The OpenCL-based FPGA accelerator data flow is as follows. First, the kernel receives the data in the input buffer; then, it stores the result in the output buffer after computing; finally, the host side reads the result into memory