This thread has been locked.

If you have a related question, please click the "Ask a related question" button in the top right corner. The newly created question will be automatically linked to this question.

IWR6843: Quantifying Memory and CPU Utilization

Part Number: IWR6843

Hi,

What is the best practice for quantifying memory usage and CPU utilization for both the DSP and ARM subsystems in the IWR6843. I see these numbers reported in TI whlte papers like this one: http://www.ti.com/lit/ug/tidue71c/tidue71c.pdf, and I would like to repeat the measurements like those in Tables 3, 4, 11, and 12, as well as measure the CPU load in the ARM core. It would be very helpful if someone can point me to the variables in the 3D People Counting Demo that were used to determine the values in the aforementioned tables.

Thanks,

Dave

  • Hi Dave,

    I will check with the code author and get back to you.

    Regards,

    Justin

  • Hi Dave,

    Apologize for the delay. Please note that the metrics in the TI Design are reported based on the 2D People Counting demo, so if you measure today on 3D you will see higher memory and MIPs utilization. Anyways, you will need to use the following functions to report these metrics:

    1. MIPs - Cycleprofiler_getTimeStamp() - this function is defined in the SDK utils.  You will find it in the DPCs before and after sections blocks to get a before and after start point. Please note that when calculating MIPs percentage utilization, the percentage changes based on frame time.
      1. Cycleprofiler_init() - must be called in main to initialize the profiler
    2. Memory - This is a bit more complicated.
      1. DSS - use Hwi_getStackInfo(), this will return data usage on the stack and cache memories. See dss_main.c line 402
      2. MSS - use DPC_ObjDetRangeHwa_MemPoolGetMaxUsage() - this is defined in the DPC itself and subtracts the starting address from the largest address used with the buffer

    Regards,

    Justin