RE: AVG programming question...

Sorry about taking so long to respond. I've been at home this week helping
take care of our new little that was born on saturday.

When I first started writting space wars and tank.. (um forget about the
second one, it's still secret) they worked great in MAME, but I had one heck
of a time getting them to run on the real hardware. I now have them working
almost 100% (just a few weird things still happening).

I don't know how similar tempest hardware is to space duel & gravitar, but
this is how I got it to work.

The ISR routine is completely independent of the vector drawing hardware and
is used only for timing (I set my drawflag every 4th entry into the ISR)
You must have a HALT instruction at the end of your vector list (0x20) or
the vector generator won't release the VRAM back to the processor. The halt
instruction will also set the halt flag (on tempest it's 0x0C00 bit 6) which
you should check before attempting to write anything to the VRAM.

The following routine should work

zoom
        bit 0x0C00
        bvc zoom ;wait for vector generator to halt
Timing
        lsr drawflag
        bcc Timing ;wait for timing flag
        dec 0x2002
        lda 0x2002
        cmp #0x18
        beq Done
        sta VGGO
        jmp zoom
Done

plus you need something like this for your ISR
ISR pha
        tya
        pha
        txa
        pha
        dec icount
        lda icount
        and #03
        bne ISR1
        inc drawflag ;set every 4th interrupt
ISR1 cli
        pla
        tax
        pla
        tay
        pla
        rti

Also, I only use VGRESET once in the RESET routine and I use VGGO only in
the main loop. That way the ISR won't fire it when you don't want it to go
(and make all the VRAM unavailable)

I have a complete API for programming the Vector Generator, it currently
works for space duel and gravitar and I've almost got it done for tempest. I
also have an API for reading controls. All you have to do is change a
compile flag and your code will run on another machine (with a few
limitations, I don't have anything for accessing the mathbox in tempest and
I'm still working on a DVG version of the VG API)

Let me know if you have any other questions and if you want, you can shoot
some of your code over to me and I can take a look at it (or course I would
keep that confidential)

-jeff

> -----Original Message-----
> From: Clay Cowgill [SMTP:ClayC@diamondmm.com]
> Sent: Thursday, April 15, 1999 11:45 AM
> To: 'vectorlist@mcfeeley.cc.utexas.edu'
> Subject: AVG programming question...
>
> Hi everyone,
>
> So this will probably be a limited audience for this message, but if any
> of you know the answer feel free to pipe up. ;-)
>
> (I'm figuring Al, Jeff, and Anders will probably be the likely
> candidates...)
>
> I'm doing my usual "write the menu system in MAME and then make it work
> on real hardware" routine with the Tempest Multigame. I've got the
> menu-system to a point I'm happy with, but alas the "real" Tempest
> hardware doesn't take so kindly to it.
>
> I've got a display at least now (it was what I thought, Al-- need to
> have the VG reset and VG go right at the end of the ISR...), but here's
> my question for any of you that have programmed on the AVG:
>
> Does the AVG do anything "funny" with read-modify-write type
> instructions accessing VRAM? I can't see any reason why it would, but
> the symptom I'm having now is that it *looks* like the folliowing
> doesn't work:
>
> VRAM is from h'2000 - h'2FFF in Tempest, so I've got something like this
> in VRAM...
>
> Starting at 0x2000:
>
> 40 80 FF 71 06 6F 37 A9 ...
>
> which basically means:
>
> 40 80 center beam
>
> FF 71 zoom factor "really small"
>
> 06 6F color register 6 (had to do a 6F to make MAME happy, I don't think
> the real hardware cares...)
>
> 37 A9 subroutine call to display the "TEMPEST" logo
>
> (Yes, I did that all by memory. Did I happen to mention that I can now
> read and write Tempest Vector Memory display lists in hex? Probably
> never come up in final Jeopardy though... :-)
>
> What I then do is basically:
>
> lda #0x01
> sta frame_done
> zoom: dec 0x2002 ; zoom in the logo - decreasing the value of
> that 0xFF in the display list above
> wait: lda frame_done
> bne wait
> lda 0x2002
> cmp #0x18
> bne zoom
> ...
>
> frame_done is a flag that is cleared by the ISR that starts the Vector
> Generator drawing. (So once a frame is drawn frame_done is 1.)
>
> What the above code does in MAME is decrement the zoom value of the
> Tempest logo (making it get larger on the screen) once per frame update
> until the zoom is equal to 0x18 after which it continues on to other
> code.
>
> What it seems to do on the *real* hardware is hang. The ISR is still
> running, but the logo never zooms in.
>
> I'm concerned that either the "dec 0x2002" isn't working (so the loop
> never exits), or maybe that since my loop flag gets set in the ISR, but
> possibly prior to the vector generator actually *finishing* drawing the
> frame, maybe I'm getting a bus contention and "dec 0x2002" isn't able to
> change the VRAM.
>
> I don't know if Tempest *has* a VG-busy bit I can read to tell when
> things are done drawing...
>
> Any thoughts/theories/solutions welcome...
>
> -Clay
Received on Thu Apr 15 15:39:38 1999