Message-ID: <11548cd5db8d1ed7a99c78d1eb84e4fd@unsw.edu.au>
Date: 2005-02-09T02:43:39Z
From: Bill Northcott
Subject: How to Speed up R on the G5
In-Reply-To: <Pine.LNX.4.62.0502081908130.15934@illuminati.stderr.org>
It would be good to know how it goes.
Bill
On 09/02/2005, at 12:09 PM, elijah wright wrote:
>>> Regarding the question of whether veclib is parallelized, I'm
>>> reasonably
>>> sure that it is. ...........
>>
>> Just to observe there is a thread about exactly these issues on
>> r-help under the title 'R on Beowulf cluster?' The discussion is
>> equally applicable to multi CPU machines such as high end Macs.
>
>
> Definitely so. [that's actually what i have, here, and have gotten
> all but a couple of those packages running under OSX, now... I'm
> hopeful that i'll get the rest worked out in the next week or so.]
>
> --elijah
>
>
>> The following expands on the packages I mentioned:
>>> look into the following knot of CRAN packages:
>>> SNOW
>>> Rmpi
>>> Rpvm
>>> rlecuyer
>>> rsprng
>>> SNOW-FT
>>> RScaLAPACK
>>> et cetera.
>>> these provide some pretty interesting bits - like a suite of
>>> parallelized
>>> apply() functions.
>>> --elijah
>>
>> Bill Northcott
>