I’ve installed irqbalance on my Omnia and placed it in /etc/rc.local to launch after boot. It seemed to work well, but monitoring /proc/interrupts tells a different story.
If I stop irqbalance, I could manually pin some of the interrupt sources to core 1 by echoing bitmasks into files under /proc/irq (e.g. SATA, ath9k, mv64xxx_i2c, USB work, but ath10k and mvneta do not). So the hardware at least supports it for some devices. As soon as I relaunch irqbalance, all IRQs go back to core 0 and stay there.
Is this a bug or a feature?
I’m on 3.11.6 currently, but AFAIR 3.11.7 showed the same behaviour (I rolled back to test something else and stayed here).