sagkit 0.0.10

A toolkit for constructing and analyzing schedule-abstraction graph in Python

SAGkit

A toolkit for constructing and analyzing schedule-abstraction graph in Python. You may also be interested in Official SAG repository (in C++), sag-go (in GO), and sag-py (in Python).

Install

You may install SAGkit from source (recommended):

git clone https://github.com/RyderCRD/sagkit

or install with pip:

pip install sagkit

or reproduce the results in our paper directly with a docker image (requires Docker):

docker pull caoruide/sagkit

Use

Change direcotry to ./sagkit (over the /src directory):

cd sagkit

Run all unit tests:

python -m unittest discover tests

Normally, all tests will pass correctly. You can then generate jobsets and build the SAGs. Change direcotry to ./src:

cd src

The jobset generator takes the following arguments:

• --ET_ratio: What percentage of jobs are ET. Default is 15.
• --utilization: What percentage of the macrocycle is the expectation of the total execution time. Default is 45.
• --jobset_folder: Which folder to save the jobsets. Default is "./jobsets/".
• --num_job: How many jobs to include in each job set. Default is 1000.
• --num_instance: How many jobsets to generate for each set of parameter combinations. Default is 1.

Generate jobsets:

python -m sagkit.jobset_generator [ET_ratio] [utilization] [jobset_folder] [num_job] [num_instance]

The SAG constructor takes the following arguments:

• --jobset_folder: Which folder to read the jobsets. Default is "./jobsets/".
• --constructor_type: What constructor(s) to use to do the construction. Default are "original,extended,hybrid".
• --save_dot: Which folder to save the dot files. Default is "./dotfiles/".
• --save_statistics: Which path to save the statistics results. Default is "./statistics.csv".

Construct SAGs:

python -m sagkit.sag_constructor [jobset_folder] [constructor_type] [save_dot] [save_statistics]

Reproduce

• Step 1: Create a folder named 'results' in the current working directory:

mkdir results

• Step 2: Reproduce Fig. 1 and 2. The .dot files corresponding to the figures ('original.dot' for Fig.1, 'hybrid.dot' for Fig.2) will appear under the /results folder after running the following command:

docker run -v "$(pwd)/results:/output" caoruide/sagkit sagkit.sag_constructor --save_dot True --jobset_folder /basic_idea/

• Step 3 (optional): Visualize Fig. 1 and 2. This step is optional because the .dot files are readable and easy to understand. Paste the contents of each .dot file to:

https://dreampuf.github.io/GraphvizOnline (you may want to access in incognito mode.)

• Step 4: Reproduce Fig. 3. Also, you may visualize the .dot files ('original.dot' for Fig. 3 (a), 'extended.dot' for Fig. 3 (b), 'hybrid.dot' for Fig. 3 (c)) following Step 3.

docker run -v "$(pwd)/results:/output" caoruide/sagkit sagkit.sag_constructor --save_dot True --jobset_folder /example1/

• Step 5: Reproduce Fig. 4. Also, you may visualize the .dot files ('original.dot' for Fig. 4 (a), 'extended.dot' for Fig. 4 (b), 'hybrid.dot' for Fig. 4 (c)) following Step 3.

docker run -v "$(pwd)/results:/output" caoruide/sagkit sagkit.sag_constructor --save_dot True --jobset_folder /example2/

• Step 6: Generate job sets:

docker run -v "$(pwd)/results:/output" caoruide/sagkit sagkit.jobset_generator --ET_ratio 0,10,15,20,30,40,50,60,70,80,90,100 --utilization 45,50,55,60,65,70,75    

• Step 7: Reproduce Fig. 5 and Tables 2-4. This step may take some time (29 hours on my machine). All numerical results corresponding to Figure 5 and Tables 2-4 will be automatically generated in the /results/statistics.csv file. The construction times may vary from Fig. 5 (a), (b), (c) depending on the computational power of the machine.

docker run -v "$(pwd)/results:/output" caoruide/sagkit sagkit.sag_constructor  

Contribute

Contributions are welcome! Please feel free to drop your issues and PRs :)