C is fastest

[Lennart Sorensen]

On Wed, Oct 14, 2009 at 03:57:12PM -0400, Christopher Browne wrote:
> On Wed, Oct 14, 2009 at 3:44 PM, Lennart Sorensen
> <lsorense-1wCw9BSqJbv44Nm34jS7GywD8/FfD2ys at public.gmane.org> wrote:
> > I do believe that if we actually want to take advantage of our now common
> > 4 and 8 way machines, we need languages that are functional not OO so
> > that the language and compiler can auto generate multithreaded code
> > to take advantage of modern machines.  Most programmers can't write
> > multithreaded code that works.  Preferably they shouldn't have to either,
> > but I do believe they have to give up OO to get there.  Perhaps OO
> > doesn't have to completely go, but certainly I don't believe java and
> > C++ stand a chance here.  Microsoft may be able to do something with C#
> > or more likely F# (which is a nice looking language, too bad I don't
> > care to write code in Visual Studio on windows).
> 
> With this I agree.
> 
> FYI, F# is a dialect of ML which is mostly like OCAML.  (Presumably
> except for some parts that aren't like OCAML; dunno what...)

Yes it does look a lot like it.

> It should be pointed out that OO has a propensity for being
> incompatible with thread-safety, and that's an increasing problem
> these days.  OO tends to bind state together with data structures, and
> requires having strict interfaces to those objects, which, with the
> greater concurrency we're seeing on many-core CPUs, leads to a lot
> more locking/blocking.

Data structures are a good thing.  Tieing the state with the data is not
such a good thing once you want to make it scalable.

> The language that seems to have gone the furthest in the opposite
> direction is Erlang.

Not one I have used.

> They basically took Prolog, dropped out backtracking, which means
> you're left with deterministic functional (e.g. - "one path") code,
> which has the massive benefit that practically all your code becomes
> re-entrant, and may be invoked many times in parallel.

Prolog I have used, and it sure is weird.  My wife has done much
more with it and some of the stuff she did I still can't understand.
Apparently prolog can do things I had never imagined, but the syntax to
do it is just weird.  Like what direction is this executing in anyhow?

> Unfortunately, they took a Prolog-ish syntax, which tends to confuse people.

It certainly would.

> It's not clear whether ML can be readily parallelized in congruent fashion.

At least it appears to have potential for it.

-- 
Len Sorensen
--
The Toronto Linux Users Group.      Meetings: http://gtalug.org/
TLUG requests: Linux topics, No HTML, wrap text below 80 columns
How to UNSUBSCRIBE: http://gtalug.org/wiki/Mailing_lists