We do not know.
In all seriousness, it would be very convenient if someone could transfer the Linpack Android application to Mono for Android, so that we can compare the performance between them.
Here is what I know:
- Currently, when you download an application on G1 for 3 days from the moment the application was launched to download the application (3), your resource will depend on the hardware. This is due to initializing the Mono runtime and loading referenced assemblies. We want to improve this in the future.
- Mono and Mono for Android, JIT everything. Meanwhile, Dalvik has a variable size JIT cache (currently 1MB on ARMv7) and an interpreter, so depending on your application it is possible (possible) that not everything will be JITed.
- Mono for Android uses JNI to invoke Android / Java code.
Then a koush performance comparison between Mono and Dalvik in 2009 . It preceded Dalvik JIT, but on identical equipment Mono spanked Dalvik.
And what does it mean? I have no idea. (Again, the Linpack port for Mono for Android will be wonderful! Tooltip hints, pushing)
However, there are recommendations for effectiveness:
- Use the screen saver while starting the application. This provides immediate feedback during application startup, which is most worrying for most users.
- If at all possible, minimize transitions between Mono and Java code. JNI will never win any performance medals, so if you can do more code in Mono without intermediate calls to Java methods, do it.
- Minimize the number of Java.Lang.Object instances that are supported simultaneously. References to GC cross-VMs can cause performance issues.
jonp source share