fir

FIR filter checked against a NumPy model

The testbench asks a NumPy model for random stimuli and the expected output of a FIR filter. VUnit's AXI-stream verification components drive the stimuli into the filter and check every output sample.

src/fir.vhd       the design: a 4-tap FIR filter
tb/tb_fir.vhd     the testbench
tb/model.py       the NumPy model

Run it

Set up the environment as described in the top-level README, then:

python run.py

Change COEFFS in tb/model.py and the test fails with a TDATA mismatch. VUnit picks a new seed, and so new stimuli, on every run; python run.py --seed repeat reruns the last one.

How it works

run.py in the repository root adds the bridge as a VUnit package:

vu.add_package("vunit-python-bridge", allow_setup=True)

The testbench loads the model and calls it. integer_array_t values cross as NumPy arrays:

import_module_from_file(join(tb_path(runner_cfg), "model.py"), "model");
stimuli := call("model.stimuli", arg(1000), arg_unsigned(get_seed(runner_cfg)));
expected := call("model.fir", arg(stimuli));