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.

TDA4VM: frame drop occurred when running app_multicam with 8 2MP-ISP cameras

Part Number: TDA4VM

Hi 

we've run app_multicam with 8 2MP-ISP@30fps camera successfully in our customer board.

 And there is no frame drop when only app_multicam is running.

After app_multicam is running. we run another apps which read I2C register peroidly by remote service at backgroud, frame drop counter increas almost every remote-service call.

performance log with multicam only:

performance log with multicam and backgroud app:

It seems HWA performance drop from 497MP/s to 487MP/s,and average execution time increas from 33333us to 34000us

What might be the cause of the performance degradation?

Regards,

Jingyu

  • Hi Jingyu,

    What exactly i2c writes are you doing? Are there multiple i2c transactions in a single rpc call? or how frequently are you calling this remote procedure?

    I am just wondering if due to some reason, interrupt processing is getting delayed, and so affecting fps. 

    But if the transactions are like once in a while, it should not affect fps. 

    Regards,

    Brijesh

  • Hi Brijesh

    Just read register of Serializer and Deserializer by I2C-6. 

    We decreased peroid to 10s,frame drop counter still increas  almost every remote-service call.

    I wonder if there is any thing we can do to verify interrupt delay?Is timestamp of capture_arrary helpful?

     

    Regards,

    Jingyu

     

  • Hi Jingyu,

    Could you please tell me how many graph parameters you have? Can you please reduce to only one graph parameter? 

    When you say decrease period to 10s, does it mean you are reading one register every 10s and still seeing frame/fps drops? If you just read it once or don't read it for few 100s, then does fps recover?

    Capture in itself seems to be running fine. Even performance for all the other nodes seem fine, running in real time, but somehow performance of the entire graph is reducing. Once you read this i2c message, do you also process this message in the same example/thread? 

    Regards,

    Brijesh