It is currently Mon Dec 18, 2017 3:39 am

All times are UTC - 7 hours





Post new topic Reply to topic  [ 7 posts ] 
Author Message
PostPosted: Mon Apr 17, 2017 9:30 pm 
Offline

Joined: Wed May 19, 2010 6:12 pm
Posts: 2433
I might as well start a new about it.

It can't be drawing it one pixel at a time because then it has only 228*312*15/7*2/256/224=5.316 cycles per pixel. It has to do at least 4 pixels at once just to move data from one place to another fast enough, but where does it get it's pixel data from then?


Top
 Profile  
 
PostPosted: Mon Apr 17, 2017 9:50 pm 
Offline

Joined: Sun Apr 13, 2008 11:12 am
Posts: 6540
Location: Seattle
The effect is often called a "rotozoom(er)". There's a writeup about how one was implemented on the contemporary and comparable-spec Atari ST here.


Top
 Profile  
 
PostPosted: Tue Apr 18, 2017 8:00 am 
Offline

Joined: Wed May 19, 2010 6:12 pm
Posts: 2433
That was interesting, but that looks like it would need at least 12 cycles per pixel.


Top
 Profile  
 
PostPosted: Tue Apr 18, 2017 8:14 am 
Offline

Joined: Sun Sep 19, 2004 11:12 pm
Posts: 19355
Location: NE Indiana, USA (NTSC)
If you can't get your rotozoomer up to full speed, render the rotation at 17 fps and use background shearing for inbetweening. This is the effect used to roll the horizon in Star Fox, and we've talked about it before.

  • On Super NES: Use mode 2, load the vertical scroll position per 8 pixels into VRAM, and use HDMA to send the horizontal position per scanline.
  • On Genesis: Load the vertical scroll position per 16 pixels into VSRAM, and load the horizontal position per scanline into VRAM.

But because shearing starts to fall apart when the angle exceeds arctan(1/8), you'll need to actually render about 16 steps.


Top
 Profile  
 
PostPosted: Tue Apr 18, 2017 8:39 am 
Offline

Joined: Wed May 19, 2010 6:12 pm
Posts: 2433
Is this how it works:

-Grab initial frame
-Use offset-per-tile and line-scroll to tilt image, while the CPU does scaling
-When it passes "arctan(1/8)" render a full affine-mapped image
-do the same thing again


Top
 Profile  
 
PostPosted: Tue Apr 18, 2017 11:57 am 
Offline

Joined: Wed May 19, 2010 6:12 pm
Posts: 2433
Scratch that idea. I think I'm pretty close at figuring this out. It involves color addition and byte aligned shearing, but with x position toggling between 0 and 1 depending on if the first pixel of a scanline is even or odd.


Top
 Profile  
 
PostPosted: Tue Nov 28, 2017 1:34 pm 
Offline

Joined: Wed May 19, 2010 6:12 pm
Posts: 2433
I think I cracked this magic trick.

The 68000 software renders a 2 pixel wide scaled vertical strip, then makes 36 palette swapped copies of itself. Then it DMAs each vertical strips of pixels from the set of 36 possible strips. The DMA source starting point of each strip takes care of the vertical shearing, where as the choice of which of the 36 strips takes care of both horizontal details and horizontal scaling. Then the last step is using line scrolling for the vertical shearing.

Is this correct?


Top
 Profile  
 
Display posts from previous:  Sort by  
Post new topic Reply to topic  [ 7 posts ] 

All times are UTC - 7 hours


Who is online

Users browsing this forum: Nioreh and 3 guests


You cannot post new topics in this forum
You cannot reply to topics in this forum
You cannot edit your posts in this forum
You cannot delete your posts in this forum
You cannot post attachments in this forum

Search for:
Jump to:  
Powered by phpBB® Forum Software © phpBB Group