Hi,
Recently I have train a network with tensorflow, and convert to dlc, it can work on DSP runtime.
But there is a conv2d layer cost about 200ms, and the filter size is 1x13.
After change the filter size to 1x9, the conv2d layer costs only 4ms.
If the filter size set to 1x10, the conv2d layer costs about 130ms.
Is it a limitation of DSP runtime? And it is not recorded on SNPE docs.
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.
