Posted on May 10, 2017 at 14:18One way is to do as few as possible in the interrupt routine itself.
Just set a flag, and do the processing in the main loop, i.e. somewhere else. Some rules for thread-safe coding still apply.
Another possibility is to prioritize interrupts - which only works if you deal with different interrupts.
If you are chasing your tail (interrupts coming faster than you can process it), you have hit a hard limit of the core.
Either optimize or move code out of the interrupt (as above), increase the core clock, or switch to a faster variant.
Or, when applicable, use DMA to collect data and reduce the interrupt rate.
... I introduced another if statement that would process the data every 3rd millisecond click instead of every millisecond and the problem disappeared.
Times of about 1 millisecond are already 'on the long side'.
That could be fine if it works - which seems not to be the case with your application.
I would try to move code out of interrupt context. And be especially careful when calling functions that involve other interrupt-based stuff, like serial output (UART, USB, SPI, ...).
