Coding Shift Registers and Delay Lines - 2023.2 English

UltraFast Design Methodology Guide for FPGAs and SoCs (UG949)

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

In general, a shift register is characterized by some or all of the following control and data signals:

  • Clock
  • Serial input
  • Asynchronous set/reset
  • Synchronous set/reset
  • Synchronous/asynchronous parallel load
  • Clock enable
  • Serial or parallel output

AMD devices contain dedicated SRL16 and SRL32 resources (integrated in LUTs). These allow efficiently implemented shift registers without using flip-flop resources. However, these elements support only LEFT shift operations, and have a limited number of I/O signals:

  • Clock
  • Clock Enable
  • Serial Data In
  • Serial Data Out

In addition, SRLs have address inputs (A3, A2, A1, A0 inputs for SRL16) determining the length of the shift register. The shift register can be a fixed static length or can be dynamically adjusted. In dynamic mode, each time a new address is applied to the address pins, the new bit position value is available on the Q output after the time delay to access the LUT.

Synchronous and asynchronous set/reset control signals are not available in the SRL primitives. However, if your RTL code includes a reset, the AMD synthesis tool infers additional logic around the SRL to provide the reset functionality.

To obtain higher clock frequency when using SRLs, AMD recommends that you implement the last stage of the shift register in the dedicated slice register. Slice registers have a better clock-to-out time than SRLs. This allows additional slack for the paths sourced by the shift register logic. Synthesis tools automatically infer this register unless this resource is instantiated or the synthesis tool is prevented from inferring this type of register because of attributes or cross-hierarchy boundary optimization restrictions. To infer the extra register, register the dynamically delayed signal separately in the RTL.

AMD recommends that you use the HDL coding styles represented in the Vivado Design Suite HDL Templates.

When using registers to obtain placement flexibility in the chip, turn off SRL inference using the following attribute:

SHREG_EXTRACT = "no"

For more information about synthesis attributes and how to specify the attributes in the HDL code, see the Vivado Design Suite User Guide: Synthesis (UG901).