Enough already. I’ve been asked a number of times what I think about Flash beeing killed by HTML 5 and I’m really getting tired of the whole discussion (which to begin with is not very interesting). That’s why I’m writing this down in a post so as to spare me from answering the same dull question once and for all.

Those seeking a quick answer to whether HTML 5 is going to kill Flash, the answer is: “No”. A slightly longer answer would go: “No, but I hope it did”.

iPhones, iPads

So the iPhone and the iPad does not support Flash, and quite likely never will. So what? At the time of writing there is no single mobile phone to be bought by consumers that support a full Flash experience (to me that is Flash 10 within the browser). So, if there’s no single mobile phone today that supports the Flash experience, how come Flash is still alive? The mobile web is still different from the desktop web, and it’s going to stay that way for a while still. People use the mobile web for different purposes and I really don’t think that to this day many have missed Flash on their mobile devices. The iPad is a different story, as the experience and use scenarios come close to the desktop web. But I think that the absence of Flash on the iPad is Apple’s loss, not Adobe’s.

Video

Whenever a major newspaper with inexperienced tech journalists write about how Flash is being killed by HTML they always refer to the video embedding capabilities of the spec. And nothing else really. Somehow someone has been able to seed the perception that Flash = web video. Within that false perception, HTML5 might actually kill Flash (in the distant future). Somebody will create a fantastic plug-in for jQuery UI that puts lovely controls around all video tags on a page, and that will become the norm. For that to happen, however, most users on the web will need to surf on a HTML5 enabled browser. Usually “most users” refers to more than 90% of all users (after which developers can start to remove any gracious fallbacks that involve Flash). If you look at the market share of IE6, which was released ages ago, you know it’ll take a while to reach that 90%.

Silverlight

The fact that Microsoft has created Silverlight and still keeps on pushing it hard is actually a case for Flash surviving the “onslaught of HTML5″. If MS felt that they could not compete with HTML5 using a plug-in technology that is far less spread than the Flash Player, they would not pour millions of dollars into it’s R&D. Nuff’ said.

But I wouldn’t mind if it did kill Flash

On the other side of this argument are Flash developers, screaming out loud things like “To arms! Flash is under attack”. WTF? Let’s imagine that HTML5 (or whatever else) would actually kill Flash.

For this to happen, there would need to be a technology or language that would be widespread amongst internet users, embraced by developers and thus superior to whatever it is that Flash is doing now. Metathesiophics (people who fear change, yes I had to look that up) aside, wouldn’t that be great? Whatever comes along, it’s bound to be programmable by something which reminds us (or is) of ECMA-Script or Java, so there’s really no steep learning curve for any Flash developer, especially as change does not come overnight, but over many years (sad but true).

So yeah, I hope something kills of Flash really quickly.

I thought of writing a post on how the Flash Player absolutely sucks on OS X. Jeah, people have been writing a number of related posts, but I didn’t think anyone actually grasped the severity of the issue. Flash 10.0 in Safari 4 on Snow Leopard (64 bit, FP running in it’s own process) actually becomes what I would say unusable. Even if you don’t do much processing, frame-rates plummet by at least a factor of four compared to a Window XP setup with the same hardware. And sure, while Firefox on OS X has run Flash sort of okeyish, it has never been able to compare against a Windows counterpart.

That is, up to now.

Tinic had a great post on changes they’ve made to how they composite Flash Player 10.1 renderings in the browser. The changes truly have made a big difference. Compared to 10.0, 10.1 screams on Safari, and is also a bit faster on Firefox.

I’ve never seen a product being talked about this much before actually beeing introduced as the iSlate/iPad/iTablet. Compared to this, the iPhone rumors were just mere whispers in the wind.

Now, can anyone really expect for the tablet to actually live up to the expectations, unless the “one more thing” is that it actually self-levitates.

We’ll have to see next wednesday.

Update: Well, it didn’t self-levitate. What a bummer.

Originally I had this idea more than a year ago, but did not start to experiment with some code untill last summer. I was about to wrap everything up this christmas, when I stumbled upon Sleep cycle alarm clock, which was released about a month ago. They’ve sold over 100.000 copies in the first month. Lazy me, could have beat them to the store had I tried a bit harder. Well, life sucks.

Anyways, I do appreciate what I learned from this “project”. It all came together much easier than Web Feeds, the first iPhone app I created one and a half years ago. Go ahead, give it a go, it’s not that expensive

Update: I pulled the app from the store. I realized that I need to figure out a better, more descriptive name than “Good Night”. Unfortunately the app name cannot be changed once it’s submitted.

Update 2: It’s available once more under the Sleep Phase wakeup clock name.

Whoa, long time no post about this subject. My apologies. This one won’t be long either, but I just had to share this freakin’ cool thing.

You might have read before that one big part of our MMOG Platform is the ability to write all code in AS3. Client-side as well as Server side. As there is no real AS3 server-side technology available we had to write our own. Our first version was up and running almost a year ago. Performance back then was OK, our tests showed us that our compiler could generate Java Byte-code from AS3 that ran as quickly as AS3 in the Flash Player, and included almost all core functionality of the AS3 scripting language.

Jump forward a year and now we have our second version of our compiler up and running (well, compiling). And boy is it fast. When you write decent code with type annotations and all, the compiler will churn out almost perfect Java Byte-code without pushing anything on the stack.

Performance?

Our simple (and yeah, a bit biased I must say) test suite with heavy calculations performed wonderfully. We ran the same test suite in the Flash Player and in it’s compiled form from the command line (compiling from the same source code of course).

All I can say is wow. Optimally, our AS3 to Java compiler produces byte code that runs 900 times faster than ActionScript in the Flash Player. 900 times.

This of course won’t hold true in the real world where we pass around closures, dynamic objects and other resource hogs. But from what we can gather it should be around 100 times faster even then. Stay tuned for some real world tests once we get to those.

I just moved, and started to compare prices for electricity. Vattenfall has a great number of options to choose from. You can choose with what method the electricity you’re buying has been produced. Nuclear, wind, water or “basic” (=nuclear, fossil, wind & water). Now, the pricing for wind generated power is a bit higher than for the other options, but what was really surprising was that the “basic” option, which is powered by fossil fuels in addition to clean energy is priced the same as electricity that has been generated 100% from water power.

My question is, why the hell does Vattenfall have an option that includes fossil fuels when there is a clean option at the same price point?

So I finally got an update to Web Feeds – an iPhone RSS reader app – send to Apple for review. It took quiet a while and yesterday I got a message that the update was rejected due to “infringement of Apple trademarks”. Huh?

I know the reviewers are very busy and probably have only minutes to test each update and look for suspicious activity. In this case their decision is rather sad. The application is preconfigured with some RSS-feeds users could be interested in, namely Huff-Post, Boing Boing, Engadget and, yes, Apple Hot News.

Now, when the app launches it fetches the feeds and looks for any FavIcons in the Feed’s domain like any web browser does. Of course the Apple Hot News feed loads up the grey-white apple FavIcon from apple.com. Enough for the reviewer to reject this update (oh, all the other 10 updates have gone through without anyone noticing, hmm).

Maybe I’ll just resubmit without the Apple Hot News feed…

NVIDIA took the big surprise out of tomorrows MAX keynote by adding a pamphlet to the MAX swag bag that everyone received when registering. It says:

Unique technical partnership between NVIDIA and Adobe engineers optimizes full pipeline acceleration of Flash 10.1 on the GPU.

This means two things: Flash Player 11 will not see the light of day tomorrow. Instead, a minor version update 10.1 is going to be releasead. This release will see full graphics pipeline rendering on the GPU. In my mind this means that everything from rendering vectors, text, transforms and Pixel Bender to compositioning will run on the GPU. That’s absolutely fantastic (if true)!

However, having only a minor revision of FP released tomorrow probably means that there won’t be many other new features. But expect to see a number of notifications regarding running FP on mobile devices. Again the NVIDIA pamphlet provided some unexpected news:

Full GPU acceleration of dynamic games and rich internet applications in the palm of your hand

Seems that GPU acceleration (and FP 10.1) is comming to your mobile (if your mobile is NVIDIA TEGRA-powered, err, which it is not…).

Stay tuned and watch the keynotes.

[Update] Shit. GPU acceletation will only be available for mobile devices and selected notebooks. Windows 7 (i.e. DiretX 11 machines) will have h.264 decoded on the GPU, but won’t support full GPU acceleration in the rendering pipeline. Mac’s and Linux boxes have even less support for the GPU. Big bummer. I really think Adobe is missing a big opportunity here. Unity3D or Torque3D will become the defacto web-based game plugin, at least for 3D games.

Anyone from up north (the european north) heading to MAX? We’re already arriving on saturday, but will probably be too exhausted to do anything that night. But do come have a few (=lots) pints with us on Saturday for some pre-conferencing

It would also be cool to join any Adobe User Group meeting that weekend. Anyone know where one’s beeing held?

Finally I decided to get myself the internet precense I deserved and got myself my own artman.fi-domain. You know, I’ve always dreaded the moment I leave Valve (yes, eventually that too will happen…) because I loose my company email-address as well. I can’t even count the number of services I’ve registered to with that specific email. Most services nowadays use the email-address as the login name, and usually the login name cannot be changed (stupid, is it not?).

After pondering over this for the past few years (and even subscribing to MobileMe to provide support for all those users who use my iPhone Web Feeds app) it was time to get an email-address which I can be certain of will follow me through the rest of my life.

Once I had the domain I thought about the same thing with my personal blog, previously hosted at valveblog.com. Again, my personal blog was tied to a company that at some point in the future I might not be working for anymore. ValveBlog started out as a co-hosted blog run my me and a few colleagues, but it clearly became apparent that I was the only one doing all the writing, so it sort of just became my personal blog over the matter of a few years.

So yeah, brand new, risen from the ashes of ValveBlog (a design that was way too old anyways) I present you with this mash-up of blog, tweets, mobile photos, weather info and delicious links.

It’s still very rough around the edges, but should get better every now and then when I have the time to play. And oh, if some Art Director from Valve would like to help me out with the background for example, I’d greatly appreciate the help.

Now even corporations who have policies against switching from IE can maybe think about upgrading to a modern browser.

I just wonder what the User-Agent is send with each HTTP-request send from this hijacked browser…

Well, it turns out everybody was wrong, and sorting with an AS3 algorithm is indeed almost twice as fast as it’s native counterpart!

Consider the following code:

The class will simply run two native sorts to test the performance on a totally random unordered and ordered Vector.<Number> and then run two sorts on the same data with a pure AS3 implementation of Shell Sort.

Shell sort has a good worst case resolve speed and works quiet well even when the origin data is already sorted or almost sorted.

The results

On my iMac native sort on the Vector with 10.000 Numbers took an average of 50 milliseconds. When sorting the already sorted Vector again, it took an average of 38 milliseconds.

The AS3 algorithm took an average of 31 milliseconds to sort the random Vector and only 15 milliseconds to sort the sorted Vector. (Update: in debug mode!)

In real life where Vectors that need to be sorted are usually not totally random, you’ll get a two-fold increase in sorting speed compared to the native algorithm.

Some things to consider, though: Shell sorting in this example gives you a head start over the native sorting algorithm that the data to be sorted are primitives. If you wanted to sort Objects (typed objects of course) based on one of their properties shellSort would no longer outperform Flash Player’s native sort, as the speed is eaten up by those numerous object property look-ups. However even then it comes very close and in programs that often sort Vectors with objects that are already almost in the right order, you’ll beat the native sorter by 10-30%.

But there’s one great type of project where speeding up sorting of Numerical Vectors will increase application performance greatly and that’s 3D. I started to dig into these algorithms after my profiler told me that 25% of all of my CPU time was going into sorting 3D polygons. Only the drawing operation was slower, but not by much.

P.S. If you want even better performance, don’t just rely on a single sorting algorithm. QuickSort is 20 times faster than Shell Sort if the Vector is already sorted, but absolutely freaking slow if the Vector sorting order is random. In 3D scenes it might be a good idea to switch between these, using Shell sort for your initial render, the user quick sort if your model rotates only slightly and thus only a few poly’s are out of place.

Update: Erik correctly pointed out that I should test the sorting algorithm in the release player instead of the debug player in debug mode, as this might affect results. And indeed it did. It made Shell sorting became much faster. A vector with 100.000 random Numbers took:

Native sort: 338 ms

Native sort on ordered Vector: 245 ms

AS3 Shell sort: 98 ms

AS3 Shell sort on ordered Vector: 49 ms

Who would have thought that interpreted code can do the job 3-5 times more quickly than a native call.

Update: I forgot to mention the license of the source code. Consider it Public Domain.

Update 2: Doh. I ported the code too quickly. inc should have been rounded, not floored. Updated the code to work properly.

Straight from the start of our MMOG Platform project we knew we wanted to create a technology which is better than anything previously available (or unavailable in the case of proprietary, closed platforms). Quite big words, so we needed a big idea.

Client synchronization and metaprogramming