[racket] (Off) Re: Purely Functional needs "One Memory/Massively Parallel"

From: Erich Rast (erich at snafu.de)
Date: Wed Feb 5 14:43:04 EST 2014

I'm not sure I understand all of your points. Anyway I believe that
purely functional programming is generally beneficial to massively
parallel processing because "pure" functions do not alter state outside
the function until the computation has ended. Purely functional
programs require less synchronization. I/O is a problem for everyone
writing parallel programs, be they purely functional or not. Exception
handling is another huge issue. (The author of ParaSail tries to deal
with it by basically eliminating the need for exceptions - not easy if
you can divide by zero...)

You can also write massively parallel programs very easily in
traditional imperative languages like Fortran if the problem lends to
straightforward parallelization like parallel matrix or array
processing, parallel data flow.

The real problem is how to automatically transfer programs that are not
devised for parallel processing and perhaps not even purely functional
into massively parallelized programs by analyzing and optimizing the
 control and data flow as a whole. Of course, functional programs
 without mutation make that easier.

 On Wed, 5 Feb 2014 10:48:50 -0600
Lawrence Bottorff <borgauf at gmail.com> wrote:

 Isn't purely functional really just waiting for
> One Memory/Massive Parallel wherein all its supposed foibles are
> moot? The whole "sort-of" functional world simply takes all the
> discretism for granted. But Isn't the purely functional paradigm
> driving us toward a day when some sort of OM/MP (virtual or real) is
> the rule?

Posted on the users mailing list.