Live data from Hacker News

Spidermonkey has passed V8 on Octane performance

robert.ocallahan.org

311–320 of 338 posts

Re: Spidermonkey has passed V8 on Octane performance

#311
post #116
post #99

Earlier quoted context omitted.

+1 for IE's touchscreen support. I keep holding out and using Chrome on my laptop for a few extensions and all of the bookmarks/passwords I have synced over the years, but IE makes Chrome look absolutely kludgy and clunky when it comes to performance and UI these days. It's a shame IE doesn't have a more approachable extensibility model...

Does IE run on anything but Windows these days? (Edit: Doesn't look like it. Alas.)

You could always scrounge up a copy of IE5 for Mac ;)

Re: Spidermonkey has passed V8 on Octane performance

#312

Earlier quoted context omitted.

> It's not an accident that Google's webservices work best (sometimes only) in Chrome. A web service that only works in Chrome? Maybe you mean web application (web service would be really odd to work in just one browser). Do you have a source for this regardless? I hadn't heard of this.

I mean the services Google provides to users in the form of web applications, yes. The terminology is sucky. As for concrete examples, Hangouts only works in non-Chrome browsers (including ones with WebRTC support) if you install a Google-provided binary blob. Which you may not be able to do. Gmail only supports offline access in Chrome (see https://support.google.com/mail/answer/6557?hl=en the "two exceptions" bit).…

This is a huge problem for the overall strength of the open web and Mozilla unfortunately is no less guilty of this. Many of the tools developed for FxOS are targeted for Gecko and wont run on other rendering engines. More and more it seems the only people actually building libs for the open web are independent developers and small shops. :(

Re: Spidermonkey has passed V8 on Octane performance

#313

Earlier quoted context omitted.

It is effectively impossible to have an unsafe memory access error in Rust Please, don't say that. Lets be reasonable. Rust is great, but Rust code crashes and segfaults too, just like any language, programs have bugs, and those bugs can cause program failure. Rust just has less of them, because it has a smart compiler. ...but it's not right to suggest that it has none. Remember: There is no way of ensuring that a ru…

I don't understand your post. If you mean 'unsafe' as in 'may misbehave' then the compiler can definitely ensure there will be no segmentation faults and there cannot be certain classes of race condition. If you mean 'unsafe' as in the keyword that lets you do dangerous things, it's trivial to check if your source has blocks marked with that keyword. There's no way to ensure the compiler is perfect, but rust itself i…

This is a misconception.

You are flat put wrong, and spreading misinformation about it doesnt help anyone.

If you use any rust, and that includes dependencies and the standard library with unsafe code, bugs in the unsafe code can and do cause segmentation faults.

Its easy to say, 'well, thats a bug in the library, not a problem with rust', but thats the same as with C++ isnt it? If you can assert any code is 100% bug free then why do we care about the nice safety features in rust?

What is true is that any 'safe' code path that never enters an unsafe block in rust proveably cannot result in certain types of failures.

BUT every rust program uses unsafe code. In the standard library. In c bindings. In 'safe' pure rust dependencies (with hidden unsafe blocks). In loading dynamic libraries.

Its completely unavoidable.

What are going to do? Vet every line of every part of every dependency in the code you use? Dont be ridiculous.

Do you use rust?

..because practically speaking it does crash. Not often, sure. ...but this falacy that rust is 'provably safe' is absolutely false. Its provably false.

Thats why people saying it is unfortunate; it makes the rust community look like a bunch of clueless fanboys.

Please stick to reality. Rust has a zero cost memory management strategy and a smart compiler that helps to prevent certain types of common errors.

We dont need to step into magical fairy land to convince people rust is good. It stands on its own merit easily enough.

Re: Spidermonkey has passed V8 on Octane performance

#314

Firefox's hopeless video and audio implementation (especially on Mac OS X) remains a deal breaker for me. It means resorting to Flash in a lot of cases, which means I use another browser. I'm glad to see strides being made in JavaScript performance, but basic browser functionality is far more important.

HTML5 video and audio in Firefox works perfectly in every site I visit. What problems do you hit? I haven't used it on Mac so perhaps the support is less good there. Some sites don't attempt to use / in Firefox for me, but that's a site issue, not a browser issue - h264 and webm both work fine.

MP4 video support is completely absent on Mac. And MP3 audio is also not supported.

Re: Spidermonkey has passed V8 on Octane performance

#315

Earlier quoted context omitted.

I don't understand your post. If you mean 'unsafe' as in 'may misbehave' then the compiler can definitely ensure there will be no segmentation faults and there cannot be certain classes of race condition. If you mean 'unsafe' as in the keyword that lets you do dangerous things, it's trivial to check if your source has blocks marked with that keyword. There's no way to ensure the compiler is perfect, but rust itself i…

This is a misconception. You are flat put wrong, and spreading misinformation about it doesnt help anyone. If you use any rust, and that includes dependencies and the standard library with unsafe code, bugs in the unsafe code can and do cause segmentation faults. Its easy to say, 'well, thats a bug in the library, not a problem with rust', but thats the same as with C++ isnt it? If you can assert any code is 100% bug…

How often has one of your Rust programs segfaulted for you, when it wasn't a bug in your own unsafe code (as opposed to that in the standard library)? For me, the count still stands at zero.

I've been writing Rust for well over a year. I like to abuse new features and I've found many compiler bugs, but my code doesn't crash at runtime.

I'm not saying it doesn't happen, and the plural of anecdote is not data, but I think you're grossly misrepresenting Rust's practical safety benefits. That you only have to trust code in unsafe blocks, rather than all the code everywhere, is a huge benefit.

Re: Spidermonkey has passed V8 on Octane performance

#316

Earlier quoted context omitted.

It has the best developer tools. You just have to be careful to not use the es6 features that aren't in stable Chrome yet.

Is there somewhere I can read about new dev tool features that have not yet made it to Chrome?

@addyosmani [1] is a Chrome DevTools engineer who often talks about new features. That's probably about the easiest way to find nightly updates without following a mailing list.

1. https://twitter.com/addyosmani

Re: Spidermonkey has passed V8 on Octane performance

#317

Earlier quoted context omitted.

I don't understand your post. If you mean 'unsafe' as in 'may misbehave' then the compiler can definitely ensure there will be no segmentation faults and there cannot be certain classes of race condition. If you mean 'unsafe' as in the keyword that lets you do dangerous things, it's trivial to check if your source has blocks marked with that keyword. There's no way to ensure the compiler is perfect, but rust itself i…

This is a misconception. You are flat put wrong, and spreading misinformation about it doesnt help anyone. If you use any rust, and that includes dependencies and the standard library with unsafe code, bugs in the unsafe code can and do cause segmentation faults. Its easy to say, 'well, thats a bug in the library, not a problem with rust', but thats the same as with C++ isnt it? If you can assert any code is 100% bug…

I guess I should have been more explicit that I was talking about the language itself (without the unsafe keyword). Not libraries written in other languages, not compilers. It's the job of theorem provers and whatnot to make those safe.

In other words I'm talking about code the programmer makes themselves.

>Its easy to say, 'well, thats a bug in the library, not a problem with rust', but thats the same as with C++ isnt it?

In C++ nothing prevents the lines I write from having memory errors. It's not the same at all.

>If you can assert any code is 100% bug free then why do we care about the nice safety features in rust?

Oh well you shouldn't do that, but you also don't have to use unsafe code willy-nilly. In C++ everything you touch is unsafe unless proven otherwise.

>What are going to do? Vet every line of every part of every dependency in the code you use?

For unsafe blocks? Sure, that's easy.

Re: Spidermonkey has passed V8 on Octane performance

#318
post #304

Earlier quoted context omitted.

SpiderMonkey itself actually has no concept of windows at all. So it doesn't know what window anything has to do with. The browser itself does, though. Some sort of "top"-like utility is in fact being worked on, for both websites and extensions.

If a script hangs, Firefox has a little popup on that tab that asks if you want to kill it or let it keep running. So something in the browser knows which tabs are JS-intensive.

Sure. That particular setup has pretty low overhead and hence also low granularity; it's not measuring intensive or not but just how long a script has been running when a timer on a background thread fires. So your script could be eating up all the CPU but just returning to the event loop every few seconds and the hung script dialog would never notice.

Re: Spidermonkey has passed V8 on Octane performance

#319
post #306

Earlier quoted context omitted.

I've been using FF for a while, but partly switched off when I got a Surface (on account of IE's better touchscreen support). But soon I'll be moving from iOS to Android, the wonderful land of being able to change your default browser. And I'll be using FF there for the bookmark syncing. Firefox on 2/3 devices ain't bad. There was a lot of hate over the australis redesign, but I like it better than Chrome.

I actually think that while its very debatable on desktop, firefox for android may be the best all around android browser. Now for surface I hear you - i was sad they discontinued the work for that. Even on touchscreen latpops.

Even with no Metro version of Firefox, the desktop version could be better. Touch scrolling in IE for desktop feels like it does on a smartphone. Firefox is jerkier and doesn't get momentum quite right.

Re: Spidermonkey has passed V8 on Octane performance

#320
post #198

Earlier quoted context omitted.

It doesn't matter. (disclaimer: googler) I see it as a hubris against the idea that Google claimed to not be evil, and the fact that evil is such a poorly defined concept, and whether or not Google meets the bar depends entirely on the values of the perceiver. Many perceivers, for instance, miss the distinction between "Don't be evil." and "Don't do evil.". The former is a mindset, strategy, and intention, while the…

What's the mindset that ends up with Chrome going out of its way to swap OK/Cancel when you enable DNT? (I don't think DNT is a real solution, but I'd guess Google wants to be able to point to data showing users don't enable it.) What's the mindset behind the broken permissions on Android? Where any app that wants to change behavior when you get a call must request permission to your IMEI and calling/called number? O…

All of these things are designers trying for one user-centric goal and having it backfire in ways that were not predicted.

None of these things say "evil" to me even remotely.

Post reply on HN