Incorrect internal_power values in generated SKY130 Liberty files

[Blueonics]

Describe the bug
The Liberty (.lib) files generated by OpenRAM contain incorrect internal_power values on the clk0 pin. The values are extremely large (e.g. 1.034875e+10 for a 224-byte SRAM, 4.603410e+10 for a 1KB SRAM) .

When read into an EDA tool (e.g. OpenROAD), the macro power dominates and reports in the tens of megawatts, making power analysis unusable.

Version
Commit ab33017fe2d5b7fa0aa730aa47c8c1811577caa8 (upstream/dev)

To Reproduce

1. Generate a Liberty file using OpenRAM with the following configuration for sky130_sram_224bytes_1rw_28x64_28:

"""
Single-port SRAM (64 x 28b)
Used for: hpdcache directory (valid+wback+dirty+fetch+tag = 28 bits, 64 sets)
"""
word_size   = 28
num_words   = 64
human_byte_size = "{:.0f}bytes".format((word_size * num_words) / 8)

write_size  = 28

num_rw_ports = 1
num_r_ports  = 0
num_w_ports  = 0

num_spare_rows = 1
num_spare_cols = 1

ports_human = '1rw'
words_per_row = 1

import os
exec(open(os.path.join(os.path.dirname(__file__), 'sky130_sram_common.py')).read())

Inspect the clk0 pin's internal_power entries in the generated .lib and observe that all four when conditions have identical rise_power and fall_power values of 1.034875e+10.

Expected behavior
The generated Liberty file should contain physically realistic internal_power values on the order of a few mW for a small SKY130 SRAM at TT/1.8V/25C.

Logs
Example snippet from sky130_sram_224bytes_1rw_28x64_28_TT_1p8V_25C.lib:

pin(clk0){
    internal_power(){
        when : "!csb0 & !web0";
        rise_power(scalar){
            values("1.034875e+10");
        }
        fall_power(scalar){
            values("1.034875e+10");
        }
    }
    internal_power(){
        when : "csb0 & !web0";
        rise_power(scalar){
            values("1.034875e+10");
        }
        fall_power(scalar){
            values("1.034875e+10");
        }
    }
    internal_power(){
        when : "!csb0 & web0";
        rise_power(scalar){
            values("1.034875e+10");
        }
        fall_power(scalar){
            values("1.034875e+10");
        }
    }
    internal_power(){
        when : "csb0 & web0";
        rise_power(scalar){
            values("1.034875e+10");
        }
        fall_power(scalar){
            values("1.034875e+10");
        }
    }
}

Resulting OpenROAD power report:

Group        Internal       Switching    Leakage      Total
Macro        2.35e+07       0.00e+00     4.41e-05     2.35e+07   100.0%

(23.5 MW reported for macro power across all SRAM instances, clearly incorrect)

Is this a bug or is it a config issue that I'm making?
Any advice is appreciated.

[mguthaus]

Yes, the sky130 library is not calibrated and needs to be calibrated. We welcome contributions if you have the time to do it.

[mguthaus]

I assume you used the analytical models in the above comment? If not, and you used spice, there might be a units bug.