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; 281 282 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; 290 291 static bool IsPseudoMirOp(Instruction::Code opcode) { 292 return static_cast<int>(opcode) >= static_cast<int>(kMirOpFirst); 293 } 294 295 static bool IsPseudoMirOp(int opcode) { 296 return opcode >= static_cast<int>(kMirOpFirst); 297 } 298 299 bool IsInvoke() const { 300 return !IsPseudoMirOp(opcode) && ((Instruction::FlagsOf(opcode) & Instruction::kInvoke) == Instruction::kInvoke); 301 } 302 303 bool IsStore() const { 304 return !IsPseudoMirOp(opcode) && ((Instruction::FlagsOf(opcode) & Instruction::kStore) == Instruction::kStore); 305 } 306 307 bool IsLoad() const { 308 return !IsPseudoMirOp(opcode) && ((Instruction::FlagsOf(opcode) & Instruction::kLoad) == Instruction::kLoad); 309 } 310 311 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 */ 318 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 */ 325 bool IsBFieldOrConstant() const { 326 return !IsPseudoMirOp(opcode) && ((Instruction::FlagsOf(opcode) & Instruction::kRegBFieldOrConstant) == Instruction::kRegBFieldOrConstant); 327 } 328 329 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 */ 339 bool Clobbers() const { 340 return !IsPseudoMirOp(opcode) && ((Instruction::FlagsOf(opcode) & Instruction::kClobber) == Instruction::kClobber); 341 } 342 343 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; 370 371 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 } 375 376 uint32_t GetStartUseIndex() const { 377 return SSARepresentation::GetStartUseIndex(dalvikInsn.opcode); 378 } 379 380 MIR* Copy(CompilationUnit *c_unit); 381 MIR* Copy(MIRGraph* mir_Graph); 382 383 static void* operator new(size_t size, ArenaAllocator* arena) { 384 return arena->Alloc(sizeof(MIR), kArenaAllocMIR); 385 } 386 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; 473 474 static void* operator new(size_t size, ArenaAllocator* arena) { 475 return arena->Alloc(sizeof(BasicBlock), kArenaAllocBB); 476 } 477 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 */ 564 BasicBlock* FindBlock(DexOffset code_offset) { 565 return FindBlock(code_offset, false, false, NULL); 566 } 567 568 const uint16_t* GetCurrentInsns() const { 569 return current_code_item_->insns_; 570 } 571 572 const uint16_t* GetInsns(int m_unit_index) const { 573 return m_units_[m_unit_index]->GetCodeItem()->insns_; 574 } 575 576 unsigned int GetNumBlocks() const { 577 return num_blocks_; 578 } 579 580 size_t GetNumDalvikInsns() const { 581 return cu_->code_item->insns_size_in_code_units_; 582 } 583 584 ArenaBitVector* GetTryBlockAddr() const { 585 return try_block_addr_; 586 } 587 588 BasicBlock* GetEntryBlock() const { 589 return entry_block_; 590 } 591 592 BasicBlock* GetExitBlock() const { 593 return exit_block_; 594 } 595 596 BasicBlock* GetBasicBlock(unsigned int block_id) const { 597 return (block_id == NullBasicBlockId) ? NULL : block_list_.Get(block_id); 598 } 599 600 size_t GetBasicBlockListCount() const { 601 return block_list_.Size(); 602 } 603 604 GrowableArray<BasicBlock*>* GetBlockList() { 605 return &block_list_; 606 } 607 608 GrowableArray<BasicBlockId>* GetDfsOrder() { 609 return dfs_order_; 610 } 611 612 GrowableArray<BasicBlockId>* GetDfsPostOrder() { 613 return dfs_post_order_; 614 } 615 616 GrowableArray<BasicBlockId>* GetDomPostOrder() { 617 return dom_post_order_traversal_; 618 } 619 620 int GetDefCount() const { 621 return def_count_; 622 } 623 624 ArenaAllocator* GetArena() { 625 return arena_; 626 } 627 628 void EnableOpcodeCounting() { 629 opcode_count_ = static_cast<int*>(arena_->Alloc(kNumPackedOpcodes * sizeof(int), 630 kArenaAllocMisc)); 631 } 632 633 void ShowOpcodeStats(); 634 635 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); 646 647 bool HasFieldAccess() const { 648 return (merged_df_flags_ & (DF_IFIELD | DF_SFIELD)) != 0u; 649 } 650 651 bool HasStaticFieldAccess() const { 652 return (merged_df_flags_ & DF_SFIELD) != 0u; 653 } 654 655 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(); 661 662 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 } 666 667 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(); 673 674 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(); 688 689 GrowableArray<BasicBlockId>* GetTopologicalSortOrder() { 690 DCHECK(topological_order_ != nullptr); 691 return topological_order_; 692 } 693 694 GrowableArray<BasicBlockId>* GetTopologicalSortOrderLoopEnds() { 695 DCHECK(topological_order_loop_ends_ != nullptr); 696 return topological_order_loop_ends_; 697 } 698 699 GrowableArray<BasicBlockId>* GetTopologicalSortOrderIndexes() { 700 DCHECK(topological_order_indexes_ != nullptr); 701 return topological_order_indexes_; 702 } 703 704 GrowableArray<std::pair<uint16_t, bool>>* GetTopologicalSortOrderLoopHeadStack() { 705 DCHECK(topological_order_loop_head_stack_ != nullptr); 706 return topological_order_loop_head_stack_; 707 } 708 709 bool IsConst(int32_t s_reg) const { 710 return is_constant_v_->IsBitSet(s_reg); 711 } 712 713 bool IsConst(RegLocation loc) const { 714 return loc.orig_sreg < 0 ? false : IsConst(loc.orig_sreg); 715 } 716 717 int32_t ConstantValue(RegLocation loc) const { 718 DCHECK(IsConst(loc)); 719 return constant_values_[loc.orig_sreg]; 720 } 721 722 int32_t ConstantValue(int32_t s_reg) const { 723 DCHECK(IsConst(s_reg)); 724 return constant_values_[s_reg]; 725 } 726 727 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 } 734 735 bool IsConstantNullRef(RegLocation loc) const { 736 return loc.ref && loc.is_const && (ConstantValue(loc) == 0); 737 } 738 739 int GetNumSSARegs() const { 740 return num_ssa_regs_; 741 } 742 743 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 } 752 753 unsigned int GetNumReachableBlocks() const { 754 return num_reachable_blocks_; 755 } 756 757 int GetUseCount(int vreg) const { 758 return use_counts_.Get(vreg); 759 } 760 761 int GetRawUseCount(int vreg) const { 762 return raw_use_counts_.Get(vreg); 763 } 764 765 int GetSSASubscript(int ssa_reg) const { 766 return ssa_subscripts_->Get(ssa_reg); 767 } 768 769 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 } 774 775 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 } 780 781 RegLocation GetDest(MIR* mir) { 782 RegLocation res = GetRawDest(mir); 783 DCHECK(!res.wide); 784 return res; 785 } 786 787 RegLocation GetSrc(MIR* mir, int num) { 788 RegLocation res = GetRawSrc(mir, num); 789 DCHECK(!res.wide); 790 return res; 791 } 792 793 RegLocation GetDestWide(MIR* mir) { 794 RegLocation res = GetRawDest(mir); 795 DCHECK(res.wide); 796 return res; 797 } 798 799 RegLocation GetSrcWide(MIR* mir, int low) { 800 RegLocation res = GetRawSrc(mir, low); 801 DCHECK(res.wide); 802 return res; 803 } 804 805 RegLocation GetBadLoc() { 806 return bad_loc; 807 } 808 809 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 */ 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 */ 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 */ 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 */ 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 */ 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); 878 879 bool MethodIsLeaf() { 880 return attributes_ & METHOD_IS_LEAF; 881 } 882 883 RegLocation GetRegLocation(int index) { 884 DCHECK((index >= 0) && (index < num_ssa_regs_)); 885 return reg_location_[index]; 886 } 887 888 RegLocation GetMethodLoc() { 889 return reg_location_[method_sreg_]; 890 } 891 892 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 } 896 897 bool IsBackwardsBranch(BasicBlock* branch_bb) { 898 return IsBackedge(branch_bb, branch_bb->taken) || IsBackedge(branch_bb, branch_bb->fall_through); 899 } 900 901 void CountBranch(DexOffset target_offset) { 902 if (target_offset <= current_offset_) { 903 backward_branches_++; 904 } else { 905 forward_branches_++; 906 } 907 } 908 909 int GetBranchCount() { 910 return backward_branches_ + forward_branches_; 911 } 912 913 // Is this vreg in the in set? 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); 959 960 bool PuntToInterpreter() { 961 return punt_to_interpreter_; 962 } 963 964 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 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