(Yes, I know, it depends. I'm just looking for a rough estimate.)
After refactoring each opcode handler into separate functions, it now runs with several subsecond delays on startup (as the various code paths are discovered and optimized, I suppose) and with no delays after a minute or so.
WebAssembly won't have this problem (neither does asm.js, but that's a different story...)
WASM should run at native speeds minus some special CPU instructions like vector stuff. So about as fast as highly optimized bytecode languages like Java, maybe 20% slower than optimized C.
Edit: an even longer list of proposed features. http://webassembly.org/docs/future-features/
Some benchmarking has been talked about before [0], the upshot being C->WASM has a slowdown around 16-25ms. That's a hell of a lot faster than JS in a lot of places.
Unfortunately, as soon as you hook into the HTML APIs, like giving JS a result, you're back to JS speeds.
One big benefit will be being able to compile from an array of languages down to WebAssembly and maintain performance.
1. Parsing text based JS is much slower than parsing something like WASM.
2. Initialising data required for programs can take significant time in JS as that data is just a program that must be parsed and run.
3. Those first two points, along with better type support should help things get fast quicker, as the parser and JIT have less work to do.
4. In general JS comes with a lot of baggage that it's extremely hard to get rid of at this point, in many ways it's not really a great foundation to build other things on.
Now, WASM seems to provide a pretty good basis for building many things low level things on, but less useful for higher level languages which might want garbage collection or other facilities. I hope we'll end up with a successor or an evolution of it being the primary thing browsers interact with and JS simply being a language supported on it.
Would it be possible to implement a garbage collector in WASM? And could it be a concurrent garbage collector, using advanced techniques such as memory barriers?
It would be cool if we could run e.g. Haskell or Go in WASM. Or even C programs which rely on the Boehm collector.
Its single-threaded performance is essentially native, but I don't have the numbers.
There's the Alan Kay attitude that we can build entire systems on top of minimal VM's, rather than sacrificing power to bake in more facilities. But even he says that too much time has been wasted by teams reinventing the wheel (or flat tire) who didn't really have the chops to do it.
The other side of that, then, is the great potential for interop that Javascript offers. We now have a worldwide platform with a built-in presentation layer and a highly-optimized interpreter with useful, reflectable data structures out of the box. (edit i.e. it's a viable platform for metaprogramming... the endgame of which is JS-in-JS (see "prepack")).
I would love to see a future in which the browser (and "personal" computer generally) remained a locus of significant computation. In practice, I suspect that highly-intensive tasks (mostly AI stuff) will continue to be done on servers, not because we lack the processing power, but because end users will have signed away the custody of all data worth processing. If that is the case, then WASM's role will be to fill a fairly marginal gap, between what JS can do, and what has to be farmed out anyway.
So ultimately I hope that WASM offers a lifeline to the computing power that individuals still have. But given the state of JS (after huge investments), it feels like starting over again.
Regardless of arguments for and against whether this is a good idea, market pressure to open floodgates to other languages will be far too high to ignore.
My prediction is that WASM will quickly evolve into a new way to containerize applications, very close to a full-fledged VM/OS. The market constantly demands more features and faster execution. The only logical conclusion is that the web browser will become an operating system.
The market wants people to want "more features and faster execution." The market doesn't care what it's selling per se. If the web-browser-as-OS seems like an inevitability in that context, it's only a side-effect, and one that could change at any time.
WASM also impacts the labor market, as you mention. But it's not, as we might like to believe, because programmers' demands for more freedom and expressive power are being honored; it's because, as monkmartinez suggests [0] (and maybe this is what you mean by "floodgates"), companies would rather not be limited to a smaller human-resource pool when developing web properties.
If I sound a little dystopian, it might be because I'm currently reading Cyber-Marx [1], an insightful book from 1999 about the "information revolution" and its relation to capital.
[0] https://news.ycombinator.com/item?id=14342983
[1] http://www.press.uillinois.edu/books/catalog/66mwg3pc9780252...
Almost all of the popular programming language have a js transpiler allowing programmers to code in their lang and use it on web.
I think the reasons why programmers aren't using their preference when coding for web is, firstly because of the lack industry approval/support and secondly the speed of development. As much I complain about writing js code, it's a lot faster to get things done compared to Rust/C++.
The browser is one of the only truly cross platform development environments around. The popularity of the browser is directly coupled to that fact in my opinion. How many developers|companies and organizations would be over the moon excited with hiring a Python|Ruby|Lua|Rust|C# developer that could talk natively to the browser? My guess is a LOT.
Looks like Docker/Unikernel Systems actually has plans to build something like this: https://github.com/linuxkit/linuxkit/tree/master/projects/mi...
Also, its relation to javascript is not any bigger than the relation to any other programming language (i.e. there is no relation). A WebAssembly VM could be made as an extension of any environment.
I recommend reading the official paper https://github.com/WebAssembly/spec/blob/master/papers/pldi2... (I learned more than about 50 "essays" that I read about WASM in the past).
WebAssembly has two representations: a binary form .WASM and a text form .WAST (with one-to-one correspondence). The browser environment can convert between them (or you can write a short piece of code to do it).
If you know C, this website may help you get familiar with WASM syntax: https://wasdk.github.io/WasmFiddle/ (when you press Build, the WASM code will appear at the bottom).
As time goes on, more pages are devoted to nailing down corner-cases, and retroactively legislating either for or against assumptions had previously been common sense until someone in the real world violated them.
WASM will see such growth in time. Think of how simple HTML was once upon a time.
Specs grow much more slowly if, in response to such non-conforming implementations of a processing stage, they just codify the common sense, and maybe give methods for detecting the non-conforming outputs and rejecting them. XHTML is not a large spec.
There's still some work to be done, but it is quite far along already.
E.g. it's great for C, but not really for Python.
Does this worry anyone else? Or I am getting worked up over nothing?
Static languages can compile to wasm. Low-level runtime implementations of other managed languages which don't mesh well with JS backends can compile to wasm.
Webasm doesn't replace JS, it complements it.
Of all the things you've ever encountered called "assembly language", how many supported GC?
That said, I wonder how the WASM runtime is sandboxed. Not having a runtime-provided memory manager makes me worry not just about memory leaks, but also about explicit shenanigans in different memory zones. It's hard to believe that a WASM runtime doesn't protect against this sort of escaping though, so it's more of an academic wondering than a real concern.
Pointers into memory are 32bit base + 32bit offset, non-wrapping, so 33-bit distances from a base pointer. Each memory access needs a check to that (relatively non-changing) pointer. Memory spaces can request expansion in multiples of 64KB, and can have configured maximum sizes. This leaves it all compatible with using the MMU to trap accesses beyond allowed memory. The example from the paper linked elsewhere is that on a 64 bit machine, you can allocate 8GB of virtual address space, which covers the entire possible addressable range of wasm instructions, and use permissions to trap out of boundness. The very notion of allocating that 8GB already encapsulates memory sandboxing, as long as the addressability limits of instructions can't be usurped (and it's formally been shown to be sound).
Would I be able to call JS functions from WASM? or pass callbacks to WASM. Because WASM sounds like it goes agains the async nature of JS and I have the feeling that is going to be very annoying to merge both worlds.
Also if I cannot call any JS func from WASM it means I cannot call WebGL or Audio so I dont know what would I need the performance if I cannot access the most performance demanding APIs.
The interface should take care of all type conversion between wasm and JS primitive types (notably varying integer sizes and JS floats). wasm's 64-bit integers will likely not be allowed across a JS<->wasm interface.
If the browser vendors were going to add another language I would have hoped it would be something so generic like LLVM instead of this WebASM which is too tied to JS.
Except Chrome and Firefox, which both support WebAssembly.
JS today is often transpiled anyway, so you have to use some mapping to the source code to do debugging (eg. source maps).
I figured the built in browser debuggers wouldn't work so well for this.
I find a lot of value in using JavaScript as a cross-platform shell. I personally do a lot of work on all of the major operating systems. I primarily (a light primary, like maybe 67%, not 90%) work on Windows, but most of my users are working on macOS. And we're all writing software that eventually has to run on Android. There is almost never a problem I can't solve using a JS project of some kind. That's the rewarding part. I feel like my software goes a lot further on the JS platform than it could have ever gone on any other.
But it's frustratingly slow. A lot of that is the poor quality of a lot of the libraries that are available. I'm not really keen on writing my own HTML templating system, but I'm afraid I might have to just to get the performance targets I want (I static-gen almost everything).
And I often feel like the language is not doing enough to help me. I love writing metaprogrammable systems, but the lack of easily accessible, standardized, advanced type information in JS is a big impediment to that. Sure, I can enumerate all of the methods an object has available to it, but there is no good way to enumerate the parameters those methods take, their return type, or even that they are methods, if I ever happen to get a reference to one of them sans the object from which it came
And yes, I still do object oriented programming. But the functional story in JavaScript is not any better. With map, filter, and reduce being methods on Array, coupled with JS's goofy OO semantics where methods can become detached from their objects if they aren't explicitly bound to them, things become frustratingly verbose to get basic functional patterns to feel good.
So a WASM that let me build better dev tools, in better languages, without losing that cross-platform capability, would be a godsend. Having it run in the browser also is just icing on the cake at this point.
But then the next obvious problem will kick in. What should we use to implement a UI. GTK? WxPython? Swing? Tkinter? QT? JavaFX? Great candidates :) It will just be great. I mean not, but it will happen, because JS and "browser tech" is considered sub par (not by me, but I keep reading this).
WebAssembly will bring in lots of innovation, opportunities and also create the next level chaos. We will look back to the tooling hell as the good old days of frontend development :)
Sounds like an easy dump from AST to text to wasm. Aside from what the Google Closure Compiler provides, I wonder what benefit, if any, this may give to ClojureScript. I'm guessing CLJS will just keep Google Closure for its optimizations, while Google Closure will output either wat files or wasm files. Unless optimizations can just be done during wat to wasm, which would be ideal for other language authors who can just translate to their intentions and not have to worry so much about fancy optimizations. (I'm rambling at this point. Sorry.)
[1]: https://developer.mozilla.org/en-US/docs/WebAssembly/Underst...
The issue is that JIT js is hard to improve performance on because it's a dynamically typed language. Even with Web Assembly, something will still need to compile down into the bytecode to run. So perf is going to depend on that step in whatever language you write in and how easily that converts and optimizes into web assembly standards.
Does this mean that strongly typed languages have a performance advantage, given that they will be able to easily compile down into optimized typed bytecode?
