3 comments

[ 2.6 ms ] story [ 16.1 ms ] thread
What amount of time does a production server spend at "peak performance"?
We discussed on reddit, but fwiw:

Under Utopian circumstances, the percentage would converge upon 100%, but in reality, we don't know. Our paper only focuses on microbenchmarks as these should be the best behaved benchmarks available.

I feel like there's more and more evidence out there that JIT in Javascript might be a huge dead end. I remember reading a couple articles about how V8 is rolling back a lot of that stuff, for example, simply because in practice it would end up causing things to be slower than they could be.