search

JuliaLang / julia

The Julia Programming Language
https://julialang.org/
MIT License
45.62k stars 5.48k forks source link

DOC: Section on distributed arrays is commented out #4059

Closed sglyon closed 11 years ago

sglyon commented 11 years ago

The section is found in lines 306-519 of the current master of julia/doc/manual/parallel-computing.rst

I am assuming this was either an accident (not too likely) or the information in the current version of the section is no longer valid. Either way it would be nice to have that section again!

ViralBShah commented 11 years ago

That was the old darray documentation, IIRC.

sglyon commented 11 years ago

Ok thanks for the info. Any plans to update that section for the new darray soon?

ViralBShah commented 11 years ago

Marking this so that we do it in the upcoming 0.2 release. The feature set is much smaller, but is reasonably usable.

sglyon commented 11 years ago

Ok great thank you.

Is there a reason many of the features were removed? Sorry for all the questions, I am a Python hacker interested in Julia and I am trying to make sure that there are enough parallel tools to make Julia a viable substitution for Python.

ViralBShah commented 11 years ago

The original DArray implementation was superseded by a new implementation that is much more general and flexible recently. We just haven't got around to adding a bunch of the old functionality, but it is really easy to do so, and this is a good forcing function as any! Anything specific you are looking for?

sglyon commented 11 years ago

Glad to hear that the new DArray isn't a step back in functionality and I will be excited to see the new docs.

In addition to simple reductions and function applications (which are implemented very nicely in Julia with @parallel and pmap :+1: ), I have done some distributed array programming. This has mostly been basic linear algebra, which I believe is handled behind the scenes in Julia after distributing an array. The other main distributed operations have used MPI (mpi4py) to populate rows or columns of matrices using fairly complex functions in parallel. An example of how to do that (populate dimensions of matrices using arbitrary functions) would be great!

ViralBShah commented 11 years ago

Actually, a number of those things should work. I will review and update.