Forums - Snapdragon Device Support Matrix

6 posts / 0 new
Last post
Snapdragon Device Support Matrix
madhavajay
Join Date: 15 Mar 18
Posts: 22
Posted: Sat, 2018-03-31 19:18
From the snpe-1.13.0 Manual 
 
Supported Snapdragon Devices
 
 
Snapdragon Device Support Matrix
Snapdragon Device CPU GPU DSP
Snapdragon 835 Yes Yes Yes (ADSP)
Snapdragon 821 Yes Yes Yes (ADSP)
Snapdragon 820 Yes Yes Yes (ADSP)
Snapdragon 660 Yes Yes Yes (CDSP)
Snapdragon 652 Yes Yes No
Snapdragon 630 Yes Yes No
Snapdragon 636 Yes Yes No
Snapdragon 625 Yes Yes No
Snapdragon 450 Yes Yes No
 
 
Running the following code on my LG G6 Android 7.0 with a Qualcomm 821 Chip:
 
final SNPE.NeuralNetworkBuilder builder = new SNPE.NeuralNetworkBuilder(application);
for (NeuralNetwork.Runtime runtime : NeuralNetwork.Runtime.values()) {
  if (builder.isRuntimeSupported(runtime)) {
    Log.v(TAG, "sNPE Runtime supported: " + runtime.ordinal + " " + runtime.name());
  }
}
03-31 23:03:22.702 29324-29324/org.tensorflow.demo V/sNPEObjectDetection: sNPE Runtime supported: 0 CPU
03-31 23:03:22.735 29324-29324/org.tensorflow.demo V/sNPEObjectDetection: sNPE Runtime supported: 1 GPU
03-31 23:03:22.736 29324-29324/org.tensorflow.demo V/sNPEObjectDetection: sNPE Runtime supported: 3 GPU_FLOAT16
 
So no NeuralNetwork.Runtime.DSP ?????
 
Trying to run the app with DSP caused it to fail saying it wasnt supported.
 
    java.lang.IllegalArgumentException: None of the selected runtime targets are supported.
 
Can someone comment?
 
  • Up0
  • Down0
jesliger
Join Date: 6 Aug 13
Posts: 75
Posted: Sun, 2018-04-01 04:50

Hi there.  The 821 supports the DSP runtime, as long as the OEM is allowing it (they sometimes disable it.).  

Please try the inception V3 example in the user's guide, which uses the command line tools (snpe-net-run) to run some images through SNPE using the DSP.  If that works, but the Java app doesnt ,then it is possible that your device has some settings that don't allow APKs access to the DSP runtime.

Feel free to provide the logcat from the failure case when you are trying your test.

 

 

  • Up0
  • Down0
madhavajay
Join Date: 15 Mar 18
Posts: 22
Posted: Sun, 2018-04-01 08:50

Its an Entel LG G6 H870 from Chile running Anroid 7. I purchased it because of the Qualcomm 821 chip.

See images here of the hardware and software.

https://imgur.com/a/WGvWD

Does anyone know if I can just flash the normal European KDZ file from LG for the H870.

I was concerned there might be baseband issues, but I assume I could always flash back to the current Chile H870 right?

Alternatively anyone know if the new LineageOS builds for the LG G6 would have the DSP enabled? Is it on a ROM level or somewhere deeper in the hardware release? Theres an Android 8.1 image in beta which I would hope has the new Android NNAPI as well but I wouldnt expect that to work off the bat, besides tests of the current TensorFlow Lite on a Pixel XL on 8.1 (as official as it gets) dont have any accel currently.

In the worst case will the GPU performance with the .dlc still be much better than the normal MobileSSD Android demo?

Could I install something like CPU-Z instead to see if DSP is enabled?

  • Up0
  • Down0
madhavajay
Join Date: 15 Mar 18
Posts: 22
Posted: Sun, 2018-04-01 09:25

CPU-Z info:

https://imgur.com/a/bj8vq

I don't see any mention of DSP?

Also I got this qualcomm Android tool:

https://play.google.com/store/apps/details?id=com.quicinc.trepn&hl=en

I dont see any DSP profiling stats:

https://imgur.com/a/33PXe

  • Up0
  • Down0
jesliger
Join Date: 6 Aug 13
Posts: 75
Posted: Sun, 2018-04-01 10:39

As previously suggested, please try the inceptionv3 example in the user's guide. The one that uses snpe-net-run.  

  • Up0
  • Down0
jesliger
Join Date: 6 Aug 13
Posts: 75
Posted: Sun, 2018-04-01 10:39

As previously suggested, please try the inceptionv3 example in the user's guide. The one that uses snpe-net-run.  

  • 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.