Hacker News Reader

Comment by afavour | Hacker News Reader

Maybe I’m missing something here but surely Node isn’t doing anything concurrently? Promises don’t execute concurrently, they just tidy up async execution. The code as given will just sequentially resolve a million promises. No wonder it looks so good. You’d need to be using workers to actually do anything concurrently.

View on HN

You're thinking of parallelism. Concurrency doesn't require them to actually be running at the same time.

Fair point. Kind of makes the comparisons between languages a little unfair, though. Go and Rust would be executing these operations in parallel, Node would not. Would make a significant difference to real world performance!

The measurement is memory, not performance. Paused/queued tasks in the sequential node version still count, and in theory could be worse since the Go and Rust ones would be consuming them in parallel and not building up the queue as much.

> The measurement is memory, not performance.

But, then they can measure memory by simply using a threat pool of size 1 and then submitting tasks to it right ? That would be the equivalent comparison for other languages.

They should launch a million NodeJS processes.

That's not entirely true. There's a thread pool of workers underneath libuv. Tasks that would block do indeed execute concurrently.

Oh, I know. But the code used in this test doesn’t utilize that thread pool at all. It just uses setTimeout.

Reply on news.ycombinator.com