Am I getting this right? The reason that the context has to be saved and then restored is that because the scheduler can not yield in the regular way while in interrupt context (as we assume there are no nested interrupts, and the yield is done through software interrupt). So we restore the context at the end of every interrupt; If a context switch was performed somewhere along the ISR, the ISR assembly wrapper will just complete the action usually done by the software interrupt used for yield. Otherwise, the context restore will do nothing. What did I get wrong there? :)
Basically, but it is not always necessary to save and restore the context each ISR, depending on the ISR, but also depending on the processor and tools. Some times it is more efficient to save everything on entry and exist, sometimes to only save when needed at the cost of some duplication (some registers getting saved twice). Also some compilers make this easier than others. Which port are you using? (processor and tools) Your comment about wrapper code narrows it down, but which one? Regards.
I’m using the SAM7 IAR port. I see when there is no reason for wrapping – essentially, when you’re not doing anything that might indirectly yield, such as locking a resource, writing to a queue, etc… BTW, this will be a very good place to admit you for the excellent, excellent work you’ve done!
Copyright (C) Amazon Web Services, Inc. or its affiliates. All rights reserved.