On 7/12/23 21:19, Dave McGuire wrote:
After getting the 5V/250A power supply going, I did quite a lot of net.archaeology to figure out ncm and get the C004s initialized.
I'm also able to exercise the whole toolchain from C source to
running a binary on one of the transputers, but only an individual one
in a TPM-4 board. After figuring out how to run ncm, I still need to determine how to access the array that ncm created. Tonight I'll be
making another cable (to a BBK-S4) and will work on it some more tomorrow.
Our activities with this fantastic machine haven't seemed to spark
much interest here, but I'll document them anyway. I've figured out, at
a basic level, some of the Parsytec-supplied tools used to create and
manage arrays within the Supercluster.
Within the Parix distribution, there's a program called mkdesc, which
takes a "shape" and dimensions (say, "8 4" or "32 1" for an 8x4 or 32x1
array) and some other stuff on the command line, and it generates a
description file which looks like this, for a 2x2 array:
$ mkdesc 2 2 parix.dsc ncm.dsc 2 2 OLDGC parsytec-fe 0 2
$ cat ncm.dsc
ROC 0 0 type: 7 ( Processor )
PROC 1 1 type: 7 ( Processor )
PROC 2 2 type: 7 ( Processor )
PROC 3 3 type: 7 ( Processor )
LINK 0 0, 0 0
LINK 0 1, 1 1
LINK 0 3, 2 3
LINK 1 0, 1 0
LINK 1 2, 3 2
LINK 1 3, 1 3
LINK 2 0, 3 0
LINK 2 1, 2 1
LINK 2 2, 2 2
LINK 3 1, 3 1
LINK 3 3, 3 3
$
Now I send that to the machine via saconfig like this:
$ iserver -sl 1 -sb saconfig.btl -s -m ncm.dsc
That creates a 2x2 array with these interconnects:
$ rspy -sl 1
# Part-rt Link0 Link1 Link2 Link3
0 T800-25 HOST ... 1-2 ...
1 T800-25 ... 2-1 0-2 4-3
2 T800-25 ... 1-1 3-2 ...
3 T800-25 4-0 ... 2-2 ...
4 T800-25 3-0 ... ... 1-3
$
The fifth processor (#0) is set up as the root, that's one of the
TPM-4 boards in the smaller card cage in the rack.
The saconfig program can also be used to generate a textual report of
the processor allocations and mappings:
$ iserver -sl 1 -sb saconfig.btl -f
Your user channel into SuperCluster: 1 (NCU channels 66, 67)
Your user channel into MultiCluster: 1 (NCU channel 66)
Free processors in System : total: 28
Cluster : 0 1 2
Free Processors : 0 12 16
Free special processor boards: 1100 1200 1300 1400
User allocation of processors:
User Id. : 1 2 3 4 5 6 7 8
Used Processors : 0 4 0 0 0 0 0 0
P r o c e s s o r M a p p i n g
Your processors are mapped to the following System processors
0 -> 0 1 -> 1 2 -> 2 3 -> 3
Your processor number -> processor number in system
N C U c o n n e c t i o n t a b l e s :
N C U 0
0 <> 66
N C U 1
2 <> 64 5 <> 1 11 <> 3 12 <> 8 14 <> 6
$
I'm enjoying figuring all of this out, but I sure wish more
documentation had been preserved.
-Dave
--
Dave McGuire, President/Curator
Large Scale Systems Museum
New Kensington, PA
--- SoupGate-Win32 v1.05
* Origin: fsxNet Usenet Gateway (21:1/5)