Each downlink AI Engine graph has four input data ports, 32 input coefficient ports, and eight output ports. A dlbf_data
PL kernel has four AXI4-Stream output ports to that plug into the four input data ports of the downlink AI Engine graph. A dlbf_coeffs
PL kernel has four AXI4-Stream output ports. So, eight dlbf_coeffs
PL kernels are required to plug into the 4*8=32 input coefficient ports. A dlbf_slave
PL kernel has one AXI4-Stream input port. So, eight dlbf_slave
PL kernels are required to plug into the eight output ports.
Remember, our system instantiates three downlink graphs, so the total number of DLBF PL kernels becomes: three dlbf_data
, 24 dlbf_coeffs
, and 24 dlbfs_slave
PL kernels.