Scheduling and cache blocking for heterogenous architectures

[psiha]

Global cache blocking values and thread scheduling don't map optimally to the new big.LITTLE systems with 'global task scheduling' (where all CPU tasks are active at the same time).