pytcl-eda

PyTCL

PyTCL allows control EDA tools directly from Python that use TCL.

Features

• It executes Python method with provided positional arguments directly as TCL procedure For example invocation of Python <object>.<name>(*args) method is like calling TCL procedure <name> {*}${args}
• Any Python value is converted to TCL value like for example Python list to TCL list
• Result from invoked TCL procedure is returned as pytcl.TCLValue that can handle any TCL value (that is represented always as string) to Python str, int, bool, float, list, dict, ...
• TCL error is returned as Python exception pytcl.TCLError
• High performance and very low (unnoticeable) overhead by using Unix domain sockets or Windows named pipes for communication between Python and TCL in streamable way (sockets/pipes are always open and ready)
• It allows to create and access TCL variables from Python side. Please see tests/test_tclsh.py for some examples
• It can work with any EDA tool. Please see tests/test_vivado.py how to use bare PyTCL class for that
• It supports Linux, macOS and Windows
• No external dependencies

Install

pip install pytcl-eda

Examples

Creating new Vivado project:

#!/usr/bin/env python3
from pathlib import Path
from pytcl import Vivado

def main() -> None:
    """Create new Vivado project."""
    hdl_dir: Path = Path.cwd() / "hdl"
    project_dir: Path = Path.cwd() / "my-awesome-project"

    with Vivado() as vivado:
        # See Vivado Design Suite Tcl Command Reference Guide (UG835) for all available Vivado TCL procedures
        # https://docs.amd.com/r/en-US/ug835-vivado-tcl-commands
        vivado.create_project(project_dir.name, project_dir)
        vivado.add_files(hdl_dir / "my_awesome_design.sv")

        synthesis_runs = list(vivado.get_runs("synth_*"))
        vivado.launch_runs(synthesis_runs)

        # wait_on_runs was introduced in Vivado 2021.2. For backward compatibility we will use wait_on_run
        # https://docs.amd.com/r/2021.2-English/ug835-vivado-tcl-commands/wait_on_runs
        # Vivado >= 2021.2 can just use: vivado.wait_on_runs(synthesis_runs)
        for run in synthesis_runs:
            vivado.wait_on_run(run)

        implementation_runs = list(vivado.get_runs("impl_*"))
        vivado.launch_runs(implementation_runs)

        for run in implementation_runs:
            vivado.wait_on_run(run)

        vivado.close_project()

if __name__ == "__main__":
    main()

To use any EDA tool where PyTCL doesn't provide neat helper classes like pytcl.Vivado you can use the pytcl.PyTCL class directly:

#!/usr/bin/env python3
from pathlib import Path
from pytcl import PyTCL

def main() -> None:
    """Create new Vivado project."""
    project_dir: Path = Path.cwd() / "my-awesome-project"

    # PyTCL offers some string placeholders {} that you can use:
    # {tcl}      -> it will insert <pytcl>/execute.tcl
    # {address}  -> it will insert Unix socket, Windows named pipe or network address
    cmd: list[str] = [
        "vivado",
        "-nojournal",
        "-notrace",
        "-nolog",
        "-mode",
        "batch",
        "-source",
        "{tcl}",
        "-tclargs",
        "{address}",
    ]

    with PyTCL(*cmd) as vivado:
        vivado.create_project(project_dir.name, project_dir)

        # Do the same magic that you would normally do in TCL

        vivado.close_project()

if __name__ == "__main__":
    main()

Architecture

stateDiagram-v2
    direction LR
    PyTCL --> connection: send()
    connection --> execute.py: string
    state tool {
        execute.py --> execute.tcl: stdin
        execute.tcl --> execute.py: stdout
    }
    execute.py --> connection: NDJSON
    connection --> PyTCL: recv()

• PyTCL will run tool with execute.tcl <address> arguments in separate process
• execute.tcl invoked by tool will invoke execute.py
• execute.py will start new connection listener on provided <address>
• PyTCL will try connect to <address> as client
• PyTCL will transform any Python method call <object>.<name>(*args) to TCL expression <name> {*}${args}
• PyTCL will send TCL expression as string to execute.py via <address>
• execute.py will print received TCL expression to standard output stdout
• execute.tcl will capture received TCL expression from execute.py using standard input stdin
• execute.tcl will evaluate received TCL expression
• execute.tcl will print result of evaluated TCL expression back to execute.py using standard output stdout in form of NDJSON {"result": "<tcl-result>", "status": <tcl-status>}
• execute.py will capture above result using standard input stdin and send it back to PyTCL
• PyTCL will return received NDJSON message as pytcl.TCLValue
• PyTCL will raise a Python exception pytcl.TCLError if received TCL status was non-zero

Development

Create Python virtual environment:

python3 -m venv .venv

Activate created Python virtual environment:

. .venv/bin/activate

Upgrade pip:

pip install --upgrade pip

Install project in editable mode with pytest:

pip install --editable .[test]

Run tests:

pytest