Re: [tlug] Fortran --> Python (was linux engineer)

[Edward Middleton <edward.middleton@example.com>]

On 06/08/2012 04:54 PM, Attila Kinali wrote:

> I meant people doing "computational science" like writing chemistry
> simulations or similar stuff. Those PhDs are 99% writing code and
> maybe 1% things of their actual field. Sure, you have to know how
> your field works to be able to write code that makes sense, but most
> of what you do is really just putting that 1% into c/c++/java/python/whatever.
> So i would expect those people to know the programming language they use
> together with the simulation system well... but i have not yet met anyone
> who does something like that who actually knows a bit about computers...

I was doing solid state physics which is probably pretty close to what 
you are talking about and I saw the other side.  i.e. highly optimized 
implementations of code that was easily outperformed by generic 
libraries because the optimized implementations couldn't take advantage 
of hardware advances like multiple processors.

To change the code would have been very difficult, because it was highly 
optimized for the specific problem and so changing it would require a 
deep understanding of the physics, the algorithms and the optimizations 
used in the implementation.

In this case it would have been much better to write in matlab and 
abusing the awesome computing power now available.

Edward