Just a very minor point on using SWAPF to push/pull without affecting STATUS.

*IF* (a VERY big IF) the push/pull is the boundary of the sole interrupt
service routine, AND the main code doesn't affect any STATUS bits, then you
can leave out STATUS from the push/pull, and push/pull W with plain MOVWF
and MOVF.  Result (for '84) can look like something I did recently:

        ORG     4               ; Interrupt vector
ISR
; Interrupt service routine
; explanation and description deleted............

        MOVWF   STAKW           ; Stack W BUT NOT S (not used in main routine)
;        SWAPF   STATUS, W       ; Pre-swap STATUS for storage
;        MOVWF   STAKS           ; Note GIE cleared by responding to interrupt

::::::::::::::::::

ENDINT
        BCF     INTCON, T0IF    ; Clear Timer interrupt that got us here
; STATUS not recovered, as not saved, as not used in main routine
;        SWAPF   STAKS, W        ; Recover (and unswap) status register
;        MOVWF   STATUS
;        SWAPF   STAKW, F        ; Preswap W
;        SWAPF   STAKW, W        ; And recover it
        MOVF    STAKW, W        ; Affects STATUS, but don't care
        RETFIE                  ; Also re-enables interupts by setting GIE

The specific application was a very simple one where everything depended on
timer ticks and the main routine was a simple loop.

Remembering the discussion a few weeks back about comments, you will note
that I've included several reminders to myself about what the code does,
and that the conventional push/pull of STATUS and W using SWAPF is not
deleted, merely commented out in case some future revision means the main
loop has to use STATUS.  No doubt some will think I've over-commented, but
I don't do very much PIC programming, and find it very easy to forget why
my old code looks the way it does.

You could argue, with much justification, that the push/pull should always
ensure that STATUS is returned in the same condition it was before the
routine - there were no overriding requirements of timing or space which
meant saving a few instructions actually mattered.  In this case it did
mean that the number of cycles lost in the prescaler (reset when the timer
was reloaded) was reduced by two, and I can envisage situations where that
might be useful.

Tim Forcer

Tim Forcer               tmf@ecs.soton.ac.uk
Department of Electronics & Computer Science
The University of Southampton, UK

The University is not responsible for my opinions