Does this mean, with the alternative model, we dont need to do broadcast ?
This is an interesting paper, http://repository.cmu.edu/cgi/viewcontent.cgi?article=3573, written by fellow alum that talks about testing GraphChi with large graphs on a just personal computer.
@bysreg It seems like you do not, because now you can just read from memory instead of having to broadcast. This is assuming all processing elements have shared access to same memory.
Does this mean, with the alternative model, we dont need to do broadcast ?
This is an interesting paper, http://repository.cmu.edu/cgi/viewcontent.cgi?article=3573, written by fellow alum that talks about testing GraphChi with large graphs on a just personal computer.
@bysreg It seems like you do not, because now you can just read from memory instead of having to broadcast. This is assuming all processing elements have shared access to same memory.