Building static c library for xtensa DSP processors

[enthusiast666]

@advaitjain
Is it possible to build static c library for xtensa DSP processors like hifi series with xtensa specific ops/kernels , so that i can link it in my application and achieve the desired results . If yes , then pls let us know on the steps.

(Note : pls don't suggest using similar to test_person_detection_test or test_micro_speech_test)

If possible , pls share with us on any available repo who have done dsp applications with tflm library built with hifi series.

Thanks in advance.

Regards ,
enthusiast 666.

[gaikwadrahul8]

Hi, @enthusiast666
I apologize for the delayed response, I understand you're looking for a concrete example, tutorial or Github Repo. Unfortunately, complete publicly available Github repo of highly optimized DSP kernels for Xtensa HiFi processors integrated with TFLM are rare. This type of code is often considered proprietary and a key competitive advantage.

However, you can refer this blog : Accelerating TensorFlow Lite Micro on Cadence Audio Digital Signal Processors and this example. As you've already created a corresponding issue in the tensorflow/tflite-micro repository (tensorflow/tflite-micro#3045) which is the most appropriate repo for this type of query, we recommend closing this issue here and continuing the discussion there. This will help consolidate the discussion and ensure your query reaches the appropriate experts within the tensorflow/tflite-micro repository, where you are most likely to receive further assistance.

Thank you for your understanding and patience.