Using Intelligent Design Runs - 2023.2 English

UltraFast Design Methodology Guide for FPGAs and SoCs (UG949)

Document ID
UG949
Release Date
2023-11-29
Version
2023.2 English

To automatically address most timing closure challenges during implementation, you can use intelligent design runs. Intelligent design runs are a special type of implementation run that leverages report_qor_suggestions, ML-based strategy predictions, and incremental compile. Intelligent design runs can run up to six iterations of place and route, which leads to a typical compile time of 3.5 times that of a standard run. However, using intelligent design runs can provide significant benefits by reducing the knowledge required to close timing and by saving hours of user analysis.

Tip: To iterate more quickly, you can extract the QoR suggestions and ML strategies from the intelligent design runs for use in a standard implementation run. If a significant design change is made, rerun the intelligent design runs to update the associated files.

Intelligent design runs comprise the following stages:

  1. Uses report_qor_suggestions to apply optimization properties to elements in the design in a predetermined order.
  2. Uses machine learning (ML) strategies to generate tool options for opt_design, place_design, phys_opt_design, and route_design that are optimized for the design.
  3. Uses a Last Mile Timing Closure feature to apply extensive effort on paths that are difficult to resolve to get the final result.

To ensure success when using intelligent design runs, follow these requirements:

  • The implementation must be project based. For non-project users, the easiest method is to create a post-synthesis netlist-based project using a pre-opt_design checkpoint.
  • The device must be from either an UltraScale or AMD UltraScale+™ - device-based family.
  • The design must have a baseline with accurate and achievable constraints.
  • All designs must comply with the recommended methodology, as reported by the report_methodology Tcl command.
  • An SLR-based floorplan might be required for SSI technology-based devices.
  • Apply only automatic implementation suggestions. Text-based suggestions or suggestions with APPLICABLE_FOR = synth_design must be applied before starting intelligent design runs.

For more information see this link in the Vivado Design Suite User Guide: Design Analysis and Closure Techniques (UG906).