Search results

It's a combination of them all. LLVM and emscripten do a lot of work, but also the JS engine on the user's machine can optimize asm.js more easily than normal JS (for example, having |0 etc. everywhere avoids JS engine overflow checks; "normal" JS has property accesses on objects which are hard...

asm.js runs fast even without those optimizations - faster than normal JS code. asm.js was inspired by seeing what already runs fast in JS engines, and improving on that. For example, asm.js was faster than normal JS on all browsers tested on box2d here...

Firefox is already 64-bit on OS X and Linux. It's just Windows that is taking a little longer (due to the usual stuff like 32-bit plugins on Windows, that a minority of users want).

Not true. First, the Firefox decline started far earlier than the recent interface change - basically since Chrome started to rise, soon after its launch. Second, Firefox usage has risen since that change, http://www.reddit.com/r/firefox/comment ... re/clt4fh1 The reason for the long Firefox...

This feels like a very unfair way to end the article. Have you seen such relatively-good under $50 devices? Then details would be interesting (and even better, a review). But if not, then just mentioning "oh, I hear other stuff exists in this space that is way better" is just speculation...

Could be. Still, there are roughly 1/3 of chrome users now on an obsolete version - is it possible so many chrome users are in large corporations with strict IT policies? That seems high, but I'm just guessing.

I think that 'has' should be a 'had'.

What if it turned out that carefree people click less on ads than serious people - would it be immoral for Facebook or Google to show content that made people slightly more serious and slightly less carefree? Because if that connection actually holds up, I think we know they're doing it already...

I think we know for a fact that Facebook (and basically every other complex social-related site on the web) has been doing such studies for many years. Every time Facebook does A/B testing on how a change to the data it presents to users affects their behavior, it's doing an "experiment". And...

First thing, that website is just beautiful - perfect design and behavior. Fast and slick. Second thing, great news all around. In particular I'm happy to hear about Web Audio coming to IE, very important for games and related things.

Exactly, it manages it's own memory for compiled C/C++ code itself, using a compiled malloc/free implementation (that, just like asm.js in general, runs close to native speed). So the JS GC is not involved at all.

See previous reply for benchmark details. Those are all CPU benchmarks (GPU performance is already very close to native already, as WebGL basically just maps to the OSes' GL driver). In asm.js code, garbage collection is negligible as no garbage is created by the compiled code whatsoever.

There are a few sources I am aware of for the 67% figure, for example https://hacks.mozilla.org/2013/12/gap-b ... mizations/ https://github.com/joelgwebber/bench2d/ ... /README.md The first is a set of benchmarks ranging for specific microbenchmarks up to realistic codebases like box2d...

I think it's unfair to call it "disingenuous" just because you think it might mean something else. But more to the point, it is in fact 100% accurate, the 60fps figure includes full 3D games like the ones seen in that video. They run just as good as the video shows, even on standard hardware...

First of all, this is done by the Internet Archive. It's a non-profit with a respected history of preserving important parts of digital history. There is no motive for profit here, which is important to keep in mind, and that should matter not just to us. Second, they explicitly mention...

Is using a proxy like this a good solution, though? The proxy is run by a single individual in Singapore. No matter how smart and honorable that person is, it's a single point of failure. Hacking that proxy gives the attacker all the activity of everyone using that addon.

Actually asm.js should be able to beat Java or similar VMs in many cases, because it is more comparable to C++ than Java or C# (and so it does not have the overhead of bounds checks on individual objects, GC, etc.). For example here asm.js beats Java on Box2D...

Those sentences are contradictory. If you use threads in PNaCl, then the code is not portable through compiling it with emscripten to asm.js anymore. So the advantage you just mention negates the previous statement. Do you have numbers on that? The best ones I am aware of show the opposite...

NetApplications and StatCounter are not measuring the same thing. One measures unique users, one measures page views. They might both be right.

There is no need to explicitly detect the "use asm" hint. It can help do AOT compilation, but v8 might be able to reach similar stable-state performance without it. It's getting closer: http://www.arewefastyet.com/#machine=11 ... asmjs-apps It is not another compiler, and doesn't work in a...