[HECnet] two dumb RSX questions

[Johnny Billquist]

On 2014-01-01 16:36, Lee Gleason wrote:


On the 11/74 you more or less had this, in the form of hotplugging
whole bus segments. One a bus segment was disconnected, you could
power it off, and remove controllers if you wanted.

  This reminds me of a conversation I overheard at an RSX Magic Session,
at a DECUS symposium in the 80s. A customer was asking Brian McCarthy, a
noted RSX developer, a lot of questions about the legendary 11/74 and
the   RSX11M+ additions that had been made to support it. They were
discussing the multiple Unibuses   (Unibusen? Unibeese?) and the software
that controls them. The user pointed out that the three letter task name
for the Host Reconfiguration Task (HRC...) was a little tortured - it
would have made more sense to be HRT. Brian said that internally they
had called it Hercules, and HRC was the three letter abbreviation they
chose for it. The user asks, "Why Hercules?". Brian   said that it took a
Hercules to wrestle the three headed dog that was the device
configuration databases involved into submission.

I think I've heard the story before. Pretty much the same thing. It would be nice to talk with Brian at some point...
And I totally understand. HRC is very complex and capable. Managing that flexibility at runtime is impressive. Current mP systems usually can not do it the way RSX can (could). There is only one thing that is tied to a specific CPU in RSX. Everything else can be done on any CPU. And that one thing is the updating of the system clock. That is done by the boot CPU. However, if the boot CPU is taken offline, I believe that task is migrated to another CPU, but I would need to check.

	Johnny