Skip to content

nvgraph.sh 1.3.0

Install from the command line:
Learn more about npm packages
$ npm install @nodef/nvgraph.sh@1.3.0
Install via package.json:
"@nodef/nvgraph.sh": "1.3.0"

About this version

CLI for nvGraph, which is a GPU-based graph analytics library written by NVIDIA, using CUDA.
🐚 Shell, πŸ“œ Files, πŸ“˜ Wiki.

This is for running nvGraph functions right from the CLI with graphs in MatrixMarket format (.mtx) directly. It just needs a x86_64 linux machine with NVIDIA GPU drivers installed. Execution time, along with the results can be saved in JSON/YAML file. The executable code is written in C++. You can install this with npm install -g nvgraph.sh.

nvGraph, as mentioned above, is a GPU-based graph analytics library written by NVIDIA. It provides four core graph algorithms: Single Source Shortest Path (SSSP), PageRank, Triangle count, and Breadth First Search (BFS) traversal. Data is loaded into the GPU in Compressed Sparse Row (CSR) format, upon which computation is performed. Here, we load the graph stored in MatrixMarket format (a text-based file format for sparse matrices) into a dynamic graph data structure in the host (CPU) memory, which is then converted to CSR format and then transferred to the device (GPU) memory. Computed results are then copied back to the host memory, and written to the output file in suitable format (JSON/YAML). Note that measured time only includes the time required for computation on the GPU.

The SSSP algorithm accepts the source vertex as an argument, and returns the shortest distance to each vertex form the source vertex. The PageRank algorithm accepts the damping factor, tolerance, and max. iterations as arguments, and returns the rank of each vertex. In addition to ranks of vertices in case of the PageRank algorithm, we like to also obtain additional analytics of the rank values, i.e., Lorenz curve, and Gini coefficient. The Triangle count algorithm, unsurprisingly, counts the number of triangles in the graph. The BFS traversal algorithm accepts the source vertex as an argument, traverses the graph in breadthwise manner, and returns the distance and predecessor of each vertex from the source vertex. Note that detailed results are written to the output file only when the full (-f) flag is provided.

Alas, nvGraph is now no more actively developed. NVIDIA started developing cuGraph a collection of graph analytics that process data found in GPU Dataframes as part of RAPIDS.

Stability: Experimental.


## Finds single source shortest path from source vertex
## β†’ returns distances
$ nvgraph sssp -o=out.json -f web-Google.mtx -s=1

## Finds pagerank of all vertices
## β†’ returns ranks
$ nvgraph pagerank -o=out.json -f web-Google.mtx -a=0.85 -t=1e-6

## Counts triangles in undirected, lower triangular graph
## β†’ returns count
$ nvgraph triangle-count -o=out.json -f web-Google.mtx

## Traverses breadth-first from source vertex
## β†’ returns distances, predecessors
$ nvgraph traversal-bfs -o=out.json -f web-Google.mtx -s=1


Index

Command Action
pagerank Finds pagerank of all vertices.
sssp Finds single source shortest path from source vertex.
traversal-bfs Traverses breadth-first from source vertex.
triangle-count Counts triangles in undirected, lower triangular graph.


References



ORG DOI

Details


Assets

  • nvgraph.sh-1.3.0.tgz

Download activity

  • Total downloads 0
  • Last 30 days 0
  • Last week 0
  • Today 0