Optimizing an effect for OCS+chip mem (finding idle hardware)

[paraj]

Just finishing up the effect for now so I can stop spamming.

I wanted to get back to amiga coding, so I fixed the screen resolution at 256x184 and optimized/simplified the one-frame version. I ended up with roughly 22 free scanlines, which isn't enough for 4 more chunky lines, but should be fine for music/whatever. Otherwise the tunnel table could be fudged more.

Most of the optimizations came from deciding on a one-frame target and dropping all the complications that came from having to run at a variable (>1) frame rate. The blitter is now started by the copper at the start of each frame and finishes before the first scan line is displayed. Meanwhile the CPU works on the data for the next frame (and instead of preparing proper data for the first frame I just fudge the counter to claim 1-frame). As suggested by chb I'm also outputting in y-scrambled format (line 0 line 4 .... line 1 line 5 ...) and moving around using bplmod.

zip with exe+source attached

EDIT: This version seems to be unstable under my "as fast as possible" setup in WinUAE 3.4.0, so as usual it's probably terribly buggy - save your work before running I only tested OCS+chip ram / cycle-exact