Forums - Running QNN with Htp backend on embedded linux

5 posts / 0 new
or Register
Last post
Running QNN with Htp backend on embedded linux
sampo.immonen
Join Date: 30 Jan 23
Posts: 3
Posted: Wed, 2023-03-15 07:27
Top

Hi,

 

Is there any examples of running QNN SDK qnn-net-run with the Htp backend on embedded linux. All the examples are shown on android.

I am working on SA8295 and the CPU and GPU backends work fine. When I try to use the Htp backend (by following the tutorial for qnn-net-run)

I get the following error: 

unsigned PD or DSPRPC_GET_DSP_INFO not supported by DSP

DspTransport.openSession qnn_open failed, 0xffffffff

 

The qnn-platform validator also does not find the required backend libraries and fails with:

PF_VALIDATOR: ERROR -2. Error while executing the sum function

 

I have tried to set the LD_LIBRARY_PATH and ADSP_LIBRARY_PATH to all directories i can think of

Thanks

 

 

  • Up3
  • Down0
Re: Running QNN with Htp backend on embedded linux
weihuan
Join Date: 12 Apr 20
Posts: 270
Posted: Sat, 2023-03-25 23:14
Top

Dear developer,

We have no plan to support QNN on embedded linux only in Android devcie.

Could you please take a look about the adb logs if error happened?

BR.

Wei

  • Up0
  • Down0
Re: Running QNN with Htp backend on embedded linux
sampo.immonen
Join Date: 30 Jan 23
Posts: 3
Posted: Mon, 2023-03-27 00:14
Top

Hi,

 

Thanks for the reply.

 

Im a bit confused because the QNN library has support for embedded linux and I can perfectly fine run models with the cpu and gpu. 

This error occurs when im trying to run things on the HTP backend.  

I also want to ask what is the preferred way to run neural network models on the SA8295 device? 

Should I use SNPE or QNN and which backend is the preferred one to use the Hexagon processors?

Thanks

 

  • Up3
  • Down0
Re: Running QNN with Htp backend on embedded linux
shuyw
Join Date: 10 Dec 23
Posts: 4
Posted: Fri, 2024-02-02 00:54
Top

Hi, is this problem solved? I met the same problem

  • Up1
  • Down0
Re: Running QNN with Htp backend on embedded linux
ahmed.bingol
Join Date: 15 Sep 23
Posts: 5
Posted: Wed, 2024-02-21 05:32
Top

Hello, 
I found that you need to quantize models for running on HTP backend at QNN documents somewhere. 

On the other hand, I was never able to run models on Linux oe.  The first tutorial in QNN SDK shows and explained which dynamic library files should be transfered to SA8295p for different platforms android, linux oe etc. And i found similar .so files found under $QNN_SDK_ROOT/2.17.0.231124/lib/aarch64-oe-linux-gcc11.2 etc. 

However, there is a huge missing information about creation  with qnn-model-lib-generator for target oe-linux devices. I debug the problem and found that makefiles for model generation such as qnn/2.17.0.231124/share/QNN/converter/Makefile.oe-linux-aarch64-gcc9.3 is missing OE compiler:

/sysroots/x86_64-oesdk-linux/usr/bin/aarch64-oe-linux/aarch64-oe-linux-g++: not found

How did you able to generate models for linux with qnn-model-lib-generator? I could not find anything in web for this compilers and setup to my system.

  • Up0
  • Down0
or Register

Opinions expressed in the content posted here are the personal opinions of the original authors, and do not necessarily reflect those of Qualcomm Incorporated or its subsidiaries (“Qualcomm”). The content is provided for informational purposes only and is not meant to be an endorsement or representation by Qualcomm or any other party. This site may also provide links or references to non-Qualcomm sites and resources. Qualcomm makes no representations, warranties, or other commitments whatsoever about any non-Qualcomm sites or third-party resources that may be referenced, accessible from, or linked to this site.