Google Groups no longer supports new Usenet posts or subscriptions. Historical content remains viewable.
Dismiss

Hacking CSA Mandelzoom to work on a B008

13 views
Skip to first unread message

Tom Stepleton

unread,
Sep 28, 2022, 7:51:56 PM9/28/22
to
There, that's more like it:
https://photos.app.goo.gl/d4HmNMHEwW8Jry6x8

As noted in the previous thread and elsewhere, the CSA Mandelbrot program won't work on a B008 as it's natively configured. The program freezes when it attempts to execute some code on the T212 installed on the board.

I've been studying the worm program that distributes the software to the Transputers, and there's a pretty easy (if kludgy) alteration to the source code that sidesteps this problem. Just above this line: https://github.com/axelmuhr/T-Mandel/blob/master/FLBOOT.TAL#L179

add these instructions:

ldl LOOPA
adc -1
cj @R3

What it's doing is pretty unsophisticated. Note in this diagram: http://transputer.net/mg/b008ug/figure/096/fig01.png
that the T212 is reached from TRAM0 (natively the host TRAM) via Link 1. Since the software distributing worm starts from TRAM0, the T212 is discovered right away and joins the Transputer network assembled by the worm.

The Transputer network is never represented completely in one place --- instead, each node records its upstream node ("BOOTIN", the neighbour that "infected" the node with the worm) and up to three downstream nodes ("BOOTOUT[0..2]"). It identifies these nodes by the memory location of their channel control word (if that's confusing to any reader, you can replace it with "link ID" to get the idea).

What my hack does is prevent Link 1 from being the downstream connection to ANY node. It's quite crude, and it prevents the network from getting the most out of some topologies, but there is no ordinary topology you can set up on a B008 in its default hardware configuration that will fail to operate --- provided you stick to 32-bit Transputers.

I'll share my notes on the worm program before long.

--Tom

Tom Stepleton

unread,
Sep 28, 2022, 7:54:15 PM9/28/22
to
On Thursday, September 29, 2022 at 12:51:56 AM UTC+1, Tom Stepleton wrote:
> What my hack does is prevent Link 1 from being the downstream connection to ANY node.

I forgot to clarify that it does this by refusing to record the channel control word for Link 1 (outbound) in BOOTOUT, regardless of whether a Transputer is reachable over that link.

--Tom
0 new messages