nRF51822 - atomic increment of a variable

RE: nRF51822 - atomic increment of a variable

Fedexx2 — Thu, 28 Sep 2017 10:52:54 GMT

I dug inside the code of CRITICAL_REGION_ENTER/EXIT and studied how it works. During the ENTER: first it call __disable_irq(), then it disable all the APP IRQs, then call __enable_irq(). and in the EXIT it does the same stuff, of coure re-enabling the APP IRQs.

So, for a very very small operation like the increment of a variable wouldn't it be better to use directly __disable_irq()/__enable_irq(), instead of having them called twice in the CRITICAL_REGION_ENTER/EXIT?

RE: nRF51822 - atomic increment of a variable

Joe Merten — Thu, 28 Aug 2014 06:55:27 GMT

I'd thought about __disable_irq() because of performance reasons (about 800ns against 25µs of critical region), but I now finally decide for critical region.

RE: nRF51822 - atomic increment of a variable

Nikita — Sat, 23 Aug 2014 17:15:19 GMT

When you call sd_nvic_critical_region_enter() it'll disable all your application interrupts but doesn't affect softdevice interrupts. Why do you want to disable all interrupts including softdevice? Softdevice interrupts won't change your variable.

I think the best way is to disable app interrupts in main on variable write.