Reference Counting for memory-slabs

zephyrproject-rtos / zephyr

Primary Git Repository for the Zephyr Project. Zephyr is a new generation, scalable, optimized, secure RTOS for multiple hardware architectures.

Apache License 2.0

10.5k stars 6.43k forks source link

When working e.g. with sensor data one may want to not pass a full sample struct around threads but rather a pointer to it. This in itself works fine with the memory-slab API. But once you start sending a sample to multiple threads e.g. a data logger and a sensor fusion module you either need to duplicate the data, which is a waste if none of the modules writes to the sample or add application logic to handle reference counting to only free the sample once all threads are done with it.

Adding a (atomic) reference counting to the memory-slab directly or replicate the module as a "reference_counted_memory_slab" would make the application code much more simple and avoid user errors.

zephyrproject-rtos / zephyr

Reference Counting for memory-slabs #71175