Awesome
ALIGN: Analog Layout, Intelligently Generated from Netlists
ALIGN is an open-source automatic layout generator for analog circuits jointly developed under the DARPA IDEA program by the University of Minnesota, Texas A&M University, and Intel Corporation.
The goal of ALIGN (Analog Layout, Intelligently Generated from Netlists) is to automatically translate an unannotated (or partially annotated) SPICE netlist of an analog circuit to a GDSII layout. The repository also releases a set of analog circuit designs.
The ALIGN flow includes the following steps:
- Circuit annotation creates a multilevel hierarchical representation of the input netlist. This representation is used to implement the circuit layout in using a hierarchical manner.
- Design rule abstraction creates a compact JSON-format representation of the design rules in a PDK. This repository provides a mock PDK based on FinFET technology (where the parameters are based on published data). These design rules are used to guide the layout and ensure DRC-correctness.
- Primitive cell generation works with primitives, i.e., blocks at the lowest level of the design hierarchy, and generates their layouts. Primitives typically contain a small number of transistor structures (each of which may be implemented using multiple fins and/or fingers). A parameterized instance of a primitive is automatically translated to a GDSII layout in this step.
- Placement and routing performs block assembly of the hierarchical blocks in the netlist and routes connections between these blocks, while obeying a set of analog layout constraints. At the end of this step, the translation of the input SPICE netlist to a GDSII layout is complete.
Documentation
Inputs
-
Circuit design inputs
A SPICE file and constraint files (optional) need to be placed in a common folder. The name of the folder, SPICE file, and top-design name should match. Some examples are provided to showcase the applications of constraints to control the layout of the design.
-
A netlist of the analog circuit in SPICE format
-
Constraint file (optional) as <hierarchy name>.const.json
-
-
Library:(SPICE format)
A basic set of libraries is predefined within ALIGN to create a hierarchical layout. Designers can modify this based on their design style.
- A basic built-in template library is provided, which is used to identify hierarchies in the design.
- More library elements can be added to the user_template library.
-
PDK: Abstracted design rules
PDK setup needs to be configured for any new technology node. We provide multiple open-source PDK options.
- A mock FinFET 14nm PDK rules file is provided, which is used by the primitive cell generator and the place and route engine.
- A new PDK can be represented using a JSON-format design rule abstraction, similar to the mock-PDK design rules file provided.
- Device definition: A basic library of device models and any PDK-specific derived device model.
- Device parameter definition: A method to translate different SPICE parameters to device size.
- Primitive cells (NMOS/ PMOS/ Resistor / Capacitor /guard ring) must be configured for any new PDK.
Outputs
- Design JSON: Final layout in JSON form which can be viewed using the ALIGN Viewer.
- Layout GDS: Final layout of the design. The output GDS can be imported into any GDSII viewer.
Getting started
Docker image
If you are an user who does not require any changes to the source code, the recommended method to use ALIGN is the docker image hosted in dockerhub as: darpaalign/align-public.
Use the image darpaalign/align-public:latest
for the latest version of ALIGN. You will need to first build a personalized image based to ensure the files generated by containers have the appropriate user/group ID and permissions.
You can do this using using the dockerfile
and build.sh
script in the install/
directory.
Detailed instructions to pull, build, and run the docker image are in docker/README.
Steps 0-3 below are to install ALIGN locally. Step 4 is to run ALIGN either locally or inside a docker container.
Step 0: Check prerequisites
The following dependencies must be met by your system:
- gcc >= 6.1.0 (For C++14 support)
- python >= 3.7 (For PEP 560 support) You may optionally install Boost & lp_solve using your distro package manager (apt, yum, etc) to save some compilation time.
Note: In case you have multiple gcc versions installed on your system, we recommend explicitly setting the compiler paths as follows:
export CC=/path/to/your/gcc
export CXX=/path/to/your/g++
Step 1: Clone the ALIGN source code to your local environment
git clone https://github.com/ALIGN-analoglayout/ALIGN-public
cd ALIGN-public
Step 2: Create a Python virtualenv and install dependencies
Note: You may choose to skip this step if you are doing a system-wide install for multiple users. Please DO NOT skip this step if you are installing for personal use and/or you are a developer.
python -m venv general
source general/bin/activate
python -m pip install pip --upgrade
source install_deps.sh
Step 3a: Install ALIGN as a USER
If you already have a working installation of Python 3.8 or above, the easiest way to install ALIGN is:
pip install -v .
Step 3b: Install ALIGN as a DEVELOPER
If you are a developer, you may wish to install ALIGN with some additional flags.
For Python/C++ developers:
python setup.py develop -v
This installs align as an editable package in your current directory. This allows you to modify python files and test them out immediately. You will still need to re-run this command to build your C++ collateral (when you are changing branches for example). More on that is below.
If you would like to build the C++ version with the debug symbols in the object files (to see the symbols in a debugger):
env BUILD_TYPE='Debug' python setup.py develop -v
The second command doesn't just install ALIGN in-place, it also caches generated object files etc. under an _skbuild
subdirectory. Re-running python setup.py install -v
commands will reuse this cache to perform an incremental build. We add the -v
or --verbose
flag to be able to see build flags in the terminal.
If you want the build type to be Release (-O3), you can issue the following three lines:
env BUILD_TYPE='Release' BUILD_TESTING='ON' python setup.py develop -v
or
env BUILD_TYPE='RelWithDebInfo' BUILD_TESTING='ON' python setup.py develop -v
Use the Release
mode if you are mostly developing in Python and don't need the C++ debugging symbols. Use the RelWithDebInfo
if you need both debug symbols and optimized code.
To debug runtime issues, run:
python -m cProfile -o stats $ALIGN_HOME/bin/schematic2layout.py $ALIGN_HOME/examples/sc_dc_dc_converter
Then in a python shell:
import pstats
from pstats import SortKey
p = pstats.Stats('stats')
p.sort_stats(SortKey.TIME).print_stats(20)
To run tests similar to the check-in and merge-to-master CI runs run:
cd $ALIGN_HOME
# Checkin
pytest -vv
CI_LEVEL='checkin' pytest -n 4 -s -vv --runnightly --placer_sa_iterations 100 -- tests/integration/
# Merge to master
CI_LEVEL='merge' pytest -n 8 -s -vv --runnightly --maxerrors=20 --placer_sa_iterations 100 -- tests/integration/ tests/pdks
Step 4: Run ALIGN
You may run the align tool using a simple command line tool named schematic2layout.py
For most common cases, you will simply run:
schematic2layout.py <NETLIST_DIR> -p <PDK_DIR> -c
For instance, to build the layout for telescopic_ota:
mkdir work && cd work
schematic2layout.py ../examples/telescopic_ota -p ../pdks/FinFET14nm_Mock_PDK/
For a full list of options supported by the tool, please use the following command:
schematic2layout.py -h
If you get an error libOsiCbc.so: cannot open shared object file
, please add ${ALIGN_HOME}/_skbuild/<OSname_Arch_PythonVer>/cmake-install/lib
to your LD_LIBRARY_PATH
.
${ALIGN_HOME}
is the path where ALIGN is installed.
For e.g.:
export LD_LIBRARY_PATH=${LD_LIBRAR_PATH}:${ALIGN_HOME}/_skbuild/linux-x86_64-3.8/cmake-install/lib
Design database:
- Examples: Contains example circuits with netlists running on CircleCI
- CircuitsDatabase: Contains benchmark circuits
Viewer :
The final output GDS can be viewed using by importing in virtuoso or any GDS viewer