Skip to content

Latest commit

 

History

History
206 lines (161 loc) · 6.57 KB

naja-edit-python-examples.md

File metadata and controls

206 lines (161 loc) · 6.57 KB
Raw

Naja Python API Examples

The Naja Python API provides powerful tools for working with hierarchical netlists. It allows users to:

  • Browse netlist data: Collect detailed information and navigate through design hierarchy seamlessly.
  • Apply ECO transformations: Make engineering changes without disrupting the design's core structure.

To get started, create a Python script with an edit function. Here’s an example setup:

from naja import snl

def edit():
    universe = snl.SNLUniverse.get()
    top = universe.getTopDesign()

    # Perform actions with the top-level design

Examples

1. Print All Design Content

This script recursively explores the design hierarchy and prints information about instances, terminals, nets, and their connections.

from naja import snl

def print_instance_tree(design):
    for ins in design.getInstances():
        print(f"Instance: {ins.getName()}")
        model = ins.getModel()
        for term in design.getTerms():
            print(f"  Terminal: {term}")
        for net in design.getNets():
            print(f"  Net: {net}")
            for bit in net.getBits():
                for component in bit.getComponents():
                    print(f"    Component: {component}")
        print_instance_tree(model)

def edit():
    universe = snl.SNLUniverse.get()
    top = universe.getTopDesign()

    print_instance_tree(top)

2. Remove Interface Buffers from an FPGA Design

The following script removes specific interface buffers (IBUF, OBUF, and BUFG) from an FPGA design.

from naja import snl

def delete_io_bufs(design):
    for ins in design.getInstances():
        model = ins.getModel()
        if model.isPrimitive():
            model_name = model.getName()
            if model_name in ['IBUF', 'BUFG']:
                input_net = ins.getInstTerm(model.getScalarTerm('I')).getNet()
                output_net = ins.getInstTerm(model.getScalarTerm('O')).getNet()
                for component in output_net.getComponents():
                    component.setNet(input_net)
                output_net.destroy()
                ins.destroy()
            elif model_name == 'OBUF':
                input_net = ins.getInstTerm(model.getScalarTerm('I')).getNet()
                output_net = ins.getInstTerm(model.getScalarTerm('O')).getNet()
                for component in input_net.getComponents():
                    component.setNet(output_net)
                input_net.destroy()
                ins.destroy()

def edit():
    universe = snl.SNLUniverse.get()
    top = universe.getTopDesign()

    delete_io_bufs(top)

3. Visualize Module Statistics

The script below aggregates statistics for all designs in a library (number of terms, nets, and instances) and visualizes them as a bar chart.

from naja import snl
import pandas as pd
import matplotlib.pyplot as plt

def plot_design_stats(library):
    data_list = []
    for design in library.getDesigns():
        nb_terms = sum(1 for _ in design.getBitTerms())
        nb_nets = sum(1 for _ in design.getBitNets())
        nb_instances = sum(1 for _ in design.getInstances())
        data_list.append({
            'design': design.getName(),
            'nb_terms': nb_terms,
            'nb_nets': nb_nets,
            'nb_instances': nb_instances
        })
    pandas_data = pd.DataFrame(data_list).set_index('design')
    plot = pandas_data.plot.bar(y=['nb_terms', 'nb_nets', 'nb_instances'], stacked=True)

    # Customize plot
    plot.set_title('Design Statistics', fontsize=16, fontweight='bold')
    plot.set_xlabel('Design Name', fontsize=12)
    plot.set_ylabel('Count', fontsize=12)

    plot_figure = plot.get_figure()
    plot_figure.tight_layout()
    plot_figure.savefig('design_stats.png')

def edit():
    universe = snl.SNLUniverse.get()
    topDesign = universe.getTopDesign()
    topLibrary = topDesign.getLibrary()

    plot_design_stats(topLibrary)

Generated Plot Example:
Design Stats

4. Compute Fanout for All Equipotentials

This example computes the fanout for all equipotentials (flat nets) in a design, generating a CSV file with the results.

import logging
from naja import snl

def edit():
    logging.basicConfig(filename='edit.log', filemode='w', level=logging.DEBUG)
    universe = snl.SNLUniverse.get()
    if universe is None:
        logging.critical('No loaded SNLUniverse')
        return 1

    top = universe.getTopDesign()
    if top is None:
        logging.critical('SNLUniverse does not contain any top SNLDesign')
        return 1

    logging.info(f'Found top design: {top}')
    primitives = []
    with open('fanout.csv', 'w') as fanout_file:
        fanout_file.write('Net,Model,Fanout\n')
        for inst in top.getInstances():
            path = snl.SNLPath(inst)
            stack = [[inst, path]]
            while stack:
                current = stack.pop()
                currentInst, currentPath = current
                for instChild in currentInst.getModel().getInstances():
                    pathChild = snl.SNLPath(currentPath, instChild)
                    instChildModel = instChild.getModel()
                    if instChildModel.isPrimitive():
                        if not (instChildModel.isConst0() or instChildModel.isConst1()):
                            primitives.append([instChild, pathChild])
                    stack.append([instChild, pathChild])

        for entry in primitives:
            inst, path = entry
            for iterm in filter(lambda term: term.getDirection() == snl.SNLTerm.Direction.Output, inst.getInstTerms()):
                if iterm.getNet():
                    ito = snl.SNLNetComponentOccurrence(path.getHeadPath(), iterm)
                    equi = snl.SNLEquipotential(ito)
                    fanout_count = sum(1 for _ in equi.getInstTermOccurrences())
                    if fanout_count > 1:
                        fanout_file.write(f'{ito},{iterm.getInstance().getModel()},{fanout_count - 1}\n')

You can process the resulting CSV and generate a histogram using the script below:

import pandas as pd
import matplotlib.pyplot as plt

df = pd.read_csv('fanout.csv')
df['Fanout'] = pd.to_numeric(df['Fanout'])
fanout_distribution = df.groupby('Fanout').size().reset_index(name='Net')

plt.bar(fanout_distribution['Fanout'], fanout_distribution['Net'], log=True)
plt.xlabel('Fanout')
plt.ylabel('Number of Nets')
plt.title('Distribution of Nets per Fanout')
plt.yscale('log')
plt.savefig('fanout_distribution.png', dpi=300)
plt.show()

Resulting Plot:
Fanout Distribution