The Cray-1 seems to use a Data General Eclipse as its control computer. The Eclipse’s OS is called RDOS – and that’s also the system the assembler is running on. (RDOS has the
PROG/A/B switch syntax described in the assembler manual, and also uses the
.SV file extension for executables.)
The good news is that there is an Eclipse simulator in
simh, and they also have an RDOS software kit.
The boot instructions are in the
simh manual, but I at least had to replace the mentions of
dkp0 (which is how
simh calls the “moving head disk”):
# install simh, then do $ eclipseemu # at least that's what it's called on debian sim> att dkp0 zrdos75.dsk sim> set tti dasher sim> boot dkp0 FILENAME? (cr) DATE (mm/dd/yy)? xx/yy/zz TIME (hh:mm:ss)? hh:mm:ss R list/e
RDOS doesn’t seem to have an on-line help system, which makes using it a bit difficult. But there’s a PDF manual on bitsavers.
So that all works but it doesn’t really bring us that much further. I guessed that the Cray OS disk might actually be (at least partly) an RDOS disk, but it isn’t:
R init dp1 DISK FORMAT ERROR: DP1
Admittedly, that was a bit naive. The COS disk belongs to a Cray X-MP and in the reference manual for that machine (bitsavers is a gold mine) there’s no mention of the Eclipse.
Said manual also contains the following section:
Once the I/O Subsystem is operating, the CRAY X-MP mainframe can be deadstarted from the Peripheral Expander magnatic tape unit or the DD-20 disk unit.
The problem is that it refers to the “I/O Subsystem (IOS) Operator’s Guide” for further information – and I couldn’t find that guide, yet.
The Cray-1 manual, on the other hand, says that
COS is loaded into memory and activated through a system startup procedure performed on the MCU (maintenance control unit). At startup, permanent datasets are reloaded or re-established on mass storage.
Is the Data General Eclipse the MCU? Probably – but probably only for the Cray-1. What’s the MCU-like thing for the Cray X-MP? Is there one?
To sum up, I can’t really say that I understand the whole auxiliary stuff around the mainframe.
So, maybe a better way to move on would be to try to manually extract the COS image that seems to be on the disk and ignore the whole IOS stuff for the moment.