3.2.1. Initiator to Target NoC Mapping
Use case 1: 1 × 1 Full Address Connection
In this configuration each pair of initiator and target is directly mapped. As shown in the figure below, the CH0 PC0 channel of the HBM controller connects to target 0 which connects to initiator 0.
The table below illustrates the mapping of HBM2E controller pseudo-channels to HBM2E DRAM channels to targets and HBM2E DRAM channels.
HBM2E IP GUI Channel | AXI4 Channel (HBMC) | HBM2E DRAM Channel |
---|---|---|
Channel 0 | ch0_u0 | Channel E |
ch0_u1 | ||
Channel 1 | ch1_u0 | Channel F |
ch1_u1 | ||
Channel 2 | ch2_u0 | Channel A |
ch2_u1 | ||
Channel 3 | ch3_u0 | Channel B |
ch3_u1 | ||
Channel 4 | ch4_u0 | Channel G |
ch4_u1 | ||
Channel 5 | ch5_u0 | Channel H |
ch5_u1 | ||
Channel 6 | ch6_u0 | Channel C |
ch6_u1 | ||
Channel 7 | ch7_u0 | Channel D |
ch7_u1 |
Use case 2: 16 × 16 Cross Bar Connection
In this configuration the address map of each initiator is configured to allow it to drive traffic to all of the 16 targets. You can use this design example configuration as a starting point for designs that have multiple masters accessing the same HBM2E pseudo-channels, for example when there is ping-pong buffering.
The design example offers options to use the AXI4-Lite access and fabric NoC with each of the NoC connectivity configurations.