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.
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.