No longer desugar the exponentiation (**) operator.

Prior to this change, the exponentiation operator was rewritten by the
parser to a call of the Math.pow builtin. However, Math.pow does not
accept BigInt arguments, while the exponentiation operator must accept
them.

This CL
- removes the parser's special treatment of ** and **=, treating them
  like any other binary op instead.
- adds a TFC builtin Exponentiate that does the right thing for
  all inputs.
- adds interpreter bytecodes Exp and ExpSmi whose handlers call the
  Exponentiate builtin. For simplicity, they currently always collect
  kAny feedback.
- adds a Turbofan operator JSExponentiate with a typed-lowering to
  the existing NumberPow and a generic-lowering to the Exponentiate
  builtin. There is currently no speculative lowering.

Note that exponentiation for BigInts is actually not implemented yet,
so we can't yet test it.

Bug: v8:6791
Change-Id: Id90914c9c3fce310ce01e715c09eaa9f294f4f8a
Reviewed-on: https://chromium-review.googlesource.com/785694Reviewed-by: Jakob Kummerow <jkummerow@chromium.org>
Reviewed-by: Sathya Gunasekaran <gsathya@chromium.org>
Reviewed-by: Yang Guo <yangguo@chromium.org>
Reviewed-by: Mythri Alle <mythria@chromium.org>
Reviewed-by: Jaroslav Sevcik <jarin@chromium.org>
Commit-Queue: Georg Neis <neis@chromium.org>
Cr-Commit-Position: refs/heads/master@{#49696}

BUILD.gn

@@ -1022,6 +1022,7 @@ v8_source_set("v8_initializers") {
     "src/builtins/builtins-iterator-gen.cc",
     "src/builtins/builtins-iterator-gen.h",
     "src/builtins/builtins-math-gen.cc",
+    "src/builtins/builtins-math-gen.h",
     "src/builtins/builtins-number-gen.cc",
     "src/builtins/builtins-object-gen.cc",
     "src/builtins/builtins-promise-gen.cc",

src/builtins/builtins-definitions.h

@@ -688,6 +688,7 @@ namespace internal {
   TFC(Multiply, BinaryOp, 1)                                                   \
   TFC(Divide, BinaryOp, 1)                                                     \
   TFC(Modulus, BinaryOp, 1)                                                    \
+  TFC(Exponentiate, BinaryOp, 1)                                               \
   TFC(BitwiseAnd, BinaryOp, 1)                                                 \
   TFC(BitwiseOr, BinaryOp, 1)                                                  \
   TFC(BitwiseXor, BinaryOp, 1)                                                 \

src/builtins/builtins-math-gen.cc

@@ -2,6 +2,8 @@
 // Use of this source code is governed by a BSD-style license that can be
 // found in the LICENSE file.
 
+#include "src/builtins/builtins-math-gen.h"
+
 #include "src/builtins/builtins-utils-gen.h"
 #include "src/builtins/builtins.h"
 #include "src/code-factory.h"
@@ -14,24 +16,6 @@ namespace internal {
 // -----------------------------------------------------------------------------
 // ES6 section 20.2.2 Function Properties of the Math Object
 
-class MathBuiltinsAssembler : public CodeStubAssembler {
- public:
-  explicit MathBuiltinsAssembler(compiler::CodeAssemblerState* state)
-      : CodeStubAssembler(state) {}
-
- protected:
-  void MathRoundingOperation(
-      Node* context, Node* x,
-      TNode<Float64T> (CodeStubAssembler::*float64op)(SloppyTNode<Float64T>));
-  void MathUnaryOperation(
-      Node* context, Node* x,
-      TNode<Float64T> (CodeStubAssembler::*float64op)(SloppyTNode<Float64T>));
-  void MathMaxMin(Node* context, Node* argc,
-                  TNode<Float64T> (CodeStubAssembler::*float64op)(
-                      SloppyTNode<Float64T>, SloppyTNode<Float64T>),
-                  double default_val);
-};
-
 // ES6 #sec-math.abs
 TF_BUILTIN(MathAbs, CodeStubAssembler) {
   Node* context = Parameter(Descriptor::kContext);
@@ -405,16 +389,19 @@ TF_BUILTIN(MathLog2, MathBuiltinsAssembler) {
   MathUnaryOperation(context, x, &CodeStubAssembler::Float64Log2);
 }
 
+CodeStubAssembler::Node* MathBuiltinsAssembler::MathPow(Node* context,
+                                                        Node* base,
+                                                        Node* exponent) {
+  Node* base_value = TruncateTaggedToFloat64(context, base);
+  Node* exponent_value = TruncateTaggedToFloat64(context, exponent);
+  Node* value = Float64Pow(base_value, exponent_value);
+  return ChangeFloat64ToTagged(value);
+}
+
 // ES6 #sec-math.pow
-TF_BUILTIN(MathPow, CodeStubAssembler) {
-  Node* context = Parameter(Descriptor::kContext);
-  Node* x = Parameter(Descriptor::kBase);
-  Node* y = Parameter(Descriptor::kExponent);
-  Node* x_value = TruncateTaggedToFloat64(context, x);
-  Node* y_value = TruncateTaggedToFloat64(context, y);
-  Node* value = Float64Pow(x_value, y_value);
-  Node* result = ChangeFloat64ToTagged(value);
-  Return(result);
+TF_BUILTIN(MathPow, MathBuiltinsAssembler) {
+  Return(MathPow(Parameter(Descriptor::kContext), Parameter(Descriptor::kBase),
+                 Parameter(Descriptor::kExponent)));
 }
 
 // ES6 #sec-math.random

src/builtins/builtins-math-gen.h

+// Copyright 2017 the V8 project authors. All rights reserved.
+// Use of this source code is governed by a BSD-style license that can be
+// found in the LICENSE file.
+
+#ifndef V8_BUILTINS_BUILTINS_MATH_GEN_H_
+#define V8_BUILTINS_BUILTINS_MATH_GEN_H_
+
+#include "src/code-stub-assembler.h"
+
+namespace v8 {
+namespace internal {
+
+class MathBuiltinsAssembler : public CodeStubAssembler {
+ public:
+  explicit MathBuiltinsAssembler(compiler::CodeAssemblerState* state)
+      : CodeStubAssembler(state) {}
+
+  Node* MathPow(Node* context, Node* base, Node* exponent);
+
+ protected:
+  void MathRoundingOperation(
+      Node* context, Node* x,
+      TNode<Float64T> (CodeStubAssembler::*float64op)(SloppyTNode<Float64T>));
+  void MathUnaryOperation(
+      Node* context, Node* x,
+      TNode<Float64T> (CodeStubAssembler::*float64op)(SloppyTNode<Float64T>));
+  void MathMaxMin(Node* context, Node* argc,
+                  TNode<Float64T> (CodeStubAssembler::*float64op)(
+                      SloppyTNode<Float64T>, SloppyTNode<Float64T>),
+                  double default_val);
+};
+
+}  // namespace internal
+}  // namespace v8
+
+#endif  // V8_BUILTINS_BUILTINS_MATH_GEN_H_

src/builtins/builtins-number-gen.cc

@@ -2,6 +2,7 @@
 // Use of this source code is governed by a BSD-style license that can be
 // found in the LICENSE file.
 
+#include "src/builtins/builtins-math-gen.h"
 #include "src/builtins/builtins-utils-gen.h"
 #include "src/builtins/builtins.h"
 #include "src/code-stub-assembler.h"
@@ -636,8 +637,11 @@ void NumberBuiltinsAssembler::BinaryOp(Label* smis, Variable* var_left,
                                        Label* bigints) {
   DCHECK_EQ(var_left->rep(), MachineRepresentation::kTagged);
   DCHECK_EQ(var_right->rep(), MachineRepresentation::kTagged);
-  DCHECK_EQ(var_left_double->rep(), MachineRepresentation::kFloat64);
-  DCHECK_EQ(var_right_double->rep(), MachineRepresentation::kFloat64);
+  DCHECK_IMPLIES(var_left_double != nullptr,
+                 var_left_double->rep() == MachineRepresentation::kFloat64);
+  DCHECK_IMPLIES(var_right_double != nullptr,
+                 var_right_double->rep() == MachineRepresentation::kFloat64);
+  DCHECK_EQ(var_left_double == nullptr, var_right_double == nullptr);
 
   Node* context = Parameter(Descriptor::kContext);
   var_left->Bind(Parameter(Descriptor::kLeft));
@@ -655,8 +659,10 @@ void NumberBuiltinsAssembler::BinaryOp(Label* smis, Variable* var_left,
 
   // At this point, var_left is a Smi but var_right is not.
   GotoIfNot(IsHeapNumber(var_right->value()), &right_not_number);
+  if (var_left_double != nullptr) {
     var_left_double->Bind(SmiToFloat64(var_left->value()));
     var_right_double->Bind(LoadHeapNumberValue(var_right->value()));
+  }
   Goto(doubles);
 
   BIND(&left_not_smi);
@@ -665,16 +671,20 @@ void NumberBuiltinsAssembler::BinaryOp(Label* smis, Variable* var_left,
     GotoIfNot(TaggedIsSmi(var_right->value()), &right_not_smi);
 
     // At this point, var_left is a HeapNumber and var_right is a Smi.
+    if (var_left_double != nullptr) {
       var_left_double->Bind(LoadHeapNumberValue(var_left->value()));
       var_right_double->Bind(SmiToFloat64(var_right->value()));
+    }
     Goto(doubles);
   }
 
   BIND(&right_not_smi);
   {
     GotoIfNot(IsHeapNumber(var_right->value()), &right_not_number);
+    if (var_left_double != nullptr) {
       var_left_double->Bind(LoadHeapNumberValue(var_left->value()));
       var_right_double->Bind(LoadHeapNumberValue(var_right->value()));
+    }
     Goto(doubles);
   }
 
@@ -970,6 +980,26 @@ TF_BUILTIN(Modulus, NumberBuiltinsAssembler) {
   }
 }
 
+TF_BUILTIN(Exponentiate, NumberBuiltinsAssembler) {
+  VARIABLE(var_left, MachineRepresentation::kTagged);
+  VARIABLE(var_right, MachineRepresentation::kTagged);
+  Label do_number_exp(this), do_bigint_exp(this);
+  Node* context = Parameter(Descriptor::kContext);
+
+  BinaryOp<Descriptor>(&do_number_exp, &var_left, &var_right, &do_number_exp,
+                       nullptr, nullptr, &do_bigint_exp);
+
+  BIND(&do_number_exp);
+  {
+    MathBuiltinsAssembler math_asm(state());
+    Return(math_asm.MathPow(context, var_left.value(), var_right.value()));
+  }
+
+  BIND(&do_bigint_exp);
+  Return(CallRuntime(Runtime::kBigIntBinaryOp, context, var_left.value(),
+                     var_right.value(), SmiConstant(Operation::kExponentiate)));
+}
+
 TF_BUILTIN(ShiftLeft, NumberBuiltinsAssembler) {
   EmitBitwiseOp<Descriptor>(Operation::kShiftLeft);
 }

src/compiler/bytecode-graph-builder.cc

@@ -2139,6 +2139,10 @@ void BytecodeGraphBuilder::VisitMod() {
   BuildBinaryOp(javascript()->Modulus());
 }
 
+void BytecodeGraphBuilder::VisitExp() {
+  BuildBinaryOp(javascript()->Exponentiate());
+}
+
 void BytecodeGraphBuilder::VisitBitwiseOr() {
   BuildBinaryOp(javascript()->BitwiseOr());
 }
@@ -2205,6 +2209,10 @@ void BytecodeGraphBuilder::VisitModSmi() {
   BuildBinaryOpWithImmediate(javascript()->Modulus());
 }
 
+void BytecodeGraphBuilder::VisitExpSmi() {
+  BuildBinaryOpWithImmediate(javascript()->Exponentiate());
+}
+
 void BytecodeGraphBuilder::VisitBitwiseOrSmi() {
   BuildBinaryOpWithImmediate(javascript()->BitwiseOr());
 }

src/compiler/js-generic-lowering.cc

@@ -60,6 +60,7 @@ REPLACE_STUB_CALL(Subtract)
 REPLACE_STUB_CALL(Multiply)
 REPLACE_STUB_CALL(Divide)
 REPLACE_STUB_CALL(Modulus)
+REPLACE_STUB_CALL(Exponentiate)
 REPLACE_STUB_CALL(BitwiseAnd)
 REPLACE_STUB_CALL(BitwiseOr)
 REPLACE_STUB_CALL(BitwiseXor)

src/compiler/js-operator.cc

@@ -577,6 +577,7 @@ CompareOperationHint CompareOperationHintOf(const Operator* op) {
   V(Multiply, Operator::kNoProperties, 2, 1)                    \
   V(Divide, Operator::kNoProperties, 2, 1)                      \
   V(Modulus, Operator::kNoProperties, 2, 1)                     \
+  V(Exponentiate, Operator::kNoProperties, 2, 1)                \
   V(BitwiseNot, Operator::kNoProperties, 1, 1)                  \
   V(Decrement, Operator::kNoProperties, 1, 1)                   \
   V(Increment, Operator::kNoProperties, 1, 1)                   \

src/compiler/js-operator.h

@@ -651,6 +651,7 @@ class V8_EXPORT_PRIVATE JSOperatorBuilder final
   const Operator* Multiply();
   const Operator* Divide();
   const Operator* Modulus();
+  const Operator* Exponentiate();
 
   const Operator* BitwiseNot();
   const Operator* Decrement();

src/compiler/js-type-hint-lowering.cc

@@ -344,6 +344,10 @@ JSTypeHintLowering::LoweringResult JSTypeHintLowering::ReduceBinaryOperation(
       }
       break;
     }
+    case IrOpcode::kJSExponentiate: {
+      // TODO(neis): Introduce a SpeculativeNumberPow operator?
+      break;
+    }
     default:
       UNREACHABLE();
       break;

src/compiler/js-typed-lowering.cc

@@ -279,6 +279,8 @@ class JSBinopReduction final {
         return simplified()->NumberDivide();
       case IrOpcode::kJSModulus:
         return simplified()->NumberModulus();
+      case IrOpcode::kJSExponentiate:
+        return simplified()->NumberPow();
       case IrOpcode::kJSBitwiseAnd:
         return simplified()->NumberBitwiseAnd();
       case IrOpcode::kJSBitwiseOr:
@@ -2123,6 +2125,7 @@ Reduction JSTypedLowering::Reduce(Node* node) {
     case IrOpcode::kJSMultiply:
     case IrOpcode::kJSDivide:
     case IrOpcode::kJSModulus:
+    case IrOpcode::kJSExponentiate:
       return ReduceNumberBinop(node);
     case IrOpcode::kJSBitwiseNot:
       return ReduceJSBitwiseNot(node);

src/compiler/opcodes.h

@@ -105,7 +105,8 @@
   V(JSSubtract)                \
   V(JSMultiply)                \
   V(JSDivide)                  \
-  V(JSModulus)
+  V(JSModulus)                 \
+  V(JSExponentiate)
 
 #define JS_SIMPLE_BINOP_LIST(V) \
   JS_COMPARE_BINOP_LIST(V)      \

src/compiler/operator-properties.cc

@@ -44,6 +44,7 @@ bool OperatorProperties::HasFrameStateInput(const Operator* op) {
     case IrOpcode::kJSMultiply:
     case IrOpcode::kJSDivide:
     case IrOpcode::kJSModulus:
+    case IrOpcode::kJSExponentiate:
 
     // Bitwise operations
     case IrOpcode::kJSBitwiseOr:

src/compiler/typer.cc

@@ -1103,6 +1103,12 @@ Type* Typer::Visitor::JSModulusTyper(Type* lhs, Type* rhs, Typer* t) {
   return NumberModulus(ToNumber(lhs, t), ToNumber(rhs, t), t);
 }
 
+// TODO(neis): Adapt all these for bigints. Why does this even work in the
+// bigint tests?
+
+Type* Typer::Visitor::JSExponentiateTyper(Type* lhs, Type* rhs, Typer* t) {
+  return Type::Number();
+}
 
 // JS unary operators.
 

src/compiler/verifier.cc

@@ -608,7 +608,9 @@ void Verifier::Visitor::Check(Node* node, const AllNodes& all) {
     case IrOpcode::kJSMultiply:
     case IrOpcode::kJSDivide:
     case IrOpcode::kJSModulus:
+    case IrOpcode::kJSExponentiate:
       // Type is Number.
+      // TODO(neis): Adapt for bigints.
       CheckTypeIs(node, Type::Number());
       break;
 

src/debug/debug-evaluate.cc

@@ -382,6 +382,8 @@ bool BytecodeHasNoSideEffect(interpreter::Bytecode bytecode) {
     case Bytecode::kDivSmi:
     case Bytecode::kMod:
     case Bytecode::kModSmi:
+    case Bytecode::kExp:
+    case Bytecode::kExpSmi:
     case Bytecode::kNegate:
     case Bytecode::kBitwiseAnd:
     case Bytecode::kBitwiseAndSmi:

src/globals.h

@@ -1305,6 +1305,7 @@ enum class Operation {
   kMultiply,
   kDivide,
   kModulus,
+  kExponentiate,
   kBitwiseAnd,
   kBitwiseOr,
   kBitwiseXor,

src/ic/binary-op-assembler.cc

@@ -529,5 +529,14 @@ Node* BinaryOpAssembler::Generate_ModulusWithFeedback(
       floatFunction, Operation::kModulus, rhs_is_smi);
 }
 
+Node* BinaryOpAssembler::Generate_ExponentiateWithFeedback(
+    Node* context, Node* base, Node* exponent, Node* slot_id,
+    Node* feedback_vector, bool rhs_is_smi) {
+  // We currently don't optimize exponentiation based on feedback.
+  Node* dummy_feedback = SmiConstant(BinaryOperationFeedback::kAny);
+  UpdateFeedback(dummy_feedback, feedback_vector, slot_id);
+  return CallBuiltin(Builtins::kExponentiate, context, base, exponent);
+}
+
 }  // namespace internal
 }  // namespace v8

src/ic/binary-op-assembler.h

@@ -42,6 +42,11 @@ class BinaryOpAssembler : public CodeStubAssembler {
                                      Node* divisor, Node* slot_id,
                                      Node* feedback_vector, bool rhs_is_smi);
 
+  Node* Generate_ExponentiateWithFeedback(Node* context, Node* dividend,
+                                          Node* divisor, Node* slot_id,
+                                          Node* feedback_vector,
+                                          bool rhs_is_smi);
+
  private:
   typedef std::function<Node*(Node*, Node*, Variable*)> SmiOperation;
   typedef std::function<Node*(Node*, Node*)> FloatOperation;

src/interpreter/bytecode-array-builder.cc

@@ -362,6 +362,9 @@ BytecodeArrayBuilder& BytecodeArrayBuilder::BinaryOperation(Token::Value op,
     case Token::Value::MOD:
       OutputMod(reg, feedback_slot);
       break;
+    case Token::Value::EXP:
+      OutputExp(reg, feedback_slot);
+      break;
     case Token::Value::BIT_OR:
       OutputBitwiseOr(reg, feedback_slot);
       break;
@@ -404,6 +407,9 @@ BytecodeArrayBuilder& BytecodeArrayBuilder::BinaryOperationSmiLiteral(
     case Token::Value::MOD:
       OutputModSmi(literal->value(), feedback_slot);
       break;
+    case Token::Value::EXP:
+      OutputExpSmi(literal->value(), feedback_slot);
+      break;
     case Token::Value::BIT_OR:
       OutputBitwiseOrSmi(literal->value(), feedback_slot);
       break;

src/interpreter/bytecodes.h

@@ -110,6 +110,7 @@ namespace interpreter {
   V(Mul, AccumulatorUse::kReadWrite, OperandType::kReg, OperandType::kIdx)     \
   V(Div, AccumulatorUse::kReadWrite, OperandType::kReg, OperandType::kIdx)     \
   V(Mod, AccumulatorUse::kReadWrite, OperandType::kReg, OperandType::kIdx)     \
+  V(Exp, AccumulatorUse::kReadWrite, OperandType::kReg, OperandType::kIdx)     \
   V(BitwiseOr, AccumulatorUse::kReadWrite, OperandType::kReg,                  \
     OperandType::kIdx)                                                         \
   V(BitwiseXor, AccumulatorUse::kReadWrite, OperandType::kReg,                 \
@@ -129,6 +130,7 @@ namespace interpreter {
   V(MulSmi, AccumulatorUse::kReadWrite, OperandType::kImm, OperandType::kIdx)  \
   V(DivSmi, AccumulatorUse::kReadWrite, OperandType::kImm, OperandType::kIdx)  \
   V(ModSmi, AccumulatorUse::kReadWrite, OperandType::kImm, OperandType::kIdx)  \
+  V(ExpSmi, AccumulatorUse::kReadWrite, OperandType::kImm, OperandType::kIdx)  \
   V(BitwiseOrSmi, AccumulatorUse::kReadWrite, OperandType::kImm,               \
     OperandType::kIdx)                                                         \
   V(BitwiseXorSmi, AccumulatorUse::kReadWrite, OperandType::kImm,              \

src/interpreter/interpreter-generator.cc

@@ -910,6 +910,13 @@ IGNITION_HANDLER(Mod, InterpreterBinaryOpAssembler) {
   BinaryOpWithFeedback(&BinaryOpAssembler::Generate_ModulusWithFeedback);
 }
 
+// Exp <src>
+//
+// Exponentiate register <src> (base) with accumulator (exponent).
+IGNITION_HANDLER(Exp, InterpreterBinaryOpAssembler) {
+  BinaryOpWithFeedback(&BinaryOpAssembler::Generate_ExponentiateWithFeedback);
+}
+
 // AddSmi <imm>
 //
 // Adds an immediate value <imm> to the value in the accumulator.
@@ -945,6 +952,14 @@ IGNITION_HANDLER(ModSmi, InterpreterBinaryOpAssembler) {
   BinaryOpSmiWithFeedback(&BinaryOpAssembler::Generate_ModulusWithFeedback);
 }
 
+// ExpSmi <imm>
+//
+// Exponentiate accumulator (base) with immediate value <imm> (exponent).
+IGNITION_HANDLER(ExpSmi, InterpreterBinaryOpAssembler) {
+  BinaryOpSmiWithFeedback(
+      &BinaryOpAssembler::Generate_ExponentiateWithFeedback);
+}
+
 class InterpreterBitwiseBinaryOpAssembler : public InterpreterAssembler {
  public:
   InterpreterBitwiseBinaryOpAssembler(CodeAssemblerState* state,

src/parsing/parser-base.h

@@ -2973,15 +2973,11 @@ ParserBase<Impl>::ParseAssignmentExpression(bool accept_IN, bool* ok) {
     impl()->SetFunctionNameFromIdentifierRef(right, expression);
   }
 
-  if (op == Token::ASSIGN_EXP) {
-    DCHECK(!is_destructuring_assignment);
-    return impl()->RewriteAssignExponentiation(expression, right, pos);
-  }
-
   DCHECK_NE(op, Token::INIT);
   ExpressionT result = factory()->NewAssignment(op, expression, right, pos);
 
   if (is_destructuring_assignment) {
+    DCHECK_NE(op, Token::ASSIGN_EXP);
     auto rewritable = factory()->NewRewritableExpression(result, scope());
     impl()->QueueDestructuringAssignmentForRewriting(rewritable);
     result = rewritable;
@@ -3131,8 +3127,6 @@ typename ParserBase<Impl>::ExpressionT ParserBase<Impl>::ParseBinaryExpression(
           // The comparison was negated - add a NOT.
           x = factory()->NewUnaryOperation(Token::NOT, x, pos);
         }
-      } else if (op == Token::EXP) {
-        x = impl()->RewriteExponentiation(x, y, pos);
       } else if (impl()->CollapseNaryExpression(&x, y, op, pos, right_range)) {
         continue;
       } else {

src/parsing/parser.cc

@@ -3845,58 +3845,6 @@ void Parser::RewriteDestructuringAssignments() {
   }
 }
 
-Expression* Parser::RewriteExponentiation(Expression* left, Expression* right,
-                                          int pos) {
-  ZoneList<Expression*>* args = new (zone()) ZoneList<Expression*>(2, zone());
-  args->Add(left, zone());
-  args->Add(right, zone());
-  return factory()->NewCallRuntime(Context::MATH_POW_INDEX, args, pos);
-}
-
-Expression* Parser::RewriteAssignExponentiation(Expression* left,
-                                                Expression* right, int pos) {
-  ZoneList<Expression*>* args = new (zone()) ZoneList<Expression*>(2, zone());
-  if (left->IsVariableProxy()) {
-    VariableProxy* lhs = left->AsVariableProxy();
-
-    Expression* result;
-    DCHECK_NOT_NULL(lhs->raw_name());
-    result = ExpressionFromIdentifier(lhs->raw_name(), lhs->position());
-    args->Add(left, zone());
-    args->Add(right, zone());
-    Expression* call =
-        factory()->NewCallRuntime(Context::MATH_POW_INDEX, args, pos);
-    return factory()->NewAssignment(Token::ASSIGN, result, call, pos);
-  } else if (left->IsProperty()) {
-    Property* prop = left->AsProperty();
-    auto temp_obj = NewTemporary(ast_value_factory()->empty_string());
-    auto temp_key = NewTemporary(ast_value_factory()->empty_string());
-    Expression* assign_obj = factory()->NewAssignment(
-        Token::ASSIGN, factory()->NewVariableProxy(temp_obj), prop->obj(),
-        kNoSourcePosition);
-    Expression* assign_key = factory()->NewAssignment(
-        Token::ASSIGN, factory()->NewVariableProxy(temp_key), prop->key(),
-        kNoSourcePosition);
-    args->Add(factory()->NewProperty(factory()->NewVariableProxy(temp_obj),
-                                     factory()->NewVariableProxy(temp_key),
-                                     left->position()),
-              zone());
-    args->Add(right, zone());
-    Expression* call =
-        factory()->NewCallRuntime(Context::MATH_POW_INDEX, args, pos);
-    Expression* target = factory()->NewProperty(
-        factory()->NewVariableProxy(temp_obj),
-        factory()->NewVariableProxy(temp_key), kNoSourcePosition);
-    Expression* assign =
-        factory()->NewAssignment(Token::ASSIGN, target, call, pos);
-    return factory()->NewBinaryOperation(
-        Token::COMMA, assign_obj,
-        factory()->NewBinaryOperation(Token::COMMA, assign_key, assign, pos),
-        pos);
-  }
-  UNREACHABLE();
-}
-
 Expression* Parser::RewriteSpreads(ArrayLiteral* lit) {
   // Array literals containing spreads are rewritten using do expressions, e.g.
   //    [1, 2, 3, ...x, 4, ...y, 5]

src/parsing/parser.h

@@ -551,11 +551,6 @@ class V8_EXPORT_PRIVATE Parser : public NON_EXPORTED_BASE(ParserBase<Parser>) {
   // Rewrite all DestructuringAssignments in the current FunctionState.
   V8_INLINE void RewriteDestructuringAssignments();
 
-  V8_INLINE Expression* RewriteExponentiation(Expression* left,
-                                              Expression* right, int pos);
-  V8_INLINE Expression* RewriteAssignExponentiation(Expression* left,
-                                                    Expression* right, int pos);
-
   Expression* RewriteSpreads(ArrayLiteral* lit);
 
   // Rewrite expressions that are not used as patterns

src/parsing/preparser.h

@@ -995,17 +995,6 @@ class PreParser : public ParserBase<PreParser> {
 
   V8_INLINE void RewriteDestructuringAssignments() {}
 
-  V8_INLINE PreParserExpression
-  RewriteExponentiation(const PreParserExpression& left,
-                        const PreParserExpression& right, int pos) {
-    return left;
-  }
-  V8_INLINE PreParserExpression
-  RewriteAssignExponentiation(const PreParserExpression& left,
-                              const PreParserExpression& right, int pos) {
-    return left;
-  }
-
   V8_INLINE void PrepareGeneratorVariables() {}
   V8_INLINE void RewriteAsyncFunctionBody(
       PreParserStatementList body, PreParserStatement block,

src/parsing/token.h

@@ -284,6 +284,8 @@ class Token {
         return Token::DIV;
       case Token::ASSIGN_MOD:
         return Token::MOD;
+      case Token::ASSIGN_EXP:
+        return Token::EXP;
       default:
         UNREACHABLE();
     }

src/runtime/runtime-bigint.cc

@@ -107,6 +107,9 @@ RUNTIME_FUNCTION(Runtime_BigIntBinaryOp) {
     case Operation::kModulus:
       result = BigInt::Remainder(left, right);
       break;
+    case Operation::kExponentiate:
+      UNIMPLEMENTED();
+      break;
     case Operation::kBitwiseAnd:
       result = BigInt::BitwiseAnd(left, right);
       break;

src/v8.gyp

@@ -200,6 +200,7 @@
         'builtins/builtins-iterator-gen.h',
         'builtins/builtins-iterator-gen.cc',
         'builtins/builtins-math-gen.cc',
+        'builtins/builtins-math-gen.h',
         'builtins/builtins-number-gen.cc',
         'builtins/builtins-object-gen.cc',
         'builtins/builtins-promise-gen.cc',

test/unittests/interpreter/bytecode-array-builder-unittest.cc

@@ -199,7 +199,8 @@ TEST_F(BytecodeArrayBuilderTest, AllBytecodesGenerated) {
       .BinaryOperation(Token::Value::SUB, reg, 2)
       .BinaryOperation(Token::Value::MUL, reg, 3)
       .BinaryOperation(Token::Value::DIV, reg, 4)
-      .BinaryOperation(Token::Value::MOD, reg, 5);
+      .BinaryOperation(Token::Value::MOD, reg, 5)
+      .BinaryOperation(Token::Value::EXP, reg, 6);
 
   // Emit bitwise operator invocations
   builder.BinaryOperation(Token::Value::BIT_OR, reg, 6)
@@ -217,6 +218,7 @@ TEST_F(BytecodeArrayBuilderTest, AllBytecodesGenerated) {
       .BinaryOperationSmiLiteral(Token::Value::MUL, Smi::FromInt(42), 2)
       .BinaryOperationSmiLiteral(Token::Value::DIV, Smi::FromInt(42), 2)
       .BinaryOperationSmiLiteral(Token::Value::MOD, Smi::FromInt(42), 2)
+      .BinaryOperationSmiLiteral(Token::Value::EXP, Smi::FromInt(42), 2)
       .BinaryOperationSmiLiteral(Token::Value::BIT_OR, Smi::FromInt(42), 2)
       .BinaryOperationSmiLiteral(Token::Value::BIT_XOR, Smi::FromInt(42), 2)
       .BinaryOperationSmiLiteral(Token::Value::BIT_AND, Smi::FromInt(42), 2)