This is an extremely simple C version of classifier/convolution dnn kernels, based on data layout and implementation from diannao paper: https://novel.ict.ac.cn/ychen/pdf/DianNao.pdf

Data is completely made up, but shouldn't matter for dense kernels. Datatypes are currently set at fp32.

Currently, Makefile is configured for example layers from VGG16. https://arxiv.org/abs/1409.1556