Monday, September 5, 2016

Backgrounds and debugging

One of the hard things. Ok, so far the ONLY hard thing about NES development is how the background tile maps work.  In theory, they're pretty simple -- a big array of memory where you write a tile number, and the PPU (the video chip) draws that tile at that position.

But there's so many little gotchas:


  • The nes addresses 4 different screens' worth of data, but by default only has ram for 2 of them. So unless you provide more RAM on the cartridge, you set up (via actual wiring on the cartridge!) mirroring where data is duplicated between screens. 
  • The screen nicely pans and wraps between screens'-worth of background data. Except that there's only 240 tiles high, so you have to do math instead of letting bytes just wrap around
  • Setting color palettes for tiles is painful -- you do it in a separate block of memory, mapped in a funky way where you map 4 tiles to 2 bits of a particular byte.  Which means you have to do odd math to work out how to change colors of backgrounds.
  • Mid-screen scroll changes are weird.  To do a header bar, many games will switch scroll positions after the header (since there aren't layered backgrounds). But switching scrolling mid-frame requires more weird math, because you can't just write to the scroll registers like normal -- you have to do weird gymnastics with the various PPU registers instead.
It's not horrible, but it's driving me a little crazy. Really, I just wish I had better debugging tools. I got really used to Stella (the Atari emulator)'s quirky debugger, and realized how awesome it was. The nintendo debuggers I've used just aren't nearly as usable, at least with my current level of experience with them.

This is a debugging diagram of the four screens of data. I'm mirroring vertically, so the top and bottom are cloned.
The left half is my game background. The right half is the header bar.
Each frame, I start by telling the NES to set the scroll to where the header bar is.
Once it's drawn, I reset the scrolling to point to the correct position on the left half.


I finally broke down today and wrote routines for my game that let me do some primitive on-screen debug messages, displaying decimal versions of values using the player score headers. That will help, I hope.

1 comment:

sverx said...

Crazy! Compared to this, the SEGA Master System's "Video Display Processor" is a breeze to drive... (well, I admit the vertical scroll has the same screen wrap problem, as the map it's only 224 pixel tall...)

NES Anguna

Well, I had a little bit of time still, while Frankengraphics is finishing up her game Project Blue, to have a little downtime on Halcyon, s...