Forums » Everything else » iOS5 canvas performance - tested

I'm just simply amazed how big of a difference iOS5 have over iOS4 - i just ran some examples I did some time ago in regular canvas and it whizzes like crazy.

Did you get to do any Box2D/ ImpactJS testing?

Did not notice the performance issues on unoptimised game that I noticed before - but still it can do much better - I think with couple of corrections I can get it running at decent framerate at all times.

If you are asking for in browser game testing - I didnt test it yet - I will in couple of hours - I wouldn't expect much from games not optimised for mobile devices though as it does get helluva framerate drop on my particle generator I written some time ago but then I generate like 800 particles at start and calculate proximities to each other and react on these proximities.

Why would the Canvas acceleration effect DirectCanvas in any way? From what I understand, it's working like IOSImpact and converting all the Canvas calls to OpenGL calls, which makes sense. Given that, acceleration of Canvas should have no meaningful impact.

Now, on a game just wrapped up in PhoneGap or using AppMobi and NOT the XDK, it should have an impact. I haven't had any time to really test that yet.

All that said, IOS5 has enough performance-enhancement code baked into it (from all I can see) that there still is probably a nice boost to DirectCanvas. I'm just quibbling about the details.

Well I can't answer your question. I ran my game that I packaged with appmobi xdk and where there was some performance pitfalls before there is none now I did nothing except changing iOS. Another thing is that new iOS does have new js engine maybe this is the reason behind the experienced performance boost on my iPad.

Ok so now I have tested 4 of my ImpactJS games (not thoroughly though) in new iOS5 safari and again - the result is jawdropping. There is an enormous boost to the performance compared to what it was before - Including the one with over 200 entities per level.

Box2D on impact.com runs @ 60FPS unless you fire loads of bullets at around 70 framerate drops to below 20FPS

We ran our HTML5 test versus our Direct Canvas test using Box2d

At 14 elements on Box2D, the framerate dropped to around 10 - before this happend around 8 elements.

In DC - We had 100 birds and the framerate was at 23-25.

We've also tested some games that are in the appstore that use AppMobi (we purchased them). There was a huge different in iOS5, but when you try comparing it to something like Private Joe, there's really no comparison. For some people though, the enhancements to Canvas will be all you need.

That's what I'm getting at. I think the changes in IOS5 have done some wonderful performance enhancements all around. My testing with DirectCanvas has shown very nice improvements also. However, I just wanted to point out that it's not because of the acceleration on Canvas itself.

What we got was a bonus to speed. We got a bonus both in JS interpretation and in Canvas. I think we may have gotten some speed in general graphics performance too. This means that some stuff to regular canvas will be better. This means that any apps using just JS (like Sencha-Touch apps) will be a lot faster (I've played with one last night, it screams now). This also means that DirectCanvas using apps seem to have gotten a double-hit of performance boost.

Hey don't get me wrong I will still most definitely use appmobi whether there is performance boost to DC or not...

@ashkettle come on who cares about the reason behind the performance boost? Do we really need to go to such depths? All I did here was a report on my first impressions on Impact running on iOS5 rather than starting full blown CSI investigation into what causes performance boosts on iOS.

The most important fact is that there is a great boost on iOS5 and that combined with AppMobi XDK makes it even easier for us HTML5 game developers to make some dough on Apple's App Store and hopefully with Android 3 also on Android Store.

You should always care about what specifically gives you a performance boost. If you just go with a black box approach, you are never going to get the best out of your code, because frankly, you won't know what exactly you can do at each point to get the fullest potential from it.

If the majority of the boost is in the javascript parser, for instance, then I can move more logic into JS and do more, if I'm processing it in JS. I could also look at optimizing my images to make the OpenGL perform better.

Conversely, if the majority of the boost is in the OpenGL rendering, then you can ignore that point, and should optimize your JS code as tightly as possible.

The first rule of optimization: Don't do it.
The Second rule of optimization: Don't do it yet.
The third rule of optimization: Get performance metrics and determine where to optimize.

An important fact is that there is a good boost on IOS5. Another important fact is that it's IOS only, so if you are looking at cross platform later on, this won't really apply. A more important fact is knowing where to optimize your code to get the performance you want and need. That fact is not yet fully realized.

Yeah whatever...

Ashkettle I think you are missing the point of this thread... All it says "HEY! IT WORKS WELL ON IOS5" and you get anal and preachy and telling me what I should care about... First rules and stuff.... Relax man... If I wanted to start the code optimisation thread I would call it this way.

BTW Android 3 also have similar acceleration in place.

Jerczu - Can't we all just get along ;)

Ashkettle - I think you're overstating a lot here. We haven't seen the big performance gains in DC you are describing. What you are missing is that we already identified the major bottlenecks (like Box2d) and moved them to the native side. We may be talking a few frames at obscure high entities, but this was also on the iPad2. We're gonna test the 3gs and iPhone 4 today. I think we'll still see the big gap on the older devices.

Not only is it iOS5, it's iOS5 :). You'd be SHOCKED the number of people still runing 3.1.3, or who will be running 4.x series.

Another thing iOS5 brought was native scrolling in the webview. Here's the thing...it's stinks. What happens is the whole webview ends up scrolling too when you hit the top/bottom of the page.

Quote from stratman05
Jerczu - Can't we all just get along ;)

Hah! This made me laugh - Thanks.

I also tested the http://html5-benchmark.com on iPad2 with iOS5 - on average it ran at around 35-40 fps sometimes jumping to 55-60 and lowest at 9-10 fps where there was a lot of particles flying - finally gaining a score of 2476

Safari was the only app running at the time.

Total CPU time - 125.62s, Total Lag:494ms.