One question that really needs to be clarified about the gather phase

[Zekai Jacob Gao]

Hey GraphLab guys,

I have a question about the gather phase. Suppose we have vertex X connected to vertices 1~100. Vertex X contains a 100*10000 matrix. Each of the vertices 1~100 only needs ONE ROW in that 100*10000 matrix. So how does GraphLab handle this in a pull-based manner? For example, for vertex 1, what I'm doing now is to fetch the whole matrix from vertex A and only use its first row. Is there a way that I can actually avoid this expense by fetching the first row only from A? And how does GraphLab implement that?

Thanks!

-Jacob

[Rong Chen]

Perhaps, you can use message parameter provided by signal() in scatter phase.
You need send different row to different neighbors, and then you also should implement ivertex_program::init() interface, which can receive the message from neighbors (the row you wanted).

BTW.
I'm not try it, and just relies on the understanding about execution engine.
I hope it can help you.


Thanks,
Rong
Institute of Parallel and Distributed Systems (IPADS)
Shanghai Jiao Tong University
http://ipads.se.sjtu.edu.cn/projects/powerlyra.html