Turbine is a Dataflow graph generator. It can generate SDFGs, CSDFGs and PCGs with initilization phases. All the graphs can be saved in the SDF3 XML format (a particular notation is considered for threshold and initilization phases).

Features

• Generate SDFG/CSDFG/PCG up to 10,000 actors (~10m for SDFG, ~2h for CSDFG/PCG).
• Normalize/denormalize dataflow graph.
• Compute initial marking.
• Compute symbolic execution of dataflow graphs.
• Generate .sdf3 xml files (see SDF3 generator)
• Generate .tur files (10 times smaller than sdf3 files and easy to write by hand)

Installation

Requirements

• networkx-1.8.1
• python-glpk-0.4.43
• glpk 4.47 or 4.48 (constraint from python-glpk)

With Fedora 17+, you can install networkx by using yum install networkx -y. Instructions to install python-glpk can be found here http://en.wikibooks.org/wiki/GLPK/Python. glpk is available at http://ftp.gnu.org/gnu/glpk/.

on UBUNTU

apt-get install python-glpk python-networkx 

Special note for python-glpk with Fedora

With fedora there is no pyversions tool (required by python-glpk). You can fix it by adding the folowing line in the swig Makefile :

`pkg-config --cflags --libs python2`

To test if the installation is ok run basic_test.py located on the root of turbine. Please, report any trouble at [email protected] .

Usage

How to run Turbine

There is no installation step, to use Turbine you just have to specify it in the PYTHONPATH environnement variable. As an example, to run the examples/compute_initial_marking.py script, in the root directory of turbine :

$ PYTHONPATH=`pwd`  ./Turbine/examples/compute_initial_marking.py
#################generated graph#################
###### Setup the SDF generation #####
###### Generate dataflow ############
name: SDF_of_test
task count: 10 arc count: 30
tot initial marking: 0
###### Compute the initial marking ##
name: SDF_of_test
task count: 10 arc count: 30
tot initial marking: 6714
....

How to generate a dataflow using Python

import turbine

param= param.parameters.Parameters()

dataflow = generate(param)
write_sdf3_file(dataflow, filename = "fileName.sdf3")

Then a SDF3 XML file filename.sdf3 will be created in current directory.

How to see the generated linear program used to compute the inital marking

param= param.parameters.Parameters()
param.set_lp_filename(test.lp)

generate(param)

A test.pl file will be created in current directory.

Several examples are avalaible in 'turbine/examples'.