Using Python

The NYU EDA servers provide a system Python installation, but pip is not available and you do not have permission to install system‑wide packages. To work around this, students should use uv, a modern, fast, user‑level Python package manager that installs entirely in your home directory. This suggestion was provided by the student Ashesh Kaji, so please thank him next time.

‘uv is a drop‑in replacement for:

  • pip
  • virtualenv
  • pipx
  • pip-tools
  • poetry (for basic workflows)

It requires no sudo access and works perfectly on the NYU servers.

1. Install uv

Log into the EDA server:

ssh <netid>@ecs03.poly.edu

Then install uv into your home directory:

curl -LsSf https://astral.sh/uv/install.sh | sh

This command installs uv as a local binary in the directory ~/.local/bin. We need to add that directory to your path. Linux has different shells, and the commands differ depending on which one you use. Check your shell with:

echo $SHELL

The output ends in either tcsh or bash. Follow the matching section below — the commands are not interchangeable. You can edit the file with any Linux editor, such as vi.

If your shell is tcsh

Add these lines to the end of ~/.tcshrc, which adds uv to your path and disables the system Python:

setenv PATH "$HOME/.local/bin:$PATH"
unsetenv PYTHONPATH

Then reload the shell configuration:

source ~/.tcshrc

If your shell is bash

Add these lines to the end of ~/.bashrc instead:

export PATH="$HOME/.local/bin:$PATH"
unset PYTHONPATH

Then reload the shell configuration:

source ~/.bashrc

Both shells

You only need to do this once. Subsequent logins will run the configuration automatically.

You can verify installation with:

uv --version

2. Install a current Python

The system Python on the EDA servers is too old. It is version 3.9, but waveflow requires 3.10 or newer. If you build the environment on the system Python, the install fails later with:

× No solution found when resolving dependencies:
╰─▶ Because the current Python version (3.9.25) does not satisfy
    Python>=3.10 and waveflow==0.1.0 depends on Python>=3.10 ...

You cannot fix this with a requirements.txt — that file lists packages, and the Python interpreter is not one of them. Instead, let uv install a newer Python for you, as shown below. This is one of the main reasons we use uv here: it installs the interpreter into your home directory and needs no administrator access.

Install Python 3.12 (the version the course material is developed against):

uv python install 3.12

This downloads a self-contained Python into your home directory. It does not touch the system Python, and no other user is affected.

3. Create a Virtual Environment

Navigate to the directory where you cloned the hwdesign repo. Generally, this is ~/hwdesign:

cd ~/hwdesign

Create the environment inside the repository, not in your home directory. The uv pip install -e . step below installs the hwdesign package from the current directory, and it will fail anywhere else with an error about a missing pyproject.toml.

Then create the environment, telling uv to use the Python you just installed:

uv venv --python 3.12

If you already created a .venv with the old system Python, uv will refuse to overwrite it. Replace it with:

uv venv --python 3.12 --clear

This creates a .venv/ folder containing a private Python environment.

Activate it:

source .venv/bin/activate.csh  # for tcsh
source .venv/bin/activate      # for bash

Your prompt should now show something like:

(.venv) <netid>@ecs03:~/hwdesign$

You can deactivate with:

deactivate

4. Install the course packages

The course uses two packages: waveflow (a general-purpose hardware modeling framework, kept in a separate repository) and hwdesign (this course’s own helpers, in this repo). See Installing the Python packages for what each one does — this page covers only the uv commands you need on the NYU servers.

From the hwdesign directory, with the virtual environment activated, install waveflow first:

uv pip install git+https://github.com/sdrangan/waveflow.git

Then install hwdesign itself as editable:

uv pip install -e .

⚠️ Run these two commands in this order. hwdesign depends on waveflow, but waveflow is distributed from GitHub rather than PyPI. If you run uv pip install -e . first, the install fails because it cannot find waveflow on PyPI. Installing waveflow first satisfies the dependency.

Do not use requirements.txt here. That file is a pre-migration snapshot of a different environment and does not list either package; the two commands above are the complete install.

Do not uv pip install pywaveflow yet. waveflow is published on PyPI under that name, but the version currently there is only a placeholder reserving the name and contains none of the actual framework. Install from GitHub as shown above until a real release is published.

Verify that both packages are importable:

python -c "import hwdesign, waveflow; print('OK')"

If this raises ModuleNotFoundError, the usual cause is that the virtual environment is not activated. If it raises an error mentioning pysilicon, that is unmigrated course material rather than a broken install — see the note at the end of Installing the Python packages.

(An earlier build of the EDA servers set a stray LD_PRELOAD that produced shared-library errors before Python even started. Those configuration files have since been removed, so this should no longer occur.)

5. Running Python

With the environment activated, run scripts with plain python:

python your_script.py

Console scripts such as sv_sim are on your path once the environment is activated:

sv_sim --source [source files] --tb [tb_files]

Everything runs inside your private environment, not the system Python.

Prefer the activated environment over uv run. Because this directory contains a pyproject.toml, uv run treats it as a project and may try to re-resolve its dependencies before running — which fails, since waveflow is not on PyPI. If you see a resolution error mentioning waveflow, activate the environment and run python directly as shown above.

6. Why We Use uv Instead of pip

The NYU EDA servers:

  • Do not include pip
  • Do not allow system‑wide package installation
  • Use a system Python that students cannot modify
  • Ship a system Python (3.9) older than the course requires (3.10+)

uv solves all of these problems:

  • Installs into your home directory
  • Requires no sudo
  • Installs Python itself, so you are not stuck with the system version
  • Manages virtual environments automatically
  • Works on macOS, Windows, and Linux
  • Is significantly faster and more reliable than pip
This site uses Just the Docs, a documentation theme for Jekyll.