Skip to main content

mflowgen: A Modular ASIC and FPGA Flow Generator

Project description

mflowgen

Documentation Status mflowgen pytest CI pypi

mflowgen is a modular flow specification and build-system generator for ASIC and FPGA design-space exploration built around sandboxed and modular nodes.

mflowgen allows you to programmatically define and parameterize a graph of nodes (i.e., sandboxes that run anything you like) with well-defined inputs and outputs. Build system files (e.g., make, ninja) are then generated which shuttle files between nodes before running them.

Key features and design philosophies:

  • Process and technology independence -- Process technology libraries and variables can be abstracted and separated from physical design scripts. Specifically, a single node called the ASIC design kit (ADK) captures this material in one place for better maintainability and access control.

  • Sandboxed and modular nodes -- Traditional ASIC flows are composed of many steps executing with fixed path dependencies. The resulting flows have low reusability across designs and technology nodes and can be confusing and monolithic. In contrast, modularity encourages reuse of the same scripts across many projects, while sandboxing makes each node self-contained and also makes the role of each node easy to understand (i.e., take these inputs and generate those outputs).

  • Programmatically defined build-system generator: A Python-based scripting interface and a simple graph API allows flexible connection and disconnection of edges, insertion and removal of nodes, and parameter space expansions. A simple graph can be specified for a quick synthesis and place-and-route spin, or a more complex graph can be built for a more aggressive chip tapeout (reusing many of the same nodes from before).

  • Run-time assertions -- Assertions can be built into each modular node and checked at run-time. Preconditions and postconditions are simply Python snippets that run before and after a node to catch unexpected situations that arise at build time. Assertions are collected and run with pytest. The mflowgen graph-building DSL can also extend a node with design-specific assertions by extending Python lists.

  • A focus on hardware design-space exploration -- Parameter expansion can be applied to nodes to quickly spin out parallel builds for design-space exploration at both smaller scales with a single parameter (e.g., sweeping clock targets) as well as at larger scales with multiple parameters (e.g., to characterize the area-energy tradeoff space of a new architectural widget with different knobs). Dependent files are shuttled to each sandbox as needed.

  • Complete freedom in defining what nodes do -- Aside from exposing precisely what the inputs and outputs are, no other restrictions are placed on what nodes do. A node may conduct an analysis pass and report a gate count. A node can also apply a transform pass to a netlist before passing it to other tools. In addition, a node can even instantiate a subgraph to implement a hierarchical flow.

mflowgen ships with a limited set of ASIC flow scripts for both open-source and commercial tools including synthesis (e.g., Synopsys DC, yosys), place and route (e.g., Cadence Innovus Foundation Flow, RePlAce, graywolf, qrouter), and signoff (e.g., Synopsys PTPX, Mentor Calibre). In addition, we include an open-source 45nm ASIC design kit (ADK) assembled from FreePDK45 version 1.4 and the NanGate Open Cell Library.

More info can be found in the documentation.


License

mflowgen is offered under the terms of the Open Source Initiative BSD 3-Clause License. More information about this license can be found here:

Project details


Download files

Download the file for your platform. If you're not sure which to choose, learn more about installing packages.

Source Distribution

mflowgen-0.7.0.tar.gz (6.9 MB view details)

Uploaded Source

Built Distribution

mflowgen-0.7.0-py3-none-any.whl (7.2 MB view details)

Uploaded Python 3

File details

Details for the file mflowgen-0.7.0.tar.gz.

File metadata

  • Download URL: mflowgen-0.7.0.tar.gz
  • Upload date:
  • Size: 6.9 MB
  • Tags: Source
  • Uploaded using Trusted Publishing? No
  • Uploaded via: twine/4.0.2 CPython/3.9.6

File hashes

Hashes for mflowgen-0.7.0.tar.gz
Algorithm Hash digest
SHA256 d31eb2e2ef19a48029ec129a8a4be6270e27af6bae0c92fb94a63d5693666c43
MD5 ad823ded7c05e9314104145c4018e8ea
BLAKE2b-256 bd360292f08f1703e9e9f2c0a17d76a1bf976092243b28a91079d69ae25b6f26

See more details on using hashes here.

File details

Details for the file mflowgen-0.7.0-py3-none-any.whl.

File metadata

  • Download URL: mflowgen-0.7.0-py3-none-any.whl
  • Upload date:
  • Size: 7.2 MB
  • Tags: Python 3
  • Uploaded using Trusted Publishing? No
  • Uploaded via: twine/4.0.2 CPython/3.9.6

File hashes

Hashes for mflowgen-0.7.0-py3-none-any.whl
Algorithm Hash digest
SHA256 cbe32477b48c97244cb1a9adfc8edc08f72a7825a8e4b984df95c55ff08d11fe
MD5 aabf71ec0e5e0080b0b1e0456b227f87
BLAKE2b-256 c1e2927448b6f861e115a51a738e80f6835df101a51f602fe7bb3e1d3ea194ae

See more details on using hashes here.

Supported by

AWS AWS Cloud computing and Security Sponsor Datadog Datadog Monitoring Fastly Fastly CDN Google Google Download Analytics Microsoft Microsoft PSF Sponsor Pingdom Pingdom Monitoring Sentry Sentry Error logging StatusPage StatusPage Status page