[regexp] Various refactors

No functional changes.

- Removed unused Isolate* argument from regexp extrefs.
- Added const where possible.
- Removed unused functions.
- Shuffled declarations for better readability.
- ...

Change-Id: I6d9093052e8de4e33e9411541a691d0bab7b20c9
Reviewed-on: https://chromium-review.googlesource.com/c/v8/v8/+/3217193
Commit-Queue: Jakob Gruber <jgruber@chromium.org>
Auto-Submit: Jakob Gruber <jgruber@chromium.org>
Reviewed-by: Patrick Thier <pthier@chromium.org>
Cr-Commit-Position: refs/heads/main@{#77316}

src/codegen/external-reference.cc

@@ -296,12 +296,6 @@ struct IsValidExternalReferenceType<Result (Class::*)(Args...)> {
     return ExternalReference(Redirect(FUNCTION_ADDR(Target)));             \
   }
 
-#define FUNCTION_REFERENCE_WITH_ISOLATE(Name, Target)                      \
-  ExternalReference ExternalReference::Name(Isolate* isolate) {            \
-    STATIC_ASSERT(IsValidExternalReferenceType<decltype(&Target)>::value); \
-    return ExternalReference(Redirect(FUNCTION_ADDR(Target)));             \
-  }
-
 #define FUNCTION_REFERENCE_WITH_TYPE(Name, Target, Type)                   \
   ExternalReference ExternalReference::Name() {                            \
     STATIC_ASSERT(IsValidExternalReferenceType<decltype(&Target)>::value); \
@@ -719,11 +713,10 @@ ExternalReference ExternalReference::invoke_accessor_getter_callback() {
 UNREACHABLE();
 #endif
 
-FUNCTION_REFERENCE_WITH_ISOLATE(re_check_stack_guard_state, re_stack_check_func)
+FUNCTION_REFERENCE(re_check_stack_guard_state, re_stack_check_func)
 #undef re_stack_check_func
 
-FUNCTION_REFERENCE_WITH_ISOLATE(re_grow_stack,
-                                NativeRegExpMacroAssembler::GrowStack)
+FUNCTION_REFERENCE(re_grow_stack, NativeRegExpMacroAssembler::GrowStack)
 
 FUNCTION_REFERENCE(re_match_for_call_from_js,
                    IrregexpInterpreter::MatchForCallFromJs)
@@ -731,15 +724,13 @@ FUNCTION_REFERENCE(re_match_for_call_from_js,
 FUNCTION_REFERENCE(re_experimental_match_for_call_from_js,
                    ExperimentalRegExp::MatchForCallFromJs)
 
-FUNCTION_REFERENCE_WITH_ISOLATE(
-    re_case_insensitive_compare_unicode,
-    NativeRegExpMacroAssembler::CaseInsensitiveCompareUnicode)
+FUNCTION_REFERENCE(re_case_insensitive_compare_unicode,
+                   NativeRegExpMacroAssembler::CaseInsensitiveCompareUnicode)
 
-FUNCTION_REFERENCE_WITH_ISOLATE(
-    re_case_insensitive_compare_non_unicode,
-    NativeRegExpMacroAssembler::CaseInsensitiveCompareNonUnicode)
+FUNCTION_REFERENCE(re_case_insensitive_compare_non_unicode,
+                   NativeRegExpMacroAssembler::CaseInsensitiveCompareNonUnicode)
 
-ExternalReference ExternalReference::re_word_character_map(Isolate* isolate) {
+ExternalReference ExternalReference::re_word_character_map() {
   return ExternalReference(
       NativeRegExpMacroAssembler::word_character_map_address());
 }
@@ -1380,7 +1371,6 @@ void abort_with_reason(int reason) {
 }
 
 #undef FUNCTION_REFERENCE
-#undef FUNCTION_REFERENCE_WITH_ISOLATE
 #undef FUNCTION_REFERENCE_WITH_TYPE
 
 }  // namespace internal

src/codegen/external-reference.h

@@ -77,14 +77,6 @@ class StatsCounter;
   V(address_of_static_offsets_vector, "OffsetsVector::static_offsets_vector")  \
   V(thread_in_wasm_flag_address_address,                                       \
     "Isolate::thread_in_wasm_flag_address_address")                            \
-  V(re_case_insensitive_compare_unicode,                                       \
-    "NativeRegExpMacroAssembler::CaseInsensitiveCompareUnicode()")             \
-  V(re_case_insensitive_compare_non_unicode,                                   \
-    "NativeRegExpMacroAssembler::CaseInsensitiveCompareNonUnicode()")          \
-  V(re_check_stack_guard_state,                                                \
-    "RegExpMacroAssembler*::CheckStackGuardState()")                           \
-  V(re_grow_stack, "NativeRegExpMacroAssembler::GrowStack()")                  \
-  V(re_word_character_map, "NativeRegExpMacroAssembler::word_character_map")   \
   V(javascript_execution_assert, "javascript_execution_assert")                \
   EXTERNAL_REFERENCE_LIST_WITH_ISOLATE_HEAP_SANDBOX(V)
 
@@ -292,6 +284,14 @@ class StatsCounter;
           "tsan_relaxed_load_function_64_bits")                                \
   V(js_finalization_registry_remove_cell_from_unregister_token_map,            \
     "JSFinalizationRegistry::RemoveCellFromUnregisterTokenMap")                \
+  V(re_case_insensitive_compare_unicode,                                       \
+    "RegExpMacroAssembler::CaseInsensitiveCompareUnicode()")                   \
+  V(re_case_insensitive_compare_non_unicode,                                   \
+    "RegExpMacroAssembler::CaseInsensitiveCompareNonUnicode()")                \
+  V(re_check_stack_guard_state,                                                \
+    "RegExpMacroAssembler*::CheckStackGuardState()")                           \
+  V(re_grow_stack, "NativeRegExpMacroAssembler::GrowStack()")                  \
+  V(re_word_character_map, "NativeRegExpMacroAssembler::word_character_map")   \
   V(re_match_for_call_from_js, "IrregexpInterpreter::MatchForCallFromJs")      \
   V(re_experimental_match_for_call_from_js,                                    \
     "ExperimentalRegExp::MatchForCallFromJs")                                  \

src/regexp/arm/regexp-macro-assembler-arm.cc

@@ -336,10 +336,9 @@ void RegExpMacroAssemblerARM::CheckNotBackReferenceIgnoreCase(
     {
       AllowExternalCallThatCantCauseGC scope(masm_.get());
       ExternalReference function =
-          unicode ? ExternalReference::re_case_insensitive_compare_unicode(
-                        isolate())
-                  : ExternalReference::re_case_insensitive_compare_non_unicode(
-                        isolate());
+          unicode
+              ? ExternalReference::re_case_insensitive_compare_unicode()
+              : ExternalReference::re_case_insensitive_compare_non_unicode();
       __ CallCFunction(function, argument_count);
     }
 
@@ -577,7 +576,7 @@ bool RegExpMacroAssemblerARM::CheckSpecialCharacterClass(base::uc16 type,
       __ cmp(current_character(), Operand('z'));
       BranchOrBacktrack(hi, on_no_match);
     }
-    ExternalReference map = ExternalReference::re_word_character_map(isolate());
+    ExternalReference map = ExternalReference::re_word_character_map();
     __ mov(r0, Operand(map));
     __ ldrb(r0, MemOperand(r0, current_character()));
     __ cmp(r0, Operand::Zero());
@@ -591,7 +590,7 @@ bool RegExpMacroAssemblerARM::CheckSpecialCharacterClass(base::uc16 type,
       __ cmp(current_character(), Operand('z'));
       __ b(hi, &done);
     }
-    ExternalReference map = ExternalReference::re_word_character_map(isolate());
+    ExternalReference map = ExternalReference::re_word_character_map();
     __ mov(r0, Operand(map));
     __ ldrb(r0, MemOperand(r0, current_character()));
     __ cmp(r0, Operand::Zero());
@@ -931,8 +930,7 @@ Handle<HeapObject> RegExpMacroAssemblerARM::GetCode(Handle<String> source) {
     static constexpr int kNumArguments = 1;
     __ PrepareCallCFunction(kNumArguments);
     __ mov(r0, Operand(ExternalReference::isolate_address(isolate())));
-    ExternalReference grow_stack =
-        ExternalReference::re_grow_stack(isolate());
+    ExternalReference grow_stack = ExternalReference::re_grow_stack();
     __ CallCFunction(grow_stack, kNumArguments);
     // If nullptr is returned, we have failed to grow the stack, and must exit
     // with a stack-overflow exception.
@@ -1127,7 +1125,7 @@ void RegExpMacroAssemblerARM::CallCheckStackGuardState() {
   __ mov(r0, sp);
 
   ExternalReference stack_guard_check =
-      ExternalReference::re_check_stack_guard_state(isolate());
+      ExternalReference::re_check_stack_guard_state();
   __ mov(ip, Operand(stack_guard_check));
 
   EmbeddedData d = EmbeddedData::FromBlob();

src/regexp/arm64/regexp-macro-assembler-arm64.cc

@@ -425,10 +425,9 @@ void RegExpMacroAssemblerARM64::CheckNotBackReferenceIgnoreCase(
     {
       AllowExternalCallThatCantCauseGC scope(masm_.get());
       ExternalReference function =
-          unicode ? ExternalReference::re_case_insensitive_compare_unicode(
-                        isolate())
-                  : ExternalReference::re_case_insensitive_compare_non_unicode(
-                        isolate());
+          unicode
+              ? ExternalReference::re_case_insensitive_compare_unicode()
+              : ExternalReference::re_case_insensitive_compare_non_unicode();
       __ CallCFunction(function, argument_count);
     }
 
@@ -663,7 +662,7 @@ bool RegExpMacroAssemblerARM64::CheckSpecialCharacterClass(base::uc16 type,
       // Table is 256 entries, so all Latin1 characters can be tested.
       CompareAndBranchOrBacktrack(current_character(), 'z', hi, on_no_match);
     }
-    ExternalReference map = ExternalReference::re_word_character_map(isolate());
+    ExternalReference map = ExternalReference::re_word_character_map();
     __ Mov(x10, map);
     __ Ldrb(w10, MemOperand(x10, current_character(), UXTW));
     CompareAndBranchOrBacktrack(w10, 0, eq, on_no_match);
@@ -676,7 +675,7 @@ bool RegExpMacroAssemblerARM64::CheckSpecialCharacterClass(base::uc16 type,
       __ Cmp(current_character(), 'z');
       __ B(hi, &done);
     }
-    ExternalReference map = ExternalReference::re_word_character_map(isolate());
+    ExternalReference map = ExternalReference::re_word_character_map();
     __ Mov(x10, map);
     __ Ldrb(w10, MemOperand(x10, current_character(), UXTW));
     CompareAndBranchOrBacktrack(w10, 0, ne, on_no_match);
@@ -1121,8 +1120,7 @@ Handle<HeapObject> RegExpMacroAssemblerARM64::GetCode(Handle<String> source) {
     // Call GrowStack(isolate)
     static constexpr int kNumArguments = 1;
     __ Mov(x0, ExternalReference::isolate_address(isolate()));
-    __ CallCFunction(ExternalReference::re_grow_stack(isolate()),
-                     kNumArguments);
+    __ CallCFunction(ExternalReference::re_grow_stack(), kNumArguments);
     // If return nullptr, we have failed to grow the stack, and must exit with
     // a stack-overflow exception.  Returning from the regexp code restores the
     // stack (sp <- fp) so we don't need to drop the link register from it
@@ -1447,7 +1445,7 @@ void RegExpMacroAssemblerARM64::CallCheckStackGuardState(Register scratch) {
 
   DCHECK_EQ(scratch, x10);
   ExternalReference check_stack_guard_state =
-      ExternalReference::re_check_stack_guard_state(isolate());
+      ExternalReference::re_check_stack_guard_state();
   __ Mov(scratch, check_stack_guard_state);
 
   __ CallBuiltin(Builtin::kDirectCEntry);

src/regexp/ia32/regexp-macro-assembler-ia32.cc

@@ -336,10 +336,9 @@ void RegExpMacroAssemblerIA32::CheckNotBackReferenceIgnoreCase(
     {
       AllowExternalCallThatCantCauseGC scope(masm_.get());
       ExternalReference compare =
-          unicode ? ExternalReference::re_case_insensitive_compare_unicode(
-                        isolate())
-                  : ExternalReference::re_case_insensitive_compare_non_unicode(
-                        isolate());
+          unicode
+              ? ExternalReference::re_case_insensitive_compare_unicode()
+              : ExternalReference::re_case_insensitive_compare_non_unicode();
       __ CallCFunction(compare, argument_count);
     }
     // Pop original values before reacting on result value.
@@ -581,8 +580,7 @@ bool RegExpMacroAssemblerIA32::CheckSpecialCharacterClass(base::uc16 type,
       BranchOrBacktrack(above, on_no_match);
     }
     DCHECK_EQ(0, word_character_map[0]);  // Character '\0' is not a word char.
-    ExternalReference word_map =
-        ExternalReference::re_word_character_map(isolate());
+    ExternalReference word_map = ExternalReference::re_word_character_map();
     __ test_b(current_character(),
               Operand(current_character(), times_1, word_map.address(),
                       RelocInfo::EXTERNAL_REFERENCE));
@@ -597,8 +595,7 @@ bool RegExpMacroAssemblerIA32::CheckSpecialCharacterClass(base::uc16 type,
       __ j(above, &done);
     }
     DCHECK_EQ(0, word_character_map[0]);  // Character '\0' is not a word char.
-    ExternalReference word_map =
-        ExternalReference::re_word_character_map(isolate());
+    ExternalReference word_map = ExternalReference::re_word_character_map();
     __ test_b(current_character(),
               Operand(current_character(), times_1, word_map.address(),
                       RelocInfo::EXTERNAL_REFERENCE));
@@ -969,8 +966,7 @@ Handle<HeapObject> RegExpMacroAssemblerIA32::GetCode(Handle<String> source) {
     __ PrepareCallCFunction(kNumArguments, ebx);
     __ mov(Operand(esp, 0 * kSystemPointerSize),
            Immediate(ExternalReference::isolate_address(isolate())));
-    __ CallCFunction(ExternalReference::re_grow_stack(isolate()),
-                     kNumArguments);
+    __ CallCFunction(ExternalReference::re_grow_stack(), kNumArguments);
     // If return nullptr, we have failed to grow the stack, and
     // must exit with a stack-overflow exception.
     __ or_(eax, eax);
@@ -1151,7 +1147,7 @@ void RegExpMacroAssemblerIA32::CallCheckStackGuardState(Register scratch) {
   __ lea(eax, Operand(esp, -kSystemPointerSize));
   __ mov(Operand(esp, 0 * kSystemPointerSize), eax);
   ExternalReference check_stack_guard =
-      ExternalReference::re_check_stack_guard_state(isolate());
+      ExternalReference::re_check_stack_guard_state();
   __ CallCFunction(check_stack_guard, num_arguments);
 }
 

src/regexp/regexp-ast.h

@@ -45,29 +45,25 @@ class RegExpVisitor {
 #undef MAKE_CASE
 };
 
-
 // A simple closed interval.
 class Interval {
  public:
   Interval() : from_(kNone), to_(kNone - 1) {}  // '- 1' for branchless size().
   Interval(int from, int to) : from_(from), to_(to) {}
   Interval Union(Interval that) {
-    if (that.from_ == kNone)
-      return *this;
-    else if (from_ == kNone)
-      return that;
-    else
-      return Interval(std::min(from_, that.from_), std::max(to_, that.to_));
+    if (that.from_ == kNone) return *this;
+    if (from_ == kNone) return that;
+    return Interval(std::min(from_, that.from_), std::max(to_, that.to_));
   }
 
-  bool Contains(int value) { return (from_ <= value) && (value <= to_); }
-  bool is_empty() { return from_ == kNone; }
+  static Interval Empty() { return Interval(); }
+
+  bool Contains(int value) const { return (from_ <= value) && (value <= to_); }
+  bool is_empty() const { return from_ == kNone; }
   int from() const { return from_; }
   int to() const { return to_; }
   int size() const { return to_ - from_ + 1; }
 
-  static Interval Empty() { return Interval(); }
-
   static constexpr int kNone = -1;
 
  private:
@@ -79,21 +75,13 @@ class Interval {
 // to to_, both ends are inclusive.
 class CharacterRange {
  public:
-  CharacterRange() : from_(0), to_(0) {}
-  // For compatibility with the CHECK_OK macro
+  CharacterRange() = default;
+  // For compatibility with the CHECK_OK macro.
   CharacterRange(void* null) { DCHECK_NULL(null); }  // NOLINT
-  V8_EXPORT_PRIVATE static void AddClassEscape(char type,
-                                               ZoneList<CharacterRange>* ranges,
-                                               Zone* zone);
-  // Add class escapes. Add case equivalent closure for \w and \W if necessary.
-  V8_EXPORT_PRIVATE static void AddClassEscape(
-      char type, ZoneList<CharacterRange>* ranges,
-      bool add_unicode_case_equivalents, Zone* zone);
-  static base::Vector<const int> GetWordBounds();
+
   static inline CharacterRange Singleton(base::uc32 value) {
     return CharacterRange(value, value);
   }
-  static constexpr int kMaxCodePoint = 0x10ffff;
   static inline CharacterRange Range(base::uc32 from, base::uc32 to) {
     DCHECK(0 <= from && to <= kMaxCodePoint);
     DCHECK(static_cast<uint32_t>(from) <= static_cast<uint32_t>(to));
@@ -102,6 +90,7 @@ class CharacterRange {
   static inline CharacterRange Everything() {
     return CharacterRange(0, kMaxCodePoint);
   }
+
   static inline ZoneList<CharacterRange>* List(Zone* zone,
                                                CharacterRange range) {
     ZoneList<CharacterRange>* list =
@@ -109,17 +98,23 @@ class CharacterRange {
     list->Add(range, zone);
     return list;
   }
-  bool Contains(base::uc32 i) { return from_ <= i && i <= to_; }
-  base::uc32 from() const { return from_; }
-  void set_from(base::uc32 value) { from_ = value; }
-  base::uc32 to() const { return to_; }
-  void set_to(base::uc32 value) { to_ = value; }
-  bool is_valid() { return from_ <= to_; }
-  bool IsEverything(base::uc32 max) { return from_ == 0 && to_ >= max; }
-  bool IsSingleton() { return (from_ == to_); }
+
+  V8_EXPORT_PRIVATE static void AddClassEscape(char type,
+                                               ZoneList<CharacterRange>* ranges,
+                                               Zone* zone);
+  // Add class escapes. Add case equivalent closure for \w and \W if necessary.
+  V8_EXPORT_PRIVATE static void AddClassEscape(
+      char type, ZoneList<CharacterRange>* ranges,
+      bool add_unicode_case_equivalents, Zone* zone);
   V8_EXPORT_PRIVATE static void AddCaseEquivalents(
       Isolate* isolate, Zone* zone, ZoneList<CharacterRange>* ranges,
       bool is_one_byte);
+
+  bool Contains(base::uc32 i) const { return from_ <= i && i <= to_; }
+  base::uc32 from() const { return from_; }
+  base::uc32 to() const { return to_; }
+  bool IsEverything(base::uc32 max) const { return from_ == 0 && to_ >= max; }
+  bool IsSingleton() const { return from_ == to_; }
   // Whether a range list is in canonical form: Ranges ordered by from value,
   // and ranges non-overlapping and non-adjacent.
   V8_EXPORT_PRIVATE static bool IsCanonical(ZoneList<CharacterRange>* ranges);
@@ -131,71 +126,22 @@ class CharacterRange {
   // Negate the contents of a character range in canonical form.
   static void Negate(ZoneList<CharacterRange>* src,
                      ZoneList<CharacterRange>* dst, Zone* zone);
-  static const int kStartMarker = (1 << 24);
-  static const int kPayloadMask = (1 << 24) - 1;
 
  private:
   CharacterRange(base::uc32 from, base::uc32 to) : from_(from), to_(to) {}
 
-  base::uc32 from_;
-  base::uc32 to_;
-};
-
-class CharacterSet final {
- public:
-  explicit CharacterSet(base::uc16 standard_set_type)
-      : ranges_(nullptr), standard_set_type_(standard_set_type) {}
-  explicit CharacterSet(ZoneList<CharacterRange>* ranges)
-      : ranges_(ranges), standard_set_type_(0) {}
-  ZoneList<CharacterRange>* ranges(Zone* zone);
-  base::uc16 standard_set_type() const { return standard_set_type_; }
-  void set_standard_set_type(base::uc16 special_set_type) {
-    standard_set_type_ = special_set_type;
-  }
-  bool is_standard() { return standard_set_type_ != 0; }
-  V8_EXPORT_PRIVATE void Canonicalize();
-
- private:
-  ZoneList<CharacterRange>* ranges_;
-  // If non-zero, the value represents a standard set (e.g., all whitespace
-  // characters) without having to expand the ranges.
-  base::uc16 standard_set_type_;
-};
-
-class TextElement final {
- public:
-  enum TextType { ATOM, CHAR_CLASS };
-
-  static TextElement Atom(RegExpAtom* atom);
-  static TextElement CharClass(RegExpCharacterClass* char_class);
-
-  int cp_offset() const { return cp_offset_; }
-  void set_cp_offset(int cp_offset) { cp_offset_ = cp_offset; }
-  int length() const;
-
-  TextType text_type() const { return text_type_; }
-
-  RegExpTree* tree() const { return tree_; }
-
-  RegExpAtom* atom() const {
-    DCHECK(text_type() == ATOM);
-    return reinterpret_cast<RegExpAtom*>(tree());
-  }
-
-  RegExpCharacterClass* char_class() const {
-    DCHECK(text_type() == CHAR_CLASS);
-    return reinterpret_cast<RegExpCharacterClass*>(tree());
-  }
-
- private:
-  TextElement(TextType text_type, RegExpTree* tree)
-      : cp_offset_(-1), text_type_(text_type), tree_(tree) {}
+  static constexpr int kMaxCodePoint = 0x10ffff;
 
-  int cp_offset_;
-  TextType text_type_;
-  RegExpTree* tree_;
+  base::uc32 from_ = 0;
+  base::uc32 to_ = 0;
 };
 
+#define DECL_BOILERPLATE(Name)                                         \
+  void* Accept(RegExpVisitor* visitor, void* data) override;           \
+  RegExpNode* ToNode(RegExpCompiler* compiler, RegExpNode* on_success) \
+      override;                                                        \
+  RegExp##Name* As##Name() override;                                   \
+  bool Is##Name() override
 
 class RegExpTree : public ZoneObject {
  public:
@@ -225,16 +171,15 @@ class RegExpTree : public ZoneObject {
 class RegExpDisjunction final : public RegExpTree {
  public:
   explicit RegExpDisjunction(ZoneList<RegExpTree*>* alternatives);
-  void* Accept(RegExpVisitor* visitor, void* data) override;
-  RegExpNode* ToNode(RegExpCompiler* compiler, RegExpNode* on_success) override;
-  RegExpDisjunction* AsDisjunction() override;
+
+  DECL_BOILERPLATE(Disjunction);
+
   Interval CaptureRegisters() override;
-  bool IsDisjunction() override;
   bool IsAnchoredAtStart() override;
   bool IsAnchoredAtEnd() override;
   int min_match() override { return min_match_; }
   int max_match() override { return max_match_; }
-  ZoneList<RegExpTree*>* alternatives() { return alternatives_; }
+  ZoneList<RegExpTree*>* alternatives() const { return alternatives_; }
 
  private:
   bool SortConsecutiveAtoms(RegExpCompiler* compiler);
@@ -249,16 +194,15 @@ class RegExpDisjunction final : public RegExpTree {
 class RegExpAlternative final : public RegExpTree {
  public:
   explicit RegExpAlternative(ZoneList<RegExpTree*>* nodes);
-  void* Accept(RegExpVisitor* visitor, void* data) override;
-  RegExpNode* ToNode(RegExpCompiler* compiler, RegExpNode* on_success) override;
-  RegExpAlternative* AsAlternative() override;
+
+  DECL_BOILERPLATE(Alternative);
+
   Interval CaptureRegisters() override;
-  bool IsAlternative() override;
   bool IsAnchoredAtStart() override;
   bool IsAnchoredAtEnd() override;
   int min_match() override { return min_match_; }
   int max_match() override { return max_match_; }
-  ZoneList<RegExpTree*>* nodes() { return nodes_; }
+  ZoneList<RegExpTree*>* nodes() const { return nodes_; }
 
  private:
   ZoneList<RegExpTree*>* nodes_;
@@ -279,10 +223,9 @@ class RegExpAssertion final : public RegExpTree {
     LAST_TYPE = NON_BOUNDARY,
   };
   explicit RegExpAssertion(AssertionType type) : assertion_type_(type) {}
-  void* Accept(RegExpVisitor* visitor, void* data) override;
-  RegExpNode* ToNode(RegExpCompiler* compiler, RegExpNode* on_success) override;
-  RegExpAssertion* AsAssertion() override;
-  bool IsAssertion() override;
+
+  DECL_BOILERPLATE(Assertion);
+
   bool IsAnchoredAtStart() override;
   bool IsAnchoredAtEnd() override;
   int min_match() override { return 0; }
@@ -293,6 +236,27 @@ class RegExpAssertion final : public RegExpTree {
   const AssertionType assertion_type_;
 };
 
+class CharacterSet final {
+ public:
+  explicit CharacterSet(base::uc16 standard_set_type)
+      : standard_set_type_(standard_set_type) {}
+  explicit CharacterSet(ZoneList<CharacterRange>* ranges) : ranges_(ranges) {}
+
+  ZoneList<CharacterRange>* ranges(Zone* zone);
+  base::uc16 standard_set_type() const { return standard_set_type_; }
+  void set_standard_set_type(base::uc16 special_set_type) {
+    standard_set_type_ = special_set_type;
+  }
+  bool is_standard() const { return standard_set_type_ != 0; }
+  V8_EXPORT_PRIVATE void Canonicalize();
+
+ private:
+  ZoneList<CharacterRange>* ranges_ = nullptr;
+  // If non-zero, the value represents a standard set (e.g., all whitespace
+  // characters) without having to expand the ranges. See the comment on top of
+  // `standard_type` below.
+  base::uc16 standard_set_type_ = 0;
+};
 
 class RegExpCharacterClass final : public RegExpTree {
  public:
@@ -318,18 +282,18 @@ class RegExpCharacterClass final : public RegExpTree {
   }
   explicit RegExpCharacterClass(base::uc16 type)
       : set_(type), character_class_flags_(CharacterClassFlags()) {}
-  void* Accept(RegExpVisitor* visitor, void* data) override;
-  RegExpNode* ToNode(RegExpCompiler* compiler, RegExpNode* on_success) override;
-  RegExpCharacterClass* AsCharacterClass() override;
-  bool IsCharacterClass() override;
+
+  DECL_BOILERPLATE(CharacterClass);
+
   bool IsTextElement() override { return true; }
   int min_match() override { return 1; }
   // The character class may match two code units for unicode regexps.
   // TODO(yangguo): we should split this class for usage in TextElement, and
   //                make max_match() dependent on the character class content.
   int max_match() override { return 2; }
+
   void AppendToText(RegExpText* text, Zone* zone) override;
-  CharacterSet character_set() { return set_; }
+
   // TODO(lrn): Remove need for complex version if is_standard that
   // recognizes a mangled standard set and just do { return set_.is_special(); }
   bool is_standard(Zone* zone);
@@ -345,7 +309,10 @@ class RegExpCharacterClass final : public RegExpTree {
   // . : non-newline
   // * : All characters, for advancing unanchored regexp
   base::uc16 standard_type() const { return set_.standard_set_type(); }
+
+  CharacterSet character_set() const { return set_; }
   ZoneList<CharacterRange>* ranges(Zone* zone) { return set_.ranges(zone); }
+
   bool is_negated() const { return (character_class_flags_ & NEGATED) != 0; }
   bool contains_split_surrogate() const {
     return (character_class_flags_ & CONTAINS_SPLIT_SURROGATE) != 0;
@@ -360,29 +327,61 @@ class RegExpCharacterClass final : public RegExpTree {
 class RegExpAtom final : public RegExpTree {
  public:
   explicit RegExpAtom(base::Vector<const base::uc16> data) : data_(data) {}
-  void* Accept(RegExpVisitor* visitor, void* data) override;
-  RegExpNode* ToNode(RegExpCompiler* compiler, RegExpNode* on_success) override;
-  RegExpAtom* AsAtom() override;
-  bool IsAtom() override;
+
+  DECL_BOILERPLATE(Atom);
+
   bool IsTextElement() override { return true; }
   int min_match() override { return data_.length(); }
   int max_match() override { return data_.length(); }
   void AppendToText(RegExpText* text, Zone* zone) override;
-  base::Vector<const base::uc16> data() { return data_; }
-  int length() { return data_.length(); }
+
+  base::Vector<const base::uc16> data() const { return data_; }
+  int length() const { return data_.length(); }
 
  private:
   base::Vector<const base::uc16> data_;
 };
 
+class TextElement final {
+ public:
+  enum TextType { ATOM, CHAR_CLASS };
+
+  static TextElement Atom(RegExpAtom* atom);
+  static TextElement CharClass(RegExpCharacterClass* char_class);
+
+  int cp_offset() const { return cp_offset_; }
+  void set_cp_offset(int cp_offset) { cp_offset_ = cp_offset; }
+  int length() const;
+
+  TextType text_type() const { return text_type_; }
+
+  RegExpTree* tree() const { return tree_; }
+
+  RegExpAtom* atom() const {
+    DCHECK(text_type() == ATOM);
+    return reinterpret_cast<RegExpAtom*>(tree());
+  }
+
+  RegExpCharacterClass* char_class() const {
+    DCHECK(text_type() == CHAR_CLASS);
+    return reinterpret_cast<RegExpCharacterClass*>(tree());
+  }
+
+ private:
+  TextElement(TextType text_type, RegExpTree* tree)
+      : cp_offset_(-1), text_type_(text_type), tree_(tree) {}
+
+  int cp_offset_;
+  TextType text_type_;
+  RegExpTree* tree_;
+};
 
 class RegExpText final : public RegExpTree {
  public:
-  explicit RegExpText(Zone* zone) : elements_(2, zone), length_(0) {}
-  void* Accept(RegExpVisitor* visitor, void* data) override;
-  RegExpNode* ToNode(RegExpCompiler* compiler, RegExpNode* on_success) override;
-  RegExpText* AsText() override;
-  bool IsText() override;
+  explicit RegExpText(Zone* zone) : elements_(2, zone) {}
+
+  DECL_BOILERPLATE(Text);
+
   bool IsTextElement() override { return true; }
   int min_match() override { return length_; }
   int max_match() override { return length_; }
@@ -395,7 +394,7 @@ class RegExpText final : public RegExpTree {
 
  private:
   ZoneList<TextElement> elements_;
-  int length_;
+  int length_ = 0;
 };
 
 
@@ -418,23 +417,22 @@ class RegExpQuantifier final : public RegExpTree {
       max_match_ = max * body->max_match();
     }
   }
-  void* Accept(RegExpVisitor* visitor, void* data) override;
-  RegExpNode* ToNode(RegExpCompiler* compiler, RegExpNode* on_success) override;
+
+  DECL_BOILERPLATE(Quantifier);
+
   static RegExpNode* ToNode(int min, int max, bool is_greedy, RegExpTree* body,
                             RegExpCompiler* compiler, RegExpNode* on_success,
                             bool not_at_start = false);
-  RegExpQuantifier* AsQuantifier() override;
   Interval CaptureRegisters() override;
-  bool IsQuantifier() override;
   int min_match() override { return min_match_; }
   int max_match() override { return max_match_; }
   int min() const { return min_; }
   int max() const { return max_; }
   QuantifierType quantifier_type() const { return quantifier_type_; }
   bool is_possessive() const { return quantifier_type_ == POSSESSIVE; }
-  bool is_non_greedy() { return quantifier_type_ == NON_GREEDY; }
+  bool is_non_greedy() const { return quantifier_type_ == NON_GREEDY; }
   bool is_greedy() const { return quantifier_type_ == GREEDY; }
-  RegExpTree* body() { return body_; }
+  RegExpTree* body() const { return body_; }
 
  private:
   RegExpTree* body_;
@@ -454,15 +452,14 @@ class RegExpCapture final : public RegExpTree {
         min_match_(0),
         max_match_(0),
         name_(nullptr) {}
-  void* Accept(RegExpVisitor* visitor, void* data) override;
-  RegExpNode* ToNode(RegExpCompiler* compiler, RegExpNode* on_success) override;
+
+  DECL_BOILERPLATE(Capture);
+
   static RegExpNode* ToNode(RegExpTree* body, int index,
                             RegExpCompiler* compiler, RegExpNode* on_success);
-  RegExpCapture* AsCapture() override;
   bool IsAnchoredAtStart() override;
   bool IsAnchoredAtEnd() override;
   Interval CaptureRegisters() override;
-  bool IsCapture() override;
   int min_match() override { return min_match_; }
   int max_match() override { return max_match_; }
   RegExpTree* body() { return body_; }
@@ -478,11 +475,11 @@ class RegExpCapture final : public RegExpTree {
   static int EndRegister(int index) { return index * 2 + 1; }
 
  private:
-  RegExpTree* body_;
+  RegExpTree* body_ = nullptr;
   int index_;
-  int min_match_;
-  int max_match_;
-  const ZoneVector<base::uc16>* name_;
+  int min_match_ = 0;
+  int max_match_ = 0;
+  const ZoneVector<base::uc16>* name_ = nullptr;
 };
 
 class RegExpGroup final : public RegExpTree {
@@ -491,19 +488,15 @@ class RegExpGroup final : public RegExpTree {
       : body_(body),
         min_match_(body->min_match()),
         max_match_(body->max_match()) {}
-  void* Accept(RegExpVisitor* visitor, void* data) override;
-  RegExpNode* ToNode(RegExpCompiler* compiler,
-                     RegExpNode* on_success) override {
-    return body_->ToNode(compiler, on_success);
-  }
-  RegExpGroup* AsGroup() override;
+
+  DECL_BOILERPLATE(Group);
+
   bool IsAnchoredAtStart() override { return body_->IsAnchoredAtStart(); }
   bool IsAnchoredAtEnd() override { return body_->IsAnchoredAtEnd(); }
-  bool IsGroup() override;
   int min_match() override { return min_match_; }
   int max_match() override { return max_match_; }
   Interval CaptureRegisters() override { return body_->CaptureRegisters(); }
-  RegExpTree* body() { return body_; }
+  RegExpTree* body() const { return body_; }
 
  private:
   RegExpTree* body_;
@@ -523,26 +516,24 @@ class RegExpLookaround final : public RegExpTree {
         capture_from_(capture_from),
         type_(type) {}
 
-  void* Accept(RegExpVisitor* visitor, void* data) override;
-  RegExpNode* ToNode(RegExpCompiler* compiler, RegExpNode* on_success) override;
-  RegExpLookaround* AsLookaround() override;
+  DECL_BOILERPLATE(Lookaround);
+
   Interval CaptureRegisters() override;
-  bool IsLookaround() override;
   bool IsAnchoredAtStart() override;
   int min_match() override { return 0; }
   int max_match() override { return 0; }
-  RegExpTree* body() { return body_; }
-  bool is_positive() { return is_positive_; }
-  int capture_count() { return capture_count_; }
-  int capture_from() { return capture_from_; }
-  Type type() { return type_; }
+  RegExpTree* body() const { return body_; }
+  bool is_positive() const { return is_positive_; }
+  int capture_count() const { return capture_count_; }
+  int capture_from() const { return capture_from_; }
+  Type type() const { return type_; }
 
   class Builder {
    public:
     Builder(bool is_positive, RegExpNode* on_success,
             int stack_pointer_register, int position_register,
             int capture_register_count = 0, int capture_register_start = 0);
-    RegExpNode* on_match_success() { return on_match_success_; }
+    RegExpNode* on_match_success() const { return on_match_success_; }
     RegExpNode* ForMatch(RegExpNode* match);
 
    private:
@@ -564,38 +555,32 @@ class RegExpLookaround final : public RegExpTree {
 
 class RegExpBackReference final : public RegExpTree {
  public:
-  explicit RegExpBackReference(RegExpFlags flags)
-      : capture_(nullptr), name_(nullptr), flags_(flags) {}
+  explicit RegExpBackReference(RegExpFlags flags) : flags_(flags) {}
   RegExpBackReference(RegExpCapture* capture, RegExpFlags flags)
-      : capture_(capture), name_(nullptr), flags_(flags) {}
-  void* Accept(RegExpVisitor* visitor, void* data) override;
-  RegExpNode* ToNode(RegExpCompiler* compiler, RegExpNode* on_success) override;
-  RegExpBackReference* AsBackReference() override;
-  bool IsBackReference() override;
+      : capture_(capture), flags_(flags) {}
+
+  DECL_BOILERPLATE(BackReference);
+
   int min_match() override { return 0; }
   // The back reference may be recursive, e.g. /(\2)(\1)/. To avoid infinite
   // recursion, we give up. Ignorance is bliss.
   int max_match() override { return kInfinity; }
-  int index() { return capture_->index(); }
-  RegExpCapture* capture() { return capture_; }
+  int index() const { return capture_->index(); }
+  RegExpCapture* capture() const { return capture_; }
   void set_capture(RegExpCapture* capture) { capture_ = capture; }
   const ZoneVector<base::uc16>* name() const { return name_; }
   void set_name(const ZoneVector<base::uc16>* name) { name_ = name; }
 
  private:
-  RegExpCapture* capture_;
-  const ZoneVector<base::uc16>* name_;
+  RegExpCapture* capture_ = nullptr;
+  const ZoneVector<base::uc16>* name_ = nullptr;
   const RegExpFlags flags_;
 };
 
 
 class RegExpEmpty final : public RegExpTree {
  public:
-  RegExpEmpty() = default;
-  void* Accept(RegExpVisitor* visitor, void* data) override;
-  RegExpNode* ToNode(RegExpCompiler* compiler, RegExpNode* on_success) override;
-  RegExpEmpty* AsEmpty() override;
-  bool IsEmpty() override;
+  DECL_BOILERPLATE(Empty);
   int min_match() override { return 0; }
   int max_match() override { return 0; }
 };
@@ -603,4 +588,6 @@ class RegExpEmpty final : public RegExpTree {
 }  // namespace internal
 }  // namespace v8
 
+#undef DECL_BOILERPLATE
+
 #endif  // V8_REGEXP_REGEXP_AST_H_

src/regexp/regexp-bytecode-generator.h

@@ -27,7 +27,7 @@ class V8_EXPORT_PRIVATE RegExpBytecodeGenerator : public RegExpMacroAssembler {
   ~RegExpBytecodeGenerator() override;
   // The byte-code interpreter checks on each push anyway.
   int stack_limit_slack() override { return 1; }
-  bool CanReadUnaligned() override { return false; }
+  bool CanReadUnaligned() const override { return false; }
   void Bind(Label* label) override;
   void AdvanceCurrentPosition(int by) override;  // Signed cp change.
   void PopCurrentPosition() override;

src/regexp/regexp-compiler-tonode.cc

@@ -845,6 +845,11 @@ RegExpNode* RegExpEmpty::ToNode(RegExpCompiler* compiler,
   return on_success;
 }
 
+RegExpNode* RegExpGroup::ToNode(RegExpCompiler* compiler,
+                                RegExpNode* on_success) {
+  return body_->ToNode(compiler, on_success);
+}
+
 RegExpLookaround::Builder::Builder(bool is_positive, RegExpNode* on_success,
                                    int stack_pointer_register,
                                    int position_register,
@@ -1116,10 +1121,6 @@ void CharacterRange::AddClassEscape(char type, ZoneList<CharacterRange>* ranges,
   }
 }
 
-base::Vector<const int> CharacterRange::GetWordBounds() {
-  return base::Vector<const int>(kWordRanges, kWordRangeCount - 1);
-}
-
 // static
 void CharacterRange::AddCaseEquivalents(Isolate* isolate, Zone* zone,
                                         ZoneList<CharacterRange>* ranges,

src/regexp/regexp-macro-assembler-tracer.h

@@ -18,7 +18,9 @@ class RegExpMacroAssemblerTracer: public RegExpMacroAssembler {
   ~RegExpMacroAssemblerTracer() override;
   void AbortedCodeGeneration() override;
   int stack_limit_slack() override { return assembler_->stack_limit_slack(); }
-  bool CanReadUnaligned() override { return assembler_->CanReadUnaligned(); }
+  bool CanReadUnaligned() const override {
+    return assembler_->CanReadUnaligned();
+  }
   void AdvanceCurrentPosition(int by) override;    // Signed cp change.
   void AdvanceRegister(int reg, int by) override;  // r[reg] += by.
   void Backtrack() override;

src/regexp/regexp-macro-assembler.cc

@@ -28,12 +28,11 @@ RegExpMacroAssembler::RegExpMacroAssembler(Isolate* isolate, Zone* zone)
       isolate_(isolate),
       zone_(zone) {}
 
-RegExpMacroAssembler::~RegExpMacroAssembler() = default;
-
 bool RegExpMacroAssembler::has_backtrack_limit() const {
   return backtrack_limit_ != JSRegExp::kNoBacktrackLimit;
 }
 
+// static
 int RegExpMacroAssembler::CaseInsensitiveCompareNonUnicode(Address byte_offset1,
                                                            Address byte_offset2,
                                                            size_t byte_length,
@@ -62,6 +61,7 @@ int RegExpMacroAssembler::CaseInsensitiveCompareNonUnicode(Address byte_offset1,
 #endif
 }
 
+// static
 int RegExpMacroAssembler::CaseInsensitiveCompareUnicode(Address byte_offset1,
                                                         Address byte_offset2,
                                                         size_t byte_length,
@@ -135,17 +135,6 @@ void RegExpMacroAssembler::LoadCurrentCharacter(int cp_offset,
                            eats_at_least);
 }
 
-bool RegExpMacroAssembler::CheckSpecialCharacterClass(base::uc16 type,
-                                                      Label* on_no_match) {
-  return false;
-}
-
-NativeRegExpMacroAssembler::NativeRegExpMacroAssembler(Isolate* isolate,
-                                                       Zone* zone)
-    : RegExpMacroAssembler(isolate, zone) {}
-
-NativeRegExpMacroAssembler::~NativeRegExpMacroAssembler() = default;
-
 void NativeRegExpMacroAssembler::LoadCurrentCharacterImpl(
     int cp_offset, Label* on_end_of_input, bool check_bounds, int characters,
     int eats_at_least) {
@@ -164,13 +153,14 @@ void NativeRegExpMacroAssembler::LoadCurrentCharacterImpl(
   LoadCurrentCharacterUnchecked(cp_offset, characters);
 }
 
-bool NativeRegExpMacroAssembler::CanReadUnaligned() {
+bool NativeRegExpMacroAssembler::CanReadUnaligned() const {
   return FLAG_enable_regexp_unaligned_accesses && !slow_safe();
 }
 
 #ifndef COMPILING_IRREGEXP_FOR_EXTERNAL_EMBEDDER
 
 // This method may only be called after an interrupt.
+// static
 int NativeRegExpMacroAssembler::CheckStackGuardState(
     Isolate* isolate, int start_index, RegExp::CallOrigin call_origin,
     Address* return_address, Code re_code, Address* subject,
@@ -298,6 +288,15 @@ int NativeRegExpMacroAssembler::Match(Handle<JSRegExp> regexp,
                  offsets_vector_length, isolate, *regexp);
 }
 
+// static
+int NativeRegExpMacroAssembler::ExecuteForTesting(
+    String input, int start_offset, const byte* input_start,
+    const byte* input_end, int* output, int output_size, Isolate* isolate,
+    JSRegExp regexp) {
+  return Execute(input, start_offset, input_start, input_end, output,
+                 output_size, isolate, regexp);
+}
+
 // Returns a {Result} sentinel, or the number of successful matches.
 // TODO(pthier): The JSRegExp object is passed to native irregexp code to match
 // the signature of the interpreter. We should get rid of JS objects passed to
@@ -380,6 +379,7 @@ const byte NativeRegExpMacroAssembler::word_character_map[] = {
 };
 // clang-format on
 
+// static
 Address NativeRegExpMacroAssembler::GrowStack(Isolate* isolate) {
   DisallowGarbageCollection no_gc;
 

src/regexp/regexp-macro-assembler.h

@@ -13,7 +13,9 @@ namespace v8 {
 namespace internal {
 
 class ByteArray;
+class JSRegExp;
 class Label;
+class String;
 
 static const base::uc32 kLeadSurrogateStart = 0xd800;
 static const base::uc32 kLeadSurrogateEnd = 0xdbff;
@@ -22,12 +24,6 @@ static const base::uc32 kTrailSurrogateEnd = 0xdfff;
 static const base::uc32 kNonBmpStart = 0x10000;
 static const base::uc32 kNonBmpEnd = 0x10ffff;
 
-struct DisjunctDecisionRow {
-  RegExpCharacterClass cc;
-  Label* on_match;
-};
-
-
 class RegExpMacroAssembler {
  public:
   // The implementation must be able to handle at least:
@@ -42,41 +38,11 @@ class RegExpMacroAssembler {
 
   static constexpr int kUseCharactersValue = -1;
 
-#define IMPLEMENTATIONS_LIST(V) \
-  V(IA32)                       \
-  V(ARM)                        \
-  V(ARM64)                      \
-  V(MIPS)                       \
-  V(LOONG64)                    \
-  V(RISCV)                      \
-  V(S390)                       \
-  V(PPC)                        \
-  V(X64)                        \
-  V(Bytecode)
-
-  enum IrregexpImplementation {
-#define V(Name) k##Name##Implementation,
-    IMPLEMENTATIONS_LIST(V)
-#undef V
-  };
-
-  inline const char* ImplementationToString(IrregexpImplementation impl) {
-    static const char* const kNames[] = {
-#define V(Name) #Name,
-        IMPLEMENTATIONS_LIST(V)
-#undef V
-    };
-    return kNames[impl];
-  }
-#undef IMPLEMENTATIONS_LIST
+  RegExpMacroAssembler(Isolate* isolate, Zone* zone);
+  virtual ~RegExpMacroAssembler() = default;
 
-  enum StackCheckFlag {
-    kNoStackLimitCheck = false,
-    kCheckStackLimit = true
-  };
+  virtual Handle<HeapObject> GetCode(Handle<String> source) = 0;
 
-  RegExpMacroAssembler(Isolate* isolate, Zone* zone);
-  virtual ~RegExpMacroAssembler();
   // This function is called when code generation is aborted, so that
   // the assembler could clean up internal data structures.
   virtual void AbortedCodeGeneration() {}
@@ -84,7 +50,8 @@ class RegExpMacroAssembler {
   // kCheckStackLimit flag to push operations (instead of kNoStackLimitCheck)
   // at least once for every stack_limit() pushes that are executed.
   virtual int stack_limit_slack() = 0;
-  virtual bool CanReadUnaligned() = 0;
+  virtual bool CanReadUnaligned() const = 0;
+
   virtual void AdvanceCurrentPosition(int by) = 0;  // Signed cp change.
   virtual void AdvanceRegister(int reg, int by) = 0;  // r[reg] += by.
   // Continues execution from the position pushed on the top of the backtrack
@@ -140,14 +107,15 @@ class RegExpMacroAssembler {
   // character. Returns false if the type of special character class does
   // not have custom support.
   // May clobber the current loaded character.
-  virtual bool CheckSpecialCharacterClass(base::uc16 type, Label* on_no_match);
+  virtual bool CheckSpecialCharacterClass(base::uc16 type, Label* on_no_match) {
+    return false;
+  }
 
   // Control-flow integrity:
   // Define a jump target and bind a label.
   virtual void BindJumpTarget(Label* label) { Bind(label); }
 
   virtual void Fail() = 0;
-  virtual Handle<HeapObject> GetCode(Handle<String> source) = 0;
   virtual void GoTo(Label* label) = 0;
   // Check whether a register is >= a given constant and go to a label if it
   // is.  Backtracks instead if the label is nullptr.
@@ -158,7 +126,6 @@ class RegExpMacroAssembler {
   // Check whether a register is == to the current position and go to a
   // label if it is.
   virtual void IfRegisterEqPos(int reg, Label* if_eq) = 0;
-  virtual IrregexpImplementation Implementation() = 0;
   V8_EXPORT_PRIVATE void LoadCurrentCharacter(
       int cp_offset, Label* on_end_of_input, bool check_bounds = true,
       int characters = 1, int eats_at_least = kUseCharactersValue);
@@ -171,6 +138,7 @@ class RegExpMacroAssembler {
   // will go to this label. Always checks the backtrack stack limit.
   virtual void PushBacktrack(Label* label) = 0;
   virtual void PushCurrentPosition() = 0;
+  enum StackCheckFlag { kNoStackLimitCheck = false, kCheckStackLimit = true };
   virtual void PushRegister(int register_index,
                             StackCheckFlag check_stack_limit) = 0;
   virtual void ReadCurrentPositionFromRegister(int reg) = 0;
@@ -183,8 +151,41 @@ class RegExpMacroAssembler {
   virtual void ClearRegisters(int reg_from, int reg_to) = 0;
   virtual void WriteStackPointerToRegister(int reg) = 0;
 
+  // Check that we are not in the middle of a surrogate pair.
+  void CheckNotInSurrogatePair(int cp_offset, Label* on_failure);
+
+#define IMPLEMENTATIONS_LIST(V) \
+  V(IA32)                       \
+  V(ARM)                        \
+  V(ARM64)                      \
+  V(MIPS)                       \
+  V(LOONG64)                    \
+  V(RISCV)                      \
+  V(S390)                       \
+  V(PPC)                        \
+  V(X64)                        \
+  V(Bytecode)
+
+  enum IrregexpImplementation {
+#define V(Name) k##Name##Implementation,
+    IMPLEMENTATIONS_LIST(V)
+#undef V
+  };
+
+  inline const char* ImplementationToString(IrregexpImplementation impl) {
+    static const char* const kNames[] = {
+#define V(Name) #Name,
+        IMPLEMENTATIONS_LIST(V)
+#undef V
+    };
+    return kNames[impl];
+  }
+#undef IMPLEMENTATIONS_LIST
+  virtual IrregexpImplementation Implementation() = 0;
+
   // Compare two-byte strings case insensitively.
-  // Called from generated RegExp code.
+  //
+  // Called from generated code.
   static int CaseInsensitiveCompareNonUnicode(Address byte_offset1,
                                               Address byte_offset2,
                                               size_t byte_length,
@@ -194,12 +195,10 @@ class RegExpMacroAssembler {
                                            size_t byte_length,
                                            Isolate* isolate);
 
-  // Check that we are not in the middle of a surrogate pair.
-  void CheckNotInSurrogatePair(int cp_offset, Label* on_failure);
 
   // Controls the generation of large inlined constants in the code.
   void set_slow_safe(bool ssc) { slow_safe_compiler_ = ssc; }
-  bool slow_safe() { return slow_safe_compiler_; }
+  bool slow_safe() const { return slow_safe_compiler_; }
 
   // Controls after how many backtracks irregexp should abort execution.  If it
   // can fall back to the experimental engine (see `set_can_fallback`), it will
@@ -223,11 +222,11 @@ class RegExpMacroAssembler {
   // Set whether the regular expression has the global flag.  Exiting due to
   // a failure in a global regexp may still mean success overall.
   inline void set_global_mode(GlobalMode mode) { global_mode_ = mode; }
-  inline bool global() { return global_mode_ != NOT_GLOBAL; }
-  inline bool global_with_zero_length_check() {
+  inline bool global() const { return global_mode_ != NOT_GLOBAL; }
+  inline bool global_with_zero_length_check() const {
     return global_mode_ == GLOBAL || global_mode_ == GLOBAL_UNICODE;
   }
-  inline bool global_unicode() { return global_mode_ == GLOBAL_UNICODE; }
+  inline bool global_unicode() const { return global_mode_ == GLOBAL_UNICODE; }
 
   Isolate* isolate() const { return isolate_; }
   Zone* zone() const { return zone_; }
@@ -272,49 +271,62 @@ class NativeRegExpMacroAssembler: public RegExpMacroAssembler {
     SMALLEST_REGEXP_RESULT = RegExp::kInternalRegExpSmallestResult,
   };
 
-  NativeRegExpMacroAssembler(Isolate* isolate, Zone* zone);
-  ~NativeRegExpMacroAssembler() override;
-  bool CanReadUnaligned() override;
+  NativeRegExpMacroAssembler(Isolate* isolate, Zone* zone)
+      : RegExpMacroAssembler(isolate, zone) {}
+  ~NativeRegExpMacroAssembler() override = default;
 
   // Returns a {Result} sentinel, or the number of successful matches.
   static int Match(Handle<JSRegExp> regexp, Handle<String> subject,
                    int* offsets_vector, int offsets_vector_length,
                    int previous_index, Isolate* isolate);
 
+  V8_EXPORT_PRIVATE static int ExecuteForTesting(String input, int start_offset,
+                                                 const byte* input_start,
+                                                 const byte* input_end,
+                                                 int* output, int output_size,
+                                                 Isolate* isolate,
+                                                 JSRegExp regexp);
+
+  bool CanReadUnaligned() const override;
+
+  void LoadCurrentCharacterImpl(int cp_offset, Label* on_end_of_input,
+                                bool check_bounds, int characters,
+                                int eats_at_least) override;
+  // Load a number of characters at the given offset from the
+  // current position, into the current-character register.
+  virtual void LoadCurrentCharacterUnchecked(int cp_offset,
+                                             int character_count) = 0;
+
   // Called from RegExp if the backtrack stack limit is hit. Tries to expand
   // the stack. Returns the new stack-pointer if successful, or returns 0 if
   // unable to grow the stack.
   // This function must not trigger a garbage collection.
+  //
+  // Called from generated code.
   static Address GrowStack(Isolate* isolate);
 
+  // Called from generated code.
   static int CheckStackGuardState(Isolate* isolate, int start_index,
                                   RegExp::CallOrigin call_origin,
                                   Address* return_address, Code re_code,
                                   Address* subject, const byte** input_start,
                                   const byte** input_end);
 
+  static Address word_character_map_address() {
+    return reinterpret_cast<Address>(&word_character_map[0]);
+  }
+
+ protected:
   // Byte map of one byte characters with a 0xff if the character is a word
   // character (digit, letter or underscore) and 0x00 otherwise.
   // Used by generated RegExp code.
   static const byte word_character_map[256];
 
-  static Address word_character_map_address() {
-    return reinterpret_cast<Address>(&word_character_map[0]);
-  }
-
+ private:
   // Returns a {Result} sentinel, or the number of successful matches.
-  V8_EXPORT_PRIVATE static int Execute(String input, int start_offset,
-                                       const byte* input_start,
-                                       const byte* input_end, int* output,
-                                       int output_size, Isolate* isolate,
-                                       JSRegExp regexp);
-  void LoadCurrentCharacterImpl(int cp_offset, Label* on_end_of_input,
-                                bool check_bounds, int characters,
-                                int eats_at_least) override;
-  // Load a number of characters at the given offset from the
-  // current position, into the current-character register.
-  virtual void LoadCurrentCharacterUnchecked(int cp_offset,
-                                             int character_count) = 0;
+  static int Execute(String input, int start_offset, const byte* input_start,
+                     const byte* input_end, int* output, int output_size,
+                     Isolate* isolate, JSRegExp regexp);
 };
 
 }  // namespace internal

src/regexp/x64/regexp-macro-assembler-x64.cc

@@ -355,10 +355,9 @@ void RegExpMacroAssemblerX64::CheckNotBackReferenceIgnoreCase(
     {
       AllowExternalCallThatCantCauseGC scope(&masm_);
       ExternalReference compare =
-          unicode ? ExternalReference::re_case_insensitive_compare_unicode(
-                        isolate())
-                  : ExternalReference::re_case_insensitive_compare_non_unicode(
-                        isolate());
+          unicode
+              ? ExternalReference::re_case_insensitive_compare_unicode()
+              : ExternalReference::re_case_insensitive_compare_non_unicode();
       __ CallCFunction(compare, num_arguments);
     }
 
@@ -619,7 +618,7 @@ bool RegExpMacroAssemblerX64::CheckSpecialCharacterClass(base::uc16 type,
       __ cmpl(current_character(), Immediate('z'));
       BranchOrBacktrack(above, on_no_match);
     }
-    __ Move(rbx, ExternalReference::re_word_character_map(isolate()));
+    __ Move(rbx, ExternalReference::re_word_character_map());
     DCHECK_EQ(0, word_character_map[0]);  // Character '\0' is not a word char.
     __ testb(Operand(rbx, current_character(), times_1, 0),
              current_character());
@@ -633,7 +632,7 @@ bool RegExpMacroAssemblerX64::CheckSpecialCharacterClass(base::uc16 type,
       __ cmpl(current_character(), Immediate('z'));
       __ j(above, &done);
     }
-    __ Move(rbx, ExternalReference::re_word_character_map(isolate()));
+    __ Move(rbx, ExternalReference::re_word_character_map());
     DCHECK_EQ(0, word_character_map[0]);  // Character '\0' is not a word char.
     __ testb(Operand(rbx, current_character(), times_1, 0),
              current_character());
@@ -1023,8 +1022,7 @@ Handle<HeapObject> RegExpMacroAssemblerX64::GetCode(Handle<String> source) {
     __ PrepareCallCFunction(kNumArguments);
     __ LoadAddress(arg_reg_1, ExternalReference::isolate_address(isolate()));
 
-    ExternalReference grow_stack =
-        ExternalReference::re_grow_stack(isolate());
+    ExternalReference grow_stack = ExternalReference::re_grow_stack();
     __ CallCFunction(grow_stack, kNumArguments);
     // If nullptr is returned, we have failed to grow the stack, and must exit
     // with a stack-overflow exception.
@@ -1229,7 +1227,7 @@ void RegExpMacroAssemblerX64::CallCheckStackGuardState() {
   __ leaq(rdi, Operand(rsp, -kSystemPointerSize));
 #endif
   ExternalReference stack_check =
-      ExternalReference::re_check_stack_guard_state(isolate());
+      ExternalReference::re_check_stack_guard_state();
   __ CallCFunction(stack_check, num_arguments);
 }
 

test/cctest/test-regexp.cc

@@ -655,7 +655,7 @@ static ArchRegExpMacroAssembler::Result Execute(JSRegExp regexp, String input,
                                                 Address input_end,
                                                 int* captures) {
   return static_cast<NativeRegExpMacroAssembler::Result>(
-      NativeRegExpMacroAssembler::Execute(
+      NativeRegExpMacroAssembler::ExecuteForTesting(
           input, start_offset, reinterpret_cast<byte*>(input_start),
           reinterpret_cast<byte*>(input_end), captures, 0, CcTest::i_isolate(),
           regexp));