One VM to Rule Them All

http://medianetwork.oracle.com/video/player/2623645003001/

by pron • 13 years ago

42 51 13 years ago

medianetwork.oracle.com

chc• 13 years ago

So now it isn't only Oracle's detractors equating it with the Dark Lord.

toolslive• 13 years ago

well, to stop people from panicking; in this context "All" means just a few.

It seems they have been building their own CLR.

pjmlp• 13 years ago

No, the idea is to eventually integrate the work started with Maxime at Sun and replace the JIT developed in C++ by a Java based one.

This work has been discussed for possible integration into the standard JDK in a post Java 8 version, as part of project Graal.

trailfox3• 13 years ago

Why would people "panic" if Oracle builds them a better VM or the tools to build a better Ruby/Python/Java VM?

It's quite different to the CLR. The CLR lets you implement any language you like, as long as it can basically be mapped to C# constructs, not that different from the JVM.

What Snoracle is working on is more generic with much higher performance.

darklajid• 13 years ago

The CLR supports things that C# the language doesn't. C# isn't the common denominator here, it's just one of the languages running on the CLR (and probably? the most well-known). VB.Net and F# are equals, not to mention all the non-MS languages that exist.

C# isn't special and doesn't represent the CLR's potential. It certainly might drive the evolution of the CLR to a point.

saurik• 13 years ago

https://news.ycombinator.com/item?id=6232240

VeejayRampay2• 13 years ago

Does this mean that we can expect faster versions of the languages laid out in the presentation (Python, Ruby, etc)?

pron7• 13 years ago

To judge by the results they present for Ruby, I assume the answer is yes.

VeejayRampay2• 13 years ago

This got me excited yeah, but the main issue is always that of the ETA. I do love Ruby, wouldn't use any other language really, but damn would a little speed boost be nice.

And of course I want it for free. Once all the goodness for JRuby, Rubinius and that Oracle stuff get shaken for a long enough time, I guess we should see the benefits in our everyday-coding.

brokenparser2• 13 years ago

In related news, Oracle still doesn't support HTML5. Back to you, pron.

moondowner2• 13 years ago

Actually, with Java EE7, they do.

http://jaxenter.com/java-ee-7-officially-launches-bringing-h... http://docs.oracle.com/javaee/7/tutorial/doc/jsf-facelets009...

brokenparser2• 13 years ago

They obviously didn't update their website to Java EE 7 then. Why wouldn't they? It's the latest and greatest thing!

owenfi• 13 years ago

I guess the one VM is Adobe Flash.

spuz2• 13 years ago

Oracle is a database. What features of HTML5 have anything to do with databases? How is this related to the article?

cowsandmilk• 13 years ago

he's talking about their website... and oracle isn't just a database company the last I checked. I mean, they make sparc servers. And VirtualBox. And Java.

rbanffy• 13 years ago

Actually, Oracle is a company famous for its relational database. Also, it's notable for its hostility towards open-source and free software.

My bet is that their website must be accessible to the corporate desktops of companies that buy Oracle stuff. That means it has to work with IE6, most likely.

trailfox3• 13 years ago

In what way is this either "related" or "news", or even accurate?

derefr7• 13 years ago

So... can I run Erlang code on it efficiently? (This question also disqualifies LLVM, the JVM, the CLR, etc. When most people say "universal VM" they really mean "universal VM for running mostly-serial code that might at most have 5 or 10 threads, and where killing and restarting the entire VM (to, e.g., spin up a newer version) isn't a big problem for the clients talking to it." This is not future-proof.)

pron7• 13 years ago

You can already run Erlang code efficiently on the JVM: https://github.com/trifork/erjang

derefr7• 13 years ago

I can't tell from the precis whether it supports hot code loading (the JVM doesn't, so it won't without heavy magic); and it definitely doesn't work in a soft-realtime manner ("In Erjang, every node runs on a single heap, and so global GC will sometimes happen.") Without these two things, you're not "running Erlang code efficiently"; you're just running a pretty-useless-for-its-intended-purpose subset of the language.

It's a bit like those benchmarks that show that a language that's a subset of Ruby, performs much faster than actual Ruby. Well, yes, that's nice--but we use Ruby precisely for all the inherently-really-slow parts. Once you add all the slow stuff, and have to check whether to branch to it at every call, your speed advantage stops being nearly so impressive.

Erlang is a painful language to code in; you don't do it because it's pretty. You do it for what the BEAM VM and the OTP libraries give you: the ability to spawn 100k processes representing, say, people's phone calls, or their MMO game sessions, and not drop any connections for something piffling like a garbage collection pass, or a bugfix. I would expect no less of any "universal" VM. (Though I might, of course, expect more. It should probably be able to do CPU-bound stuff like large matrix multiplications without dragging the rest of the VM down, for example.)

ZoFreX• 13 years ago

While I agree with the other points, the JVM can cope with a lot more than "5 or 10 threads".

derefr7• 13 years ago

Sure, I didn't mean "what the JVM can do in extremis"; just "what the JVM's designers imagined the average-case use of the VM to be--the one worth optimizing and creating fast-paths for."

Most sequential programs that rely on native threads, whether in C or Java or whatever, just use 5-10 "big" threads, where each thread is still multiplexing work on its own large corral of data. In a game, for example, you'll see one "graphics thread", one "network thread", one "AI thread", etc. Not "one graphics thread per voxel"; not "one network thread per peer connection"; not "one AI thread per NPC". These designs are bad, in a traditional sequentially-oriented VM, precisely because they will blow up on you in the worst case (e.g. when you look to the horizon on a Minecraft server.)

sixbrx• 13 years ago

can I run Erlang code on it efficiently? (This question also disqualifies LLVM

Are you saying LLVM can't be used to efficiently implement Erlang-style processes/tasks? I thought Rust was aiming to be able to handle a large number of lightweight tasks, and its based on LLVM?

derefr7• 13 years ago

Erlang actually uses LLVM as a JIT, but it does so by doing the same layering-on of semantics that I talk about here (https://news.ycombinator.com/item?id=6359716). By itself, LLVM IR has neither a "send" instruction, nor a "call" instruction with yield-semantics, and LLVM offers no tail-call elimination support. The JIT has to do a lot of pre-munging (dynamic recompilation) to produce code that LLVM can treat "blindly" as sequential machine-code. (LLVM also has a specific calling-convention implemented just for Erlang code, similar to the one it uses for Haskell code, since both generally avoid mutable state, which confuses LLVM's regular register-allocator.)

Rust is a lot more like Go, with its multitasking primitives being based on channels (shared-memory atomic data structures) rather than message-passing-by-copying, and each actor/thread of execution allowing mutable state within itself, as long as its mutations don't leak across actor boundaries. In other words, it doesn't really change the semantics overmuch from that of the non-NUMA multicore Von Neumann machine (e.g. x86) it's running on; it just does "green threads" the traditional way, with some elegant builtins for getting data from one to another. Rust's and Go's semantics are both isomorphic to x86/LLVM/JVM/CLR semantics, as far as I can tell; you can translate directly from one to the other without any overhead from maintaining an implicit VM state machine (in est, a monad) across the compiled code chunks.

m_ram• 13 years ago

I read "Oracle" and "VM" in the title and immediately thought that VirtualBox was going to be OpenOffice/MySQL-ized. I don't know how VirtualBox fits into Oracle's strategy, but I'm glad that they're not messing with it.

nodata• 13 years ago

I read Oracle in the title, and immediately closed the tab.

trailfox3• 13 years ago

Always good to see such well reasoned technical discussion here on HN. You make a particularly well-reasoned point about the text in the title and how it affects the technical aspects of the VM.

a3n• 13 years ago

Me too, and I wonder what I'll do when Oracle finally works it's way that far down the alphabet. Docker/LXC? LibreBox?

gary4gar• 13 years ago

Interesting. Wondering, if the ruby implementation they wrote for demo available? Want to play around with it. How do I install it?

jaydub• 13 years ago

... and in the darkness (of Oracle) bind them!

SEJeff• 13 years ago

Isn't think supposedly what parrot[1] is supposed to be for? Granted the perl 6 stuff under parrot is practically a laughing stock amongst much of the tech community, but they are trying to make something awesome.

[1] http://www.parrot.org/

chromatic• 13 years ago

That was one of the goals of Parrot, but that P6 mess eventually chased off almost all of Parrot's developers:

http://www.modernperlbooks.com/mt/2013/02/goodnight-parrot.h...

icefox• 13 years ago

What about vmkit?

One VM to Rule Them All

Related Stories