Related project: GPUGraphs.jl

[gdalle]

Hi @albertomercurio,
Awesome work on this package! You just earned your second star :)
I have been working with @AntoineBut on hardware-agnostic sparse linear algebra, in the context of graph algorithms. He has developed GPUGraphs.jl, where some of the basic kernels you have here are extended for more matrix formats and binary operations (to enable GraphBLAS algorithms).
Perhaps it would make sense to join our efforts in order to avoid duplication? I am also looking for a postdoc to continue Antoine's work on this and related topics, in case you know anyone interested!

[albertomercurio]

Hi @gdalle,

Thank you for the information. I'm absolutely open to joining our efforts!

[AntoineBut]

Opened #11 to bring the SELL format here. I think the most logical thing to do is to define and implement different sparse arrays structs and basic operations in this package, and implement GraphBLAS-specific functions on top of these structs in another package. This makes it much easier to follow the usual arrays interface here.