Zephyr RTOS system-wide runtime visualizer via SWD probe!
Take a broader look on your Zephyr application with a non-heavy, small footprint, Kconfig-only thread stats analyser.
To be able to properly analyze your Zephyr app, your ELF binary must be compiled and flashed with just some Kconfig extra options:
## prj.conf
CONFIG_INIT_STACKS=y # For stack watermarks
CONFIG_THREAD_NAME=y # For thread names
CONFIG_THREAD_MONITOR=y # For thread search
CONFIG_THREAD_STACK_INFO=y # For thread information
CONFIG_THREAD_RUNTIME_STATS=y # For CPU usage tracking
CONFIG_SYS_HEAP_RUNTIME_STATS=y # For heap runtime stats, if needed
You will also need a debug probe capable of SWD debugging, currently only Segger JLink and PyOCD compatible probes are supported, but more are to come!
Then just run ZViewTUI (a terminal front-end for ZView) on your Zephyr env:
(.venv) $ python zview_tui.py -e path/to/your/elf --mcu YOUR_MCU_TAG_NAME --runner YOUR_RUNNERNote
The MCU tag name is the name attributed to your MCU for runner connection, and is optional on some runners.
ZView attempts to be a neat tool for those who need to debug applications without the need of a shell/UART for stack statistics. The intent is to achieve minimal footprint and behaviour branching for the binary, avoiding statistics stdout and processing threads.
The keypoint is to take advantage of the debug probes ability to read runtime memory without CPU halting from the APB bus, since ELF file contains all information needed about the kernel objects, we just scan through the stack list and runtime analyse watermarks and CPU usages, so your source code can be kept free from debug implementations.
Note
When tracking CPU usage, the idle thread expresses its usage in true CPU activity, all other threads have thier
CPU usage shown in percentage relative to the non-idle time. This means that if a thread foo has 50% CPU usage,
and idle has 40%, foo is consuming 50% of the remaining 60% of non-idle CPU usage, effectively consuming 30% of
the total CPU execution. This is useful since then it is fast to understand how your threads are behaving in relation
to other active threads, and how much of the CPU is effectively being used.
ZViewTUI acts as a TUI (who would guess?), so you navigate with the arrow keys UP and DOWN from the default view:
To track CPU usage for a thread (and return to all threads) just hit ENTER over it:
Note
NEW: It is also possible to sort the data by pressing S, and invert the sorting by pressing I. To access the heap runtime visualization, press H.
Listening to community feedback, ZView is ongoing some new features development, such as, and not limited to:
- Heap visualization
- Extra thread information (like number of context switches and so)
- Full Zephyr integration through West (maybe an in-tree west command?)
- Live global variables tracking
- IDK, feel free to open an issue if you feel like this has some potential!