If the soft PWM driver misses an IRQ then the next CC value (computed as CC[x] += delay) could already be behind the timer counter value, and hence the next compare IRQ won't fire until the timer counter value wraps around.
This needs some testing because the timing of the PWM will now be slightly different compared to before, since the next event will be computed from the current counter value plus delay, rather than the counter value at the time of the previous IRQ. But I can't really think of any other way to fix the issue of missed IRQs.
If the soft PWM driver misses an IRQ then the next CC value (computed as CC[x] += delay) could already be behind the timer counter value, and hence the next compare IRQ won't fire until the timer counter value wraps around.
This needs some testing because the timing of the PWM will now be slightly different compared to before, since the next event will be computed from the current counter value plus delay, rather than the counter value at the time of the previous IRQ. But I can't really think of any other way to fix the issue of missed IRQs.
Should fix #600