Date   : Sat, 11 Mar 2006 21:51:59 +0000
From   : Fragula <fragula@...>
Subject: Re: Warning: Sad case on list!

Hi Jonathan!

I happened again across the technical description of the tube interrupt
ops at
http://www.nvg.ntnu.no/bbc/doc/Master%20512%20Technical%20Guide/m512tech04.htm


<winces@... of link>

But it seems to make sense (Not that I can remember where NMIVEC is or
anything) because ths host coulnd't run in a tight loop when its busy.
Drawing, getting FDC or ADLC interrupts.

Hmm. AFAICS there is no cast-in-stone reason the parasite could not
spinpoll during disk/net ops though, or for other reasons, poll on its
internal interrupt for events.

So.. I'll keep an open mind.

Cheers!

M.






Jonathan Graham Harston wrote:

>>Message-ID: <440A29A3.9020101@...>
> 
>  
> Fragula <fragula@...> wrote:
> 
>>Have to question the interrupt bit though, Its a helluva long time ago,
>>and I really should really check the circuit diagram, but I'll live
>>dangerously on this one. Feel free to flame me if I'm wrong.
> 
>  
> See http://mdfs.net/Software/Tube for various Tube documentation,
> source and resources.
>  
> 
>>There are a couple of tricks the 2Ps code does.. First the stuff in the
>>host (DNFS or whatever) does its initial stuff, the 2P initialises, and
>>uploads (the 2p being the initiator) replacement interrupt handlers over
>>to the host, trapping IRQs and so I have read, NMIs too.
> 
>  
> The host just runs a tight polling loop waiting for instruction
> fron the client. The client never interupts the host. The host
> interupts the client to transfer blocks of data between the two
> systems.
>  


-- 
(Sorry if I top-quote occasionally, the b@...*rds make me use LookOut in
work.)