https://github.com/hardbyte/anonlink-cuda

https://github.com/hardbyte/anonlink-cuda

Last synced: about 1 year ago
JSON representation

• Host: GitHub
• URL: https://github.com/hardbyte/anonlink-cuda
• Owner: hardbyte
• Created: 2019-03-27T07:24:12.000Z (about 7 years ago)
• Default Branch: master
• Last Pushed: 2019-03-29T03:13:55.000Z (about 7 years ago)
• Last Synced: 2025-02-28T15:55:29.182Z (about 1 year ago)
• Language: Python
• Size: 17.6 KB
• Stars: 0
• Watchers: 2
• Forks: 0
• Open Issues: 1
• Metadata Files:
  • Readme: README.md

Awesome Lists containing this project

README

          

An early attempt at running anonlink similarity scores on the GPU using `cupy`.

## Running 

To run you will need cuda v8.0 or later. Install deps with:

    python -m pipenv shell

Run the main test with:

    python cudadice.py

### Running with the profiler

`nvprof -f -o pyprof.nvprof python cudadice.py`

Then open pyprof.nvprof in NVIDIA Visual Profiler 

## Things to consider

* need to validate against anonlink/cpu that it is correct

* merge sorting the returned edges on the CPU while the GPU is busy

* streaming

# Benchmarking

Single CPU using [anonlink](https://github.com/n1analytics/anonlink):

50 M cmp/s

Current speed on a GTX 1080:

1.3 B cmp/s including data transfer