NoC Slave Unit - 1.0 English

Versal Adaptive SoC Programmable Network on Chip and Integrated Memory Controller 1.0 LogiCORE IP Product Guide (PG313)

Document ID
PG313
Release Date
2023-11-01
Version
1.0 English

The NoC slave unit (NSU) is the egress point to the NoC. The NSU provides:

  • Conversion of NoC packetized data (NPD) to and from AXI protocol data.
  • Asynchronous clock domain crossing and rate-matching between the AXI slave and the NoC.
  • AXI exclusive access handling.
  • Configurable AXI interface widths of 32, 64, 128, 256, or 512 bits. AXI4-Stream interfaces support 128, 256, and 512-bit widths. AXI data width is configured via parameter propagation from the connected IP.
  • Support for AXI4 and AXI4-Stream.
  • AXI-ID compression.
  • Programmable support for 32 outstanding read and 32 outstanding write transactions.
  • 32 entry interleaved read tracker and 32 entry chop-merge write tracker.
  • Support for two virtual read and two virtual write channels.
  • Response to QoS control.
  • Two versions of the NoC slave unit block are available:
    • A fully featured version used for all interface constructs except for memory.
    • A memory specialized version omitting the AXI interface. This unit is integrated in the hard memory controllers at the NoC endpoints of the device.
Figure 1. NoC Slave Unit

The main function of the NSU is to receive and respond to packets from the NoC that are addressed to the NSU packet interface and intended to go to the attached AXI slave.

As shown in the previous figure, the NSU logic de-packetizes the received NoC data packets and converts them into AXI transactions. The re-created AXI transaction passes through the buffered asynchronous data crossing and rate-matching (fast to slow) logic to the AXI master interface where the request is processed and a response accepted. The packetizing block receives AXI responses through the clock domain crossing and rate-matching logic (between the slow and fast domains). The read responses are buffered before forwarding to minimize bubbles (stalls) in the read responses.