22 Sep
2013
22 Sep
'13
11:19 a.m.
On Sun, 22 Sep 2013, Chuck Guzis wrote:
Ah, we've been spoiled by fast hardware doing
pretty much nothing. For
text-mode data entry and POS and such, you don't need much CPU power (or
memory, for that matter). We used a 3.5MHz 8085 and 64K to service 5
users--and that included word processing and the usual suite of business
apps (AR/AP/GL). We eventually added 128K of mapped memory, but most
users got by just fine with 64K.
"With Windoze, you need a multi-GHz multi-core processor to key in a memo"
Keep in mind that you have software-assisted inefficiency.
run-time pseudo-code was very compact and, in fact,
outran Microsoft
compiled BASIC in just about every test.
As benchmarks go, that is certainly "damning with faint praise"!
The most common use in the early days for BASCOM was just to hide, or at
least obfuscate, the source code.
overheard at a Comdex party (~1986):
Microsoft programmer: What language is your program written in?
Vendor: That information is proprietary.
Microsoft programmer: Oh. WHICH BASIC compiler?
BASCOM did not seem to provide much speed improvement over interpreted
BASIC.
When I was teaching FORTRAN, using the IBM/Microsoft compiler (first
version), it seemed rather slow. So, I keyed in a simple "Sieve Of
Erastothanes" benchmark. It was, indeed, slower than their interpreted
BASIC! Even when I let the BASIC version default to floating point!
(prime number algorithms was a homework assignment)
There's a huge difference between general-purpose
OS software such as
CP/M and MP/M and a written-for-purpose OS. Incidentally, we eventually
went to a 80186/80286 combination CPU running Xenix--and found to our
dismay that it didn't run much faster in multi-user mode than the old
8085 system.
software [in]efficiency always trumps hardware.
Remember when the people in the business discussed
made-for-purpose
microcomputer systems. That is, you'd have an accounting "appliance"
like your toaster?
In many cases, many of the people proposing such systems didn't have your
level of expertise, and were just pushing vertical bundling, such as
a PenTabs Computer (Vector Graphic, preloaded with accounting software)
That seems to have vanished. We now have telephones
that aim to do
everything.
and how well?
A classic rotary dial landline has much better sound quality. (not that
the dialing mechanism is relevant, other than to knock "progress")
Does generalization always win out over
specialization? Should my
subcompact car also serve as a 10-ton dump truck?
That's why I have a station wagon (Prius V). NOT really an improvement
over when I had a Willys FC170, a Honda S600 sports car, a motorcycle, and
a handful of Honda N600s for in town use (ability to nose into a half a
parking space and heft the rear end in was important - the "SmartCar" ad
driving over the curb doesn't impress me).
For my office, I used a 5160 with Lotus for order processing and
accounting, a 5160 with PC-Write for phone log (PC-Tools was nice, until I
found that its replacement of the critical error handler with "ALWAYS
IGNORE" had cost me sectors within my phone log!), generic 386s for
compiling and assembling (made a noticeable difference with "The Sales Tax
Genie"), a 5160 for Cordata Laser (which refused to run in 286 or above),
and a relatively complete selection of TRS80s, Apples, Amiga, 16K 5150,
and up, for program testing. Numerous KVM switches. Yes, my chairs
had wheels.
"Replace all of those with one FAST machine"??!? NO WAY. Compiling and
assembling was the ONLY place that "performance improvements" were
helpful.
We had a college administrator who was insisting on wanting to replace the
CIS department computer lab (48 80286 PCs with an average of 25 students
using the lab a time) with FOUR Pentium machines. Yeah, THAT would help.
People insist on using "performance numbers" INSTEAD of suitability to
task.
--
Grumpy Ol' Fred cisin at xenosoft.com
http://www.xenosoft.com/FPUIB