Re: [linux] Re: News from the EFF
- On Thu, Apr 04, 2013 at 01:54:49AM -0000, thad_floryan wrote:
> [...]On the subject of numerical computing and languages. I learnt something,
> which confirms the overstriking, really a stupid design. MATLAB is
> a far, FAR better language for numerical computing:
don't use PERL/Python/ruby/php for applications that are going to do
many computations. They're just plain slow. Use Java/c#/c/c++, something
that's not dynamically typed. On the other hand, if it's not going to be
doing much in the way of numbers and is going to do text comparisons
then it doesn't really matter.
- --- In firstname.lastname@example.org, ed <ed@...> wrote:
> On Thu, Apr 04, 2013 at 01:54:49AM -0000, thad_floryan wrote:
> > [...]
> > which confirms the overstriking, really a stupid design. MATLAB
> > is a far, FAR better language for numerical computing:
> On the subject of numerical computing and languages. I learnt
> something, don't use PERL/Python/ruby/php for applications that are
> going to do many computations. They're just plain slow.
Correct -- they're executing interpretively which, by definition, is
slow. I'd add any variant of "basic" to that list, too, except for the
few basic that actually compile machine code (e.g., one that I wrote
and another that I have on my 3B1 systems.
> Use Java/c#/c/c++, something that's not dynamically typed.Correct because "generally" it's compiled code running at hardware
speed. But see below for an exception, one of my languages.
> On the other hand, if it's not going to be doing much in the way ofIt depends. If something is one-off then it really doesn't matter,
> numbers and is going to do text comparisons then it doesn't really
but if it's going into "production" usage then the implementation
language choice is very important.
I've designed and written many commercial compilers over the years
which have been used by the US government and military and Fortune
1000 companies. Most of those were integrated into DBMS systems that
I also authored and sold to the same clients, and they generated hard
machine code for even simple queries (e.g., "if A > B") and extremely
complex (and portable) programs 1000s of lines long.
I came up with a new idea generating Machine Independent Code (MIC)
and all tests I've run so far show it to be in many cases faster than
hand-coded assembly language. The MIC is also immediately portable
to numerous hardware platforms (big-/little-endian, fixed or variable
word sizes, etc.) and I'm still working on this concept just to keep
A little while back I posted several previews of how the above works;
see the following for examples:
Sat Aug 15, 2009 6:50 pm
"Dell Linux netbooks with ARM vs. SheevaPlug with ARM"
Sat Jun 25, 2011 6:29 pm
"OT: Preview of something I've been working on"
The 25-June-2011 article shows more detail. I've already compiled
programs of 10,000+ lines in this language and compared them to C
equivalents and the runtimes are very close. I've even written a
hangman game in this language and it works fine. :-)
Note: a fixed-width font is necessary to view the above 2 articles;
I'll post a "howto" later today (didn't have time last night).
I'm still working on the system but I haven't given it the priority
it deserves. I thought I'd have more time now that I'm "retired" but
I'm discovering that's not the case since I have so many interests
and there's only 24 hours in each day and even I do need to get some
sleep occasionally. :-)