Intel® Hyperflex™ Architecture High-Performance Design Handbook

ID 683353
Date 10/04/2021
Public

A newer version of this document is available.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

Document Table of Contents

5.2.10. Critical Chains with Dual Clock Memories

Hyper-Retiming does not retime registers through dual clock memories. Therefore, the Compiler can report a functional block between two dual clock FIFOs or memories, as the critical chain. The report specifies a limiting reason of Insufficient Registers, even after Fast Forward compile.

If the limiting reason is Insufficient Registers, and the chain is between dual clock memories, you can add pipeline stages to the functional block. Alternatively, add a bank of registers in the RTL, and then allow the Compiler to balance the registers. Refer to the Hyper-Pipelining (Add Pipeline Registers), Add Pipeline Stages and Remove Asynchronous Resets, and Appendix A: Parameterizable Pipeline Modules for a pipelining techniques and examples.

A functional block between two single-clock FIFOs is not affected by this behavior, because the FIFO memories are single-clock. The Compiler can retime registers across a single-clock memory. Additionally, a functional block between a dual-clock FIFO and registered device I/Os is not affected by this behavior, because the Fast Forward Compile can pull registers into the functional block through the registers at the device I/Os.