Has anyone come across the same issue?
To me, from QNN version 2.17, optimization level 3 (aka O3 or "O":3 specified in config json) has severe perf regression compared with 2.16. Same thing happens to the latest QNN 2.18.
According to my experiment, the regression happens to all kinds of models, e.g. resnet, ssd, transformers..
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.
Also, even O1 can deliver better perf than O3, which is very wrierd.