[Beowulf] Stroustrup regarding multicore

[Tim Cutts]

On 28 Aug 2008, at 3:34 pm, saville at comcast.net wrote:

> Ladies and Gentleman,
>
> I get a ton of emails I have to delete.
>
> I respectfully request we keep the emails limited to beowulf-related  
> topic. Please take your conversations about scientists-programmers  
> offline.

While I agree that it's got a bit out of hand, I would assert that  
parts of this discussion *have* been relevant to people building and  
running Beowulf systems, because these systems are close to useless  
unless programmed properly, no matter how smart we've been designing  
and building the clusters we provide.

Many of us do have political issues to do with recruitment and skills  
balances.  It's actually quite useful to discuss these issues, as long  
as we don't get too bogged down in flame wars.  These issues are very  
relevant to successful beowulf computing; as others have said getting  
this wrong can cause performance problems of two or more orders of  
magnitude, which is much worse than the difference between (say) using  
infiniband and GigE for the interconnect.  So one could argue that  
getting the programming culture right is actually *more* important  
than some of the technical details we like to get so bogged down in.

Tim


-- 
 The Wellcome Trust Sanger Institute is operated by Genome Research 
 Limited, a charity registered in England with number 1021457 and a 
 company registered in England with number 2742969, whose registered 
 office is 215 Euston Road, London, NW1 2BE.