f226350fcb
PR-URL: https://github.com/nodejs/node/pull/47251 Reviewed-By: Yagiz Nizipli <yagiz@nizipli.com> Reviewed-By: Jiawen Geng <technicalcute@gmail.com> Reviewed-By: Rafael Gonzaga <rafael.nunu@hotmail.com> Reviewed-By: Richard Lau <rlau@redhat.com>
28 lines
1.3 KiB
JavaScript
28 lines
1.3 KiB
JavaScript
// Copyright 2023 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: --maglev --allow-natives-syntax
|
|
|
|
function f(x) {
|
|
x++; // Making sure {x} has a Float64 alternative
|
|
let phi = x || 0; // Creating a Phi whose inputs are Smi and Float64. Phi
|
|
// untagging will make this a Float64 phi.
|
|
phi + 1; // Float64 operation with {phi}, which will cause insert a
|
|
// CheckedNumberToFloat64(phi). After phi untagging, the
|
|
// CheckedNumberToFloat64 will be killed, resulting in the addition
|
|
// to take {phi} directly as input.
|
|
phi << 4; // Int operation with {phi}, which will produce a
|
|
// TruncateNumberToInt32(phi). TruncateNumberToInt32 doesn't check
|
|
// if the input is a Number, and call into runtime if necessary to
|
|
// to the truncation, so it can't deopt. After Phi untagging, this
|
|
// should be replaced by a Float64->Int32 truncation
|
|
// (TruncateFloat64ToInt32), which shouldn't deopt either (because
|
|
// we can't replace a non-deopting node by a deopting one).
|
|
}
|
|
|
|
%PrepareFunctionForOptimization(f);
|
|
f(1.5);
|
|
%OptimizeMaglevOnNextCall(f);
|
|
f(1.5);
|