Apple Looking to Boost Safari JavaScript Performance with New Accelerator Upgrades

[mrxak]

What I want most from a browser these days is a low memory footprint per tab and stability. I'm sure JavaScript is plenty fast enough and won't really be all that noticeable. I'm not against performance gains, even slight ones, but in terms of improvements I'd like to see them devote more resources to support for very large numbers of tabs.

 

[Parasprite]

I know what everyone is thinking... can it run Crysis?

What is this, 2011?

 

[MikhailT]

What I want most from a browser these days is a low memory footprint per tab and stability. I'm sure JavaScript is plenty fast enough and won't really be all that noticeable. I'm not against performance gains, even slight ones, but in terms of improvements I'd like to see them devote more resources to support for very large numbers of tabs.

Not going to happen, memory is cheap, and there are reasons for the increased memory consumption. Each tab is now its own sandboxes process, which means almost nothing is shared between tabs, and that automatically means the memory consumption will remain high for good. Each tab has to process several things like CSS, HTML, and JavaScript, all of which takes up space in memory.

 

[MacVista]

It is unclear whether Apple is targeting OS X 10.10 for these JavaScript improvements

Wanna bet??
I can see it coming:
System Requirements:
OS X 10.10
Macintosh Mid 2014 or later*

*All others, will be slowed down via automatic update or via lack of patch

 

[mdriftmeyer]

The entire code base is being C++11 transitioned and LLVM/Clang enabled. Moving to the LLVM backed JIT Javascript backend will improve matters greatly.

The continued additions of OpenCL enabled parts and OpenMP 4.0 when it lands in LLVM proper will also improve the state of WebKit.

 

[MacLC]

On my machine, Snow Leopard with Safari 5 is faster than Mavericks with Safari 7.

Though I guess talking about going back to SL is more taboo than talking about something being "snappy".

 

[Piggie]

I don't know if real or being sarcastic. Not enough emoticons used.

Oh no. 100% serious.

I'm not sure as I cannot remember exactly the details hence me asking.

I thought it was something about Apple, in iOS kept special (routines let's call them, thought that is the wrong term) locked down and away from 3rd party browser devs, so they could never get the speed that safari could offer users.

Can someone please clarify this point and if it's still true?


Here you go. See I'm not just making things up....................

http://gregstechblog.blogspot.co.uk/2013/05/the-problem-with-3rd-party-ios-browsers.html

Apple hasn't allowed these third party developers to use their engine so they can continue to claim Safari is the fastest browser for iOS. Of course it's true, but it's an artificial victory.

 

[MacSince1990]

I'll wait for Firefox 31 which will be out the day after tomorrow.

And still superior to both Chrome and Safari >_>

 

[Sasparilla]

Any improvement to Safari is welcome. It is my favorite browser and although it is notably faster with version 7 it still has this issue where it will randomly hang and I must quit and reopen it to regain performance...

I see this in Mountain Lion as well and its annoying enough that I'm considering changing from Safari as my default browser on my Mac (haven't been pushed over the edge yet - but you'd think Apple would fix this, its a problem other browsers don't manifest in the same environment).

I'll take other improvements as well of course.

 

[McDave]

Apple should end these browser wars by extending the A8 instruction set to execute HTML5, CSS & JS on silicon. I don't get why Apple haven't explored a custom instruction set for core iOS/OSX technologies.

 

[andreiru]

Excellent, javascipt is what's dragging it down performance wise as well as being a big battery hog

 

[Traverse]

I see this in Mountain Lion as well and its annoying enough that I'm considering changing from Safari as my default browser on my Mac (haven't been pushed over the edge yet - but you'd think Apple would fix this, its a problem other browsers don't manifest in the same environment).

I'll take other improvements as well of course.

Yes it is annoying, but I don't want to switch. I am Leary of Chrome and Firefox's OS X integration is just laughable. At least Google Chrome stays updated with the latest OS 10 features, like rubber bands scrolling, notifications, etc.

 

[cjbryce]

The only people who think/say that only heard it from someone else and have no idea what they are talking about.

Ok, I'm sorry. My obscure reference to the (seriously) bad film (released 1980) was obviously a tad too obscure

For the record Flash (as in Adobe nee Macromedia) was a tool I "created" with, briefly, about 15 years ago. Now I voluntarily block it in on all OSs.

----------

And still works perfectly on PCs today

Full screen, 1080p Flash, very low system load. Never a problem

Ok, I'm sorry. My obscure reference to the (seriously) bad film (released 1980) was obviously a tad too obscure

For the record Flash (as in Adobe nee Macromedia) was a tool I "created" with, briefly, about 15 years ago. Now I voluntarily block it in on all OSs.

 

[G4DP]

Not going to happen, memory is cheap, and there are reasons for the increased memory consumption. Each tab is now its own sandboxes process, which means almost nothing is shared between tabs, and that automatically means the memory consumption will remain high for good. Each tab has to process several things like CSS, HTML, and JavaScript, all of which takes up space in memory.

Other browsers manage 100 tabs fine without requiring GB's of RAM. So why can't Safari. The same reason all of Apples own consumer software is terrible. Pee poor coding.

Any improvements are most welcome, just a shame they are 5 years behind.

 

[Alenore]

But who will set the standard?

 

[Count Blah]

Can apple please fix safari on my iPad Air? Having two tabs refresh, with no other apps running, is really pathetic and a horrible use experience.

 

[wizard]

Any improvement to Safari is welcome. It is my favorite browser and although it is notably faster with version 7 it still has this issue where it will randomly hang and I must quit and reopen it to regain performance.

In one of those random things I downloaded WebKit last week to try it out. Not having done so in probably a year or more, I was shocked it is markedly faster; this on an old 2008 MBP. So I got on the WebKit blog to see what is up and frankly there is a lot happening to make Safari snappy. I could start a list but the blogs explain it in more detail then I could here. To put it bluntly there is more happening than JavaScript improvements.

BTW, that Safari icon matches the Gold iPhone 5S!

I do wonder if these improvements can be ported to iOS devices. It would be neat to see LLVM shipped as a base package that all apps could use in iOS. Apple needs better support for scripting and other features that LLVM could help with.

----------

What I want most from a browser these days is a low memory footprint per tab and stability. I'm sure JavaScript is plenty fast enough and won't really be all that noticeable. I'm not against performance gains, even slight ones, but in terms of improvements I'd like to see them devote more resources to support for very large numbers of tabs.

I think you will be pleased with what they are doing in WebKit. The WebKit blog has some interesting charts that show the results if reducing executable size and other improvements.

As noted I was running a WebKit nightly last week and was seeing very good behavior for that short run. That isn't all day testing but the machine was heavily loaded at the time. I'm really hoping these improvements make it into the next Mac OS release and hopefully the iOS release. I was impressed to say the least, not bad at all performance wise for a developers release.

----------

The entire code base is being C++11 transitioned and LLVM/Clang enabled. Moving to the LLVM backed JIT Javascript backend will improve matters greatly.

The continued additions of OpenCL enabled parts and OpenMP 4.0 when it lands in LLVM proper will also improve the state of WebKit.

I urge anybody that uses Safari and is interested in the next release to read the blog. I think its fair to call current WebKit a major overhaul, as you note they are using new technology and are refactoring the code to use better practices. If everything I saw in last weeks WebKit makes it into the next Safari release I don't see anybody being disappointed.

 

[Traverse]

Image

That is funny, and true!

----------

In one of those random things I downloaded WebKit last week to try it out. Not having done so in probably a year or more, I was shocked it is markedly faster; this on an old 2008 MBP. So I got on the WebKit blog to see what is up and frankly there is a lot happening to make Safari snappy. I could start a list but the blogs explain it in more detail then I could here. To put it bluntly there is more happening than JavaScript improvements.

You got that from webkit.org?

Is this just like having a second Safari, but one that is bugger but updated far more often?

 

[wizard]

Apple should end these browser wars by extending the A8 instruction set to execute HTML5, CSS & JS on silicon. I don't get why Apple haven't explored a custom instruction set for core iOS/OSX technologies.

How do you know they haven't? Think about it, they went all in on their own ARM hardware, they may have future plans we don't know about. While not new instructions Apple has demonstrated a willingness to leverage the architecture to their advantage, for example by packing data into pointers. So I could see them adding instructions for their specific needs in the future.

However if they did do that, I doubt Apples custom instructions would have anything to do with JavaScript or other features directly. It isn't at all impossible because Apple has been patenting processor features for some time now and some of those are designed to improve the execution of apps and the OS. So I'd have to say Apple has explorers CPU customization to accelerate certain things. Do an Apple patent search for the lays few years and see what is up.

 

[Michael Goff]

That is funny, and true!

----------



You got that from webkit.org?

Is this just like having a second Safari, but one that is bugger but updated far more often?

I really wouldn't say it's buggier, but it definitely is updated more frequently. At times, I've had the old Safari crash tabs more often than a fully updated WebKit Nightly.

 

[wizard]

That is funny, and true!

----------



You got that from webkit.org?

Yeah. Search for WebKit nightlies.

Is this just like having a second Safari, but one that is bugger but updated far more often?

It is like having a development version of Safari. Understand what that means if you go and download it. By virtue of what they are, development snap shots, you may not have a stable version at all or it might be reasonably robust. There are no assurances as to functionality. So if you aren't willing to experiment don't bother as you are basically a tester.

On the flip side Webkits nightlies do not displace the system Safari so if you are doing important things it is best to fire up the system supplied Safari. Since the quality or usability varies nightly, you may find many releases are not worth the trouble.

I hope this is balanced. The most important thing to realize here is that the software is not stable.

 

[bumblebritches5]

The reason I don't use Safari is due to the iCloud Keychain messages that keep popping up. If I press the "?" on the message box for the Keychain,it also crashes Safari. Anyone know how to stop this popup?

Dudeeee seriously if you go to a login page and have autofill turned on it will literally ask a dozen times if it has permission to autofill, it's insane.

 

[wizard]

I really wouldn't say it's buggier, but it definitely is updated more frequently. At times, I've had the old Safari crash tabs more often than a fully updated WebKit Nightly.

So true! The important thing is for people to realize that you are basically testing software under development and as such there will be problems. If run into nightlies that I couldn't even use. It isn't a big deal if your expectations are aligned properly with the idea of being a tester.

 

[bumblebritches5]

This is good news. I had high hopes for HTML5. But when Google decided to fork HTML5, I was concerned that Safari might fall behind. Or at the least, we'd be end up with the old browser wars where web pages would looks completely different from browser to browser. I wish they'd just all play nice together. With Apple incorporating support for Mozilla's JS, maybe there's hope...

HTML is a standard, you're talking about WebKit, the rendering engine that google forked about a year ago into it's own project called Blink...

----------

"FTLJIT"? Faster Than Light Just In Time?

That's just a wild guess, but knowing Apple's penchant for modest, understated marketing..

I thought the same exact thing, but apparently it stands for "Fourth Stage LLVM", but I don't believe that for a minute, it stands for Faster than Light adn they're just lying to seem modest.

----------

very low system load

lol

 

[wizard]

I actually do think it's now enabled in the Webkit nightlies without any cli options. I just ran google Octane and the nightly was a whopping 49% faster than Safari 7.0.3 and even 3% faster than Chrome 34. This was not the case a month or so ago when I last tested.

This is going to make a much bigger difference when it lands in Mobile Safari.

Well something is up because I was experiencing notably better performance. It was impressive after going more than a year without a nightly installed.

I totally agree that mobile Safari needs these improvements. I'm just not sure all hardware has the CPU chops to pull off these improvements. Further iOS devices need more RAM.