| |
Rudi
Account closed
Registered: May 2010
Posts: 125 |
Multiplexing sprites?
Hi,
Thanks for all the help ive gotten so far here on csdb.
Now I've runned into another problem.
How the *** does sprite multiplexing work?
I just need the basic idea. I tried to read Cadaver and other tuts, but they seemed too advanced for my taste in continuing.
What my goal is basically to display 64 sprites that stands still. So it will display a checkerboard, hence the sprites only need one of two colors. and they are filled with either of the color. The sprites are displayed beneath the characters so i can either have black and white chess-pieces (using $d800) this works fine with 8 sprites! but now i need to fill the entire checkerboard.
(edit: the entire board is in singlecolor, not multicolor btw)
I have not yet figured out how it works in theory? I just need to have interrupts on each row and display the sprites.
I have a feeling it would work something like this:
01234567
12345670
01234567
12345670
01234567
12345670
01234567
12345670
sprite 0 has darkgray, sprite 1 gray color, sprite 2 darkgray, ..etc.. so it will show a checkerboard in the end..
is it as simple as just setting an interrupt and move the sprite or do one need something else to take into consideration as well? |
|
... 42 posts hidden. Click here to view all posts....
|
| |
Oswald
Registered: Apr 2002
Posts: 5094 |
Quote: Quoting Oswaldy stretch would solve the 16 vs 21 pixel problem :) but I'd rather use X expanded sprites so you dont need all 8, otherwise sprite and character row dma will steal all your cycles on badlines. Not quite ALL cycles, if i'm not mistaken.
Out of the otherwise available 20+3 CPU cycles on a PAL badline, only 16 are taken by all sprite DMA.
There remain 3 CPU cycles in the right border (the first of those being the one to switch $d016 for open sideborders) before sprite 0 DMA begins and 4 CPU cycles in the left border, after sprite 7 DMA has finished.
Switching $d018 for new sprite patterns after a badline is possible using read-modify-write instructions to get into the 3-cycle gap.
yeah tried once, didnt know about the RMW trick, I was either 1 cycle soon or 1 cycle late. Didnt help with my dislike towards coding raster effects :) |
| |
Krill
Registered: Apr 2002
Posts: 2980 |
Quoting FreshNope.
6510 is still in idle from the badline and before the first sprite fetch there are exactly 3 cycles: BA will be kept low and the CPU will have to wait for the whole sprite DMA to finish. On a PAL machine right after a badline you can't do anything before sprite 0 fetch, you'll lose 19 cycles.
Moreover, this means that you can't open side borders on a badline if sprite 0 is active. Okay, you're probably right there. Still 4 cycles left to switch $d018. I certainly have coded multiplexers which didn't display any glitches when moving a seamless wall of sprites vertically across badlines, thanks to RMW instructions. :) |
| |
Fresh
Registered: Jan 2005
Posts: 101 |
Yes, it's doable and you don't even need a RMW instruction. You can put a STx $d018 before BL so that its third (R) cycle happens when BA goes low: at the end of sprite DMA the CPU will execute the last W cycle.
Beside the strict timing, the other annoying thing is that you need to prepare a second charpage to switch $d018 transparently. |
| |
Krill
Registered: Apr 2002
Posts: 2980 |
Quoting FreshYes, it's doable and you don't even need a RMW instruction. You can put a STx $d018 before BL so that its third (R) cycle happens when BA goes low: at the end of sprite DMA the CPU will execute the last W cycle.
Beside the strict timing, the other annoying thing is that you need to prepare a second charpage to switch $d018 transparently. That's a minor nuisance in the world of C-64 raster coding :D
For some reason, a plain STx didn't work for me. But it's a while ago now, rastercode was more trial&error than educated guesses back then, to me, anyways :) |
| |
Skate
Registered: Jul 2003
Posts: 494 |
trial&error style coding is more valuable. it's all about "wow! somehow it worked, wohoo!" feeling. :) |
| |
Flavioweb
Registered: Nov 2011
Posts: 463 |
Quoting FreshYes, it's doable and you don't even need a RMW instruction. You can put a STx $d018 before BL so that its third (R) cycle happens when BA goes low: at the end of sprite DMA the CPU will execute the last W cycle.
Beside the strict timing, the other annoying thing is that you need to prepare a second charpage to switch $d018 transparently.
You need to set D018 before BA goes low if sprite 0 is on, otherwise BA still low until sprite 0 data is fetched on pal machines... |
| |
Fresh
Registered: Jan 2005
Posts: 101 |
This way you would end up changing sprite definitions one line earlier.
If you need to change them on the line after BL (which is the one giving problems) and you are working with 8 sprites, you must switch $d018 on the only available cycle: cycle 11, which is "located" after preceeding sprite DMA and before badline DMA.
In short, changing the sprite definition on the line after a BL can be done by using a normal STx so that its fourth cycle falls exactly on cycle 11 *before* badline. |
| |
Fungus
Registered: Sep 2002
Posts: 686 |
@Krill now do wall of sprites moving with d011 scroll changing too. :D
Looking forward to that GAME. :) |
| |
Oswald
Registered: Apr 2002
Posts: 5094 |
I dont think it can be done with stx, I have tried it with a stable raster and it was either too soon or too late with 1 cycle difference. |
| |
chatGPZ
Registered: Dec 2001
Posts: 11386 |
fungus: thats fairly easy using the overlapping data "trick" :) |
| Previous - 1 | 2 | 3 | 4 | 5 | 6 - Next |
