A parallelism library for Common Lisp.
Common Lisp



cl-parallel is an API for Common Lisp that's designed to make parallelism easy. cl-parallel is based on bordeaux-threads, so it should work on most threaded Common Lisps. That being said, I've only tested it with SBCL on Fedora Linux and Mac OS X.

##Installation (requires quicklisp)##

Now that this project is in Quicklisp, simply run:

(ql:quickload 'cl-parallel)

##Making cl-parallel available##

Once cl-parallel is installed, you can now access all of its functions, which are in the PARALLEL package.

To use cl-parallel in your packages, try

(defpackage :your-package
  (:use :parallel)

If you don't want to import all symbols exported by cl-parallel, try

(defpackage :your-package
  (:import-from :parallel #:par-map-reduce)

which imports only par-map-reduce into your pakage.


###Low Level 'future' API###

  • future - given an arbitrary number of forms, spawn a thread to calculate the value of the implicit progn made up of the forms.

  • future-p - given a form, test whether it's a future.

  • future-finished-p - given a form, return T iff it's a future that has finished executing.

  • realize - if the argument is a future, block until its thread is joined; otherwise, return the argument.

  • realize-if-finished - given a future, realize it iff it has finished executing. If the argument is not a future, it is returned.

  • #! (read macro) - synonym for realize.

###Actual parallelism API###

All of the below functions takes the optional keywords :max-threads (default 4) and :sleep-time (default 0):

  • par-calls - given an arbitrary number of forms, evaluate them all in parallel.

  • par-map - given a function, and a list, apply that function to every in the list.

  • par-some - given a predicate and list, return T if any of the elements in the list satisfy the predicate; NIL otherwise.

  • par-every - given a predicate and list, return T if all of the elements in the list satisfy the predicate; NIL otherwise.

  • par-find-if - given a predicate and list, return an element in the list that satisfies the predicate; NIL otherwise.

  • par-find - given an item and list, return the item if it is in the list; NIL otherwise.

  • par-map-chunked - same as par-map, but there is another keyword parameter, :chunk-size. This parameter denotes the size of the chunks that will be processed in parallel. This function is often more efficient than par-map.

  • par-map-reduce - given a mapping function, a reducing function, an initial value, and a list, map the first function across the list in parallel and collects the result using the reducing function. Note that this function will call the reducing function in no specific order.