Re:Asm-Dom – WebAssembly Virtual DOM
by pizlonator on Sunday May 21, 2017 @10:31AM
(#14387466)
That's a pretty good assessment.
I would note that emscripten's sdk comes with a really great malloc. If you're programming in C++ with a great malloc then your memory allocation performance will be better than what you'd get on the JS side. Be weary of microbenchmarks that try to say otherwise because the reason why malloc beats GCs for real shit is that it scales better. For example, GCs get slower if the heap shape gets weird, while malloc doesn't care about that.
Finally, WebKit's GC doesn't implement the nursery as an arena. I don't think that's such a popular GC technique anymore, particularly now that using non-compacting generational GCs is becoming so common.
I would note that emscripten's sdk comes with a really great malloc. If you're programming in C++ with a great malloc then your memory allocation performance will be better than what you'd get on the JS side. Be weary of microbenchmarks that try to say otherwise because the reason why malloc beats GCs for real shit is that it scales better. For example, GCs get slower if the heap shape gets weird, while malloc doesn't care about that.
Finally, WebKit's GC doesn't implement the nursery as an arena. I don't think that's such a popular GC technique anymore, particularly now that using non-compacting generational GCs is becoming so common.