On Thu, 2011-03-03 at 13:48 +0000, Kevin Wright wrote:
[ . . . ]
> No. He's saying that C and Fortran have characteristics that make them
> ideal for that particular domain. This doesn't in any way imply that a
> dynamic language isn't more suitable in a different domain.

Well I wouldn't call C and Fortran ideal for the task they are being
used for.  Nor is C++.  Or D, Go, etc.

Interestingly the scientific community are moving to a situation where
they use Python to write all the GUI and coordination code managing the
C, C++ and Fortran code that does all the grunt computation.  cf. SciPy,
http://www.scipy.org/

> He is saying, however, that there are domains where java isn't the
> best choice.

Indeed.  There are some people that claim that JVMs can run code Java
faster than C++ or Fortran code runs for the same problem -- the
rationale being that a JIT optimizes the code that is actually executing
whereas an AoT optimizer has to cover all bases.  However until there is
real objective data, it is all just conjecture.

Of course, most of the "Big Science" codes were written in the 1960s and
1970s and no-one has the resources to rewrite the codes in sensible
languages.  Well until Chapel, X10, Fortress, C++0x with Just::Thread
Pro, etc. are ready for production use there is no point, you just have
very bright Fortran compiler folk making 50 year old sequential codes
run in parallel -- with the help of OpenMP and MPI.  The observation is
that Java is not in the frame for these rewrites which is why Python has
become the dynamic language of choice: it's all native code, so no place
for a JVM.

However, I digress . . .  ;-)

-- 
Russel.
=============================================================================
Dr Russel Winder      t: +44 20 7585 2200   voip: sip:[email protected]
41 Buckmaster Road    m: +44 7770 465 077   xmpp: [email protected]
London SW11 1EN, UK   w: www.russel.org.uk  skype: russel_winder

Attachment: signature.asc
Description: This is a digitally signed message part

Reply via email to