Menace tutorial AGA fix

[spud]

Hi all

Tinkering with asm (writing a game - slowly) and making use of the startup/shutdown framework provided by the Menace tutorial all those years ago I wonder if anyone can explain what is required to fix the source code to work on AGA?

Right now it runs fine when ECS or OCS is selected as the chipset, but AGA gives a corrupt screen.

Whether the startup code provided by the tutorial is or isn't a good example of such code I don't know, but it would be good to know why it doesn't work and how to correct it. It should hopefully further understanding of the architecture which is quite overwhelming when coming from a z80 coding perspective !

Cheers

[Methanoid]

One for Galahad methinks!!

[deicidal]

Try here

http://zakalwe.fi/~shd/amiga-cracking/agafix-v2.html

[spud]

That looks like a useful resource, thanks

Done so much reading around since I posted the original question and managed to pull together code and put it in the right (?) place to seemingly fix it to display on AGA.

Running this before I install the copperlist:
sub.l a1,a1
move.l GraphicsBase,a6
jsr _LVOLoadView(a6) ; Blank the screen
jsr _LVOWaitTOF(a6)
jsr _LVOWaitTOF(a6)

and aligning the actual copperlist with
cnop 0,8

Why does it work... more reading needed to determine that. Information overload?!

[Leffmann]

On AGA, register FMODE is usually set to instruct the bitplane DMA to fetch data in chunks of 64 bits, so when all other bitplane parameters are set up for 16 bits like in OCS/ECS, you get these distorted graphics.

LoadView(NULL) fixes things because it sets up a Copper program that sets FMODE, BPLCON2 and BPLCON3 (and probably some other registers) to OCS/ECS compatible values. There's never any need to align the Copper program on 8 bytes though, a plain even address will do.

I think calling LoadView like this will also enable internal display output if you have RTG in your system, so you should keep this in your program.

[StingRay]

Originally Posted by Leffmann

I think calling LoadView like this will also enable internal display output if you have RTG in your system, so you should keep this in your program.

That's correct. Without the LoadView() call the display won't be switched to the RTG screen which means you won't see anything if you only have 1 monitor connected to your Amiga.

[spud]

Thanks guys

[Galahad/FLT]

Originally Posted by spud

Thanks guys

First bit of code to try in your setup routine would be

move.w #0,$dff1fc

This clears the Fmode burst modus, which 'should' correct the screen.

[StingRay]

Originally Posted by Galahad/FLT

First bit of code to try in your setup routine would be

move.w #0,$dff1fc

This clears the Fmode burst modus, which 'should' correct the screen.

The LoadView(0) call does this (and more) as Leffmann said.

[spud]

$dff1fc is presumaby the f-mode register then? that is a "new" register that isn't featured in the O/ECS? which would explain why I couldn't find it in the hardware ref manual, yet there is lots of talk about it. Sometimes the blindingly obvious is staring you in the face, but you can't see it.

Anyway, I tested moving 0 into it, and I see that works too.The copperlist didn't mind not being 64bit aligned either. Thanks again for the help.

I'll provide a demo of what I'm up to when it progresses beyond routine testing - at the moment I've got a software sprite moving about (not using the blitter - good (if simple) excercise in programming I think), platforms blitted into place, dual playfield, some copper background goodies and now it works under AGA. Lots more to learn, though. More question to follow? Maybe

[StingRay]

Yes, it's the fetch mode register that was introduced with AGA. And yes, just setting the fetch mode register to 0 will work too but the LoadView approach is a much better one, especially if you want full compatibility with "alien" screen modes.

[Galahad/FLT]

Originally Posted by StingRay

The LoadView(0) call does this (and more) as Leffmann said.

I know that, but its simple code to track down his screen problem and actually diagnose it as an fmode problem

[freehand]

del.