On Thu, Dec 12, 2013 at 05:39:12PM +0000, Catalin Marinas wrote:
If we can't really guarantee the effect of this patch, I would rather keep it in the LSK kernel only until the scheduler is fixed (can this be treated as a performance issue independent of the power-aware scheduling? We could get it merged quicker).
My understanding is that the behaviour is reasonably well understood, it's just not great in all situations (hence all the energy aware scheduler work) but then the default scheduler behaviour isn't that good either and possibly worse (hence all the energy aware scheduler work).
I think if we're that worried about problems that might be caused by doing this we should remove the equivalent ARMv7 code to keep the two in parity in terms of behaviour - it's going to be enough work getting big.LITTLE working well without introducing software only differences between ARMv7 and ARMv8 big.LITTLE. If you're dead set on that then I can do patches for that.