Versal Adaptive SoC System Integration and Validation Methodology Guide (UG1388) - 2024.1 English

Document ID: UG1388
Release Date: 2024-06-19
Version: 2024.1 English

Introduction
About the Versal Adaptive SoC Design Methodology
Navigating Content by Design Process
About This Guide
Simulation Flows
Logic Simulation
Logic Simulation Using SystemC Models
HLS Simulation
AI Engine Simulation
Embedded Software Simulation
Hardware Emulation
NoC Simulation
CIPS Verification IP
Design Closure
Timing Closure
Understanding Timing Closure Criteria
Checking for Valid Constraints
Checking for Positive Timing Slacks
Understanding Timing Reports
Checking That Your Design is Properly Constrained
Fixing Issues Flagged by check_timing
No Clock and Unconstrained Internal Endpoints
Generated Clocks
Loops and Latch Loops
No Input/Output Delays and Partial Input/Output Delays
Multiple Clocks
Fixing Issues Flagged by report_methodology
Methodology DRCs with Impact on Timing Closure
Methodology DRCs with Impact on Signoff Quality and Hardware Stability
Other Timing Methodology DRCs
Assessing the Maximum Frequency of the Design
Baselining the Design
Defining Baseline Constraints
Identifying Which Clocks Must Be Created
Verifying That No Clocks Are Missing
Constraining Clock Domain Crossings
Reviewing Clock Relationships
Identifying Clock Pairs without Common Primary Clocks
Identifying Tight Timing Requirements
Constraining Both Primary and Generated Clocks at the Same Time
Limiting I/O Constraints and Timing Exceptions
Evaluating Design WNS After Each Step
Post-Synthesis and Post-Logic Optimization
Pre- and Post-Placement
Pre- and Post-Physical Optimization
Pre- and Post-Route
Baselining and Timing Constraints Validation Procedure
Analyzing and Resolving Timing Violations
Identifying Timing Violations Root Cause
Reviewing Timing Slack
Using the Design Analysis Report
Analyze Path Characteristics
Review the Logic Level and Routes Distribution Tables
Datapath Delay and Logic Levels
Clock Skew and Uncertainty
Reducing Logic Delay
Optimizing Regular Fabric Paths
Optimizing Paths with Dedicated Blocks and Macro Primitives
Reducing Net Delay Caused by Physical Constraints
Reducing Net Delay Caused by Congestion
Congestion Area and Level Definition
Congestion Level Ranges
Interconnect Congestion Level in the Device Window
Congestion in the Placer Log
Congestion in the Router Log
Report Design Analysis Congestion Report
Report Design Analysis Complexity Report
Reducing Clock Skew
Using Synchronous Clocks
Adding Timing Exceptions between Asynchronous Clocks
Applying Common Techniques for Reducing Clock Skew
Improving Skew in Versal Devices
Reducing Clock Delay in Versal Devices
Reducing Clock Uncertainty
Using MBUFGCE to Reduce Clock Uncertainty
Applying Common Timing Closure Techniques
Improving the Netlist with Block-Level Synthesis Strategies
Improving Logic Levels
Reducing Control Sets
Follow Control Set Guidelines
Reduce the Number of Control Sets
Optimizing High Fanout Nets
Allow Register Replication
Promote High Fanout Nets to Global Routing
Use Physical Optimization
Prioritize Critical Logic Using the group_path Command
Fixing Large Hold Violations Prior to Routing
Addressing Congestion
Lower Device Utilization
Balance SLR Utilization for SSI Devices
Use Alternate Placer and Router Directives
Turn Off Cross-Boundary Optimization
Disable LUT Combining
Limit High-Fanout Nets in Congested Areas
Use Cell Bloating
Tuning the Compilation Flow
Using Strategies and Directives
ML Strategies
Predefined Strategies
Custom Strategies
Using Optimization Iterations
Overconstraining the Design
Considering Floorplan
Grouping Critical Logic
Reusing Placement Results
Using Incremental Implementation
Choose a High Quality Reference Checkpoint
Select Incremental Implementation Directives for High Reuse Mode
Avoid Floorplanning and Overconstraining
XPIO-PL Interface Techniques for Timing
AI Engine-PL Interface Techniques for Timing
SSI Technology Considerations
Using Hard SLR Floorplan Constraints
Using Soft SLR Floorplan Constraints
Using SLR Crossing Registers
Using Auto-Pipelining for SLR Crossings
Clustering Logic
Using IMUX Registers to Improve Timing to Hard Macros
Using Intelligent Design Runs
Power Closure
Estimating Power Throughout the Flow
Recommended Power Constraints
Minimum Recommended Constraints
Additional Recommended Constraints
Best Practices for Accurate Power Analysis
Reviewing the Design Power Distribution After Running Power Analysis
Power Timing Slack
DRC Closure
System Performance Closure
Analyzing System Performance for Platform-Based Designs
Analyzing System Performance in Simulation
Analyzing AI Engine Performance in Simulation
Analyzing PL Kernel Performance in Simulation
Using Hardware Emulation to Analyze System Performance
Analyzing System Performance in Hardware
Measuring Performance with RTL or HLS Monitors
Measuring AI Engine Performance
Measuring Performance with AI Engine Run Time Event APIs
Measuring Performance with Event Trace Debug
Debugging AI Engine-PL Performance on Silicon
Improving Performance in the PS
Improving Performance in the PL
Reducing Latency
Addressing Floorplanning Impact on Performance
Improving NoC Performance
Design Considerations
Ensuring Proper Communication Between Master and Slave
Improving QoS and Bandwidth
Addressing Suboptimal Solutions
Taking NoC Performance Measurements
Memory-First Method
Analyzing Memory Subsystems
Analyzing Memory Channel Saturation
Analyzing NoC Agent Performance
Measuring Master Performance
Measuring the Master Injection Rate
Using Exclusive Routing Groups
Splitting Traffic at the Master
Analyzing Master Transactions
Improving Performance in the AI Engine
Improving Performance Through the CPM and PL PCIE
Configuration and Debug
Creating the Device Image
Configuration
Debugging
Debugging Configuration
Debugging Board Bring Up
BootROM Debug
Platform Loader and Manager Debug
ATF and U-Boot Debug
Linux Kernel Debug
Debugging the System
JTAG Status and Error Status
Rails Voltage Status
Debugging Power
Debugging the PS/PMC
PS/PMC Peripheral Resets
PS/PMC Clocks
Debugging the NoC
Debugging the DDR Memory Controllers
Debugging DDR Memory Controller Performance
Debugging the DDR Memory Controller Interface
Debugging PCIe
Debugging the AI Engine
Debugging with SmartLynq+
Debugging the PL
AXI4 Debug Hub Connectivity
Using ILA Cores
Probing the Design
Choosing Debug Nets
Retaining Names of Debug Probe Nets Using MARK_DEBUG
ILA Core and Timing Considerations
ILA Core Designs with High-Speed Clocks
Using VIO Cores
VIO Core Considerations
Debugging AXI Interfaces in Vivado Hardware Manager
Using IBERT GTY, GTYP, and GTM for Transceiver Link Characterization
Running Debug-Related DRCs
Generating AXI Transactions
Modifying the Implemented Netlist to Replace Existing Debug Probes
Inserting, Deleting, or Editing ILA Cores on an Implemented Netlist
Connecting a Net to a Free External Pin Using Post-Route ECO
Using Remote Debugging
Debugging the Software
Target Processor
Operating System
Application Software/Kernel Debug
Functionality/Performance Debug
Validation
Block and IP Validation
Generator and Checkers
AI Engine Design Validation
Functional Validation
Performance Validation
System Validation
Design Debug
Additional Resources and Legal Notices
Finding Additional Documentation
Support Resources
References
Revision History
Please Read: Important Legal Notices

Back to home page