JARMOLOWSKI%ESDSDF.decnet@crdgw1.ge.com (07/19/89)
We have a system that connects a real time I/O subsystem to a sun by using
a VME bus extension. Here is what our hardware guru has to say about Sun's
VME bus.
- Tom
jarmolowski%esdsdf.decnet@re-crd.arpa
**********************************************************************
The SUN 4/110 VMEbus like most of the SUN VME's has certain restrictions,
anomalies, and outright violations. The SUN bus arbiter only allows a
single level arbiter. This unlike other VME boards is not strappable it's
hard wired to BR3/BG3. Therefore any other VME Master must use BR3. The
SUN assumes it is the MASTER of the system so it will make use of the VME
spec that says a master can release the bus whenever it is ready, this
could be awhile. Also, being the CPU and ARBITER the SUN won't issue a
bus request if the BBSY signal is not active, this is a direct violation
of spec.
Timing is where SUN took the most exceptions to the spec (which on the SUN
3/xxx was the VME rev B, I'm not positive if the 4's are at rev C.1 or if
only the newest SPARC stations are). SUN has a time-out restriction as
well. It occurs when the SUN requests the bus from another MASTER. The
SUN 3/110 will wait 767us, 3/260 waits 190us, however the sun 4/110 waits
~10us. This is a problem if you do any real-time processing. If you reach
this time-out the sun then violates the VME spec once again.
scenario:
- real time CPU drives BBSY
- sun issues BR3
- time elapses = 10us
- real time CPU releases BBSY
- sun issues BBSY for 70ns (spec=90ns)
sun releases BR3 (spec says BBSY MUST be held
held for 30ns after BR3 is release)
result:
other bus devices (in my case a HVE VME-VME RPTR)
are still waiting for sun to use the bus.
The SUN's VME interrupts are also fixed on the board, so don't plan to use
the ones SUN ALLOCATES. I don't use them so I don't know which ones they
are.
There are other things that I found but I can't remember them all. We do
use the HVE Halversa 2000 VME to VME repeater.