VCSV Genie

This repo details methods for parsing transient simulation VCSV files from Cadence Virtuoso.

Installation

Install via pip:

pip install vcsvgenie

https://pypi.org/project/vcsvgenie/

Installation from source (using Poetry)

1. Clone this repo

git clone https://github.com/FSharp4/vcsvgenie.git
cd vcsvgenie

2. Install to an environment

# If you haven't installed poetry:
pip install poetry

# Install project + dependencies to virtual environment
poetry install

3. Copy over a vcsv file of your choice, and use vcsvgenie to process the file, extracting propagation delays, printing traces, etc.

Installation from source

1. Clone this repo

git clone https://github.com/FSharp4/vcsvgenie.git
cd vcsvgenie

2. Build this repo

pip install build
python -m build

3. Install the repo

pip install -e .

4. Copy over a vcsv file of your choice, and use vcsvgenie to process the file, extracting propagation delays, printing traces, etc.

Example files can be seen in vcsvgenie/_dev.

Remark: In order for VCSV genie to automatically parse your file for propagations, you need to specify which signals are 'outputs' and which signals are inputs.

• Signals must be 'outputs' for propagation delay calculations to be performed on them with respect to inputs.
• Within Virtuoso Schematics/Layouts, these signals may be inputOutput or intermediate.

To specify which signals are inputs/outputs, supply lists of signal names to the input and output constructor arguments for TransientResultSpecification. (See Usage for an example).)

Functionality

Transient Analysis

Users can:

• Create pandas dataframes or numpy arrays from VCSV files
• Create $(x, y)$ dataseries of individual waveforms (note that all timestamp $x$ vectors are the same in a VCSV file, and are not uniformly spaced)
• Create collections of waveforms from specifications (vcsvgenie.transient_waveform.TransientResult)
• Recognize signal buses using caret notation (i.e., A<3:0>) by specifying the individual signals in the TransientResultSpecification
• Digitize and tabulate signal bus data
• Identify and measure critical path(s) propagation delay(s) via vcsvgenie.transient_waveform.critical_propagation_delays and quasicritical_propagation_delays
• Estimate global worst-case delays via vcsvgenie.transient_waveform.estimate_global_critical_delay

DC Analysis

Users can:

• Load waveforms from a DC analysis into a rudimentary DC Results object (no functionality yet)
• Measure Read SRAM Noise Margin
• Measure Write SRAM Noise Margin

Usage

General Example:

from pathlib import Path
from pprint import pprint

from vcsvgenie.read import read_vcsv
from vcsvgenie.transient_waveform import TransientResultSpecification, average_propagation_delays_by_category,
    maximum_propagation_delays_by_category
from vcsvgenie.waveform import construct_waveforms

path = Path("example.vcsv")
dataframe, titles = read_vcsv(path)
waveforms = construct_waveforms(dataframe, titles)

specification = TransientResultSpecification(
    inputs=[
        '/A<3>', '/A<2>', '/A<1>', '/A<0', '/B<3>', '/B<2>', '/B<1>', '/B<0>', 'Clk'
    ],
    outputs=['/z<7>', '/z<6>', '/z<5>', '/z<4>', '/z<3>', '/z<2>', '/z<1>', '/z<0>'],
    clock_period=1e-9,
    logic_threshold=0.5  # volts
)

results = specification.interpret(waveforms)
results.find_transitions()
results.find_propagations()

averages = average_propagation_delays_by_category(results.propagations)
# pprint(results.propagations)
# pprint(averages)

maxima = maximum_propagation_delays_by_category(results.propagations)
# pprint(maxima)

results.digitize()
results.resolve_buses()
bus_data = results.tabulate_bus_data()
bus_data.to_csv("bus_data.csv")

results.plot(separate=True)

Retrieving observed worst-case delays:

from pprint import pprint

n = 10 # number of delays to retrieve
quasicritical_delays = transient_waveform.quasicritical_propagation_delays(results.propagations, n)
pprint([str(quasicritical) for quasicritical in quasicritical_delays])

Estimating global worst-case delays (in simulations on random inputs):

maxtpd_idx, maxtpd = transient_waveform.find_max_delay_trend(results.delays)
transient_waveform.plot_max_delay_trend(maxtpd_idx, maxtpd, show=True)

global_estimation = transient_waveform.estimate_global_critical_delay(maxtpd_idx, maxtpd, thres_samp=thres_samp)

Dependencies

• numpy
• matplotlib
• pandas
• sortedcontainers
• (optional): jupyter

Contribution

This code targets python 3.11 syntax.

This codebase follows semver guidelines:

• Patch versions indicate bug fixes or documentation adjustments with no intentional functionality changes
• Minor versions indicate functionality changes or additions with no breaking changes
• Major versions indicate breaking functionality changes