Real time task statistics in the uxTaskGetSystemState()

[RomkaE]

Problem: Statistics collection API provide only AVERAGE statistics since kernel startup and don't calculate real time statistics.

Hi guys,
I would like to start a discussion on this issue and make changes to the Kernel to expand the API for collecting real time task statistics.

Example 1: the task was blocked for 10 minutes, and on event woke up for 1 second and worked for 500ms. The current implementation calculates statistics as 100x0.5/10x60 = 0.083(3)%. In integer representation it is equal to 0. But real time statistics for the last second is equal 50%, or for the last 10 seconds is equal 5%.

Example 2: one task is created only for initialization. For 5 seconds the task used 100% of CPU time, after initialization the task is deleted. In this case the sum of statistics of the remaining tasks (including the IDLE task) is less than 100%.

I found a couple of topics with similar discussions but in a different context:
Issue #307
Forum: How to reset vTaskGetRunTimeStats?

I've been using task statistics collection in every my projects with FreeRTOS to analyze and optimize CPU performance and memory consumption. I use uxTaskGetSystemState() and my own implementation of formatted output via UART.
To calculate real time statistics I use the following changes: commit f7ac81d
But these changes cause other statistic APIs to not work correctly. And I understand that I can't create a pull request with such changes.

And the last point as an argument in favor of real time statistics: couple screenshots from production project with real time statistics

In the screenshots, the statistics update time is 1 sec. The screenshots show peak loads on the processor. With average statistics calculations, the peak load is not visible.

[cookpate]

Thanks for the insight!

[amazonKamath]

@RomkaE Thanks for the opening the issue here, but in order to get community feedback and involvement, we would request you to create this as a forum post and close the github enhancement issue here. Thank you.

[RomkaE]

@amazonKamath ok. Thanks.