Intel Agilex® 7 M-Series FPGA Network-on-Chip (NoC) User Guide

Download PDF
ID 768844
Date 12/04/2023
Public

A newer version of this document is available. Customers should click here to go to the newest version.

Document Table of Contents
Document Table of Contents x
Answers to Top FAQs 1. Network-on-Chip (NoC) Overview 2. Hard Memory NoC in Intel Agilex® 7 M-Series FPGAs 3. NoC Design Flow in Intel® Quartus® Prime Pro Edition 4. NoC Real-time Performance Monitoring 5. Simulating NoC Designs 6. NoC Power Estimation 7. Hard Memory NoC IP Reference 8. Document Revision History of Intel Agilex® 7 M-Series FPGA Network-on-Chip (NoC) User Guide
1. Network-on-Chip (NoC) Overview x
1.1. Introduction to Intel Agilex® 7 M-Series FPGAs 1.2. Terminology for Intel Agilex® 7 M-Series FPGAs 1.3. General NoC Architecture and Applications
1.3. General NoC Architecture and Applications x
1.3.1. General NoC Architecture 1.3.2. Example NoC Applications
1.3.2. Example NoC Applications x
1.3.2.1. Parallel Computing NoC Application 1.3.2.2. SmartNIC NoC Application
2. Hard Memory NoC in Intel Agilex® 7 M-Series FPGAs x
2.1. High-Level Architecture 2.2. NoC Segments 2.3. NoC Switch and Link Detail 2.4. Fabric NoC 2.5. NoC Protocol Support 2.6. NoC Design Considerations
2.2. NoC Segments x
2.2.1. UIB Segments 2.2.2. GPIO-B Segments 2.2.3. GPIO-B and HPS Segment 2.2.4. SDM Segment 2.2.5. PLL and SSM Segment
2.5. NoC Protocol Support x
2.5.1. AXI4 Protocol Support 2.5.2. AXI4 Handshaking Support 2.5.3. Quality of Service (QoS) Support 2.5.4. Transaction Ordering Support 2.5.5. AXI4-Lite Protocol Support
2.6. NoC Design Considerations x
2.6.1. Determining the Number of NoC Targets 2.6.2. Determining the Number of NoC Initiators 2.6.3. Fabric NoC Considerations 2.6.4. Latency Considerations 2.6.5. Initiator and Target Bandwidth Considerations 2.6.6. Horizontal Bandwidth Considerations 2.6.7. GPIO-B Bypass Mode and Initiators
3. NoC Design Flow in Intel® Quartus® Prime Pro Edition x
3.1. Hard Memory NoC Design Flow Overview 3.2. Using NoC Example Designs 3.3. NoC Building Blocks 3.4. Connecting NoC IP 3.5. Making NoC Logical Assignments 3.6. Making NoC Physical Assignments Using Interface Planner 3.7. Compiling the NoC Design
3.1. Hard Memory NoC Design Flow Overview x
3.1.1. NoC Design Flow Options
3.3. NoC Building Blocks x
3.3.1. NoC Initiators for Hard Processor Systems 3.3.2. NoC Targets for Hard Processor Systems 3.3.3. NoC Initiators for Fabric AXI4 Managers 3.3.4. NoC Targets for Fabric AXI4 Managers 3.3.5. NoC Clock Control
3.4. Connecting NoC IP x
3.4.1. General NoC IP Connectivity Guidelines 3.4.2. Connectivity Guidelines: NoC Initiators for Fabric AXI4 Managers 3.4.3. Connectivity Guidelines: NoC Targets for Fabric AXI4 Managers 3.4.4. Connectivity Guidelines: NoC Clock Control 3.4.5. Connectivity Guidelines: NoC Initiators for HPS 3.4.6. Connectivity Guidelines: NoC Targets for HPS
3.4.1. General NoC IP Connectivity Guidelines x
3.4.1.1. Connecting NoC IP and Assigning Base Addresses in the Platform Designer Connection Flow 3.4.1.2. Connecting NoC IP and Assigning Base Addresses in the NoC Assignment Editor Connection Flow
3.5. Making NoC Logical Assignments x
3.5.1. Creating NoC Assignments for Compilation 3.5.2. Using the NoC Assignment Editor 3.5.3. Troubleshooting NoC Assignment Editor 3.5.4. NoC Connection and Addressing Examples
3.5.2. Using the NoC Assignment Editor x
3.5.2.1. Step 1: Make Group Assignments in the NoC Assignment Editor 3.5.2.2. Step 2: Make Connection Assignments in the NoC Assignment Editor 3.5.2.3. Step 3: Make Attribute Assignments in the NoC Assignment Editor 3.5.2.4. Step 4: Validate Logical NoC Assignments and Generate Simulation File
3.5.4. NoC Connection and Addressing Examples x
3.5.4.1. Example 1: External Memory Interface with 1 AXI4 Initiator and 1 AXI4-Lite Initiator 3.5.4.2. Example 2: Two External Memory Interfaces with One AXI4 Initiator and One AXI4-Lite Initiator 3.5.4.3. Example 3: One External Memory Interfaces with Two AXI4 Initiators and One AXI4-Lite Initiator 3.5.4.4. Example 4: Two High-Bandwidth Memory Pseudo-Channels with Two AXI4 Initiators (Crossbar) and Shared AXI4-Lite Initiator 3.5.4.5. Example 5: Hard Processor System with Two External Memory Interfaces
3.5.4.1. Example 1: External Memory Interface with 1 AXI4 Initiator and 1 AXI4-Lite Initiator x
3.5.4.1.1. Example 1: Platform Designer Connection Flow 3.5.4.1.2. Example 1: NoC Assignment Editor Connection Flow
3.5.4.2. Example 2: Two External Memory Interfaces with One AXI4 Initiator and One AXI4-Lite Initiator x
3.5.4.2.1. Example 2: Platform Designer Connection Flow 3.5.4.2.2. Example 2: NoC Assignment Editor Connection Flow
3.5.4.3. Example 3: One External Memory Interfaces with Two AXI4 Initiators and One AXI4-Lite Initiator x
3.5.4.3.1. Example 3: Platform Designer Connection Flow 3.5.4.3.2. Example 3: NoC Assignment Editor Connection Flow
3.5.4.4. Example 4: Two High-Bandwidth Memory Pseudo-Channels with Two AXI4 Initiators (Crossbar) and Shared AXI4-Lite Initiator x
3.5.4.4.1. Example 4: Platform Designer Connection Flow 3.5.4.4.2. Example 4: NoC Assignment Editor Connection Flow
3.5.4.5. Example 5: Hard Processor System with Two External Memory Interfaces x
3.5.4.5.1. Example 5: Platform Designer Connection Flow 3.5.4.5.2. Example 5: NoC Assignment Editor Connection Flow
3.6. Making NoC Physical Assignments Using Interface Planner x
3.6.1. Using Interface Planner 3.6.2. Recommended Placement Order for NoC Elements in Interface Planner 3.6.3. Hard Memory NoC Locations in Interface Planner 3.6.4. NoC Performance Reports in Interface Planner
3.7. Compiling the NoC Design x
3.7.1. Fitter NoC Reports 3.7.2. Viewing NoC Elements in Chip Planner
5. Simulating NoC Designs x
5.1. Generating a Simulation Registration Include File (Platform Designer Connection Flow) 5.2. Generating a Simulation Registration Include File (NoC Assignment Editor Connection Flow) 5.3. Contents of Simulation Registration Include File
6. NoC Power Estimation x
6.1. Using the Intel FPGA PTC to Estimate NoC Power
7. Hard Memory NoC IP Reference x
7.1. NoC Initiator Intel FPGA IP 7.2. NoC Clock Control Intel FPGA IP
7.1. NoC Initiator Intel FPGA IP x
7.1.1. NoC Initiator Intel FPGA IP Parameters 7.1.2. NoC Initiator Intel FPGA IP Interfaces
7.1.2. NoC Initiator Intel FPGA IP Interfaces x
7.1.2.1. NoC Initiator Intel FPGA IP AXI4/AXI4-Lite Interfaces 7.1.2.2. NoC Initiator AXI4 User Interface Signals 7.1.2.3. NoC Initiator Intel FPGA IP AXI4-Lite User Interface Signals 7.1.2.4. NoC Initiator Intel FPGA IP Clock and Reset Signals 7.1.2.5. NoC Initiator Intel FPGA IP Platform Designer-Only Signals
7.2. NoC Clock Control Intel FPGA IP x
7.2.1. NoC Clock Control Intel FPGA IP Parameters 7.2.2. NoC Clock Control Intel FPGA IP Interfaces 7.2.3. NoC Clock Control Intel FPGA IP Platform Designer-only Signals
Answers to Top FAQs
1. Network-on-Chip (NoC) Overview
2. Hard Memory NoC in Intel Agilex® 7 M-Series FPGAs
3. NoC Design Flow in Intel® Quartus® Prime Pro Edition
4. NoC Real-time Performance Monitoring
5. Simulating NoC Designs
6. NoC Power Estimation
7. Hard Memory NoC IP Reference
8. Document Revision History of Intel Agilex® 7 M-Series FPGA Network-on-Chip (NoC) User Guide

2.5.3. Quality of Service (QoS) Support

Quality of Service (QoS) is a technique that the hard memory NoC uses to provide control of arbitration choices that satisfy performance requirements. You can also use QoS to prioritize some traffic. You can change this prioritization dynamically, or set this prioritization during configuration. To achieve system goals, system architects can use the QoS settings to specify the relative priority of traffic flows.

One can categorize initiator generated traffic into the following groups, according to sensitivity to latency:

Table 2.  Initiator-Generated Traffic Types
Traffic Type Description
Real-Time NoC initiators in this group are very sensitive to long-term and short-term latency, and fail when their buffers become empty. Examples include video display traffic and real-time applications. A video display engine must be able to fetch graphic data within a bounded amount of time to display images correctly on a monitor. Failing to meet real-time requirements may result in corrupt pixels or degradation of image sharpness or loss of screen synchronization.
Latency Sensitive The performance of this application type is directly linked to the latency of access. For example, in a CPU, processing can stop for many cycles when there is a cache miss.
Best Effort NoC initiators in this group can tolerate delays that other traffic causes and are insensitive to latency. These NoC initiators consume the remaining bandwidth, and must not interfere with real-time or latency-sensitive traffic. For example, file transfer or file download applications can stall without compromising the experience.

AXI QoS determines the associated NoC QoS. Urgency level is the priority of traffic within the NoC subsystem. A system can have multiple priority levels for traffic going to and from the memory.

You can configure the NoC Initiator Intel FPGA IP to use AXI4 QoS signals to determine the associated NoC QoS, or to send all NoC traffic with a fixed priority. If you configure the NoC Initiator Intel FPGA IP to use a fixed NoC priority, you can specify separate priorities for read and write traffic.

You can choose to specify the QoS Generator priority levels when parameterizing the NoC Initiator Intel FPGA IP. The priority levels are not run-time programmable. For more information on the QoS Generator in the NoC Initiator Intel FPGA IP, refer to NoC Initiator Intel FPGA IP.

Figure 12. Options for Generating Quality of Service: QoS Generator vs. AxQOS


Table 3. Memory Priority Versus QoS Settings shows the mapping of QoS signals to the urgency value of the corresponding packets. The NoC subsystem has four priority levels (urgency = 0, 1, 2 or 3). Within the NoC subsystem, traffic with priority level zero (0) has the lowest priority, while traffic with priority level 3 has the highest priority. While the external memory interface Hard Memory Controller supports up to four priority levels, HBM2e Hard Memory Controller supports only two priority levels. Best effort traffic should have a default urgency level of zero (0), with higher urgency levels reserved for real-time and latency-sensitive traffic. Note that only the top two bits of the AXI command QoS (arqos or awqos) map to a NoC quality-of-service option. The remaining bits are unused.

Table 3.  Memory Priority Versus QoS Settings
QoS[1:0] Urgency Level External Memory Priority Level HBM2e Priority Level

2’b00

0

0

0

2’b01

1

1

0

2’b10

2

2

1

2’b11

3

3

1
Level Two Title