Copper Wait Problem

[sandruzzo]

I've got some problem to get position below 200px window. To get that I'll do what HW sugget: wait last line and do another wait. But it seems don't work.
here my code:

dc.w $ffdf,$fffe
dc.w $1001,$fffe
dc.w $0180,$0fff

Some bugs on winuae? Maybe after that end copper list must be different? I use the classic

dc.w $ffff,$fffe

My examples is running on Aga. I did some test on ocs/ecs and seems to works fine.

This issue is present when I activate Burst Aga mode to 64bit:

dc.w $01fc,%1111

[Toni Wilen]

100% not enough information! (My crystal ball says bitplane dma steals cycles from copper)

[sandruzzo]

Originally Posted by Toni Wilen

100% not enough information! (My crystal ball says bitplane dma steals cycles from copper)

if I turn of burst mode it works. It seems a bug since I change bitplane pointer after that copper wait, and nothing is happening!

Do you want the exe? I can provide code too

[Toni Wilen]

Executable (that needs nothing else) is the best way to confirm it.

[sandruzzo]

Originally Posted by Toni Wilen

Executable (that needs nothing else) is the best way to confirm it.

Were I send it?

[Toni Wilen]

Attach file? Upload to the zone? Email? Your choice

[sandruzzo]

Originally Posted by Toni Wilen

Attach file? Upload to the zone? Email? Your choice

email , if I may

[Toni Wilen]

My crystal ball guess was correct. Real A1200 also shows identical image.

DDFSTOP is far too large for AGA higher fetch modes, causing bitplane DMA to steal nearly all cycles at the end of scanline. Vertical position wraps around to 0 before copper has had time to wake up -> copper keeps waiting..

You can use DMA debugger to check it ("v -1" in debugger to enable it). Then v <vpos> <hpos> to see line's activity.

[sandruzzo]

Originally Posted by Toni Wilen

My crystal ball guess was correct. Real A1200 also shows identical image.

DDFSTOP is far too large for AGA higher fetch modes, causing bitplane DMA to steal nearly all cycles at the end of scanline. Vertical position wraps around to 0 before copper has had time to wake up -> copper keeps waiting..

You can use DMA debugger to check it ("v -1" in debugger to enable it). Then v <vpos> <hpos> to see line's activity.

So I cant' use pal screen? I have to use only ntsc? Is there any way to do that?

[Toni Wilen]

Originally Posted by sandruzzo

So I cant' use pal screen? I have to use only ntsc? Is there any way to do that?

No, your screen is too wide, not too tall and wasting DMA cycles. You have extra 64 pixels completely hidden in right border.. Just reduce DDFSTOP, probably something like $c0 or so is correct. (not $d0!)

[sandruzzo]

Originally Posted by Toni Wilen

No, your screen is too wide, not too tall and wasting DMA cycles. You have extra 64 pixels completely hidden in right border.. Just reduce DDFSTOP, probably something like $c0 or so is correct. (not $d0!)

ok. I'll try, thanks for help

[Toni Wilen]

Even easier way: reduce DDFSTOP until screen becomes corrupted, then compensate it with modulo. Done

[sandruzzo]

Originally Posted by Toni Wilen

No, your screen is too wide, not too tall and wasting DMA cycles. You have extra 64 pixels completely hidden in right border.. Just reduce DDFSTOP, probably something like $c0 or so is correct. (not $d0!)

Using somenting like b8 is fine

[sandruzzo]

Originally Posted by Toni Wilen

Even easier way: reduce DDFSTOP until screen becomes corrupted, then compensate it with modulo. Done

Thanks for helping me. I left there some broken code