A framework for modeling dataflow applications and their execution on MPSoC platforms.
It is recommended to install mocasin within a virtual environment. Currently we support Python versions 3.7, 3.8, 3.9, and 3.10. You can create a new environment with
virtualenv -p python3 ~/virtualenvs/mocasin
Note that you can adjust the path to your needs. You can activate the previously created environment as follows:
source ~/virtualenvs/mocasin/bin/activate
It is recomended to use the latest version of pip. To ensure your version is up to date, simply run the following command:
python -m pip install --upgrade pip
Then you can install mocasin and all its dependencies from inside the mocasin root directory:
pip install .
If you plan to modify the mocasin code base, you should install the package in development mode:
pip install -e ."[dev]"
This ensures that any modifications you make, will have an immediate effect to your installed mocasin package.
You can now execute the mocasin command:
mocasin help
Depending on your OS and your version of pip, you might need to install further
dependencies in order to install Mocasin (or more accurately, to install some
of the python packages Mocasin depends on). On Ubuntu, you might need to
install libboost-all-dev and lua5.3-dev.
mocasin uses hydra for its runtime configuration. The mocasin
package provides a basic set of general configuration files.
You can create your own configuration files and extend or override the default
ones. mocasin automatically loads all configuration files located in the local conf/ directory. If that is not sufficient for your needs you can set the environment variable export MOCASIN_CONF_PATH=<path1>:<path2>:<...> to contain more paths
mocasin can execute one of several tasks. It expects the first argument to be a
valid task and all the remaining arguments are passed to hydra. The help
task we used above gives an overview about all available tasks.
graph_to_dot is a simple task that takes a dataflow graph and produces a dot file
that represents this graph. There are a view SDF and task graphs available
within the examples directory. See examples/sdf3/ and examples/tgff.
For instance you can visualize an SDF application like this:
mocasin graph_to_dot graph=sdf3_reader sdf3.file=<abs_path>/examples/sdf3/medium_acyclic.xml
or a task graph application like this:
mocasin graph_to_dot graph=tgff_reader tgff.directory=<abs_path>/examples/tgff/e3s-0.9/ tgff.file=auto-indust-cords.tgff
Note that it is currently required for path to be absolute. We are working on a fix that enables use of relative paths.
The above commands will generate an *.dot output file. This file can be
viewed for instance with the xdot command line tool. Note that hydra
outomatically changes the CWD and places the generated output in a
subdirctory depending on the current time: outputs/<date>/<time>/.
Hydra allows to print the whole configuration tree using the -c
argument. This gives you an overview of the parameters available for the task.
mocasin graph_to_dot graph=sdf3_reader -c job
One of mocasin's most important tasks is simulate. It simulates the exectution
of an application on a given platform. For this, we need also a mapping
that maps dataflow nodes to processors in the platform and traces that describe
the runtime behavior of the application. For instance:
mocasin simulate graph=sdf3_reader trace=sdf3_reader platform=odroid mapper=random sdf3.file=<abs_path>/examples/sdf3/medium_acyclic.xml
simulates the execution of the specified SDF3 application on the Odroid platform using a random mapping.
The simulation task also produces a simulation trace that can be viewed in
Google Chrome's built-in trace viewer. On default, the generated trace is
called trace.json and placed in the output directory
outputs/<date>/<time>/. This can be overwritten using the hydra parameter
simtrace.file. To view the trace, open about:https://tracing in Google Chrome
and load the previously generated trace file
mocasin comes with a set of tests. You can run them simply by running pytest. By default this will run all the tests in the repository. Alternatively, you can also give pytest a path, limiting the tests it searches for.
See Style Guide
mocasin supports bash autocompletion. To activate bash completion, run:
eval "$(cat misc/bash-autocomplete.sh)"
For completion in zsh, you first need to add
autoload -Uz bashcompinit && bashcompinit
to your .zshrc configuration file.
- Christian Menard, Andrés Goens, Gerald Hempel, Robert Khasanov, Julian Robledo, Felix Teweleitt, and Jeronimo Castrillon. 2021. Mocasin—Rapid Prototyping of Rapid Prototyping Tools: A Framework for Exploring New Approaches in Mapping Software to Heterogeneous Multi-cores. In DroneSE and RAPIDO 2021, System Engineering for constrained embedded systems, January 18-20, 2021, Virtual event. ACM, New York, NY, USA, 8 pages. (PDF, Video Presentation, doi)