John R. Sowden wrote:

> 
> In order to use these old computers, we need more software written in
> assembler.  When the x86 based computer is more than 85% of the hardware
> platforms in service today, I think writing for 'cross-platform' is a red
> herring.  An effort could be made to write libraries in assembler, and use
> assembler for the portions of programs that are repetitive, and speed
> sensitive.  This would bring these older PCs back to life.  32-64 MB PCs
> could be used for applications like OOo.  Most of the newer features would
> not be available, but the core features which are used the most would be
> there.  This effort would eliminate the so-called 'digital divide'.  If I
> can
> go to a thrift store and buy a 150Mhz Pentium with 32 MB of RAM  2 GB hard
> drive for $35.00, and buy a cd off the net with linux and a tiny OOo for
> $10.00, there is no excuse for the 'low income' population segment to not
> be
> skilled in the use of the core OOo features.  This would open up job
> opportunities and the use of these computers in elementary/high schools.
> 

I retired from a job three years ago where I was mostly working with
mainframes running Fortran programs. For a few applications, we still
maintained Assembler subroutines as they were up to ten times faster than
similar Fortran routines from which the compiler had produced
fully-optimised pseudo-assembler code. The problem with assembler routines
was that there were too few programmers skilled in assembler. 

In my time, I found that similar speed improvements could be by checking the
Fortran code was written efficiently. On one occasion, I found that
replacing a frequently-used implied do-loop in data input [READ
(ARRAY(I),I=1,J)  where J was a large variable] with a small subroutine,
made the program run ten times faster. This sort of code probably has no
relevance to more modern languages but I suspect that an experienced
programmer in those languages could find a few samples of inefficient code.

I found that a large number of programmers where I worked were only
interested in getting a program working. Worrying about how fast it ran or
how much space the thing gobbled up was a bit beneath them. Even something
as simple as using the full optimisation of the compiler when producing a
final product was not contemplated, although this simple act would produce
a two- or three-fold increase in speed. I wonder if this attitude also
exists in programmers for the PC.

I admit that I my programming experience may not be directly relevant to
languages used for writing programs for the PC but I've seen descriptions
of bug-fixes and coding improvements in PC products that suggested to me
that in some cases programming standards were a decade or two behind what
I'd been used to.

-- 
Graham P Davis
Bracknell, Berks., UK
Send e-mails to "newsman" as mails to "newsboy" are ignored.

---------------------------------------------------------------------
To unsubscribe, e-mail: [EMAIL PROTECTED]
For additional commands, e-mail: [EMAIL PROTECTED]

Reply via email to