Path: utzoo!mnetor!uunet!lll-winken!lll-tis!ames!pasteur!ucbvax!LOYVAX.BITNET!PGOETZ
From: PGOETZ@LOYVAX.BITNET
Newsgroups: comp.sys.apple
Subject: 60 Hz interrupt mod
Message-ID: <8804211832.aa02388@SMOKE.BRL.ARPA>
Date: 21 Apr 88 20:57:00 GMT
Sender: daemon@ucbvax.BERKELEY.EDU
Organization: The Internet
Lines: 65


   Derived from "A Clock Interrupt for Your Apple" by Charles Putney,
published in MICRO No. 62 July 1983, p. 36-41.

   This is written for the Apple II and II+.  NOT the IIe, which I have heard
can generate interrupts itself anyway.  Chips might have different
numbers/locations in other models, or might not be present.  The old Euro-Apple
has different counters (because European screens have a different scan rate),
so this mod won't work on them either.

   Basically, this mod steals a periodic signal off the board & sends it to
the 6502's IRQ line.  Counters D11, D12, D13, & D14 generate the horizontal
byte position & vertical line position on the screen (to feed the proper video
info out).  Connect pin 11 of D11 (the chip to the right of the middle row of
RAM) to Pin 4 of H7/8 (the 6502).  30-gauge wire-wrap wire is best.  Lift the
pins enough to insert the wire along with the chip.  This way you can remove
the wire easily.
   This provides ~60 interrupts per second (59.92 on Putney's Apple).  Putney
claims this will add only 2.9% interrupt-processing overhead to a running
program.
   He suggests several applications: keep track of time, as a keyboard buffer,
or as a print spooler.  I think you could also implement multi-tasking, though
you might have to swap page 0 in & out between tasks.  The article contains a
clock program and keyboard buffer.  I can't send them out because they're
copyrighted. (For any lawyers out there: What happened to MICRO's copyrights
since they went bankrupt?  Did they revert to the authors?)
   The IRQ (Interrupt ReQuest) line, as opposed to the NMI (Non-Maskable
Interrupt) line, only registers an interrupt if the interrupt disable bit (bit
2) of the processor status byte is clear.  Clear it with a CLI ($58), set it
with a SEI ($78).
   If this request is received, the processor jumps thru a vector at $FFFE,
which points to $FA86 in the II and $FA40 in the II+.  This routine saves the
accumulator (A) in $45?.  Then it jumps through $3FE to your interrupt handler.

Your installer should:
1. Point $3FE at your handler.
3. SEI (Putney's program does this: LDA LO+1; EOR #$20; STA LO+1; LO: SEI.
That way the first call does a CLI & the second call does a SEI.)
4. RTS

Your handler should:
1. Save the X, Y, and P registers.  The A reg is saved in $45 (not a good place,
I think it could cause trouble with DOS) by $FA40.  DON'T save them using
the monitor IOSAVE routine, because internal functions use the IOSAVE routine,
and interrupting them would screw them up.
2. Do whatever.
3. Restore X and Y, load A from $45, load the processor status (P) register.
4. RTI (not RTS).

   Payton mentions that a keyboard buffer, which clears $C010's hi-bit, will
not work with programs which check $C010 for a keypress.  He must mean $C000,
because checking $C010 would tell the apple to throw away the last keypress.
If a program looks like LDA $C000; BMI NO_KEYPRESS, and your keyboard
buffer interrupt program is busy buffering keypresses & accessing $C010
to clear the hibit of $C000, the program will miss many keypresses, the ratio
depending on how fast it is.
   Note that if you write an interrupt handler which takes more than 1/60 of
a second, you will be caught in an infinite loop.

Charles Putney's address is
18 Quinns Road
Shankill
County Dublin, Ireland

Phil Goetz
PGOETZ@LOYVAX.bitnet