Hmm... Interesting proposal, but I am still up against the same challenge. time
for interrupt plus latency in and latency out is still there, and I would need to
verify that I am not trampling the main routine.

I was thinking that I could run an open loop timer with no interrupt, then run the
interrupt, and compare the increment rate (similar to processor benchmarking).
Then the ratio would tell me the time for the interrupt (we know the period).

Chris Eddy

Andrew Kunz wrote:

> Why not use one of the on-board timers?  That's what they are there for.
>
> Andy
>
> Chris Eddy <ceddy@NB.NET> on 03/20/2000 06:11:45 AM
>
> Please respond to pic microcontroller discussion list <PICLIST@MITVMA.MIT.EDU>
>
>  To:      PICLIST@MITVMA.MIT.EDU
>
>  cc:      (bcc: Andrew Kunz/TDI_NOTES)
>
>  Subject: [ot] interrupt latency under DOS?
>
> I have a turbo C program running under datalight RomDOS.  I have tricked
> the clock into interrupting much faster than normal, and dividing down
> to call the regular clock (a common trick in DOS).  Alas, I need to go
> even faster.  Knowing that the 486/Pentium architecture has cache
> running and pipelining and crap like that, I cannot rely solely on
> setting a pin and resetting at the end of interrupt for a duration.
> (The power of a PIC!) Does anyone have worst case figures on a Pentium
> or 486's interrupt latency? Any tricks to confirm that the interrupt is
> short enough?
>
> Thanks
> Chris Eddy