FMODE 0 to FMODE 1

[Auscoder]

Hey all

I am trying to move my game to AGA from ECS version, and would like to set the FMODE now higher than the Normal Fetch. I am trying just $01 for now...

I am seeing visual error when setting $01, this is looking much like what would be a bad modulo.... All works well with Normal Fetch.

I have ensured the following which I can find is required for FMODE $01

Bitmap data located on an 8 byte aligned memory address.
Bitmap data bytes width is multiple of 8 bytes
Also modulos are multiple of 8 bytes

Cooper list is also located on 8 byte aligned memory address

DDFSTART is #$30
DDFSTOP is #$D0

Not sure where else to look? Is there something I missed that is still needed.

Changing FMODE back to 0 with all above changes, things are looking fine .

[mcgeezer]

You have to adjust the modulo depending on what fetch mode you are in, try setting your bitplane modulo but with -6 or -4.

There is a calculation but i dont have it to hand atm.

[Auscoder]

Originally Posted by mcgeezer

You have to adjust the modulo depending on what fetch mode you are in, try setting your bitplane modulo but with -6 or -4.

There is a calculation but i dont have it to hand atm.

Thanks mcgeezer, I tried a few variations, positive and negative, a few made things look a bit better, some much worse. actually -2 came closest. Looks like that at least one bitplane is right like so, but following bitplanes are weirdly offset. Almost like following bitplane pointers are being set too late...

I realized that having original style horizontal scrolling running in FMODE 1 that would no longer be appropriate and could be causing issues?

[roondar]

This is from memory, so it could be wrong but here goes: AFAIK the modulo for FMODE 4x is -8 and FMODE 2x is -4.

I remember having similar issues when I set up an FMODE 4x screen and it turned out that my bitplane data was incorrectly aligned. My first step would be to check if the individual bitplane pointers are all aligned to 4 bytes (for FMODE 2x, or 8 bytes if using FMODE 4x).

In my case it turned out they were not because I added a scrolling offset to the bitplane pointers. Remember, when horizontally scrolling an FMODE 2x/4x screen you need to scroll for 32/64 pixels rather than just 16 and update the bitplane pointers by 4 or 8 bytes for the new offsets rather than just 2 bytes.

Another error I made was not updating the Blitter modulo's (I had made my bitmap wider because of the new alignment, but I forgot to update the constants I used for the Blitter). Last but not least, I had the wrong DDFSTOP, which also caused issues that looked like wrong modulo values.

[Auscoder]

Originally Posted by roondar

This is from memory, so it could be wrong but here goes: AFAIK the modulo for FMODE 4x is -8 and FMODE 2x is -4.

I remember having similar issues when I set up an FMODE 4x screen and it turned out that my bitplane data was incorrectly aligned. My first step would be to check if the individual bitplane pointers are all aligned to 4 bytes (or 8 bytes if using FMODE 4x).

In my case it turned out they were not because I added a scrolling offset to the bitplane pointers. Remember, when horizontally scrolling an FMODE 2x/4x screen you need to scroll for 32/64 pixels rather than just 16 and update the bitplane pointers by 4 or 8 bytes for the new offsets rather than just 2 bytes.

Another error I made was not updating the Blitter modulo's (I had made my bitmap wider because of the new alignment, but I forgot to update the constants I used for the Blitter). Last but not least, I had the wrong DDFSTOP, which also caused issues that looked like wrong modulo values.

Thanks Roondar, there are some nice options for me to go through tomorrow evening. Its getting a little late to be doing much more here now.

You probably are on to it with the scroll offset . I am adjusting byte offsets on even (but multiple of 2) bytes.. So I will look in to that. For now I will just disable original scroll code completely and reset any offsets to 0. I had hoped old scroll code would continue to function "as is".

Back to it tomorrow night. Thanks for the info chaps.

[StingRay]

Set your DDFSTOP to $c8, this will fix your display problems too without having to mess with the modulos.

[Toni Wilen]

Yeah, you only waste DMA time (and possible trigger bitplane DMA to refresh slot conflicts = screen gets very corrupted) if you try to fix the display by adjusting modulo instead of using correct DDFSTOP.

[ross]

Originally Posted by Toni Wilen

Yeah, you only waste DMA time (and possible trigger bitplane DMA to refresh slot conflicts = screen gets very corrupted) if you try to fix the display by adjusting modulo instead of using correct DDFSTOP.

Is it possible to trigger a WinUAE HALT condition if the refresh slots are invaded? (not only in log)
I remember that my real Amiga alway crashed when I accidentally made these conditions (with code in chip or fake fast ram).
Just for a even more real emulation and to make the programmer conscious if the glitches are due to other reasons

[roondar]

Huh... I need to go and change my DDFSTOP setting in AGA code then. I could've sworn I read the "you need to change modulo to -4/-8 for AGA FMODE 2x/4x" in several places and always assumed that was just the way it was.

Good to learn something new, so thanks for the info.

[Auscoder]

Originally Posted by roondar

This is from memory, so it could be wrong but here goes: AFAIK the modulo for FMODE 4x is -8 and FMODE 2x is -4.

In my case it turned out they were not because I added a scrolling offset to the bitplane pointers. Remember, when horizontally scrolling an FMODE 2x/4x screen you need to scroll for 32/64 pixels rather than just 16 and update the bitplane pointers by 4 or 8 bytes for the new offsets rather than just 2 bytes.

Headsup, Thanks Roondar, it was exactly this. In conjunction I had some padding being added at runtime to the bitplane pointers to avoid having to do small amounts of clipping on blits. Adding some validation to where I wrote the bitplane pointers picked it right up.

All good now! Thanks for the learned tips.

[Auscoder]

Originally Posted by StingRay

Set your DDFSTOP to $c8, this will fix your display problems too without having to mess with the modulos.

Thank you for this advice, I have update DDFSTOP, but did still have to update Modulos. I don't quite understand these registers and how they work properly, so it's time to do some reading.

[roondar]

No problem, glad to be of assistance