opt_design - 2024.1 English

Vivado Design Suite Tcl Command Reference Guide (UG835)

Document ID
UG835
Release Date
2024-05-30
Version
2024.1 English

Optimize the current netlist. This will perform the retarget, propconst, sweep and bram_power_opt optimizations by default.

Syntax

opt_design [‑retarget] [‑propconst] [‑sweep] [‑bram_power_opt] [‑remap]
    [‑aggressive_remap] [‑resynth_remap] [‑resynth_area]
    [‑resynth_seq_area] [‑directive <arg>] [‑muxf_remap]
    [‑hier_fanout_limit <arg>] [‑bufg_opt] [‑mbufg_opt]
    [‑shift_register_opt] [‑dsp_register_opt] [‑srl_remap_modes <arg>]
    [‑control_set_merge] [‑control_set_opt] [‑merge_equivalent_drivers]
    [‑carry_remap] [‑debug_log] [‑property_opt_only] [‑quiet] [‑verbose]

Usage

Name Description
[-retarget] Retarget
[-propconst] Propagate constants across leaf-level instances
[-sweep] Remove unconnected leaf-level instances
[-bram_power_opt] Perform Block RAM power optimizations
[-remap] Remap logic optimally in LUTs
[-aggressive_remap] High effort remap optimization
[-resynth_remap] High effort re-synthesis to reduce the depth of logic in critical timing paths
[-resynth_area] Resynthesis
[-resynth_seq_area] Resynthesis (with Sequential optimizations)
[-directive] Mode of behavior (directive) for this command. Please refer to Arguments section of this help for values for this option Default: Default
[-muxf_remap] Optimize all MuxFx cells to LUT3
[-hier_fanout_limit] Replicate by module with threshold N
[-bufg_opt] Insert, Merge and Split BUFGs
[-mbufg_opt] Merge BUFGs to MBUFGs
[-shift_register_opt] Pull register stage from shift register
[-dsp_register_opt] Push/Pull Registers out of a DSP
[-srl_remap_modes] Remap shift registers to flops or flops to shift registers
[-control_set_merge] Merge all equivalent control set drivers to a single driver
[-control_set_opt] Optimize control set, Versal only
[-merge_equivalent_drivers] Merge all LUT,Flop equivalent driver replications
[-carry_remap] Remap carries into luts
[-debug_log] Show debug message
[-property_opt_only] Do targeted optimizations on tagged cells
[-quiet] Ignore command errors
[-verbose] Suspend message limits during command execution

Categories

Tools

Description

Optimizes a design netlist for the target part. Optimization provides improvements to synthesized netlists from third-party tools, or for netlists that are not optimized during synthesis.

Run this command after synthesis but prior to implementation to optimize the design and simplify the netlist before placing and routing the design. Assign the DONT_TOUCH property to prevent the optimization of individual cells.

Tip: To see what actions opt_design is taking in optimizing your design, you can use the -verbose option to get a more detailed transcript of the process. This can help you in understanding and debugging some of the changes made to your design.

The opt_design command performs the following optimizations by default:

  • Retarget
  • Constant Propagation
  • Sweep
  • Global Buffer (BUFG) optimizations
  • DSP Register optimizations
  • Shift-Register Logic optimizations
  • Block RAM Power optimizations
  • Implement MIG cores
  • Implement Debug cores
Important: Using command-line options for specific optimizations results in opt_design performing only the specified optimizations and disabling all others, even the ones that are usually performed by default.

To perform LUT Remapping, you must specify -remap.

To perform area-based re-synthesis, you must specify -resynth_area, or -directive ExploreArea.

To perform sequential area-based re-synthesis, you must specify -resynth_seq_area, or -directive ExploreSequentialArea.

Arguments

-retarget - (Optional) Retarget one type of block to another when retargetting the design from one device family to another. For example, retarget instantiated MUXCY or XORCY components into a CARRY4 block; or retarget DCM to MMCM. The retarget optimization also absorbs inverters into downstream logic where possible.

Note: The -retarget argument is optional, as are the other optimizations. However this optimization is run by default unless explicitly overridden by another optimization.

-propconst - (Optional) Propagate constant inputs through the circuit, resulting in a simplified netlist. Propagation of constants can eliminate redundant combinational logic from the netlist.

-sweep - (Optional) Remove unnecessary logic, removing loadless cells and nets. Additional optimization can be performed such as tie-off, register replication driving multiple OBUF, IBUF/OBUF insertion, retargeting dual port RAM to single port RAM, and optimizing IDELAYCTRL by grouping IO.

-bram_power_opt - (Optional) Enables power optimization on Block RAM cells. Changes the WRITE_MODE on unread ports of true dual-port RAMs to NO_CHANGE, and applies intelligent clock gating to Block RAM outputs.

Note: Specific Block RAM cells can be excluded from this optimization using the set_power_opt command.

-remap - (Optional) Remap the design to combine multiple LUTs into a single LUT to reduce the depth of the logic.

-aggressive_remap - (Optional) Similar to the -remap option to reduce LUT logic depth but more exhaustive. This may reduce more LUT levels at the expense of longer optimization runtime.

-resynth_area - (Optional) Perform re-synthesis in area mode to reduce the number of LUTs.

-resynth_seq_area - (Optional) Perform re-synthesis to reduce both combinational and sequential logic. Performs a superset of the optimization provided by -resynth_area.

-resynth_remap - (Optional) Remap the design to improve the critical paths in timing-driven mode by performing re-synthesis to reduce the depth of logic. This timing-based approach will replicate LUT’s with fanout and collapse smaller LUT’s into bigger functions at the expense of longer optimization runtime.

-directive <arg> - (Optional) Direct the mode of optimization with specific design objectives. Only one directive can be specified for a single opt_design command, and values are case-sensitive. Supported values include:

  • Explore - Run additional optimizations to improve results.
  • ExploreArea - Run Explore with resynth_area, to reduce the number of LUTs.
  • ExploreWithRemap - Run Explore with aggresive_remap optimization to compress logic levels.
  • ExploreSequentialArea - Run Explore with resynth_seq_area optimization to reduce registers and related combinational logic.
  • RuntimeOptimized - Runs Default without bram_power_opt. Run the fewest optimizations, trading optimization results for faster runtime.
  • RQS - Instructs opt_design to select the opt_design directive specified by the report_qor_suggestion strategy suggestion. Requires an RQS file with a strategy suggestion to be read in prior to calling this directive. See report_qor_suggestions -help for more details.
  • Default - Run the default optimization.

Refer to the Vivado Design Suite User Guide: Implementation (UG904) for more information on the effects of each directive.

Note: The -directive option controls the overall optimization strategy, and is not compatible with any specific optimization options. It can only be used with -debug_log, -quiet and -verbose.

-muxf_remap - (Optional) Convert MUXFs to LUT3s opportunistically when it can potentially improve route-ability of the design without affecting timing.

-hier_fanout_limit <arg> - (Optional) Net drivers with fanout greater than the specified limit (<arg>) will be replicated according to the logical hierarchy. For each hierarchical instance driven by the high-fanout net, if the fanout within the hierarchy is greater than the specified limit, then the net within the hierarchy is driven by a replica of the driver of the high-fanout net. The minimum value for hier_fanout_limit is 512.

-bufg_opt - (Optional) Perform various optimizations related to global buffers (BUFG/BUFGCE): Insert a buffer on unbuffered clock nets (fanout > 30). Insert BUFGs on high fanout nets where the control set pin count on the net exceeds the tool threshold. Perform load-splitting when a high-fanout net drives both combinational and sequential logic: the combinational portion bypasses the BUFG because the added delay is too large.

Tip: The phys_opt_design command can be used to optimize the combinational portion, while the sequential portion can be driven by BUFG.

-mbufg_opt - (Optional) MBUFG optimization transforms parallel clock buffers driven by a common driver, or clock modifying blocks (CMB) such as MMCM, DPLL, or XPLL, to MBUFG.

-shift_register_opt - (Optional) For high-fanout nets originating from an SRL, add output register pipeline stages to improve timing prior to the high-fanout net replication. This optimization is performed as part of the default optimization.

-dsp_register_opt - (Optional) Pull registers out of a DSP, or push registers into a DSP, to improve timing and resource utilization as needed. This optimization is performed as part of the default optimization.

-srl_remap_modes <args> - (Optional) Many designs have timing closure issues related to deep SRLs, and long register chains. This option provides both an automatic rule to trigger remapping between FF chains and SRLs, and also a manual rule to specify when to convert a FF chain to an SRL, or an SRL to a FF chain. Automatic remapping starts with the longest FF chains and converts those to SRLs when needed to reduce FF utilization, and takes the shallowest SRLs and converts those to FF chains when needed to reduce LUT utilization. Manual remapping takes FF chains in excess of specified lengths and remaps them to SRLs, and takes SRLs with specified depths and remaps them to FF chains. The automatic and manual modes cannot both be used in the same opt_design command pass. The following shows the syntax for the automatic and manual rules:

  • Automatic:
    • -srl_remap_modes {{target_ff_util <util> target_lutram_util <util>}}: Where <util> is a number from 0 to 100 expressed as a percentage of utilization. When either FF or LUTs exceed the specified utilization, the tool remaps as needed. When both utilization rates are exceeded neither is remapped.
  • Manual:
    • -srl_remap_modes {{min_depth_ffs_to_srl <val>}{max_depth_srl_to_ffs <val>}}: Where <val> specifies the depth of an SRL or length of a FF chain. When a FF chain is greater than the specified min value, the chain is remapped to an SRL with the required depth. If an SRL has depth less than or equal to the specified max value, the SRL is remapped to a FF chain.

-control_set_merge - (Optional) Reduce the drivers of logically-equivalent control signals to a single driver. This is like a reverse fanout replication, and results in nets that are better suited for module-based replication.

-control_set_opt - (Optional) Reduce the control set structure complexity and provide more flexibility to placer by converting smaller control sets into bigger group. This switch introduce new LUT logic in front of flop D pins to preserve functionality. The optimization is enabled by default in all explore strategies.

Explore/ExploreArea/ExploreSequentialArea/ExploreWithRemap

-merge_equivalent_drivers - (Optional) This option merges equivalent drivers on both control signals and non-control signals, to reduce logic in the design.

-carry_remap - (Optional) Remap carry signals into LUTs.

-debug_log - (Optional) Generate log file for debugging purposes.

-property_opt_only - (Optional) Indicates that the opt_design command should be driven only by properties on objects in the design, and not by the options defined as part of the command syntax. Refer to the Vivado Design Suite Properties Reference Guide (UG912) for more information on specific properties. The following properties trigger property-only optimizations: CARRY_REMAP, CONTROL_SET_REMAP, EQUIVALENT_DRIVER_OPT, LUT_REMAP, MUXF_REMAP, REG_TO_SRL, SRL_STAGES_TO_REG_INPUT, SRL_STAGES_TO_REG_OUTPUT, SRL_TO_REG.

Tip: When -property_opt_only is specified, the tool will return an error if additional optimizations have been specified in the opt_design command.
-quiet - (Optional) Execute the command quietly, returning no messages from the command. The command also returns TCL_OK regardless of any errors encountered during execution.
Note: Any errors encountered on the command-line, while launching the command, will be returned. Only errors occurring inside the command will be trapped.

-verbose - (Optional) Temporarily override any message limits and return all messages from this command. This option displays detailed information about the logic that is affected by each optimization.

Examples

The following example performs all four default optimizations: retarget, constant propagation, sweep, and Block RAM power optimization. The command returns detailed results with the -verbose switch:

opt_design -verbose

This example excludes specific Block RAM cells from power optimization using the set_power_opt command, and then runs opt_design with the four default optimizations:

set_power_opt -exclude_cells [get_cells \
   -filter {PRIMITIVE_TYPE =~ BMEM.*.*} \
   -of_objects [get_pins -leaf -filter {DIRECTION == IN} \
   -of_objects [get_nets -of_objects [get_pins clock/bufgctrl_clk_mld/O]]]]
opt_design

The following example performs the sweep and retarget optimizations:

opt_design -sweep -retarget
Note: Because -sweep and -retarget are expressly enabled in the prior example, -propconst optimization and -bram_power_opt are implicitly disabled.

The following example directs the opt_design command to use various algorithms to achieve potentially better results:

opt_design -directive Explore

The following example directs the opt_design command to use various algorithms to achieve potentially better results, while focusing on area reduction:

opt_design -directive ExploreArea

This example uses the -srl_remap_modes argument with the automatic option to convert FF chains into SRLs when FF utilization is greater than 20%, or SRLs into FF chains when SRL utilization is over 50%:

-srl_remap_modes {{target_ff_util 20 target_lutram_util 50}}
Tip: In this case, the srl_remap is only performed if either specified utilization rate is exceeded, but NOT if both are exceeded.

The following example uses the -srl_remap_modes argument with the manual option to convert FF chains with length greater than 5 to SRLs, and SRLs with depth less than or equal to 8 to FF chains:

-srl_remap_modes {{min_depth_ffs_to_srl 5}{max_depth_srl_to_ffs 8}}