MAGICAL

MAGICAL: Machine Generated Analog IC Layout

This is the top-level MAGICAL flow repository. In MAGICAL, we maintain seperate components, such as constraint generation, placement and routing, in different repository. And we integrate each component through top-level python flow.

This project is currently still under active development.

Dependency

• Python 3.7 and additional packages

  • Refer to Dockerfile.

• Boost

  • Need to install and visible for linking. Required version at least 1.62.

• Flex

  • Need to install, required by Limbo package

• Zlib

  • Required by Limbo package

• Limbo

  • Required latest

• LPSolve 5.5

  • Required version 5.5

• Lemon 1.3.1

  • Required version 1.3.1

How to clone

To clone the repository and submodules, go to the path to download the repository.

# clone the repository 
git clone https://github.com/magical-eda/MAGICAL.git
git submodule init
git submodule update

How to build

Two options are provided for building: with and without Docker. You can also build from source (NOT RECOMMENDED) resolving the required dependancies first.

Build with Docker

You can use the Docker container to avoid building all the dependencies yourself.

1. Install Docker on Linux.
2. Navigate to the repository.

   cd MAGICAL
   

3. Get the docker container with either of the following options.
   • Option 1 (Recommended): pull from the cloud jayl940712/magical.

   docker pull jayl940712/magical:latest
   

   • Option 2: build the container.

   docker build . --file Dockerfile --tag magical:latest
   

4. Run the docker container

   docker run -it -v $(pwd):/MAGICAL jayl940712/magical:latest bash
   

   Or if you used option 2 to build the container

   docker run -it -v $(pwd):/MAGICAL magical:latest bash
   

How to run

Benchmark circuit examples are under examples/

All technology related parameters including benchmark circuit sizing are samples and not related to any proprietary PDK information.

Benchmark circuits currently includes: 1 adc, 1 comparator, 3 ota

To run the benchmark circuits

cd /MAGICAL/examples/BENCH/ (ex. adc1)
source run.sh

The output layout gdsii files: BENCH/TOP_CIRCUIT.route.gds (ex. adc1/xxx.route.gds)

Note: currently adc2 have routing issues.

Custom layout constraint inputs

The automatic symmetry constraint generation is currently embedded into the flow. To ensure circuit functionality it is ideal that designers provide constraints to guide the placement and routing.

A sample device and net symmetry constraint is given for adc1. These files should also be the output for the current automatic symmetry constraint generation flow.

Sample symmetry device constraint file: examples/adc1/CTDSM_TOP.sym

Sample symmetry net constraint file: examples/adc1/CTDSM_TOP.symnet

Device symmetry constraints

Device symmetry constraints greatly affect the placement solution and output layout quality. Currently we only consider symmetry groups, symmetry device pairs and self-symmetric device constraints.

Symmetry group: A group of symmetry device pairs and self-symmetric devices that share the same symmetry axis.

Symmetry device pair: Two devices that are reflection symmetric with respect to a symmetry axis (usually vertical).

Self-symmetry device: A single device that is reflection symmetric with itself respect to a symmetry axis.

Net symmetry constraints

Similar to device symmetry constraints, we consider symmetry net pairs and self-symmetry net constraints.

Symmetry net pair: Two nets that are reflection symmetric with respect to a symmetry axis (usually vertical). For a valid constraint, the corresponding pins of the two nets must be reflective symmetric with a axix.

Self-symmetry net: A single net that is reflection symmetric with itself respect to a symmetry axis.

License

BSD 3-Clause