PDKMaster 0.9.2.post1

ASIC PDK Manager and Design Framework

Description

PDK Master is a tool to manage PDKs for ASIC design and a framework for designing circuits and layouts in those technologies. It is a Python framework under heavy development and with an unstable API.

Release history

• v0.9.2: support new namespace in coriolis export
• v0.9.1: fix coriolis export, primitive type values are not python types.
• v0.9.0: release notes
• v0.3.0: release notes
• v0.2.1: add klayout dependency
• v0.2.0: release notes
• v0.1: release notes

Overview

Currently no documentation is available, the documentation is planned to be added as part of the stabilization of the PDKMaster API. To show the things PDKMaster wants to solve here an overview of the different parts of the current PDKMaster code base:

• pdkmaster: the top python module
  • _util.py: some helper functions and classes
  • technology: this submodule handles the description of an ASIC technology with final target to allow describe that in one python file.

    • property_.py: base class to represent properties on operations that can be done on them.

    • rule.py: abstract base class to represent a rule object, e.g. a condition on properties that has to be fulfilled to be manufacturable.

    • mask.py: class to represent the photolithography masks used in ASIC production and the properties on them. The latter are then used to define design rules.

    • edge.py: class representing mask edges and it's properties to be used in design rules.

    • wafer.py: object to represent a (silicon) wafer that is the start of processing and that is auto-connected to some device ports.

    • net.py: class representing a net, e.g. o group of conductors in a circuit that are connected together.

    • primitive.py: classes for all possible device primitives available in a technology. This goes from low-level interconnect to transistors. As indication of the content here the exported classes are given:

      __all__ = ["Marker", "Auxiliary", "ExtraProcess",
                 "Implant", "Well",
                 "Insulator", "WaferWire", "GateWire", "MetalWire", "TopMetalWire",
                 "Via", "PadOpening",
                 "Resistor", "Diode",
                 "MOSFETGate", "MOSFET", "Bipolar",
                 "Spacing",
                 "UnusedPrimitiveError", "UnconnectedPrimitiveError"]
      

      The object attibutes defined by these classes are used to derive mask design rules.

    • technology_.py: class to define the capability of a certain technology: all support devices, the masks needed to define them and the rules for making circuit in this technology.

  • dispatch: helper classes inspired by the Visitor design pattern.
  • design: support code for making circuits compliant with a given technology.
    • circuit.py: defines a factory that allows to generate objects of the Circuit class using devices from a given technology.
    • layout.py: classes to define layout compliant with a given technology and a factory to generate layouts for a given circuit that are technology design rule compliant.
    • library.py: contains:
      • Cell class: representing several possible circuit representations and layouts of a block with the same function
      • Library class: represent a collections of cells
  • io: submodule to import and export technology data, circuit and layouts. It also allows to interface with external tools.
    • parsing: submodule to parse setup files for other EDA tools and extract data to build a PDKMaster technology object based on this data.
      • skill_grammar.py: modgrammar based parser for SKILL(-like) files. SKILL is the Cadence bastardized version of Lisp.
      • tf.py, display.py, layermap.py, assura.py: classes for representing Cadence EDA files based on the SKILL grammar.
    • pdkmaster: support code to export a PDKMaster technology as Python souce code; main targeted use case to use the parsing submodule to extract data from NDA covered PDK and generate PDKMaster Technology object without needing to disctribute NDA covered data.
    • coriolis: support code to generate Coriolis technology setup, cells and libraries from PDKMaster objects.
    • klayout: support code to generate klayout Technology setup for PDKMaster Technology object including DRC and LVS scripts.
    • spice: support code to convert PDKMaster Circuit to use in SPICE simulations; currently PySpice is used to interface with SPICE.

The current code base has been gradually grown to allow to do a 0.18µm prototype layout of the NLNet sponsored Libre-SOC project. It has known (and unknown) inconsistencies and shortcomings. A full revision of the current API is planned without any backwards guarantees whatsoever. As this is an open source project it is meant to be used by other people but one should be aware that user code using PDKMaster has a high chance of breaking with each commit done to the repository in the near future.

Installation

All dependencies for installation should be available so PDKMaster should be able to installed by a simple pip command.

% pip install PDKMaster

To install build dependencies:

% pip install -r dev-requirements.txt

More in depth discussion of different pip use case scenarios is out of the scope of this document.

Run time dependencies:

• modgrammar
• shapely, descartes (deprecated way of plotting in notebooks)
• PySpice >= 1.5
• klayout (pypi, home): for pdkmaster.io.klayout

Licensing Rationale

Open source projects and it's surrounding community can only strive when improvements to the code or application of the code itself are released to the public and allowed to be used by them. Copyleft type license are using the licensing terms to guarantee this actually happens and no party uses a 'take-but-don't-give-back' approach. The GNU Aferro General Public License V3.0 is used as main license for the code in this project as it is a copyleft type license that is also applicable to cloud services without binary distribution of the code.

One of the problems with a strict copyleft license is that it can introduce incompatibilities with code released under other open source licenses. In order to improve compatibility and thus also reusablity the code in this repo is multi-licensed. Multi-licensing under established open source licenses was preferred over custom extension of licenses.

The GNU General Public License V2.0 was added as optional license in order to allow derived code not to be bound by the anti-tivoization clauses introduced in the GNU General Public License Version 3. The latter was not deemed necessary for this project and the addition was done to increase compatibility with some corporate policies.

The CERN Open Hardware Licence Version 2 - Strongly Reciprocal is included as it is a copyleft license specifically targeted for hardware but incompatible with the GPL licenses.

The Apache License Version 2.0 is included to maximize compatiblity with existing open source code. One is supposed to not use it to avoid having to release one's own derived code to the public. If you plan development of a project in a proprietary way, one is kindly asked to not derive one's code from this project's code.

In future the list of allowed licenses may be reduced. A reason could be that such an action is deemed necessary by the project maintainers to encourage open sourcing of derived code.

Analog to how the object files and the executables generated by the gcc compiler are not necessarily goverened by the GPL license, the multi-licensing applies only to code derived from code in this repository. Output files generated through the use of the code in this repository are not by default bound to the multi-licensing requirements of this project's code.

Copyright

The code in this repo is copyrighted by all the contributors to it; git is used to track this copyright.