Simultaneous use of OpenCL(GPU) and TensorFlow Lite(NPU)

[tetsuro-okuyama]

Hi,

I want to use OpenCL and Tensorflow lite in the same process.

But my program freezes after calling clEnqueueMapBuffer on i.MX 8Mplus.
Is it possible to use OpenCL and tflite (NPU) at the same time?

The outline of the program is as follows:

1. Initialize OpenCL and loading Tensorflow lite model(using NPU).
2. Call clEnqueueNDRangeKernel and clEnqueueMapBuffer for executing OpenCL
 ->Does not return from clEnqueueMapBuffer

Are there any restrictions on simultaneous use?
Also, if I need any settings, please let me know.


Best regards,

[Bio_TICFSL]

Hello Tetsuro,

There is no restriction to use OpenCL and TensorFlow lite, but I do not see that clEnqueueNDRangeKernel has the data for NPU, so please send us your test code to validate.

Regards