Skip to content

orfs-build.md

Latest commit

 

History

History
209 lines (148 loc) · 5.95 KB

orfs-build.md

File metadata and controls

209 lines (148 loc) · 5.95 KB

Building OpenROAD with ORFS

Most people will not need to build OpenROAD from source. I encourage you to use the ORFS Installation to avoid this unless you need to modify the source code.

You can build OpenROAD from source either in a Docker image or locally if you have a supported machine. If you are using WSL on a Windows machine, you should use the local method. The full instructions are at ORFS, but this is a document of what has worked for us.

BEWARE, if you have local versions of tools installed, they will get priority over the compiled versions in the Docker image. This includes Yosys, OpenSTA, and OpenROAD. If you took CSE 125/225, you may have Yosys installed locally!

Clone the repository

All steps need to get the ORFS repository:

git clone https://github.com/The-OpenROAD-Project/OpenROAD-flow-scripts.git
cd OpepNROAD-flow-scripts

Unless otherwise specified, all commands should be run from the root of this repo.

Docker

Dependencies

The dependencies are installed inside the Docker image, so you do not need to install them. You do nee to install docker, though.

Building the Docker image

You only need to run the build command:

./build_openroad.sh

as Docker is the default build method.

To add debug symbols in the docker method, you must modify the tools/OpenROAD/docker/Dockerfile.builder file to add the CMake arguments to the build command like this:

RUN ./etc/Build.sh -compiler=${compiler} -threads=${numThreads} -deps-prefixes-file=${depsPrefixFile} -cmake="-DCMAKE_BUILD_TYPE=DEBUG"

NOTE the --openroad-args argument for ./build_openroad.sh is not passed to the Docker build scripts, so you cannot enable debug like the local build method.

Using OpenROAD in Docker

This is very similar to the ORFS docker image that you used in the walkthrough except that you need to specify a local docker image and tag that are shown in the final steps of your compilation:

#25 naming to docker.io/openroad/flow-ubuntu22.04-builder:6cd62b

I make a modified version of the runorfs.sh script (called runbuilder.sh) that uses this image and a general tag:

#!/bin/bash
TAG="${1:-latest}"
echo "Running OpenROAD flow with tag: ${TAG}"
docker run --rm -it \
  -u $(id -u ${USER}):$(id -g ${USER}) \
  -v $(pwd)/flow:/OpenROAD-flow-scripts/flow \
  -e DISPLAY=${DISPLAY} \
  -v /tmp/.X11-unix:/tmp/.X11-unix \
  -v ${HOME}/.Xauthority:/.Xauthority \
  --network host \
  --security-opt seccomp=unconfined \
  docker.io/openroad/flow-ubuntu22.04-builder:${TAG}

Specify the tag to run this version:

./runbuilder.sh 6cd62b

Local

Dependencies

You need to run the commands in setup.sh to install dependencies. This does three things. You can run this script if you are root access, but I break down each step since some do not require root access.

  1. (NO ROOT) It recursively clones the repsositories needed to build OpenROAD:
git submodule update --init --recursive

(Note, you could have run --recursive when you clone the repository as well, but this is a good way to ensure you have the latest submodules.)

  1. (NEEDS ROOT) Install system dependencies:
sudo ./etc/DependencyInstaller.sh -base

This is a script that comes with ORFS. It will install the dependencies assuming that you have a supported OS.

  1. (NO ROOT) Install the other common depdencies:
./etc/DependencyInstaller.sh -common -prefix="./dependencies"

This builds things such as specific versions of SWIG, cmake, etc. in the subdirectory "dependencies".

Building the code

tl;dr My command line looks like this:

source dev_env.sh
./build_openroad.sh --no_init --openroad-args "-DCMAKE_BUILD_TYPE=DEBUG" --local

The dev environment ensures that you use the dependencies in "./dependencies". To compile the code locally (not in Docker), you specify --local.

To add debug symbols, you can specify the --openroad-args flag with and argument to cmake.

The --no_init flag is used to not re-initialize the submodules. In addition to this, you can specify to use a particular repository and branch of code with: --or_repo <REPO> --or_branch <BRANCH>. By default this is the ORFS repo and branch. I usually manually clone it and use the no init option.

./build_openroad.sh --local

This will update submodules to the version needed by the current commit.

Using OpenROAD

To set your path with the newly built openroad, you should source the environment script:

source env.sh

You can then see that you are using the correct version of OpenROAD by running:

which openroad

which should point to

OpenROAD-flow-scripts/tools/install/OpenROAD/bin/openroad

OpenROAD Regression Tests

You can run regression tests for OpenROAD overall by doing:

cd tools/OpenROAD/tests
# For a single test
./regression gcd_nangate45
# For all the tests (very slow!)
./regression.sh

or just:

openroad gcd_nangate45.tcl

Module Regression Tests

You can run regression tests for a specific submodule like this:

cd tools/OpenROAD/src/rsz/tests
# Run all the tests, with 10 threads
./regression -j 10
# Run the tests that match the regex
./regression -R repair_setup
# Run a regression test TCL directly
openroad repair_setup1.tcl

The correct log output of a regression test is saved with the extension ".ok" and the correct Verilog or DEF is saved with ".vok" and ".defok", respectively. To determine correctness, the final result is compared with these. A simple diff is usually used unless the test has equivalence checking enabled.

The outputs of the regression test will be saved in the results subdirectory. The log is called <TEST>-tcl.log and the diff with the ".ok" log is in <TEST>-tcl.diff. If there is a Verilog or DEF output, it is saved with the extension <TEST>_out-tcl.v (or def).