Mailing list for all users of the OCaml language and system.
 help / color / mirror / Atom feed
* [Caml-list] [ANN]: Parmap 0.9.9, and call for contributors
@ 2012-06-06 13:41 Roberto Di Cosmo
  0 siblings, 0 replies; only message in thread
From: Roberto Di Cosmo @ 2012-06-06 13:41 UTC (permalink / raw)
  To: caml-list; +Cc: roberto, marcod

Dear all,
       a few lines to announce the (much improved) version 
0.9.9 of Parmap, the minimalistic library which can be useful
to exploit your multicore processor with minimal modifications
to your OCaml programs.

A full description is in the README file, as well as several
examples, and an article on Parmap has just appeared here:
http://www.sciencedirect.com/science/article/pii/S1877050912003237

Main changes
------------

 - the new combinator pariter iterates over a sequence; since
   no result is returned to the main process, this code is
   faster than calling parmap and ignoring the result
 
 - new functions that provide as argument the index of the
   element of the sequence: parmapi, parmapifold, 
   array_parmapi, array_float_parmapi

 - it is now possible to set the default number of cores used
   when calling the combinators without the optional parameter
   ncores

 - eliminated dependencies on external libraries: this code
   should now build out of the box on any machine

 - added configure support for use with legacy OCaml 3.11
   and tested to build correctly with OCaml 4.00

 - initial support for Oasis packaging

In a nutshell
-------------

If you want to use your many cores to accelerate an operation which 
happens to be a map, fold or map/fold (map-reduce), just use Parmap's
parmap, parfold and parmapfold primitives in place of the standard
List.map and friends; you can specify the number of subprocesses to
use with the optional parameter ncores, and the size of granularity
of the paralle computation with the optional parameter chunksize.

For example, in the classical Mandelbrot present in the example
directory, the line 

  Parmap.parmap ~ncores: !ncores ~chunksize: !chunksize pixel (Parmap.L tasks)

allows to spawn ncores separate processes, each working on sublists of 
size chunksize of the tasks list.

How to get it
-------------

Project home: https://gitorious.org/parmap

To compile and install:

 git clone git://gitorious.org/parmap/parmap.git
 git checkout pipes
 aclocal -I m4
 autoconf
 ./configure
 make
 make install

How to help
-----------

We would use some help on 

 - writing unit tests, both for functionality, and
   speedup estimation

 - adding support for automatically determining
   the number of available physical cores on the
   different unix-like operating systems (can be tricky)

Enjoy

-- Marco Danelutto and Roberto Di Cosmo

P.S.: many thanks to Francois Berenger for extensive testing, to Jerome
Maloberti for GODI packaging, and to Markus Weissman for suggestion the
default for ncores.






^ permalink raw reply	[flat|nested] only message in thread

only message in thread, other threads:[~2012-06-06 13:37 UTC | newest]

Thread overview: (only message) (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2012-06-06 13:41 [Caml-list] [ANN]: Parmap 0.9.9, and call for contributors Roberto Di Cosmo

This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox