On 03/23/2015 05:31 PM, Carsten Lechte wrote:
On 23/03/15 19:44, Dave Angel wrote:
I'll give you a worse version. Back in the day I had occasion to
write a simple program
in a language which had no add or subtract. It could only increment
and decrement
indices.
Oh yes, the olden days. I have not quite lived through the punch card
era, but I do want to bring to your attention a recent blog post, not
about Fibonacci, but about the Mandelbrot set. On an IBM 1401. Read it
and count your blessings.
<http://www.righto.com/2015/03/12-minute-mandelbrot-fractals-on-50.html>
Imagine the 123 pass python interpreter for this device...
Thanks for the trip down memory lane. My fondest memories from my first
decade of development was microcoding an all decimal floating point
package, on hardware that had a bcd alu (8 bits wide).
To optimize multiply, I precalculated 2, 10, and 20 times the original
values, and for a given column either added or subtracted depending on
whether the multiplier digit was less than or greater than 5.
(the 10 and 20 were just a half-byte shift, but it was still quicker to
do it once on the multiplicand than repeatedly on the result. If I were
doing it now, I'd investigate doing the multiplier in two passes, one
for the odd digits and one for the even ones.)
My first real assignment at that company (1973) I did use core memory,
but it was outboard, in what we called a "ramdisk." 8K of core memory
in a suitcase-sized box, cabled to our calculator, so I could do
overlays from it into main, semiconductor-based ram. The project was a
satellite navigation system (to produce latitude and longitude for
shipboard use) that had to run in 2k of (ROM) code space plus 1.5k of
data space. While developing, code and data had to share that 1.5k, so
I wrote a little overlay manager. Burning those EPROMS took about 20
minutes, so the ramdisk was really quite useful.
--
DaveA
--
https://mail.python.org/mailman/listinfo/python-list
