Scribd Logo
Upload

Welcome to Scribd!

  • Common Commands in ICC2 2 Place Stage

    Uploaded by

    RAZ
    1K views5 pages

    Original Title

    blog.csdn.net-Common commands in ICC2 2 place stage

    Copyright

    © © All Rights Reserved

    Available Formats

    PDF, TXT or read online from Scribd

    Did you find this document useful?

    Is this content inappropriate?

    Report this Document

    Copyright:

    © All Rights Reserved
    Download as PDF, TXT or read online from Scribd
    Flag for inappropriate content
    Download as pdf or txt
    1K views5 pages

    Common Commands in ICC2 2 Place Stage

    Uploaded by

    RAZ

    Copyright:

    © All Rights Reserved
    Download as PDF, TXT or read online from Scribd
    Flag for inappropriate content
    Download as pdf or txt
    of 5

    Common commands in ICC2 (2) place stage

    blog.csdn.net/m0_61544122/article/details/124320320

    Copyright statement: This article is an original blogger article, following the CC 4.0 BY-SA
    copyright agreement, please attach the original source link and this statement for
    reprinting.
    Link to this article: https://blog.csdn.net/m0_61544122/article/details/124320320


    ICC2
    The column includes this content

    38 articles
    51 subscriptions

    subscribe column
    One, the process of place_opt

    place_opt is a packaged command that contains:  

    1. Corase placement This step is to place the standard cell of the timing driven according
    to the timing requirements. At this time, the placement of the standard units is still illegal,
    that is, the standard units are not on the row, and they may all overlap.  

    Command: initial_place  

    SPG flow uses DCG netlist and def, and can skip initial_place and initial_drc.

    set_app_options -name

    place_opt.flow.do_spg -value true

    Two pass placement can optimize the timing driven and congestion driven in this step,
    and preprocess the high fanout net.

    set_app_options -name place_opt.initial_drc.global_route_based -value true

    set_app_options -name place_opt.initial_place.two_pass -value true

    2. HFN solves fanout This step will first remove buffer trees, that is, buffers and inverters
    that the tool considers redundant. Then start to solve fanout net and logic drc . It is worth
    noting that this stage does not solve the ideal network network.  

    Command: initial_drc  

    3. Preliminary timing optimization This step is the timing optimization before the most
    refine placement.  

    Command: initial_opto   

    1/5
    4. Final placement This step is the most timing & congestion driven placement. After this
    step, the positions of all standard units are legal, and there will be no cases where the
    standard units are not on the row or have overlaps .  

    Command: final_place  

    5. The final timing optimization This step is based on the results of final placement for
    further timing optimization, this process will also involve the rearrangement of cells.  

    Command: final_opto

    Second, the relevant settings of the place stage

    1. early clock

    Doing place_opt of CTS will effectively improve the timing

    set_app_options -name place_opt.flow.clock.aware_placement -value true

    set_app_options -name place_opt.flow.trial_clock_tree -value true

    set_app_options -name place_opt.flow.optimize_icgs -value true

    #Using optimize icg does not need to set the above two app options.

    create_routing_rule icc2rm_2w2s \

    -default_reference_rule -multiplier_width 2 -multiplier_spacing 2

    set_routing_rule -rule icc2rm_2w2s [get_nets -physical_context -of [get_clock_tree_pins]]

    # For early clock, it is better to specify cts buffer or inverter, and ndr rule, which is more
    consistent with the cts stage.

    2. global route

    set_app_options -name place_opt.initial_drc.global_route_based -value 1

    set_app_options -name place_opt.initial_place.two_pass -value true

    # global route combined with two pass placement can reduce congestion and improve
    timing

    3. Other options can improve timing

    2/5
    set_app_options -name place_opt.flow.optimize_ndr -value true

    set_app_options -name place_opt.flow.do_path_opt -value true

    set_app_options -name place_opt.flow.enable_ccd -value false

    # It is recommended to be consistent with the clock_opt ccd switch state

    set_app_options -name place.coarse.icg_auto_bound -value true

    # Set bounds for icg and the registers it drives, and putting them together can improve the
    violation of reg2clkGate.

    set_app_options -name opt.buffering.enable_rebuffering -value true

    3. create bound

    Limit the area of ​the specified cell place

    create_bound -name boundname -boundary {{lx ly} {ux uy}}

    add_to_bound boundname [get_flat_cells**]

    set_size_only [get_flat_cells **]

    Even if there is a bound, cells will still be copied and placed outside the bound during
    optimization. Using size only allows the constrained cells to only be placed in the bound.

    4. missing def

    Do not add scan def when doing some small modules without dft or create placement try
    floorplan in the floorplan stage

    set_app_options -name \

    place.coarose.continue_on_missing_scandef -value true

    5.spacing rule

    remove_placement_spacing_rules -all

    set_placement_spacing_label -name X -side both -lib_cells [get_lib_cells */*]

    set_placement_spacing_rule -label {XX} {1 1}

    set_placement_spacing_rule -label {X SNPS_BOUNDARY} {1 1}

    If you add filler to generate 1x spacing, you can use create_stdcell:fillers -lib_cells
    */filler_list -rules no_1x

    6. Magnet placement

    3/5
    Preplace the pin/port related logic of the fixed instance.

    magnet_placement\

    -move_fixed //Soft constraints, optimization can still be moved \

    -mark_fixed //hard constraint, fixed after magnet \

    -stop_on_sequential_cells [get_cells xxx] //magnet trace to stop sequential devices,


    including sequential devices, there is also a stop_by** \

    -logic_level 1 //Select the level of magnet.

    7.keep module port

    Prevent some module ports from being optimized during the optimization process.

    set_freeze_port -all [get_cells -filter "is_hierarchical == true"]

    8. Consider the place of preroute

    set_app_options -name place.legalize.enable_advanced_legalizer -value true

    set_app_options -name place.legalize.enable_advanced_prerouted_net_check -value true

    set_app_options -name place.legalize.enable_prerouted_net_check -value true //Considering


    the pin shape or blockage inside the std cell, place can make the std cell avoid places that
    will generate drc, such as near pg via

    set_app_options -name plan.pgroute.honor_std_cell_drc -value true

    9. legalize

    legalize_placement \

    -cell [get_cells xx] \

    -moveable_distance 10 //The cells within the specified cell10um range can be moved 

    set_app_options -name \

    place.legalize.report_displacement_limit -value 1000 //report upper limit of 1000

    set_app_options -name \

    place.legalize.stream_place -value true //More instance small displacements instead of large


    displacements

    set_app_options -name \

    place.legalize.optimize_orientations -value true //Change direction to reduce displacement

    4/5
    Shilu Building

    WeChat public account

    IC Backend Study Guide

    5/5

    You might also like