Hi,

On Wed, 15 Nov 2006, Carl Cerecke wrote:
> That paper I referred to is online:
> 
> http://www.eecs.berkeley.edu/Pubs/TechRpts/2006/EECS-2006-1.html
> 
> The fact that the author is from one of the top three CS departments
> in North America (the other two being Stanford and MIT) lends his
> paper some extra weight, I think.
> 
Where the guy is from is useful to know, but that will not stop me from 
thinking about what he wrote, and evaluating it.

He does note that "For applications with intrinsic nondeterminacy, such as 
servers, concurrent database access, or competition for resources, this is 
appropriate". I think he means is that there are times when threads are 
the natural thing to use.

However, he is clearly saying that when solving a problem such as a matrix 
convolution, or FFT, threads are not appropriate. I do tend to agree here 
- writing such threaded code is harder.

The arguement that it is almost impossible to write bug free thread code, 
so don't write threaded code, is interesting.. If you read the article on 
the development of code for the space shuttle,
http://www.fastcompany.com/online/06/writestuff.html
the last three versions of 420 000 lines of code had one bug in it.
similar commercial code has 5000 errors. Since even the shuttle code has 
at least one bug in it, does that mean we should not write normal code 
(cause it has bugs anyhow)......

===============================================================

Maybe the real need is not thread free code, or new languages, is that we 
desparately need analysis tools to find bugs.

Or do we desparately need software developers who clearly express their
code so that both computers and humans can understand and follow it.

Or do we need good software designs, so that the bug rate is low?

My personal view is that we won't be switching away from C/C++ anytime 
soon, so we are going to have to work towards better design and better 
implementation - and better documentation.


Derek.

-- 
Derek Smithies Ph.D.
IndraNet Technologies Ltd.
Email: [EMAIL PROTECTED]
ph +64 3 365 6485
Web: http://www.indranet-technologies.com/

Reply via email to