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: C71 model inference stuck

Part Number: TDA4VM


The application creates a total of 6 threads, and each thread independently loads a model for model inference. In some cases, the inference will get stuck, and all 6 threads will be stuck. In response to this situation, we did a preliminary investigation and added a log to the firmware of the C7x. If the firmware is working normally, the log will continue to output. After the stuck problem reappears, the log of the C7x will also stop outputting. Therefore, we guess that it may be stuck when calling the dsp function in the firmware. But the deeper reason cannot be determined. 
Other information that may be helpful is as follows: 
1. The hardware is tda4VM, and the tidl version is 7.03 
2. The system is qnx 
3. The 6 threads in the application program use C7x resources in parallel, and the utilization rate of C7x resources has reached 100% 
4. The stuck recurrence rate is very low, and it may occur once after thousands of hours of operation. 
5. The model does not have a custom operator, so there is no custom modification of the C7x firmware