Comment by The_Colonel
9 months ago
> The longer something can stay a one-person project, the better! Nothing kills creativity, innovation, and velocity faster than having to make every decision by committee.
One person can get surprisingly far, but there's a limit beyond which no single human will scale. Getting to the v8 performance is IMHO such an example. You might be OK with a browser which has a noticeably subpar performance, but it will likely stifle mainstream adoption (which again, might be OK for you and that's fine).
> Getting to the v8 performance is IMHO such an example.
There's no doubt in my mind that Andreas could achieve that by himself. He's worked professionally on webkit, and implemented a JS interpreter, a JS bytecode interpreter, and a JS JIT all by himself after all. Also let's not forget that V8 is open-source, all their optimizations are available for others to see and implement.
But to be clear this isn't a one man project, he hired a few contributors to work full time on it. Sure, it's a small team, but as said in sibling comments a small team has much more velocity.
Seems like Andreas doesn't have the same delusions of ridiculing the army of (pretty smart) Chrome/V8 devs by doing the same job just on his own. His own goal is to achieve the performance of non-JITed JavaScriptCore - i.e. an optimizing interpreter.
Matching V8's perf would be infeasible, but couldn't a small team get within an order of magnitude of V8's perf for a decent chunk of websites? How much slower is Fabrice Bellard's QuickJS?
1 reply →
Andreas isn’t targeting V8’s JIT performance. The goal is to be roughly in line with WebKit’s performance with the JIT turned off.
The theory is that JS JIT compilers don’t actually improve real world performance on the majority of websites. This was apparently per the advice of the authors of Chrome’s and Safari’s JITs.