[openbeos] Re: Multiprocessing,Mutithreading - Linux vs Openbeos

  • From: Michael Phipps <mphipps1@xxxxxxxxxxxxxxxx>
  • To: openbeos@xxxxxxxxxxxxx
  • Date: Wed, 12 Nov 2003 22:19:49 -0500

This walks the quasi-off topic line, but I think that the orignial poster does 
deserve a response. :-)

In general, scientific applications are batch applications. They start up, are 
CPU bound, run for hours and are often times easily divided up into multiple 
processes (SETI at home exemplifies all of this). 

This is, traditionally, what server OS's are good at. They emphasize throughput 
(more CPU time spent on processing) over reponsiveness (less human time spent 
waiting for a prompt). This is a tradeoff - there are better and worse 
implementations, but, overall, it is a tradeoff. 

Batch processes, generally, belong on servers.

Additionally, supporting scientific computing is non-trivial. Much of it is 
still written in Fortran. Yes, Fortran. Many Fortran compilers are still more 
efficient at generating code (esp math code) than C/C++ compilers. 
Additionally, a lot of older scientists learned it years back. Scientific 
computing needs a great deal of other support, as well. Numeric stuff, etc. 
While BeOS/OBOS *COULD* be modified to support it, it would be a major 

Sorry if that disappoints you. :-/

On 2003-11-10 at 23:24:18 [-0500], Eric MARTIN wrote:
> I have a question about openbeos.
> Is it possible that  openbeos would be better than linux at clustering
> because of  pervasive multhreading and strong SMP(symetrical
> multiprocessing)  ?
> Could it be a contender for supercomputing clusters in the future ?
> I am thinking this OS would be great for scientific supercomputers because
> of this advantage.
> Look at the Viginia Tech supercomputer.  it uses mac os x.
> http://www.chaosmint.com/mac/vt-supercomputer/
> _________________________________________________________________
> MSN Shopping upgraded for the holidays!  Snappier product search...
> http://shopping.msn.com

Other related posts: