Copper Loops & Horizontal Waits

[DanielAllsopp]

Hey!

I see there's already a thread on horizontal waits in a copper list, I think this slightly different, but please forgive me if I'm going over old ground here. Basically, I'm following this example from ross about repeating a copper list, which I think I understand (sort of).

Here's a link to ross's example, and here's what happens when I attempt to implement the technique in my code:



What I really need to happen is to have just the name changed to white, and then after the name, the palette changes back to green. What I do sort of figure out is that the more copper commands you have, waiting horizontally, the less time you have to change the colour.

My copper list looks like this (please assume COP1LCH and COP2LCH are set correctly etc.):

  dc.w $6901,$ff00 ; wait for line $50
  dc.w $ef81,$ffff ; skip loop for lines >=$f0
  dc.w COPJMP2,$0000 ; copjmp2
  dc.w $ffff,$fffe

CooperLoop:
  dc.w BPLCON3,$0000
  dc.w COLOR24,$0072
  dc.w COLOR25,$02c5
  dc.w BPLCON3,$0200
  dc.w COLOR24,$00b2
  dc.w COLOR25,$023a

  dc.w $0053,$00fe
  dc.w BPLCON3,$0000
  dc.w COLOR24,$0ddd
  dc.w COLOR25,$0eee
  dc.w BPLCON3,$0200
  dc.w COLOR24,$0000
  dc.w COLOR25,$0666

  dc.w $0085,$00fe
  dc.w BPLCON3,$0000
  dc.w COLOR24,$0072
  dc.w COLOR25,$02c5
  dc.w BPLCON3,$0200
  dc.w COLOR24,$00b2
  dc.w COLOR25,$023a

  dc.w $ff81,$ffff ; skip loop for lines >=$f0
  dc.w COPJMP2,$0000 ; copjmp2

  dc.w $ffff,$fffe

At first I thought it was working, but then I noticed the '1' and '2' are white but should be green, and then the white starts getting changed at a totally different location. There are 6 bitplanes in use here.

Am I trying to be too clever here with horizontal waits, and should I just load a separate white font for those characters? The copper is some beast to tame mind!

Thanks for any help!
Daniel

[roondar]

Ah, there's a trick to it. Due to the Copper using the top bit of the wait command for the status of the Blitter as well as rasterline positions over $80, waiting on horizontal positions only works as you do it until you hit rasterline $80.

At that point, the wait has to change to dc.w $8053,$00fe/dc.w $8085,$00fe. Then, at rasterline $100, it has to change back to dc.w $0053,$00fe/dc.w $0085,$00fe. Basically, you can't mask bit 15 with the Copper wait mask - so you need to set it as needed for the raster position used.

See here for more info:
http://amigadev.elowar.com/read/ADCD.../node005A.html and
http://amigadev.elowar.com/read/ADCD.../node005B.html

[ross]

That's what my code here does:

.w8	cmpi.w	#$7f80,6(a5)
	bls.b	.w8
	bchg	#7,(a0)

I advise you to 'split' the copper list in two for cases <$80 and >=$80.
I think I wrote examples, maybe..

[DanielAllsopp]

Meh! I've been looking at this all day today and I just don't have enough time to change the two colours I need. Ignoring the > scanline $80 issue for now, I can test setting the two colours by using the background colour. I'm simulating the timing of setting two colours here by setting the same colour twice:

CooperLoop:
  dc.w $0001,$00fe
  dc.w BPLCON3,$0000
  dc.w COLOR00,$00f0
  dc.w COLOR00,$00f0
  dc.w BPLCON3,$0200
  dc.w COLOR00,$00f0
  dc.w COLOR00,$00f0

  ;-----

  dc.w $0053,$00fe
  dc.w BPLCON3,$0000
  dc.w COLOR00,$0fff
  dc.w COLOR00,$0fff
  dc.w BPLCON3,$0200
  dc.w COLOR00,$0fff
  dc.w COLOR00,$0fff

  ;-----

  dc.w $0085,$00fe
  dc.w BPLCON3,$0000
  dc.w COLOR00,$00f0
  dc.w COLOR00,$00f0
  dc.w BPLCON3,$0200
  dc.w COLOR00,$00f0
  dc.w COLOR00,$00f0

  ;-----

  dc.w BPLCON3,$0000
  dc.w COLOR01,$0000
  dc.w COLOR00,$0000
  dc.w BPLCON3,$0200
  dc.w COLOR01,$0000
  dc.w COLOR00,$0000

which produces this output, all nicely lined up!



However, if I set the second COLOR00 to a different colour to the first COLOR00 which will actually visualise the time taken to set a colour, then we can see the issue become apparent:

CooperLoop:
  dc.w $0001,$00fe
  dc.w BPLCON3,$0000
  dc.w COLOR00,$0f00
  dc.w COLOR00,$00f0
  dc.w BPLCON3,$0200
  dc.w COLOR00,$0f00
  dc.w COLOR00,$00f0

  ;-----

  dc.w $0053,$00fe
  dc.w BPLCON3,$0000
  dc.w COLOR00,$000f
  dc.w COLOR00,$0fff
  dc.w BPLCON3,$0200
  dc.w COLOR00,$000f
  dc.w COLOR00,$0fff

  ;-----

  dc.w $0085,$00fe
  dc.w BPLCON3,$0000
  dc.w COLOR00,$0f00
  dc.w COLOR00,$00f0
  dc.w BPLCON3,$0200
  dc.w COLOR00,$0f00
  dc.w COLOR00,$00f0

  ;-----

  dc.w BPLCON3,$0000
  dc.w COLOR01,$0000
  dc.w COLOR00,$0000
  dc.w BPLCON3,$0200
  dc.w COLOR01,$0000
  dc.w COLOR00,$0000

Blue and white both need to be set with enough time between the number, and the name for both colours to be ready before the name is output! That's not going to be possible in this case.

So, I guess my solution is to use a separate font at ~12k which uses different colour indexes in the palette, keep the name written in green text, or use a single colour font.

Never mind, thanks for the help, and at least I know a little bit more about the copper now

[mcgeezer]

Use BPLCON4 mate.

[DanielAllsopp]

Originally Posted by mcgeezer

Use BPLCON4 mate.

Ahhh, something to look at today. I've briefly read up on BPLCON4 but not in any great detail, so that's my next port of call!

Cheers!

[DanielAllsopp]

Ah, you little beauty mcgeezer...

dc.w $0005,$00fe
dc.w BPLCON4,$0011

dc.w $0059,$00fe
dc.w BPLCON4,$0411

dc.w $0089,$00fe
dc.w BPLCON4,$0011

dc.w $00e1,$00fe

Now I'm not 100% sure why this is working, and I don't like writing code when I don't know what it does, but I think by setting the second byte value in that register to a value of $0004, or $0411 for the whole word, shifts the colour selection to the left by 4. Tested with 6 and indeed, the text changes to yellow!



Of course, this has altered my COLOR00 value, which is now pointing to whatever is in COLOR04 so all I needed to do was change COLOR04 (which is unused anyway) to $0000 and everything looks hunky dory!

Now... onto the next bit of this challenge, the > scanline $80

[DanielAllsopp]

Done, and with 15 minutes to spare before I turn into a pumpkin, aka a forced cut off time of 12am to help keep my sanity!



Thanks again for everyone's help with this one, your names are up in lights!

On to the next Trello ticket tomorrow then!