Hello all
First time post on here, apologies if not properly on topic.
As part of a project to get HP Access running using the hp2100 simulator on a 
plug computer (sheevaplug), I have been tweeking the throttle parameters of the 
iop and main cpu on my Access setup to get best performance out of the single 
core ARM5 host processor. For those of you who dont know the HP Access setup 
requires two hp2100 CPUs configured as a main processor and as an IO processor 
and talking to each other through an interconnect which is simulated on two tcp 
ports. In the course of tweaking the throttle parameter  I found the following:

If I run up the main and the iop manually and then use ctrl-e to get into simh 
and set throttle 45% on both the iop and the main, then this works just fine 
and the processes duly only consume 45% of host cpu.

However, if I put a set throttle 45% in the startup scripts, asmain and asiop, 
the iop stubbornly takes 96+% host cpu and the main gets the rest. 

Finally, if I start asmain manually but do not start asiop, I note that the 
main cpu runs at 45% host cpu. Next I start asiop, and as soon as the 
interconnect goes live the iop runs up to 96+% of host and the main cpu drops 
to 4% or less.

I have confirmed the throttle settings using a show throttle just after the set 
in each case, and placed the set just before the run. I have also confirmed the 
hp2100 process are running at the same host priority.

So it seems that if you throttle the cpu before the IOP interconnect goes up 
the throttle is either oddly interpreted or ignored and it seems to effect both 
main and iop. 

Anyone have a similar experience or any idea what is going on here? 

Other oddities seem to occur if you ctrl-e asmain or asiop and then immediately 
cont. The throttle changes, but not how you might expect. Would be good to fix 
it, as then I can get the plug computer to act as an appliance. A, literally, 
plug in and go HP-Access system, if you like.

Regards

Quentin

Sent from my iPad
_______________________________________________
Simh mailing list
[email protected]
http://mailman.trailing-edge.com/mailman/listinfo/simh

Reply via email to