Welcome to the home repository of Graspan-G.
Graspan-G is a scalable GPU-based out-of-core graph system for interprocedural static analysis.
This Readme (under revision) provides a how-to-use guide for Graspan-GPU.
A full description of Graspan and its GPU backend can be found on our TOCS'21 paper.
Ensure that you have a recent version(10.2 in our project) of the CUDA Toolkit and the Boost library installed in your system.
First, download the entire Graspan-G source code into your machine. Next, edit the src/makefile to set the paths to the CUDA include files and lib files in your machine. Finally, run the makefile in the src folder using make. Graspan-G should now be compiled and ready to run.
Graspan-G needs two input files:
- A program graph on which Graspan-G can perform computations;
- A grammar file which describes how the computations are to be performed.
A sample input is inside the src/inputFiles folder.
After getting the graph and grammar file, run Graspan-G by entering the following command and monitor the progress of the computation on the screen:
../bin/comp <graph_file> <grammar_file> <filter_flag>
Note that the filter flag is a switch to optimize the performance of dataflow analysis rules only. 0 means off optimization, 1 means on optimization. In other analysis, the flag should be set to 0.
- Zhiqiang Zuo - Assistant Professor, Nanjing University
- Yiyu Zhang - PhD Student, Nanjing University
- Shenming Lu - Master Student, Nanjing University
- Harry Xu - Associate Professor, UCLA