https://github.com/metonymic-smokey/par-a-graph

Parallelising Graph Algorithms
https://github.com/metonymic-smokey/par-a-graph

Last synced: about 2 months ago
JSON representation

Parallelising Graph Algorithms

• Host: GitHub
• URL: https://github.com/metonymic-smokey/par-a-graph
• Owner: metonymic-smokey
• License: apache-2.0
• Created: 2021-03-31T12:58:05.000Z (almost 4 years ago)
• Default Branch: main
• Last Pushed: 2021-06-26T14:43:44.000Z (over 3 years ago)
• Last Synced: 2024-05-01T15:54:47.190Z (9 months ago)
• Language: Go
• Homepage:
• Size: 5.39 MB
• Stars: 9
• Watchers: 2
• Forks: 0
• Open Issues: 0
• Metadata Files:
  • Readme: README.md
  • License: LICENSE

Awesome Lists containing this project

README

        
# `par-a-graph`

Parallel implementations of graph algorithms using Go. Parallelization is performed both on CPU (using goroutines) and GPU (using CUDA).

## Algorithms

### Pagerank

Parallel implementation of Pagerank. This was done as part of the project for the course "Heterogeneous Parallelism" (UE18CS342) at PES University.

#### CPU

Details can be found in the [report slides](docs/report-presentation.pdf) and the [README](./cpu).

Summary:

 - Fixed **partitioning** based parallelization using goroutines

 - Tested and **benchmarked on 5 datasets** of varying sizes and characteristics (number of nodes, edges, average degree, density, etc.)

 - **Up to 4.2x speedup** on the Stanford Web graph (on a system with 4 cores, 8 threads)

#### CUDA

TODO: rough implementation can be found in the `cuda_pr` branch.

## Directory Structure

 - [`cpu`](./cpu): CPU-based parallel implementation of Pagerank

 - [`datasets`](./datasets)

 - [`images`](./images): images for README, etc.

 - [`docs`](./docs): documents related to the project - presentation slides, etc.

## License

[Aapche 2.0 license](LICENSE)