Benedikt Meurer

Not just that, it also has serious correctness regressions. We're currently reverting most of the async/Promise performance improvements for Node 10, so performance-wise it should soon be the same as...

There's no clear decision what to do with the `async_hooks`. I guess for Node 10 you'll just need to revert to an alternative if it's a serious problem for now.

Chromium DevTools did indeed support ANSI color codes, but not very well until recently. I've recently started to take inventory of what would be a reasonable subset to support for...

Those are great suggestions, Henry, thanks a lot! The current version is mostly a one-shot prototype. Ideally the benchmark payloads would be created and driven by experts like you, who...

@jdalton Wanna submit a PR? 😎

That's exactly the thing: The benchmark should work unbundled, just by running `node src/cli`, since that's the closest configuration to real-world we get, since tools are also run unbundled normally...

The main problem with all this magic is that the benchmark is now dominated by the `fs` mock, some use of the `vm` module and a lot of other unrelated...

All browsers, unless there's a good reason to exclude them (i.e. like old internet explorer versions don't need to be supported). But since the tools have to run on Node...

The value shown is the geometric mean of the average runs/sec of the individual tests. An earlier version was computing the geometric mean of the times of the individual tests,...

I guess we could remove the first N samples and do the average/stddev solely on the remaining samples. iframe/realm is not really an option here, since Node doesn't support realms...