Repeat after me: "Premature optimization is the root of all evil"

Marlow on 2007/04/03 12:02 of Bus Factor == 1

Day 2, Monday, was a working day, so I had no real time to look at this until the evening. During my drive home I figured that I could do with a collision grid to speed up collision detection - anything to reduce the number of collision tests.

So, when I finally got time to myself in the evening, I sat down and wrote a nice collision grid class. I even unit tested it to make sure that it did what it said on the tin. And then I fired it up to see what would happen. And guess what. It was slower! I took some timings and found that the overhead of maintaining the grid was just too much compared with iterating through a list and comparing a few thousand rectangles.

So day 2 ended with a whimper.

I feel a bit better about it now, as this morning I got up early and did a quick and rather more successful spike before heading off to work, ending up with a reasonably competent implementation of doors.

(log in to comment)

Comments

simono on 2007/04/03 12:27:

agree+1 if something already (somewhat) works in my game, i force myself to NOT tinker with that part again, unless it brings gameplay-value.

alex on 2007/04/03 15:00:

With my last pyweek entry (nelly) I used a uniform grid, but it took quite a bit of tweaking of the number of rows and columns for it to work efficiently; I ended up using wide columns but narrow rows (perhaps because this suited the general shape of many of the placeables). I say: don't throw it out yet, you may find use for it later down the track.

Marlow on 2007/04/03 21:23:

I've definitely kept the code. It certainly reduces the number of collision tests that are necessary, but rect/rect tests aren't all that expensive and that's all I'm going for.

You'd think I'd know better by now though.