CMSIS-NN: Optimized convolution for 16x8 data type
Authors: Sebastian-Larsson, oscarandersson8218 (GitHub) 1. Optimized function is added for input data type of int16 and filter data type of int8. The main constraint is that the kernel volume is < 512. 2. Unit tests are added as well. Change-Id: I9f0f3067a275c8bf5f9263ebe76cd9a1033c7793
Loading
Please sign in to comment