FPGA-based Deep Learning Inference Accelerators: Where Are We Standing?