The worst kind of crabgrass

I was experimenting with a shader effect that would make grass billboards grow bigger if they were farther from the camera, so lots of faraway terrain could be covered with relatively few grass billboards.

That was the idea anyway. I overshot it a little:

I had the scaling effect set so dramatically that as grass spawned further from the camera, it would gain back more than that distance by growing bigger. So it crept back towards the camera by getting huge.

I say "crept", but it wasn't a very subtle process.

Things didn't get much better from there.

Bugs! In *my* grass?

I'm still preparing the post on grass blades. Part of the slowness is because I'm trying to go through each step incrementally, and show screenshots of the improvement. Because this means commenting out and tweaking parts of the grass code, I've ended up rewriting it a few times. The latest rewrite made the code simpler and saved some math, but it introduced an interesting bug. See if you can spot what's wrong:

If you guessed "those weird streaks shouldn't be there", then consider yourself a programmer, because they should almost definitely probably not be there.

Although they are pretty dramatically placed...

And they look kind of exciting...

I got too close. I guess this is my life now.

Fixing/improving the grass illumination

As part of some mundane tech debt cleanup, I ended up tweaking the code that creates and places grass billboards (also: Playform has 3D grass now. Post on that soon). In doing that, I noticed that when a "tuft" of grass was placed on top of a polygon, the grass code was recalculating the surface normal. That information's already available from the polygon structure, so it's definitely a little silly to be recalculating it when it's sitting right there.

But the worse part is that I was doing it badly. The surface normal was just a cross-product of the polygon's sides. The surface normals stored in the polygon struct itself are averaged vertex normals, which make for smoother lighting. And they're sitting right there.

So long story short I took a little time out and made the code interpolate the vertex normals instead. Here's what the changed normals look like with grass debugging shaders:

(Before)

(After)

And in some non-debug shots:

(Before)

(After)

The grass itself has changed too, so it's hard to see the difference in the foreground. But the slope in the background really shows improvement.