tree e6946f0e221dd6bb0c18c797d35b62e0aa7fddf5
parent ca74563c11195dafa4403cc306b10423e01f6be2
author Michael Brooks <mbrooksx@users.noreply.github.com> 1690933218 -0700
committer GitHub <noreply@github.com> 1690933218 +0000
gpgsig -----BEGIN PGP SIGNATURE-----
 
 wsBcBAABCAAQBQJkyZfiCRBK7hj4Ov3rIwAA9t4IAG/IWm9cJRyL5ndGU7iD+ADe
 oPUh1Z3UJ3rMwlXwJzVKBab/lPyah27077wDdChe+ev2cFQMVB+Dk4/Uy3BsACe0
 +Kj9zeq4eRgOQ1g+24m2xmL14eGgYwF/ziKi+9BtR7zk5SoIwTQk6VUAedox0NV9
 jEzNVR1cgKQVjk/CBYS4LJ3FZoZtb4vbOW96rHwqp1xMdk7ftRpbIoOJSd+7DKHU
 kUzH6i5NI7yP+fr3NABE55lulJai1HTKwwdfefjs7NV8SYvnKQOGthlDYuGZYnXe
 PGjDH1bdcL3lPoJEvvWMMUBXjYEbK2cwUtqZiz9PwmHlTOzMs1gBKmNxi/lveUY=
 =FWXb
 -----END PGP SIGNATURE-----
 

Add Q16x8 Depthwise Conv Support (#2140)

Adds support for 16-bit activations + 8-bit weights for depthwise convolution in the reference kernel. Uses 64-bit bias to match TFLite. Also adds passthrough to the q16x8 reference kernel for Xtensa, CEVA, and ARC (CMSIS already has it's own implementation).

Tested:
depthwise_conv_test

BUG=2141