This is the code for Triangle Counting submitted at Graph Challenge 2019.

Compile the code

make

Dataset

For generating the binary of datasets, we use the converter from:
    https://github.com/huyang1988/TC/blob/master/README.md
Some example datasets are provided in the data folder of this repository. For using different repository, provide the path to the dataset.
i.e. for p2p08 dataset, "data/p2p08/"

How to use the converter?

1. Go to TC/gConv/ directory. 
    'make gConvu'
2. Go to TC/graph_converter/undirected_csr/  directory. 
    'make'
 Copy gConvu and tuple_to_undirected_csr.bin to a folder. 
3. Download the Adjacency MMIO file for datasets from graphchallenge website [https://snap.stanford.edu/data/] to the same folder.
4. ./converter.sh <MMIO_file>

Multi-GPU

We use jsrun to run multi-gpu version of the code on Summit.
https://www.olcf.ornl.gov/for-users/system-user-guides/summit/summit-user-guide/#running-jobs

Configuration

Change arguments in Makefile.

Input arguments

1. Folder name containing the binary file of dataset.
2. Total number of process
3. Number of Threads per Block (Minimum 32)
4. Number of Blocks
5. Number of Buckets for hashing (Limit: 256)
6. Block-based (1) or Warp-based(0)
7. Degree based workload partition (1) or Index based partition(0)

Output Format

1. Name of the dataset
2. Vertex count
3. Edge count
4. Triangles count
5. Max Time
6. Min Time
7. TEPS rate based upon max time
8. Total number of process

Contact

Please send me a email if you have any questions: santosh.pandey2222@gmail.com