Any Chance of GPU Acceleration?

[Yihong Liu]

Greetings! As is known to many practitioners, GPU device can accelerate matrix computations, and there're many application of SNA powered by CUDA on github or project websites such as 

gunrock/gunrock

Adam27X/hybrid_BC

fvella/MGBC

GPU-based Fast Analysis of Networks (GPU-FAN)

However, the question is that many of them(almost ALL) are not quite user-friendly, or in another word, don't have a good python wrapper like networkx.....HAHAHA

Is there any chance for networkx developers to implement GPU acceleration of networkx?

Have a nice day! ;-)

[Allan Leander Rostock Hansen]

It's a good idea, but I think NetworkX is implemented in pure Python, so I don't think that a CUDA or OpenGL backed implementation is likely anytime soon.

Some use of Numba seems more likely, and maybe some Cython... but since NetworkX can use anything hashable as nodes, I guess it's hard to increase speed without sacrificing backwards compatibility?

[James Rubino]

Graph-Tool supports parallel processing.

Creating a networkx compatiple api interface would be nice.

--
You received this message because you are subscribed to the Google Groups "networkx-discuss" group.
To unsubscribe from this group and stop receiving emails from it, send an email to networkx-discuss+unsubscribe@googlegroups.com.
To post to this group, send email to networkx-discuss@googlegroups.com.
Visit this group at https://groups.google.com/group/networkx-discuss.
For more options, visit https://groups.google.com/d/optout.