scripts

Test Scripts and Stimuli Generation

This directory contains Python scripts and ModelSim/Questa DO files for automated test stimuli generation and simulation execution.

Table of Contents

• Overview
• Directory Contents
• Test Script Architecture
• Creating Test Scripts
• Running Scripts
• Related Documentation

Overview

The test script infrastructure automates the generation of cycle-accurate test stimuli for VHDL component verification.

Key features:

• Python-based stimuli generation
• Automated .dat file creation for VHDL testbenches
• ModelSim/Questa DO files for compilation and simulation
• Waveform configuration scripts
• Reference model integration

Script types:

• *_test.py - Python stimuli generators
• sim_*.do - Simulation execution scripts
• wave_*.do - Waveform display configuration

Directory Contents

scripts/
├── README.md          # This file
│
├── Python Stimuli Generators:
├── *_test.py          # Component-specific test generators
├── layer_test.py      # Dense layer tests
├── conv_layer_test.py # Convolutional layer tests
├── neuron_test.py     # Neuron tests
├── network_test.py    # Network-level tests
│
├── ModelSim/Questa Scripts:
├── sim_*.do           # Simulation execution
└── wave_*.do          # Waveform configuration

Test Script Architecture

All test scripts follow a consistent pattern for predictable and maintainable test generation.

Standard Structure

Test scripts conceptualize each DUT port as a Python list of values. Lists are constructed using comprehensions and loops, then formatted into .dat files.

Data Flow Pipeline

Python Script                  DAT File              VHDL Testbench
-------------                  --------              --------------
                              +----------+
d = [d1, d2, d3, d4, d5]      | ex.dat   |           read file line
w = [w1, w2, w3, w4, w5]      |          |           by line
o = [o1, o2, o3, o4, o5]      | d1 w1 o1 |
                        -->   | d2 w2 o2 |    -->    apply d, w to DUT
                              | d3 w3 o3 |           verify o output
                              | d4 w4 o4 |
                              | d5 w5 o5 |
                              +----------+

DAT File Format

Structure:

• One line per simulation cycle
• Space-separated values
• Columns correspond to DUT ports in order
• Values in hexadecimal or decimal format

Example:

# Input_d Input_w Expected_output
00000001 00000002 00000003
00000004 00000005 00000009
00000007 00000008 0000000F

Python Script Pattern

Typical script structure:

# 1. Define test parameters
data_width = 10
num_samples = 100

# 2. Generate input stimuli
input_d = [i for i in range(num_samples)]
input_w = [i * 2 for i in range(num_samples)]

# 3. Calculate expected outputs
output = [d + w for d, w in zip(input_d, input_w)]

# 4. Format and write DAT file
with open('test_data.dat', 'w') as f:
    for d, w, o in zip(input_d, input_w, output):
        f.write(f"{d:08x} {w:08x} {o:08x}\n")

Creating Test Scripts

Step-by-Step Guide

1. Identify DUT ports:

   • List all inputs and outputs
   • Determine data widths and formats

2. Create Python script:

   # component_test.py
   import numpy as np
   
   # Generate test vectors
   inputs = generate_test_inputs()
   expected = calculate_expected_outputs(inputs)
   
   # Write DAT file
   write_dat_file("component_test.dat", inputs, expected)

3. Create DO file:

   # sim_component.do
   vcom ../source/component.vhd
   vcom component_tb.vhd
   vsim component_tb
   do wave_component.do
   run -all

4. Test and validate:

   • Run Python script to generate stimuli
   • Execute DO file in ModelSim/Questa
   • Verify outputs match expectations

Naming Conventions

Consistent naming ensures easy navigation:

File Type	Pattern	Example
Python test	<component>_test.py	neuron_test.py
DAT file	<component>_test.dat	neuron_test.dat
Simulation DO	sim_<component>.do	sim_neuron.do
Waveform DO	wave_<component>.do	wave_neuron.do
Testbench	<component>_tb.vhd	neuron_tb.vhd

Running Scripts

Generating Stimuli

# From scripts/ directory
python neuron_test.py
python layer_test.py

Running Simulations

# In ModelSim/Questa console
cd simulation/scripts
do sim_neuron.do
do sim_layer.do

Complete Workflow

# 1. Generate test data
python component_test.py

# 2. Run simulation
# In ModelSim/Questa:
do sim_component.do

# 3. View results
# Waveforms appear in GUI
# Console shows pass/fail status

Related Documentation

Parent directory:

• ../README.md - Simulation environment overview

Component specifications:

• ../../source/README.md - Layer architectures and specifications

Project overview:

• ../../README.md - Complete workflow and quick start