3 ms·

Pretty sure v8 gets a few orders of magnitude more attention and development resources than jvm and graal.

dpratt 6y ago
You're probably right about Graal, but there is no way that V8 has significantly more effort put into it than the JVM. At best, they're roughly similar, but I suspect that the JVM has quite a bit more development time and effort put into it.
- The_Colonel 6y ago
  Yep, just work on all those advanced JVM garbage collectors with various performance characteristics and trade offs is unmatched elsewhere in the industry.
- arghwhat 6y ago
  I did not say that jvm historically hasn't gotten a lot of resources, but any innovation done there has been common knowledge for compiler developers for ages. V8 gets to benefit from all that research, while simultaneously having far more resources (now as well as for the past many years) to apply it, as well as do more of their own.
  - dpratt 6y ago
    Perhaps I’m misunderstanding you, it seems like you’re implying that the JVM, as of today, isn’t the source of a lot of research or innovation. If that’s true, you’re incorrect. There’s a _gigantic_ amount of money, time and people involved with it, arguably more than there ever has been, and it’s only getting bigger. While you’re correct that some of the knowledge that came out of building the implementation of the JVM has found its way into V8, there’s still cutting edge work being done into extremely sophisticated JIT complication (with graal, arguably the best in the world) and the recent work on GC is frankly stunning. With recent JVM versions running zgc, it’s typical to see sub-millisecond GC pauses, and that’s only going to improve as they add more collection outside of safepoints and adding generational logic to it. Net/net, while you’re correct that there’s a lot of innovation and resources behind v8 today and they’re doing fantastic work as well, don’t discount the JVM. Yes, Java itself isn’t a hugely interesting language, but the JVM is still arguably one of the most impressive bits of software ever produced and continues to be so.
    - wenc 6y ago
      > There’s a _gigantic_ amount of money, time and people involved with it, arguably more than there ever has been, and it’s only getting bigger. Just wondering why this is the case? Who are the sponsors for GC research? (HFT outfits?)
pjmlp 6y ago
Pretty sure you don't know how many JVM implementations are out there since around 2000. By the way does V8 already handle TB sized heaps with 1ms pause? Or is able to scale down into embedded devices with hundred KB, while respecting real time constraints? Or able to have an implementation in JS itself? Or does V8 cache JIT code across sessions and improves it via additional PGO data gathered across the sessions history? I am not speaking of the same implementation here, rather a taste of what is out there on the Java world.
- arghwhat 6y ago
  You are speaking of many different JVM implementations which are all specialized for one purpose. This is good and all, but splits finite resources even further without a shared benefit, and these variants have as much to do with each other as V8 has to do with Spidermonkey and JScript.
  - pjmlp 6y ago
    Competition is what drives progress. Monocultures stagnate and deple resources as everyone is happy with good enough.

You're probably right about Graal, but there is no way that V8 has significantly more effort put into it than the JVM. At best, they're roughly similar, but I suspect that the JVM has quite a bit more development time and effort put into it.

The_Colonel 6y ago
Yep, just work on all those advanced JVM garbage collectors with various performance characteristics and trade offs is unmatched elsewhere in the industry.
arghwhat 6y ago
I did not say that jvm historically hasn't gotten a lot of resources, but any innovation done there has been common knowledge for compiler developers for ages. V8 gets to benefit from all that research, while simultaneously having far more resources (now as well as for the past many years) to apply it, as well as do more of their own.
- dpratt 6y ago
  Perhaps I’m misunderstanding you, it seems like you’re implying that the JVM, as of today, isn’t the source of a lot of research or innovation. If that’s true, you’re incorrect. There’s a _gigantic_ amount of money, time and people involved with it, arguably more than there ever has been, and it’s only getting bigger. While you’re correct that some of the knowledge that came out of building the implementation of the JVM has found its way into V8, there’s still cutting edge work being done into extremely sophisticated JIT complication (with graal, arguably the best in the world) and the recent work on GC is frankly stunning. With recent JVM versions running zgc, it’s typical to see sub-millisecond GC pauses, and that’s only going to improve as they add more collection outside of safepoints and adding generational logic to it. Net/net, while you’re correct that there’s a lot of innovation and resources behind v8 today and they’re doing fantastic work as well, don’t discount the JVM. Yes, Java itself isn’t a hugely interesting language, but the JVM is still arguably one of the most impressive bits of software ever produced and continues to be so.
  - wenc 6y ago
    > There’s a _gigantic_ amount of money, time and people involved with it, arguably more than there ever has been, and it’s only getting bigger. Just wondering why this is the case? Who are the sponsors for GC research? (HFT outfits?)

pjmlp 6y ago

Pretty sure you don't know how many JVM implementations are out there since around 2000. By the way does V8 already handle TB sized heaps with 1ms pause? Or is able to scale down into embedded devices with hundred KB, while respecting real time constraints? Or able to have an implementation in JS itself? Or does V8 cache JIT code across sessions and improves it via additional PGO data gathered across the sessions history? I am not speaking of the same implementation here, rather a taste of what is out there on the Java world.

arghwhat 6y ago
You are speaking of many different JVM implementations which are all specialized for one purpose. This is good and all, but splits finite resources even further without a shared benefit, and these variants have as much to do with each other as V8 has to do with Spidermonkey and JScript.
- pjmlp 6y ago
  Competition is what drives progress. Monocultures stagnate and deple resources as everyone is happy with good enough.