README.md

IOBufs

This is the code for I/O-Efficient Butterfly Counting at Scale.

Organization

The code of IOBufs is at DynamicBatch fold.

Environment

CUDA Toolkit 11.6; g++ 7.5.0;

Usage

compile the code $ make -j run the code $ ./butterfly.bin /data/dataset/dataset/livejournal/ GPU 100 edge-centric 12000000000 108 -1 adaptiveRecy 32 blockForSmallWorkload

Input parameter is

1. input graph folder
2. platform: CPU or GPU
3. partition strategy: radix, random or range
4. variant: edge-centric or wedge-centric
5. memory size
6. thread num (CPU)/ block num(GPU)
7. batch num (automatically generated when setting -1)
8. Hashtable recycle option for GPU: adaptiveRecy, scanHashtableRecy or scanWedgeRecy
9. subwarp size for block-level parallelism in GPU
10. benchmark warp and block for small workload when setting blockForSmallWorkload