8 Oct
2022
8 Oct
'22
7:08 p.m.
*From:* Jon Elson via cctalk <cctalk(a)classiccmp.org>
*Subject:* [cctalk] Re: Bendix G-15 Restoration
*Date:* Friday, October 7, 2022 at 7:10 PM
*To:* ben via cctalk <cctalk(a)classiccmp.org>
*Cc:* Jon Elson <elson(a)pico-systems.com>
On 10/7/22 18:14, ben via cctalk wrote:
With respect to timing, it depends what you mean by "bit clock." The
drum rotated at 1800 rpm and there were 3596 bits recorded along its
circumference (124 words of 29 bits), so that works out to 107,880
bits/s. That is a bit period of 9.3 µs, which is what the G-15 Theory of
Operation manual calls a "pulse period." A pulse period, however, was
divided by the clock chassis into phases of 1 µs, 6.3 µs, and 2 µs, so
the internal frequency of the logic was higher, if somewhat irregular.
That said, Jon is absolutely right about replacing the drum with some
sort of random-access memory -- you would need to extensively redesign
the logic of the machine to take advantage of the new memory. You would
also have to modify a lot of the software. Timing dependencies aside,
G-15 instructions didn't have addresses -- they had "timing numbers"
that effectively told the hardware how long to wait before reading or
writing a word on the drum.
Paul
On 2022-10-07 1:09 p.m., paul.kimpel--- via
cctalk wrote:
Tghe G-15 was a serial computer with an 90 KHz bit clock. The entire
organization of teh computer revolved around the drum (pun intended).
There was an optimizer that organized instructions around the drum so
that the next instruction came up on the read head just as the last
instruction finished. Without tearing the entire machine apart and
redesigning the logic, core would not make it faster.
The PDP-8S did have core memory, and for a bit serial computer, it was
fairly fast.
Jon
We'd all like to see the ALGO compiler, but
be forewarned -- it's
something like 14 passes on paper tape, with intermediate results
punched on paper tape. I understand it's a bit more convenient to
use if you have magnetic tape drives, but it's still going to be
slow -- there's only so much you can do with 2K words of memory.
Trying to hide the fact the drum makes it slow.
Did any one ever replace the drum with core memory, on the early
serial computers?
Ben.