NetworkX is Fast Now- Zero Code Change Acceleration
summary
Claim: zero code change acceleration lets NetworkX dispatch algorithms to backends like GraphBLAS and cuGraph via an environment variable and a plugin system. Benchmarks show connectedness dropping from ~7.5s to <1s, degree centrality from ~1.3s to ~0.74s, page rank from ~54s to under 1s for some backends, and betweenness from ~21 minutes to ~6s. Backends expose extra algorithms, caching avoids repeated conversion, and missing-implementation falls back or errors if unsupported.
AI-written summary · about
chapters (13)
- 0:00Introduction
- 1:09Why NetworkX is fast
- 2:34Graph basics and uses
- 5:48Key graph questions
- 9:16Why NetworkX is ideal
- 12:48Scaling challenges
- 14:14Performance without rewrites
- 16:00Zero code acceleration
- 23:46Live GPU demo
- 30:15Caching and backend limits
- 32:22Dispatching integration
- 33:45NetworkX as graph DB
- 34:47Next steps
related talks
31mFast NetworkX and How Accelerated Backends Are Changing Graph Analytics
32mZero Code Change GPU-Powered Graph Analytics with NetworkX and cuGraph
24mZero Code Change Acceleration: familiar interfaces and high performance
22mSpeed up! Incremental Compilation with Nx
39mAccelerate large-scale machine learning with NP on MXNet
33mCNI-Agnostic Network Performance Accelerator with eBPF
27mSpeed up Python with hardware accelerators
01h 29mWhen Nanoseconds Matter: Ultrafast Trading Systems in C++