Intel® Hyperflex™ Architecture High-Performance Design Handbook

ID 683353
Date 12/08/2023
Public

A newer version of this document is available.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

Document Table of Contents

2.3.2.1. Pipelining at Variable Latency Locations

Commonly, FPGA designs include some locations that are insensitive to additional latency, such as at clock domain boundaries, connections between major functional blocks, and false paths. Best design practices recommend adding pipeline stages at clock domain boundaries or between major functional blocks to improve timing. However, adding excessive pipeline stages can also bloat area usage, and increase routing congestion.

The current version of the Intel® Quartus® Prime software includes new features to help improve timing performance for design paths that are insensitive to additional latency. The Hyper-Retimer can now automatically add pipeline stages on false paths that you tag as latency-insensitive, and also insert the appropriate number of pipeline stages at the registers you specify. The Hyper-Retimer retimes the added registers into timing-critical parts of the design. The number of pipeline stages that the Hyper-Retimer adds can change for each compilation, or any time you change the design.

Note:
  • If you do not specify latency-insensitive false paths or use autopipelining, the Hyper-Retimer output netlist is cycle-equivalent to your RTL.
  • If you specify latency-insensitive false paths or use autopipelining, the Hyper-Retimer output netlist is not cycle-equivalent to your RTL. Therefore, your simulation and verification environments must accommodate variations in the circuit latency to use these techniques.