Artifact for PPoPP'24 paper

Exploiting Fine-Grained Redundancy in Set-Centric Graph Pattern Mining.

1.1. Download datasets.

• Download datasets. Datasets are available here.

• Download single dataset(cit-Patents.mtx).

cd datasets
bash fetch_cit-Patents.sh

• Batch download datasets.

cd datasets
bash fetch_all_data.sh

1.2. Compile implementation.

mkdir build && cd build && cmake .. && make -j10 && cd ..

2. Run initial test experiment with basic APIs.

2.1. Run baseline G²Miner.

Usage: ./build/test_G2Miner <graph_path> <pattern_name>
Support Patterns(pattern_name): K-Clique-Finding(CF4,CF5,CF6,CF7), Motif-Counting(MC4), Triangle-Counting(TC), Pattern-Enumeration(P1,P2,P3,P4,P5,P6,P7,P8)

./build/test_G2Miner ./datasets/youtube.mtx CF4
./build/test_G2Miner ./datasets/cit-Patents.mtx MC4
./build/test_G2Miner ./datasets/cit-Patents.mtx TC
./build/test_G2Miner ./datasets/cit-Patents.mtx P1

2.2. Run GraphFold's Pattern-Enumeration.

Usage: ./build/test_patternEnum <graph_path> <pattern_name>
Support Patterns(pattern_name): Pattern-Enumeration(P1,P2,P3,P4,P5,P6,P7,P8)

./build/test_patternEnum  ./datasets/cit-Patents.mtx P1

2.3. Run GraphFold's K-Clique-Finding.

Usage: ./build/test_cliquefinding <graph_path>
Support Patterns(pattern_name): K-Clique-Finding(4,5,6,7)

./build/test_cliquefinding  ./datasets/cit-Patents.mtx 5

2.4. Run GraphFold's Motif-Counting.

Usage: Usage: ./build/test_motifcounting <graph_path>

Support Patterns(pattern_name): Motif-Counting(4)

./build/test_motifcounting  ./datasets/cit-Patents.mtx 4

2.5. Run GraphFold's Triangle-Counting.

Usage: ./build/test_trianglecounting <graph_path>

Support Patterns(pattern_name): Triangle-Counting()

./build/test_trianglecounting  ./datasets/cit-Patents.mtx

2.6 Run GraphFold on multi-GPUs.

Usage: ./build/test_$BIN_NAME$_multigpu <graph_path> <pattern_name> <num_gpus>
Support all patterns on multi-GPUs, and support number of GPUs(1,2,3,4,5,6,7,8).

./build/test_patternEnum_multigpu ./datasets/cit-Patents.mtx P5 4

3. Reproduce the major results from paper.

3.1 Compare with G²Miner on V100 GPU. (Table 2.The runtime of GraphFold vs. G²Miner)

bash run_table2_graphfold.sh

Note that the results of GraphFold can be found at "table2_graphfold.csv". Set DATA_PATH(where the graph datasets dir), GRAPH_LIST(which graph runs), GPU_ID(which GPU use), OUTPUT_NAME(where results output), PATTERN_LIST(which pattern runs).

bash run_table2_g2miner.sh

Note that the results of G²Miner can be found at "table2_g2miner.csv". Set DATA_PATH(where the graph datasets dir), GRAPH_LIST(which graph runs), GPU_ID(which GPU use), OUTPUT_NAME(where results output), PATTERN_LIST(which pattern runs).

3.2 Compare with G²Miner on V100 GPU. (Table 3.4.5)

bash run_multi_tables_graphfold.sh

Note that the results of GraphFold can be found at "tablexx(3,4,5)_graphfold.csv". Set DATA_PATH(where the graph datasets dir), OUTPUT_NAME(where results output).

bash run_multi_tables_g2miner.sh

Note that the results of G²Miner can be found at "tablexx(3,4,5)_g2miner.csv". Set DATA_PATH(where the graph datasets dir), OUTPUT_NAME(where results output).

3.3 Compare with G²Miner on 8*V100 GPU. (Figure7.)

bash run_fig7_graphfold.sh

Note that the results of GraphFold can be found at "figure7_graphfold.csv". Set DATA_PATH(where the graph datasets dir), GRAPH_LIST(which graph runs), GPU_LIST(number of GPUs), OUTPUT_NAME(where results output), PATTERN_LIST(which pattern runs).

bash run_fig7_g2miner.sh

Note that the results of G²Miner can be found at "figure7_g2miner.csv". Set DATA_PATH(where the graph datasets dir), GRAPH_LIST(which graph runs), GPU_LIST(number of GPUs), OUTPUT_NAME(where results output), PATTERN_LIST(which pattern runs).

4. Run GraphFold in docker.

4.1 Launch the GraphFold docker

cd docker 
./build.sh

4.2 Launch the GraphFold docker and recompile,

• The compiled exectuable will be located under GPM-artifact/.

cd docker 
./launch.sh
cd GPM-artifact && mkdir build && cd build && cmake .. && make -j10