ARM – AXI3 / AXI4 INTERCONNECT
Start Date: Please contact us
Course Overview
AXI v4 is the interconnect used by Cortex-A and Cortex-R ARM CPUs, see the outlines of these courses
Prerequisite:
• Knowledge of an interconnect, such as IBM CoreConnect or ARM AHB is recommended
Course Goals and Objectives
• This course details first the AXI3 protocol
• New signals present in AXI4 are then described
• The course explains the AXI4 stream protocol and indicates in which case this simplified protocol is suitable
• AXI4-lite protocol is described
• The NIC-301 interconnect IP is studied, clarifying synthesis options as well as software QOS parameterizing
• AXI Coherency Extensions (ACE) new channels are explained through an overall introduction to snooping
• The CCI-400 interconnect IP is described, highlighting the purpose of ACE-lite ports
Course Outline:
First day
1. AXI3 PROTOCOL [4-hour]
• ARM AMBA versions
• Address and data channels
• Basic read and write transactions, pipelining, data reordering
• Global signals, clocking, low power handshake interface
• Detailing address channel signals
– Clarifying cache related signals, allocate policies
– Explaining bufferable vs not bufferable attribute
• Detailing data channel signals
– Using write strobe signals, managing unaligned transfers
• Detailing response signals
• Ordering model, composite transaction ID
• Managing exclusive resources, local and global monitors
• AXI Master parameter list
• AXI Slave parameter list
2. AXI4 NEW SIGNALS [1/2-hour]
• AXI4 slave response dependencies
• QoS signalling, defining a per-transaction priority
• Multiple region signalling
• User signals, implementation example in Cortex-A9
3. AXI4 LITE [1/2-hour]
• Simpler control register-style interface
• Bursts of 1 data beat
• Signal list, highlighting the AXI4 non supported signals
• Conversion, protection and detection
4. AXI4 STREAM [1-hour]
• Objectives of this new protocol
• Byte definition: data byte, null byte, position byte
• Byte stream example
• Signal list
• Merging and packing
• Downsizing / upsizing
• Packet transfer
• Source and destination signaling
5. NIC-301 AXI3 INTERCONNECT [2-hour]
• Block diagram
• Bus conversion wrappers
• TrustZone support
• Programmable features, QoS
• Arbitration algorithms: Round Robin, Least Recently Granted
• Slave and master interface options
• Programmer’s model
Second day
6. INTRODUCTION TO CACHE AND TLB COHERENCY [2-hour]
• Cache organization
• Explaining the need for coherency
• Software coherency
• Hardware coherency
• Translation Lookaside Buffer
• Implementing an I/O MMU
7. AXI COHERENCY EXTENSION [3-hour]
• Signals added to the traditional 5 channels
• Explaining what is a shareability domain
• Using barriers, related ARM instructions
• The three additional channels
• ACE new transactions: explaining through sequences their utilization
– Non-cached transactions
– Shareable read transactions
– Shareable write transactions
– Write-back transactions
– Cache maintenance transactions
• Distributed virtual memory
• ACE-lite subset
8. CCI-400 AXI4 ACE INTERCONNECT [2-hour]
• SoC architecture example
• CCI-400 features, implementation in a Big/little system
• Speculative fetch
• Performance monitoring unit
• TrustZone support
• Error responses
• QoS value arbitration and propagation
• Regulation of outstanding transactions
• QoS value based on latency measurement
9. Summary
