Very interesting article on PhoneDog:
We’ve been tackling this very issue lately while trying to get the optimal output from 4 core hardware. Android is certainly behind in the game, and custom thread treatments are necessary to make use of available processing power. One of the challenges is proper allocation of threads for operations which won’t block other segments from completing. All too often the addition of hardware has proven minimal due to inadequate handling. Although many times overlooked, bottlenecks from memory bus speed, transfers and external hardware interfaces play an important role in the optimization for multi-core software. By optimizing for a given platform based on the true criteria for the application, we’ve been successful in getting true 4X scaling on quad core mobile systems.
See how we did it here: