IMX8M plus: NNRT segmentation fault

[keithmok]

Segmentation fault on /usr/lib/libnnrt.so.1.1.9 when running the attached tflite model using tensorflow lite on IMX8M plus using NPU.. 

Procedure : Use latest NXP release for IMX8M plus
Download Google tensorflow prebuilt benchmark tools for linux_aarch64 from https://www.tensorflow.org/lite/performance/measurement.

Run /home/root/linux_aarch64_benchmark_model --graph=ai_benchmark/memnet_quant.tflite --use_nnapi=true

GDB shows  it segmentation fault in /usr/lib/libnnrt.so.1.1.9 .

[Thread debugging using libthread_db enabled]

Using host libthread_db library "/lib/libthread_db.so.1".

STARTING!

Log parameter values verbosely: [0]

Graph: [ai_benchmark/memnet_quant.tflite]

Use NNAPI: [1]

NNAPI accelerators available: [vsi-npu]

Loaded model ai_benchmark/memnet_quant.tflite

INFO: Created TensorFlow Lite delegate for NNAPI.

Explicitly applied NNAPI delegate, and the model graph will be completely executed by the delegate.

The input model file size (MB): 0.656504

Initialized session in 87.803ms.

Running benchmark for at least 1 iterations and at least 0.5 seconds but terminate if exceeding 150 seconds.

[New Thread 0xffffe5b7c1c0 (LWP 3343)]

[New Thread 0xffffe537b1c0 (LWP 3344)]

Thread 1 "linux_aarch64_b" received signal SIGSEGV, Segmentation fault.

0x0000fffff7ac89bc in nnrt::op::Operand::echo(unsigned int) const () from /usr/lib/libnnrt.so.1.1.9

[jimmychan]

I think you should use the tool for ARM.

[keithmok]

Did you read and know what you are talking bout ?

I was using prebuilt tensorflow lite benchmark tool for Linux ARM 64 bits.

[jimmychan]

"benchmark tools for linux_x86-64" in your question. So, I was thinking you are using the tool not for ARM.

[keithmok]

That's a typ (I corrected that), but pls read the full log and command I post:
Run /home/root/linux_aarch64_benchmark_model
aarch64 is clearly a  linux arm 64 bits binary.