1804964 - Experiment using CanGC allocations directly for JSDependentString::new_

Status: NEW

(Core :: JavaScript Engine, task, P2)

Description

[Ted Campbell [:tcampbell]]

When I remove this in JSDependentString::new_ I see 5% improves for a simple String.substring microbenchmark. This is higher than I expected and I wonder if there are more cases where we are better off just doing the rooting in the first place.

let lefty  = "AAAAAAAAAAAAAAAAAAAAAAAAAAAAAA";
let result = [0, 1, 2, 3, 4, 5, 6, 7];

let start = Date.now();
for (var i = 0; i < 50000000; ++i) {
	result[i & 8] = lefty.substring(20, 28);
}
let end = Date.now();

console.log(end - start);

Needs further experimentation on larger benchmarks