3 ms·

Byte code vs native code? I have a hard time believing that anyone would make an Android game without the NDK (without which you're basically CPU starved - Andr

Byte code vs native code? I have a hard time believing that anyone would make an Android game without the NDK (without which you're basically CPU starved - Android isn't the platform for heavy CPU work), but maybe Angry Birds is Dalvik-only?

mdaniel 16y ago
I have it on good authority[1] that Angry Birds is a native-inclusive app. Also, I wouldn't look down my nose at Dalvik-only apps. The JIT (which, AFAIK, is the target of the litigation) gets pretty close to optimal code the longer it runs - and theoretically games run the same code a lot more than your average application. In fact, my EUR0.02 is that native-inclusive apps (just like any Java apps which rely on native libraries) are worse, in my opinion, because it limits the number of platforms that one can execute the apps upon. And as a huge Linux fan, current Mac user, and user of x64 Windows at work, I can assure you that I get left out in the cold a LOT. It's one level of bad to tie your app only to Windows, which coincidentally includes 90% (last metric I heard) of the world. It's another thing entirely to say, "oh, sorry, your Android isn't the same as my Android: too bad for you." 1 = unzip -l AngryBirds_1.3.5.apk|grep -i \\.so
- CountSessine 16y ago
  Also, I wouldn't look down my nose at Dalvik-only apps. The JIT (which, AFAIK, is the target of the litigation) gets pretty close to optimal code the longer it runs - and theoretically games run the same code a lot more than your average application That's interesting - are you sure about that? There's a big difference between a static translation JIT and the more modern and memory-intensive progressively-optimizing JIT's that are used outside mobile. My understanding was that Dalvik's JIT was a one-time-only translater. Its difficult to find any comprehensive benchmarks of the 2.2 JIT - there are plenty of benchmarks showing the 2.1 interpreter eating CPU cycles compared to native code (think 20-to-1), and Google themselves claimed that the 2.2 JIT would mean a 2-5x improvement in CPU efficiency, but I'm still waiting for a set of comprehensive cross-platform benchmarks.

mdaniel 16y ago

I have it on good authority[1] that Angry Birds is a native-inclusive app. Also, I wouldn't look down my nose at Dalvik-only apps. The JIT (which, AFAIK, is the target of the litigation) gets pretty close to optimal code the longer it runs - and theoretically games run the same code a lot more than your average application. In fact, my EUR0.02 is that native-inclusive apps (just like any Java apps which rely on native libraries) are worse, in my opinion, because it limits the number of platforms that one can execute the apps upon. And as a huge Linux fan, current Mac user, and user of x64 Windows at work, I can assure you that I get left out in the cold a LOT. It's one level of bad to tie your app only to Windows, which coincidentally includes 90% (last metric I heard) of the world. It's another thing entirely to say, "oh, sorry, your Android isn't the same as my Android: too bad for you." 1 = unzip -l AngryBirds_1.3.5.apk|grep -i \\.so

CountSessine 16y ago
Also, I wouldn't look down my nose at Dalvik-only apps. The JIT (which, AFAIK, is the target of the litigation) gets pretty close to optimal code the longer it runs - and theoretically games run the same code a lot more than your average application That's interesting - are you sure about that? There's a big difference between a static translation JIT and the more modern and memory-intensive progressively-optimizing JIT's that are used outside mobile. My understanding was that Dalvik's JIT was a one-time-only translater. Its difficult to find any comprehensive benchmarks of the 2.2 JIT - there are plenty of benchmarks showing the 2.1 interpreter eating CPU cycles compared to native code (think 20-to-1), and Google themselves claimed that the 2.2 JIT would mean a 2-5x improvement in CPU efficiency, but I'm still waiting for a set of comprehensive cross-platform benchmarks.