Doc about multithreading vs multiprocess

[latot]

Hi hi, I'm little confused about this, is like there is mixed doc.

Lets say the basic difference, multithreading shares the memory with the main process, so we can edit variables in the main process, while in multiprocess, every function is executed in a isolated process, so we can't edit the vars in the main session, but we can use the available copy in the child process to work with them.

Actually the plans to use in future talk about childs in r sessions, so is probable a multiprocess, while in

https://rdrr.io/cran/future/man/future.options.html

We can see an option, future.fork.multithreading.enable that talks about multithreading, so..., I'm not sure...

I think would be great have a place in the doc to tell us, how to use mulithreading, and how to use multiprocess.

Thx!.

[HenrikBengtsson]

Thanks for the feedback.

Everything in futureverse uses multi_processing_. That's actually the only way you can parallelize R code. Multi_threading_ is not supported - to use multi_threading_ one has to implement it in C, C++, ..., which is a whole different game and basically outside of R.

The R option future.fork.multithreading.enable is meant to control whether forked parallel process is allowed to use multithreading or not. When options(future.fork.multithreading.enable = FALSE), it uses low-level tools configure settings so that a single thread is used. This is very low-level and was added to avoid certain crashes in certain environments.

I'll see if I can clarify this in the documentation. If you have suggestions on where in the docs to make it clear that the futureverse is dealing with multiprocessing and not multithreading, please let me know.

[latot]

:O Thx!, now is a lot more clear, to me is sad we can't use multithreading, would helps for some process......

Thx!

[latot]

after thinking about this, maybe would be useful the next things:

• replace multithreading word with multiprocessing
• add in the plans, a description to says is multiprocessing
• add in README a very short/simple explanation that in R we can only do multiprocessing, not multithreading

That should clarify and avoid confusion in the future.

Thx!

[HenrikBengtsson]

• replace multithreading word with multiprocessing

I've checked again, I cannot find any place where "thread" is used when it should be "process". Let me know if you know such a mistake.

• add in the plans, a description to says is multiprocessing

I'd say this is already the case, e.g.

• multisession is titled "Create a multisession future whose value will be resolved asynchronously in a parallel R session" and talks about, for instance, "Evaluate futures in parallel on the local machine via as many background processes as available to the current R process"

• cluster is titled "Create a cluster future whose value will be resolved asynchronously in a parallel process" and talks about, for instance, "Evaluate futures via two background R processes on the local machine"

• multicore is titled "Create a multicore future whose value will be resolved asynchronously in a forked parallel process" ...

• add in README a very short/simple explanation that in R we can only do multiprocessing, not multithreading

I'll see if I can add this somewhere.