What you practice
Out-of-order and interleaving: observe-only routing theory (3.5). See how transaction IDs support out-of-order read responses and interleaved write beats. It is in Out-of-Order and Interleaving — Part 3 · AXI4 Full RTL Design of the AXI Masterclass. Burst, outstanding, and a BRAM memory bridge.
AXI only counts when the handshake simulates. RTL Studio is the Verilog and SystemVerilog IDE where you edit the RTL and prove each beat on a VCD.
What you will learn
This lesson explains how AXI4 Full supports out-of-order read responses and interleaved write data beats using transaction IDs, enabling high-throughput memory systems without strict ordering constraints.
Learning objectives
- Understand how
aridtags enable read responses to return in a different order than issued. - Recognize how
awidtags allow write data beats from multiple transactions to interleave on the W channel. - Trace how
ridandbidecho the original transaction IDs for correct master-side correlation. - Identify FIFO-based routing structures (
arid_fifo,awid_fifo) that preserve ordering metadata.
Theory
AXI4 allows the slave to complete read and write bursts in any order, as long as each response carries the correct ID tag.
Read responses use rid to echo the original arid, so the master can reassemble data even if responses arrive out-of-order. Write responses use bid to echo the original awid, ensuring the master knows which write completed.
The slave maintains per-ID FIFOs (arid_fifo, awid_fifo) to track outstanding transactions and route responses correctly.
| Signal | Purpose | Source | Echoed by |
|---|---|---|---|
arid | Read request ID | Master (AR channel) | rid on R channel |
awid | Write request ID | Master (AW channel) | bid on B channel |
Interleaving is possible because AXI4 channels are independent: W beats from different awid values can interleave, and R beats from different arid values can return in arbitrary order.
Practice
- Open RtlStudio and select Lesson 3.5 in the Examples list.
- In Explore, open
golden/rtl/axi_full_mem.svand read the comments onarid_fifo,rid_route, andawid_fifo. - Open
student/rtl/axi_full_mem.svand complete thearid_fifopush andrid_route__TODO__placeholders. - Open
golden/tb/tb_top.svto see how a single AW burst triggers one B response. - Review the
arid_fifoandawid_fifologic to understand ID-based routing.
This is an observe-only lesson, so there is no Simulation or Synthesis run.
Summary
This lesson reinforced how AXI4 uses transaction IDs—arid on the address channel and rid on the data channel—to enable flexible out-of-order response routing for reads. It also explained how write data beats can interleave on the W channel via awid tags, with bid on the B channel preserving transaction completion order. Students observed how FIFOs and routing logic in the reference RTL support these behaviors without requiring active implementation. The key takeaway is that AXI4’s ID-based routing decouples request ordering from response ordering, improving system efficiency and throughput.
Run it in the Verilog IDE
- Open the lesson in the IDE and find the valid/ready or AXI channel ports in the editor.
- Simulate with Verilator. Confirm the beat on the VCD, not only in the transcript.
- Check that data moves only on the beats this lesson describes, and that backpressure holds the transfer.
- Synthesize the RTL master, slave, or buffer with Yosys. Leave the UVM bench out of synthesis.
The lesson is done when the simulation matches Out-of-order and interleaving: observe-only routing theory. Change the source, run the Verilog simulator again, and read the waveform without leaving the IDE. Simulation guide · Yosys guide.