How to change parameter of cell when replace library

[Nguyenthehuy1]

Hello everyone,

I want to replace the original library with my library. Still, I can not find anything about the timing, power, or ... and where the original library inserts the file library (with .lib)?

Thank you Best regards Huy

[tangxifan]

Hi @Nguyenthehuy1 We do not need a .lib file when running OpenFPGA. In the openfpga architecture file, you may define the name of circuit models using the name of standard cell or any other cell from your library. Then the output netlists are mapped to your technology library.

Consider the example: https://openfpga.readthedocs.io/en/master/manual/arch_lang/circuit_model_examples/#standard-cell-multiplexer If you have a MUX standard cell, e.g., MUX2D1, you can name the circuit model as

<circuit_model type="mux" name="MUX2D1" prefix="MUX2D1">

[Nguyenthehuy1]

Thank you so much for responding to me,

1. As you said, we do not need a .lib file, so do we need a .db file? If not, how openFPGA architecture can understand timings or delays..., because “verilog_neslist” is just about function, especially the “delay_matrix” characterize parameter, i only searched “pb_type_grapth_annotations.cpp” file to solve the “delay_matrix” parameter, where i can find function to calculate “delay_matrix” parameter?

2. and if I change “vdd, pn_ratio.. in technology_library in K6_fac_N10_adder_chanin_40nm_openfaga.xml”, where I can find this change (output)?

Thank you

[tangxifan]

You do not need a .db file. We do not handle delay or timing through .db file. We expect designers to back-annotate timing to our architecture file. See the delay parameters in the VPR architecture description:

https://docs.verilogtorouting.org/en/latest/arch/reference/#tag-%3Cdelay_constantmax=

The VDD, pn ratio etc. are required for FPGA-SPICE. If you are not using it, just leave dummy values.

[Nguyenthehuy1]

Thank you for responding to me,

what about delay_matrix values? What do you rely on to determine those values? In a .lib file, we usually have rise and fall time to determine its pointing to n x n matrix, for example: (if index_1 = 0.01 (position 1 of index 1), and index_2 = 0.0005 (position 1 of index 2) => we have the value: 1x1 (position of value))

index_1("0.0100000000, 0.0230505800, 0.0531329300, 0.1224745000, 0.2823108000, 0.6507428000, 1.5000000000"); index_2("0.0005000000, 0.0012265480, 0.0030088420, 0.0073809800, 0.0181062600, 0.0444164000, 0.1089577000"); values("0.0335820000, 0.0324435000, 0.0295460000, 0.0224904000, 0.0052049000, -0.0370580000, -0.1406482000", \ "0.0330832000, 0.0319430000, 0.0290717000, 0.0220002000, 0.0046753000, -0.0375541000, -0.1412124000", \ "0.0325117000, 0.0313173000, 0.0285060000, 0.0214064000, 0.0040939000, -0.0381634000, -0.1417224000", \ "0.0319710000, 0.0307878000, 0.0279193000, 0.0208373000, 0.0035437000, -0.0386644000, -0.1422429000", \ "0.0314140000, 0.0302145000, 0.0274018000, 0.0205042000, 0.0033479000, -0.0388361000, -0.1425158000", \ "0.0309295000, 0.0297733000, 0.0269733000, 0.0200110000, 0.0026365000, -0.0393728000, -0.1428567000", \ "0.0334235000, 0.0323313000, 0.0294552000, 0.0222344000, 0.0049136000, -0.0379047000, -0.1425150000");

[tangxifan]

@Nguyenthehuy1 See my answer in #1221