wavekit
English | 中文
Wavekit is a Python library for high-level digital waveform analysis.
It loads signals from waveform files as clock-aligned Waveform objects,
allowing users to query and operate on them at the cycle and transaction levels,
making complex hardware behavior easier to analyze.
Quick links: Online playground ·
AI integration
Features
- Flexible signal queries: find and batch-load related signals from
hierarchical waveform data using multiple path-matching options.
- Cycle-level analysis: use a range of waveform operations on clock-sampled
data to analyze cycle-based behavior such as interface backpressure and FIFO
occupancy.
- Transaction-level analysis: use temporal pattern matching to describe
signal relationships across multiple clock cycles for protocol analysis,
transaction extraction, and latency measurement.
- Multi-format waveform support: load and analyze VCD, FST, and FSDB files
using the same API.
Installation
python -m pip install wavekit
wavekit supports Python 3.9 and newer. FSDB support requires the Verdi NPI
runtime (libNPI.so); see the installation guide for
setup details.
Quick start
Calculate the occupancy of a FIFO from a VCD waveform:
import numpy as np
from wavekit import VcdReader
with VcdReader('simulation.vcd') as r:
fifo = r['tb.u_fifo']
with r.clock_domain(clock='tb.clk'):
w_ptr = fifo['w_ptr'].w
r_ptr = fifo['r_ptr'].w
depth = 16
occupancy = (w_ptr + depth - r_ptr) % depth
print('Average occupancy:', np.mean(occupancy.value))
See the first waveform tutorial for a
complete runnable FIFO example.
Documentation
See the documentation for the full reference and guides:
Development
Install the development dependencies and run the checks with Poetry:
poetry install
poetry run pytest
poetry run ruff check .
poetry run ruff format --check .
poetry run mypy
Build the documentation locally with Python 3.10 or newer:
poetry install --with docs
poetry run zensical build --clean --strict
See Contributing for contribution guidelines.
License
This project is licensed under the MIT License. See LICENSE.
评论