   BasicBlock* Copy(MIRGraph* mir_graph);
434 
435   /**
436    * @brief Reset the optimization_flags field of each MIR.
437    */
438   void ResetOptimizationFlags(uint16_t reset_flags);
439 
440   /**
441    * @brief Hide the BasicBlock.
442    * @details Set it to kDalvikByteCode, set hidden to true, remove all MIRs,
443    *          remove itself from any predecessor edges, remove itself from any
444    *          child's predecessor growable array.
445    */
446   void Hide(CompilationUnit* c_unit);
447 
448   /**
449    * @brief Is ssa_reg the last SSA definition of that VR in the block?
450    */
451   bool IsSSALiveOut(const CompilationUnit* c_unit, int ssa_reg);
452 
453   /**
454    * @brief Replace the edge going to old_bb to now go towards new_bb.
455    */
456   bool ReplaceChild(BasicBlockId old_bb, BasicBlockId new_bb);
457 
458   /**
459    * @brief Update the predecessor growable array from old_pred to new_pred.
460    */
461   void UpdatePredecessor(BasicBlockId old_pred, BasicBlockId new_pred);
462 
463   /**
464    * @brief Used to obtain the next MIR that follows unconditionally.
465    * @details The implementation does not guarantee that a MIR does not
466    * follow even if this method returns nullptr.
467    * @param mir_graph the MIRGraph.
468    * @param current The MIR for which to find an unconditional follower.
469    * @return Returns the following MIR if one can be found.
470    */
471   MIR* GetNextUnconditionalMir(MIRGraph* mir_graph, MIR* current);
472   bool IsExceptionBlock() const;
473 
newBasicBlock474   static void* operator new(size_t size, ArenaAllocator* arena) {
475     return arena->Alloc(sizeof(BasicBlock), kArenaAllocBB);
476   }
deleteBasicBlock477   static void operator delete(void* p) {}  // Nop.
478 };
479 
480 /*
481  * The "blocks" field in "successor_block_list" points to an array of elements with the type
482  * "SuccessorBlockInfo".  For catch blocks, key is type index for the exception.  For switch
483  * blocks, key is the case value.
484  */
485 struct SuccessorBlockInfo {
486   BasicBlockId block;
487   int key;
488 };
489 
490 /**
491  * @class ChildBlockIterator
492  * @brief Enable an easy iteration of the children.
493  */
494 class ChildBlockIterator {
495  public:
496   /**
497    * @brief Constructs a child iterator.
498    * @param bb The basic whose children we need to iterate through.
499    * @param mir_graph The MIRGraph used to get the basic block during iteration.
500    */
501   ChildBlockIterator(BasicBlock* bb, MIRGraph* mir_graph);
502   BasicBlock* Next();
503 
504  private:
505   BasicBlock* basic_block_;
506   MIRGraph* mir_graph_;
507   bool visited_fallthrough_;
508   bool visited_taken_;
509   bool have_successors_;
510   GrowableArray<SuccessorBlockInfo*>::Iterator successor_iter_;
511 };
512 
513 /*
514  * Collection of information describing an invoke, and the destination of
515  * the subsequent MOVE_RESULT (if applicable).  Collected as a unit to enable
516  * more efficient invoke code generation.
517  */
518 struct CallInfo {
519   int num_arg_words;    // Note: word count, not arg count.
520   RegLocation* args;    // One for each word of arguments.
521   RegLocation result;   // Eventual target of MOVE_RESULT.
522   int opt_flags;
523   InvokeType type;
524   uint32_t dex_idx;
525   uint32_t index;       // Method idx for invokes, type idx for FilledNewArray.
526   uintptr_t direct_code;
527   uintptr_t direct_method;
528   RegLocation target;    // Target of following move_result.
529   bool skip_this;
530   bool is_range;
531   DexOffset offset;      // Offset in code units.
532   MIR* mir;
533 };
534 
535 
536 const RegLocation bad_loc = {kLocDalvikFrame, 0, 0, 0, 0, 0, 0, 0, 0, RegStorage(), INVALID_SREG,
537                              INVALID_SREG};
538 
539 class MIRGraph {
540  public:
541   MIRGraph(CompilationUnit* cu, ArenaAllocator* arena);
542   ~MIRGraph();
543 
544   /*
545    * Examine the graph to determine whether it's worthwile to spend the time compiling
546    * this method.
547    */
548   bool SkipCompilation(std::string* skip_message);
549 
550   /*
551    * Should we skip the compilation of this method based on its name?
552    */
553   bool SkipCompilationByName(const std::string& methodname);
554 
555   /*
556    * Parse dex method and add MIR at current insert point.  Returns id (which is
557    * actually the index of the method in the m_units_ array).
558    */
559   void InlineMethod(const DexFile::CodeItem* code_item, uint32_t access_flags,
560                     InvokeType invoke_type, uint16_t class_def_idx,
561                     uint32_t method_idx, jobject class_loader, const DexFile& dex_file);
562 
563   /* Find existing block */
FindBlock(DexOffset code_offset)564   BasicBlock* FindBlock(DexOffset code_offset) {
565     return FindBlock(code_offset, false, false, NULL);
566   }
567 
GetCurrentInsns()568   const uint16_t* GetCurrentInsns() const {
569     return current_code_item_->insns_;
570   }
571 
GetInsns(int m_unit_index)572   const uint16_t* GetInsns(int m_unit_index) const {
573     return m_units_[m_unit_index]->GetCodeItem()->insns_;
574   }
575 
GetNumBlocks()576   unsigned int GetNumBlocks() const {
577     return num_blocks_;
578   }
579 
GetNumDalvikInsns()580   size_t GetNumDalvikInsns() const {
581     return cu_->code_item->insns_size_in_code_units_;
582   }
583 
GetTryBlockAddr()584   ArenaBitVector* GetTryBlockAddr() const {
585     return try_block_addr_;
586   }
587 
GetEntryBlock()588   BasicBlock* GetEntryBlock() const {
589     return entry_block_;
590   }
591 
GetExitBlock()592   BasicBlock* GetExitBlock() const {
593     return exit_block_;
594   }
595 
GetBasicBlock(unsigned int block_id)596   BasicBlock* GetBasicBlock(unsigned int block_id) const {
597     return (block_id == NullBasicBlockId) ? NULL : block_list_.Get(block_id);
598   }
599 
GetBasicBlockListCount()600   size_t GetBasicBlockListCount() const {
601     return block_list_.Size();
602   }
603 
GetBlockList()604   GrowableArray<BasicBlock*>* GetBlockList() {
605     return &block_list_;
606   }
607 
GetDfsOrder()608   GrowableArray<BasicBlockId>* GetDfsOrder() {
609     return dfs_order_;
610   }
611 
GetDfsPostOrder()612   GrowableArray<BasicBlockId>* GetDfsPostOrder() {
613     return dfs_post_order_;
614   }
615 
GetDomPostOrder()616   GrowableArray<BasicBlockId>* GetDomPostOrder() {
617     return dom_post_order_traversal_;
618   }
619 
GetDefCount()620   int GetDefCount() const {
621     return def_count_;
622   }
623 
GetArena()624   ArenaAllocator* GetArena() {
625     return arena_;
626   }
627 
EnableOpcodeCounting()628   void EnableOpcodeCounting() {
629     opcode_count_ = static_cast<int*>(arena_->Alloc(kNumPackedOpcodes * sizeof(int),
630                                                     kArenaAllocMisc));
631   }
632 
633   void ShowOpcodeStats();
634 
GetCurrentDexCompilationUnit()635   DexCompilationUnit* GetCurrentDexCompilationUnit() const {
636     return m_units_[current_method_];
637   }
638 
639   /**
640    * @brief Dump a CFG into a dot file format.
641    * @param dir_prefix the directory the file will be created in.
642    * @param all_blocks does the dumper use all the basic blocks or use the reachable blocks.
643    * @param suffix does the filename require a suffix or not (default = nullptr).
644    */
645   void DumpCFG(const char* dir_prefix, bool all_blocks, const char* suffix = nullptr);
646 
HasFieldAccess()647   bool HasFieldAccess() const {
648     return (merged_df_flags_ & (DF_IFIELD | DF_SFIELD)) != 0u;
649   }
650 
HasStaticFieldAccess()651   bool HasStaticFieldAccess() const {
652     return (merged_df_flags_ & DF_SFIELD) != 0u;
653   }
654 
HasInvokes()655   bool HasInvokes() const {
656     // NOTE: These formats include the rare filled-new-array/range.
657     return (merged_df_flags_ & (DF_FORMAT_35C | DF_FORMAT_3RC)) != 0u;
658   }
659 
660   void DoCacheFieldLoweringInfo();
661 
GetIFieldLoweringInfo(MIR * mir)662   const MirIFieldLoweringInfo& GetIFieldLoweringInfo(MIR* mir) const {
663     DCHECK_LT(mir->meta.ifield_lowering_info, ifield_lowering_infos_.Size());
664     return ifield_lowering_infos_.GetRawStorage()[mir->meta.ifield_lowering_info];
665   }
666 
GetSFieldLoweringInfo(MIR * mir)667   const MirSFieldLoweringInfo& GetSFieldLoweringInfo(MIR* mir) const {
668     DCHECK_LT(mir->meta.sfield_lowering_info, sfield_lowering_infos_.Size());
669     return sfield_lowering_infos_.GetRawStorage()[mir->meta.sfield_lowering_info];
670   }
671 
672   void DoCacheMethodLoweringInfo();
673 
GetMethodLoweringInfo(MIR * mir)674   const MirMethodLoweringInfo& GetMethodLoweringInfo(MIR* mir) {
675     DCHECK_LT(mir->meta.method_lowering_info, method_lowering_infos_.Size());
676     return method_lowering_infos_.GetRawStorage()[mir->meta.method_lowering_info];
677   }
678 
679   void ComputeInlineIFieldLoweringInfo(uint16_t field_idx, MIR* invoke, MIR* iget_or_iput);
680 
681   void InitRegLocations();
682 
683   void RemapRegLocations();
684 
685   void DumpRegLocTable(RegLocation* table, int count);
686 
687   void BasicBlockOptimization();
688 
GetTopologicalSortOrder()689   GrowableArray<BasicBlockId>* GetTopologicalSortOrder() {
690     DCHECK(topological_order_ != nullptr);
691     return topological_order_;
692   }
693 
GetTopologicalSortOrderLoopEnds()694   GrowableArray<BasicBlockId>* GetTopologicalSortOrderLoopEnds() {
695     DCHECK(topological_order_loop_ends_ != nullptr);
696     return topological_order_loop_ends_;
697   }
698 
GetTopologicalSortOrderIndexes()699   GrowableArray<BasicBlockId>* GetTopologicalSortOrderIndexes() {
700     DCHECK(topological_order_indexes_ != nullptr);
701     return topological_order_indexes_;
702   }
703 
GetTopologicalSortOrderLoopHeadStack()704   GrowableArray<std::pair<uint16_t, bool>>* GetTopologicalSortOrderLoopHeadStack() {
705     DCHECK(topological_order_loop_head_stack_ != nullptr);
706     return topological_order_loop_head_stack_;
707   }
708 
IsConst(int32_t s_reg)709   bool IsConst(int32_t s_reg) const {
710     return is_constant_v_->IsBitSet(s_reg);
711   }
712 
IsConst(RegLocation loc)713   bool IsConst(RegLocation loc) const {
714     return loc.orig_sreg < 0 ? false : IsConst(loc.orig_sreg);
715   }
716 
ConstantValue(RegLocation loc)717   int32_t ConstantValue(RegLocation loc) const {
718     DCHECK(IsConst(loc));
719     return constant_values_[loc.orig_sreg];
720   }
721 
ConstantValue(int32_t s_reg)722   int32_t ConstantValue(int32_t s_reg) const {
723     DCHECK(IsConst(s_reg));
724     return constant_values_[s_reg];
725   }
726 
ConstantValueWide(RegLocation loc)727   int64_t ConstantValueWide(RegLocation loc) const {
728     DCHECK(IsConst(loc));
729     DCHECK(!loc.high_word);  // Do not allow asking for the high partner.
730     DCHECK_LT(loc.orig_sreg + 1, GetNumSSARegs());
731     return (static_cast<int64_t>(constant_values_[loc.orig_sreg + 1]) << 32) |
732         Low32Bits(static_cast<int64_t>(constant_values_[loc.orig_sreg]));
733   }
734 
IsConstantNullRef(RegLocation loc)735   bool IsConstantNullRef(RegLocation loc) const {
736     return loc.ref && loc.is_const && (ConstantValue(loc) == 0);
737   }
738 
GetNumSSARegs()739   int GetNumSSARegs() const {
740     return num_ssa_regs_;
741   }
742 
SetNumSSARegs(int new_num)743   void SetNumSSARegs(int new_num) {
744      /*
745       * TODO: It's theoretically possible to exceed 32767, though any cases which did
746       * would be filtered out with current settings.  When orig_sreg field is removed
747       * from RegLocation, expand s_reg_low to handle all possible cases and remove DCHECK().
748       */
749     CHECK_EQ(new_num, static_cast<int16_t>(new_num));
750     num_ssa_regs_ = new_num;
751   }
752 
GetNumReachableBlocks()753   unsigned int GetNumReachableBlocks() const {
754     return num_reachable_blocks_;
755   }
756 
GetUseCount(int vreg)757   int GetUseCount(int vreg) const {
758     return use_counts_.Get(vreg);
759   }
760 
GetRawUseCount(int vreg)761   int GetRawUseCount(int vreg) const {
762     return raw_use_counts_.Get(vreg);
763   }
764 
GetSSASubscript(int ssa_reg)765   int GetSSASubscript(int ssa_reg) const {
766     return ssa_subscripts_->Get(ssa_reg);
767   }
768 
GetRawSrc(MIR * mir,int num)769   RegLocation GetRawSrc(MIR* mir, int num) {
770     DCHECK(num < mir->ssa_rep->num_uses);
771     RegLocation res = reg_location_[mir->ssa_rep->uses[num]];
772     return res;
773   }
774 
GetRawDest(MIR * mir)775   RegLocation GetRawDest(MIR* mir) {
776     DCHECK_GT(mir->ssa_rep->num_defs, 0);
777     RegLocation res = reg_location_[mir->ssa_rep->defs[0]];
778     return res;
779   }
780 
GetDest(MIR * mir)781   RegLocation GetDest(MIR* mir) {
782     RegLocation res = GetRawDest(mir);
783     DCHECK(!res.wide);
784     return res;
785   }
786 
GetSrc(MIR * mir,int num)787   RegLocation GetSrc(MIR* mir, int num) {
788     RegLocation res = GetRawSrc(mir, num);
789     DCHECK(!res.wide);
790     return res;
791   }
792 
GetDestWide(MIR * mir)793   RegLocation GetDestWide(MIR* mir) {
794     RegLocation res = GetRawDest(mir);
795     DCHECK(res.wide);
796     return res;
797   }
798 
GetSrcWide(MIR * mir,int low)799   RegLocation GetSrcWide(MIR* mir, int low) {
800     RegLocation res = GetRawSrc(mir, low);
801     DCHECK(res.wide);
802     return res;
803   }
804 
GetBadLoc()805   RegLocation GetBadLoc() {
806     return bad_loc;
807   }
808 
GetMethodSReg()809   int GetMethodSReg() const {
810     return method_sreg_;
811   }
812 
813   /**
814    * @brief Used to obtain the number of compiler temporaries being used.
815    * @return Returns the number of compiler temporaries.
816    */
GetNumUsedCompilerTemps()817   size_t GetNumUsedCompilerTemps() const {
818     size_t total_num_temps = compiler_temps_.Size();
819     DCHECK_LE(num_non_special_compiler_temps_, total_num_temps);
820     return total_num_temps;
821   }
822 
823   /**
824    * @brief Used to obtain the number of non-special compiler temporaries being used.
825    * @return Returns the number of non-special compiler temporaries.
826    */
GetNumNonSpecialCompilerTemps()827   size_t GetNumNonSpecialCompilerTemps() const {
828     return num_non_special_compiler_temps_;
829   }
830 
831   /**
832    * @brief Used to set the total number of available non-special compiler temporaries.
833    * @details Can fail setting the new max if there are more temps being used than the new_max.
834    * @param new_max The new maximum number of non-special compiler temporaries.
835    * @return Returns true if the max was set and false if failed to set.
836    */
SetMaxAvailableNonSpecialCompilerTemps(size_t new_max)837   bool SetMaxAvailableNonSpecialCompilerTemps(size_t new_max) {
838     if (new_max < GetNumNonSpecialCompilerTemps()) {
839       return false;
840     } else {
841       max_available_non_special_compiler_temps_ = new_max;
842       return true;
843     }
844   }
845 
846   /**
847    * @brief Provides the number of non-special compiler temps available.
848    * @details Even if this returns zero, special compiler temps are guaranteed to be available.
849    * @return Returns the number of available temps.
850    */
851   size_t GetNumAvailableNonSpecialCompilerTemps();
852 
853   /**
854    * @brief Used to obtain an existing compiler temporary.
855    * @param index The index of the temporary which must be strictly less than the
856    * number of temporaries.
857    * @return Returns the temporary that was asked for.
858    */
GetCompilerTemp(size_t index)859   CompilerTemp* GetCompilerTemp(size_t index) const {
860     return compiler_temps_.Get(index);
861   }
862 
863   /**
864    * @brief Used to obtain the maximum number of compiler temporaries that can be requested.
865    * @return Returns the maximum number of compiler temporaries, whether used or not.
866    */
GetMaxPossibleCompilerTemps()867   size_t GetMaxPossibleCompilerTemps() const {
868     return max_available_special_compiler_temps_ + max_available_non_special_compiler_temps_;
869   }
870 
871   /**
872    * @brief Used to obtain a new unique compiler temporary.
873    * @param ct_type Type of compiler temporary requested.
874    * @param wide Whether we should allocate a wide temporary.
875    * @return Returns the newly created compiler temporary.
876    */
877   CompilerTemp* GetNewCompilerTemp(CompilerTempType ct_type, bool wide);
878 
MethodIsLeaf()879   bool MethodIsLeaf() {
880     return attributes_ & METHOD_IS_LEAF;
881   }
882 
GetRegLocation(int index)883   RegLocation GetRegLocation(int index) {
884     DCHECK((index >= 0) && (index < num_ssa_regs_));
885     return reg_location_[index];
886   }
887 
GetMethodLoc()888   RegLocation GetMethodLoc() {
889     return reg_location_[method_sreg_];
890   }
891 
IsBackedge(BasicBlock * branch_bb,BasicBlockId target_bb_id)892   bool IsBackedge(BasicBlock* branch_bb, BasicBlockId target_bb_id) {
893     return ((target_bb_id != NullBasicBlockId) &&
894             (GetBasicBlock(target_bb_id)->start_offset <= branch_bb->start_offset));
895   }
896 
IsBackwardsBranch(BasicBlock * branch_bb)897   bool IsBackwardsBranch(BasicBlock* branch_bb) {
898     return IsBackedge(branch_bb, branch_bb->taken) || IsBackedge(branch_bb, branch_bb->fall_through);
899   }
900 
CountBranch(DexOffset target_offset)901   void CountBranch(DexOffset target_offset) {
902     if (target_offset <= current_offset_) {
903       backward_branches_++;
904     } else {
905       forward_branches_++;
906     }
907   }
908 
GetBranchCount()909   int GetBranchCount() {
910     return backward_branches_ + forward_branches_;
911   }
912 
913   // Is this vreg in the in set?
IsInVReg(int vreg)914   bool IsInVReg(int vreg) {
915     return (vreg >= cu_->num_regs);
916   }
917 
918   void DumpCheckStats();
919   MIR* FindMoveResult(BasicBlock* bb, MIR* mir);
920   int SRegToVReg(int ssa_reg) const;
921   void VerifyDataflow();
922   void CheckForDominanceFrontier(BasicBlock* dom_bb, const BasicBlock* succ_bb);
923   void EliminateNullChecksAndInferTypesStart();
924   bool EliminateNullChecksAndInferTypes(BasicBlock* bb);
925   void EliminateNullChecksAndInferTypesEnd();
926   bool EliminateClassInitChecksGate();
927   bool EliminateClassInitChecks(BasicBlock* bb);
928   void EliminateClassInitChecksEnd();
929   bool ApplyGlobalValueNumberingGate();
930   bool ApplyGlobalValueNumbering(BasicBlock* bb);
931   void ApplyGlobalValueNumberingEnd();
932   /*
933    * Type inference handling helpers.  Because Dalvik's bytecode is not fully typed,
934    * we have to do some work to figure out the sreg type.  For some operations it is
935    * clear based on the opcode (i.e. ADD_FLOAT v0, v1, v2), but for others (MOVE), we
936    * may never know the "real" type.
937    *
938    * We perform the type inference operation by using an iterative  walk over
939    * the graph, propagating types "defined" by typed opcodes to uses and defs in
940    * non-typed opcodes (such as MOVE).  The Setxx(index) helpers are used to set defined
941    * types on typed opcodes (such as ADD_INT).  The Setxx(index, is_xx) form is used to
942    * propagate types through non-typed opcodes such as PHI and MOVE.  The is_xx flag
943    * tells whether our guess of the type is based on a previously typed definition.
944    * If so, the defined type takes precedence.  Note that it's possible to have the same sreg
945    * show multiple defined types because dx treats constants as untyped bit patterns.
946    * The return value of the Setxx() helpers says whether or not the Setxx() action changed
947    * the current guess, and is used to know when to terminate the iterative walk.
948    */
949   bool SetFp(int index, bool is_fp);
950   bool SetFp(int index);
951   bool SetCore(int index, bool is_core);
952   bool SetCore(int index);
953   bool SetRef(int index, bool is_ref);
954   bool SetRef(int index);
955   bool SetWide(int index, bool is_wide);
956   bool SetWide(int index);
957   bool SetHigh(int index, bool is_high);
958   bool SetHigh(int index);
959 
PuntToInterpreter()960   bool PuntToInterpreter() {
961     return punt_to_interpreter_;
962   }
963 
SetPuntToInterpreter(bool val)964   void SetPuntToInterpreter(bool val) {
965     punt_to_interpreter_ = val;
966   }
967 
968   char* GetDalvikDisassembly(const MIR* mir);
969   void ReplaceSpecialChars(std::string& str);
970   std::string GetSSAName(int ssa_reg);
971   std::string GetSSANameWithConst(int ssa_reg, bool singles_only);
972   void GetBlockName(BasicBlock* bb, char* name);
973   const char* GetShortyFromTargetIdx(int);
974   void DumpMIRGraph();
975   CallInfo* NewMemCallInfo(BasicBlock* bb, MIR* mir, InvokeType type, bool is_range);
976   BasicBlock* NewMemBB(BBType block_type, int block_id);
977   MIR* NewMIR();
978   MIR* AdvanceMIR(BasicBlock** p_bb, MIR* mir);
979   BasicBlock* NextDominatedBlock(BasicBlock* bb);
980   bool LayoutBlocks(BasicBlock* bb);
981   void ComputeTopologicalSortOrder();
982   BasicBlock* CreateNewBB(BBType block_type);
983 
984   bool InlineSpecialMethodsGate();
985   void InlineSpecialMethodsStart();
986   void InlineSpecialMethods(BasicBlock* bb);
987   void InlineSpecialMethodsEnd();
988 
989   /**
990    * @brief Perform the initial preparation for the Method Uses.
991    */
992   void InitializeMethodUses();
993 
994   /**
995    * @brief Perform the initial preparation for the Constant Propagation.
996    */
997   void InitializeConstantPropagation();
998 
999   /**
1000    * @brief Perform the initial preparation for the SSA Transformation.
1001    */
1002   void SSATransformationStart();
1003 
1004   /**
1005    * @brief Insert a the operands for the Phi nodes.
1006    * @param bb the considered BasicBlock.
1007    * @return true
1008    */
1009   bool InsertPhiNodeOperands(BasicBlock* bb);
1010 
1011   /**
1012    * @brief Perform the cleanup after the SSA Transformation.
1013    */
1014   void SSATransformationEnd();
1015 
1016   /**
1017    * @brief Perform constant propagation on a BasicBlock.
1018    * @param bb the considered BasicBlock.
1019    */
1020   void DoConstantPropagation(BasicBlock* bb);
1021 
1022   /**
1023    * @brief Count the uses in the BasicBlock
1024    * @param bb the BasicBlock
1025    */
1026   void CountUses(struct BasicBlock* bb);
1027 
1028   static uint64_t GetDataFlowAttributes(Instruction::Code opcode);
1029   static uint64_t GetDataFlowAttributes(MIR* mir);
1030 
1031   /**
1032    * @brief Combine BasicBlocks
1033    * @param the BasicBlock we are considering
1034    */
1035   void CombineBlocks(BasicBlock* bb);
1036 
1037   void ClearAllVisitedFlags();
1038 
1039   void AllocateSSAUseData(MIR *mir, int num_uses);
1040   void AllocateSSADefData(MIR *mir, int num_defs);
1041   void CalculateBasicBlockInformation();
1042   void InitializeBasicBlockData();
1043   void ComputeDFSOrders();
1044   void ComputeDefBlockMatrix();
1045   void ComputeDominators();
1046   void CompilerInitializeSSAConversion();
1047   void InsertPhiNodes();
1048   void DoDFSPreOrderSSARename(BasicBlock* block);
1049 
1050   /*
1051    * IsDebugBuild sanity check: keep track of the Dex PCs for catch entries so that later on
1052    * we can verify that all catch entries have native PC entries.
1053    */
1054   std::set<uint32_t> catches_;
1055 
1056   // TODO: make these private.
1057   RegLocation* reg_location_;                               // Map SSA names to location.
1058   ArenaSafeMap<unsigned int, unsigned int> block_id_map_;   // Block collapse lookup cache.
1059 
1060   static const char* extended_mir_op_names_[kMirOpLast - kMirOpFirst];
1061   static const uint32_t analysis_attributes_[kMirOpLast];
1062 
1063   void HandleSSADef(int* defs, int dalvik_reg, int reg_index);
1064   bool InferTypeAndSize(BasicBlock* bb, MIR* mir, bool changed);
1065 
1066   // Used for removing redudant suspend tests
AppendGenSuspendTestList(BasicBlock * bb)1067   void AppendGenSuspendTestList(BasicBlock* bb) {
1068     if (gen_suspend_test_list_.Size() == 0 ||
1069         gen_suspend_test_list_.Get(gen_suspend_test_list_.Size() - 1) != bb) {
1070       gen_suspend_test_list_.Insert(bb);
1071     }
1072   }
1073 
1074   /* This is used to check if there is already a method call dominating the
1075    * source basic block of a backedge and being dominated by the target basic
1076    * block of the backedge.
1077    */
1078   bool HasSuspendTestBetween(BasicBlock* source, BasicBlockId target_id);
1079 
1080  protected:
1081   int FindCommonParent(int block1, int block2);
1082   void ComputeSuccLineIn(ArenaBitVector* dest, const ArenaBitVector* src1,
1083                          const ArenaBitVector* src2);
1084   void HandleLiveInUse(ArenaBitVector* use_v, ArenaBitVector* def_v,
1085                        ArenaBitVector* live_in_v, int dalvik_reg_id);
1086   void HandleDef(ArenaBitVector* def_v, int dalvik_reg_id);
1087   void HandleExtended(ArenaBitVector* use_v, ArenaBitVector* def_v,
1088                       ArenaBitVector* live_in_v,
1089                       const MIR::DecodedInstruction& d_insn);
1090   bool DoSSAConversion(BasicBlock* bb);
1091   bool InvokeUsesMethodStar(MIR* mir);
1092   int ParseInsn(const uint16_t* code_ptr, MIR::DecodedInstruction* decoded_instruction);
1093   bool ContentIsInsn(const uint16_t* code_ptr);
1094   BasicBlock* SplitBlock(DexOffset code_offset, BasicBlock* orig_block,
1095                          BasicBlock** immed_pred_block_p);
1096   BasicBlock* FindBlock(DexOffset code_offset, bool split, bool create,
1097                         BasicBlock** immed_pred_block_p);
1098   void ProcessTryCatchBlocks();
1099   bool IsBadMonitorExitCatch(NarrowDexOffset monitor_exit_offset, NarrowDexOffset catch_offset);
1100   BasicBlock* ProcessCanBranch(BasicBlock* cur_block, MIR* insn, DexOffset cur_offset, int width,
1101                                int flags, const uint16_t* code_ptr, const uint16_t* code_end);
1102   BasicBlock* ProcessCanSwitch(BasicBlock* cur_block, MIR* insn, DexOffset cur_offset, int width,
1103                                int flags);
1104   BasicBlock* ProcessCanThrow(BasicBlock* cur_block, MIR* insn, DexOffset cur_offset, int width,
1105                               int flags, ArenaBitVector* try_block_addr, const uint16_t* code_ptr,
1106                               const uint16_t* code_end);
1107   int AddNewSReg(int v_reg);
1108   void HandleSSAUse(int* uses, int dalvik_reg, int reg_index);
1109   void DataFlowSSAFormat35C(MIR* mir);
1110   void DataFlowSSAFormat3RC(MIR* mir);
1111   void DataFlowSSAFormatExtended(MIR* mir);
1112   bool FindLocalLiveIn(BasicBlock* bb);
1113   bool VerifyPredInfo(BasicBlock* bb);
1114   BasicBlock* NeedsVisit(BasicBlock* bb);
1115   BasicBlock* NextUnvisitedSuccessor(BasicBlock* bb);
1116   void MarkPreOrder(BasicBlock* bb);
1117   void RecordDFSOrders(BasicBlock* bb);
1118   void ComputeDomPostOrderTraversal(BasicBlock* bb);
1119   void SetConstant(int32_t ssa_reg, int value);
1120   void SetConstantWide(int ssa_reg, int64_t value);
1121   int GetSSAUseCount(int s_reg);
1122   bool BasicBlockOpt(BasicBlock* bb);
1123   bool BuildExtendedBBList(struct BasicBlock* bb);
1124   bool FillDefBlockMatrix(BasicBlock* bb);
1125   void InitializeDominationInfo(BasicBlock* bb);
1126   bool ComputeblockIDom(BasicBlock* bb);
1127   bool ComputeBlockDominators(BasicBlock* bb);
1128   bool SetDominators(BasicBlock* bb);
1129   bool ComputeBlockLiveIns(BasicBlock* bb);
1130   bool ComputeDominanceFrontier(BasicBlock* bb);
1131 
1132   void CountChecks(BasicBlock* bb);
1133   void AnalyzeBlock(BasicBlock* bb, struct MethodStats* stats);
1134   bool ComputeSkipCompilation(struct MethodStats* stats, bool skip_default,
1135                               std::string* skip_message);
1136 
1137   CompilationUnit* const cu_;
1138   GrowableArray<int>* ssa_base_vregs_;
1139   GrowableArray<int>* ssa_subscripts_;
1140   // Map original Dalvik virtual reg i to the current SSA name.
1141   int* vreg_to_ssa_map_;            // length == method->registers_size
1142   int* ssa_last_defs_;              // length == method->registers_size
1143   ArenaBitVector* is_constant_v_;   // length == num_ssa_reg
1144   int* constant_values_;            // length == num_ssa_reg
1145   // Use counts of ssa names.
1146   GrowableArray<uint32_t> use_counts_;      // Weighted by nesting depth
1147   GrowableArray<uint32_t> raw_use_counts_;  // Not weighted
1148   unsigned int num_reachable_blocks_;
1149   unsigned int max_num_reachable_blocks_;
1150   GrowableArray<BasicBlockId>* dfs_order_;
1151   GrowableArray<BasicBlockId>* dfs_post_order_;
1152   GrowableArray<BasicBlockId>* dom_post_order_traversal_;
1153   GrowableArray<BasicBlockId>* topological_order_;
1154   // Indexes in topological_order_ need to be only as big as the BasicBlockId.
1155   COMPILE_ASSERT(sizeof(BasicBlockId) == sizeof(uint16_t), assuming_16_bit_BasicBlockId);
1156   // For each loop head, remember the past-the-end index of the end of the loop. 0 if not loop head.
1157   GrowableArray<uint16_t>* topological_order_loop_ends_;
1158   // Map BB ids to topological_order_ indexes. 0xffff if not included (hidden or null block).
1159   GrowableArray<uint16_t>* topological_order_indexes_;
1160   // Stack of the loop head indexes and recalculation flags for RepeatingTopologicalSortIterator.
1161   GrowableArray<std::pair<uint16_t, bool>>* topological_order_loop_head_stack_;
1162   int* i_dom_list_;
1163   ArenaBitVector** def_block_matrix_;    // num_dalvik_register x num_blocks.
1164   std::unique_ptr<ScopedArenaAllocator> temp_scoped_alloc_;
1165   uint16_t* temp_insn_data_;
1166   uint32_t temp_bit_vector_size_;
1167   ArenaBitVector* temp_bit_vector_;
1168   std::unique_ptr<GlobalValueNumbering> temp_gvn_;
1169   static const int kInvalidEntry = -1;
1170   GrowableArray<BasicBlock*> block_list_;
1171   ArenaBitVector* try_block_addr_;
1172   BasicBlock* entry_block_;
1173   BasicBlock* exit_block_;
1174   unsigned int num_blocks_;
1175   const DexFile::CodeItem* current_code_item_;
1176   GrowableArray<uint16_t> dex_pc_to_block_map_;  // FindBlock lookup cache.
1177   ArenaVector<DexCompilationUnit*> m_units_;     // List of methods included in this graph
1178   typedef std::pair<int, int> MIRLocation;       // Insert point, (m_unit_ index, offset)
1179   ArenaVector<MIRLocation> method_stack_;        // Include stack
1180   int current_method_;
1181   DexOffset current_offset_;                     // Offset in code units
1182   int def_count_;                                // Used to estimate size of ssa name storage.
1183   int* opcode_count_;                            // Dex opcode coverage stats.
1184   int num_ssa_regs_;                             // Number of names following SSA transformation.
1185   ArenaVector<BasicBlockId> extended_basic_blocks_;  // Heads of block "traces".
1186   int method_sreg_;
1187   unsigned int attributes_;
1188   Checkstats* checkstats_;
1189   ArenaAllocator* arena_;
1190   int backward_branches_;
1191   int forward_branches_;
1192   GrowableArray<CompilerTemp*> compiler_temps_;
1193   size_t num_non_special_compiler_temps_;
1194   size_t max_available_non_special_compiler_temps_;
1195   size_t max_available_special_compiler_temps_;
1196   bool punt_to_interpreter_;                    // Difficult or not worthwhile - just interpret.
1197   uint64_t merged_df_flags_;
1198   GrowableArray<MirIFieldLoweringInfo> ifield_lowering_infos_;
1199   GrowableArray<MirSFieldLoweringInfo> sfield_lowering_infos_;
1200   GrowableArray<MirMethodLoweringInfo> method_lowering_infos_;
1201   static const uint64_t oat_data_flow_attributes_[kMirOpLast];
1202   GrowableArray<BasicBlock*> gen_suspend_test_list_;  // List of blocks containing suspend tests
1203 
1204   friend class ClassInitCheckEliminationTest;
1205   friend class GlobalValueNumberingTest;
1206   friend class LocalValueNumberingTest;
1207   friend class TopologicalSortOrderTest;
1208 };
1209 
1210 }  // namespace art
1211 
1212 #endif  // ART_COMPILER_DEX_MIR_GRAPH_H_
1213