This repository supports the PAPAA Croucher Summer School 2016 held at the University of Hong Kong between 18th to 22nd July 2016.

1. caffe-ref/ contains a Caffe specification for the Lenet5 model we use for this set of labs.
2. cpp-ref/ contains a flat C++ implementation of the MNIST/Lenet5 convolutional network layers. Not directly performance optimized. Only for illustrative/teaching use.
3. opencl-src/ contains the set of labs implementing OpenCL version of the MNIST/Lenet5 convolutional network layers. Each lab targets some combination of hardware platform (CPU, GPU, and FPGA) and platform-specific optimizations.
4. docs/ contains a set of PDFs for Intel/NVIDIA/Xilinx/Altera OpenCL optimization guides.