Forums - Low perfornace on Snapdragon 855 CPU

3 posts / 0 new
Last post
Low perfornace on Snapdragon 855 CPU
prophett0527
Join Date: 11 Dec 18
Posts: 3
Posted: Wed, 2019-05-22 05:59

Hello,

I am developing apps on Galaxy S10+ (Snapdragon 855).

The inference performance of my deep learning model (Mobilenet V2)  is as below:

GPU: 5 ~ 15 ms

CPU: 120 ~ 140 ms

 

Also, same deep learning model implemented by Tensorflow Lite (CPU) is considerably faster than SNPE.

TFLite CPU: 20 ~ 40 ms

 

My code is almost the same with the example app in SNPE SDK.

Why is the performance gap between SNPE and TFLite on CPU?

Thanks in advance.

 

  • Up0
  • Down0
prophett0527
Join Date: 11 Dec 18
Posts: 3
Posted: Wed, 2019-05-22 19:09

BTW, quantization is NOT used here.

TFlite only use CPU. GPU or NNAPI is not used.

  • Up0
  • Down0
zhaoyangstar
Join Date: 14 Apr 19
Posts: 23
Posted: Wed, 2019-05-22 20:43

Yes. I also met the samilar case. I have to see SNPE CPU runtime is very slow. I think maybe Qualcomm didnot do much optimization on CPU runtime. BTW, SNPE CPU runtime can only running on one thread, not support multi-threads.

  • 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.