tree 2f140de8c73482411f2148a7d33914046f90eb7b
parent b3e02969ca3c9b782ef5134e96a2c175d230660f
author Naveen Dodda <ndodda@google.com> 1722899326 +0000
committer Naveen Dodda <ndodda@google.com> 1724780967 +0000

Specialized Conv kernel to handle input tensors

*This kernel is helpful to reduce the bottleneck due to input conv layer for 3d inputs.
*Kernel handles nxnx3 inputs with a 3x3 filter
*reduces cycles count by ~51x

Change-Id: I7978508f32f8974c1236ba00c828a0d04f88c8e9