----------
> From: John Payson <supercat@MCS.NET>
> To: PICLIST@MITVMA.MIT.EDU
> Subject: Re: TMR0 interrupt (PIC16c84 and PWM)
> Date: Monday, 19 May 1997 09:58
>
> > Just for a start, DON'T re-enable GIE inside the ISR.  The RETFIE
> > instruction does this automagically, and, if you have had an int whilst
> > processing the isr, you'll blow your stack really quickly this way.
This
> > is the sort of thing a simulator will probably never find, BTW, as its
very
> > time-sensitive.
>
> As a slight clarification, never enable GIE within an ISR unless you are
> absolutely sure you know what you're doing *AND* you have counted up the
> stack overhead.  I have on a couple of programs used a short ISR which
ran,
> essentially, this:
>
>         [ save W and PSW ]
>         [ do a little bit of work]
>         [ restore W and PSW ]
>         decfsz  counter,f
>          retfie
>         bsf     gie
>         [ save W and PSW to another pair of locations]
>         [ set up counter again]
>         [ do 'a lot' more work (more than a timer-ticks' worth ]
>         [ restore PSW and W]
>         retfie
>
> My smaller interrupt routine does not make any sub-calls.  My larger one
> does make one level of calls, but they are "guarded" with disable-timer
> interrupt and enable-timer interrupt instructions.
>
> For applications where an ISR has to do periodic "thinks" but mostly just
> keeps a counter or two, I think this approach is an excellent one.  Has
> anyone else used it?

That's an interesting exception to the rule.  Must make a note of it.
Almost reminds me of Windows background processing <g>

MikeS
<mikesmith_oz@relaymail.net>