On 2015-05-18 17:46, John Forecast wrote:
CASTOR:: was the M+ group’s 11/74 housed in it’s own machine room in ZK. Since
the machine hardware could be relatively easily reconfigured, POLLUX:: was
reserved for when the system was running as 2 independent nodes (3+1 or 2+2
CPUs). The DECnet group had a dual processor system with node name ELROND::.
I used CASTOR:: standalone a couple of times while debugging the MP changes to
the DECnet kernel code until ELROND:: became stable. Eventually, ELROND::
became the timesharing system for the DECnet-RSX group.
The lack of cache-coherency and the limited work-arounds (cache-bypass or cache
flushing) meant that all the DECnet kernel code, both drivers and protocol
stack, ran with cache-bypass enabled on its data PDRs. So the networking code
on an MP system had lower peak performance than on a normal 11/70 although you
would never notice the difference in regular operation.
Thanks for the corrections and clarifications, John. I think I've heard
this before, but my memory is constantly failing me.
Also, nice work! The DECnet code works just fine in MP.
Johnny
--
Johnny Billquist || "I'm on a bus
|| on a psychedelic trip
email: b...@softjar.se || Reading murder books
pdp is alive! || tryin' to stay hip" - B. Idol
