[SunRescue] How FAST is a SUN IPC

Dave McGuire mcguire at neurotica.com
Thu Jul 6 19:59:39 CDT 2000


On July 6, Fleet Captain Druaga wrote:
> I know from experience that you can't use two _separate_ frame buffers in an
> IPC.  You can however discount the integrated BW frame buffer if you have
> another one installed as the SBUS one "should" take over control of the
> video. However I have heard of people using more than one SBUS frame buffer
> in Sparc 1 and above systems.

  My intention is not to argue with you...but this is definitely
incorrect.  The onboard framebuffer sits on the sbus in a "virtual"
(electrical) sense, even though it's not a separate card with an sbus
connector.  The cards coexist with their own address spaces, and it's
up to the operating system's device drivers to properly probe &
initialize them upon boot.

  Laziness prevents me from actually doing it, but I was originally
going to prove a point by driving to my friend's house a few miles
from here and replying to this email on his IPC that has three
monitors on it. :)

  I might respectfully suggest that your machine, if it didn't behave
correctly with multiple heads, might have been misconfigured or had
some sort of weird revision conflict.  If you want to hack on this
again at some point, I'd be happy to help you track down what was
preventing your other framebuffer from operating.

  What does happen here, though, which can be confusing at first, is
that the onboard framebuffer is in a "higher numbered" sbus slot, in
the virtual sense...any framebuffers in "lower" numbered slots
(i.e. any of the real, physical sbus slots) will override it FOR THE
BOOT ROM OUTPUT and bootup messages and such.  The boot rom will use
the first framebuffer it finds and recognizes for its output if it is
configured to use a framebuffer instead of a serial console.

> They're also good for dedicated uses.  For instance I have two running at
> separate houses running NetBSD (www.netbsd.org) and bnetd (www.bnetd.org)
> and doing fine as standalone Battlenet(tm) servers for local LAN games.
> (The bnetd source took me about 2 hours to make and compile on the
> IPC(25mhz) compared to about 45 minutes on a faster IPX(40mhz))

  Quite true...unless there are folks here who have much better
connectivity than I do, in the case of network services provided to
the outside world, the bottleneck will *always* be the leased line.
Always.  All but the oldest of SPARC machines can saturate a 10mbps
ethernet all by themselves.


             -Dave McGuire





More information about the rescue mailing list