Commit 3118f473 authored by Ross McIlroy's avatar Ross McIlroy Committed by Commit Bot

[TurboFan] Fix type checks for lowering SpeculativeNumberBinop.

Ensure we only lower SpeculativeNumberBinops to a pure operator for
non-string plain primitives. Previously we could lower if a value might be
the-hole, however this would fail a CHECK in ConvertInputsToNumber which
expects a plain primitive.

BUG=chromium:772420

Change-Id: I0c755d10db7afd9cabfb638eca5662d70dfc8d51
Reviewed-on: https://chromium-review.googlesource.com/715717
Commit-Queue: Ross McIlroy <rmcilroy@chromium.org>
Reviewed-by: 's avatarJaroslav Sevcik <jarin@chromium.org>
Cr-Commit-Position: refs/heads/master@{#48649}
parent 269b35a3
......@@ -532,7 +532,7 @@ Reduction JSTypedLowering::ReduceSpeculativeNumberBinop(Node* node) {
NumberOperationHint hint = NumberOperationHintOf(node->op());
if ((hint == NumberOperationHint::kNumber ||
hint == NumberOperationHint::kNumberOrOddball) &&
r.BothInputsAre(Type::NumberOrOddball())) {
r.BothInputsAre(Type::NumberOrUndefinedOrNullOrBoolean())) {
r.ConvertInputsToNumber();
return r.ChangeToPureOperator(r.NumberOpFromSpeculativeNumberOp(),
Type::Number());
......
......@@ -170,6 +170,8 @@ namespace compiler {
kHole) \
V(NumberOrString, kNumber | kString) \
V(NumberOrUndefined, kNumber | kUndefined) \
V(NumberOrUndefinedOrNullOrBoolean, \
kNumber | kNullOrUndefined | kBoolean) \
V(PlainPrimitive, kNumberOrString | kBoolean | \
kNullOrUndefined) \
V(Primitive, kSymbol | kPlainPrimitive) \
......
// 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.
// Flags: --allow-natives-syntax
function foo(arg) {
var value;
// None of the branches of this switch are ever taken, but
// the sequence means value could be the hole.
switch (arg) {
case 1:
let let_var = 1;
case 2:
value = let_var;
}
// Speculative number binop with NumberOrOddball feedback.
// Shouldn't be optimized to pure operator since value's phi
// could theoretically be the hole (we would have already thrown a
// reference error in case 2 above if so, but TF typing still
// thinks it could be the hole).
return value * undefined;
}
foo(3);
foo(3);
%OptimizeFunctionOnNextCall(foo);
foo(3);
Markdown is supported
0% or
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment