Farseer and double precision

Question

I've been searching around the internet for several hours to find some Farseer implementation which uses double precision, no results. My game is going to have big, open world and I need that precision much.

Where to find such a Farseer implementation or any other Box2D implementation accessible in XNA and Monogame?

For now my options are:

Keep on looking
Modify Farseer (seems the best, so far)
Apply offsets to bodies, so their positions will never exceed the float capacity (this might be problematic, because I could want to have different parts of the world loaded simultaneously - like player 1 and player 2).

Thanks in advance.

What exact issue are you having with the current `float` implementation? I'm not sure what they'll say, but gamedev.stackexchange.com might have a better opinion. — Nate, Jan 10 '14 at 18:21
@Nate Problem is, that my world may need big positions, with big precision like `0.0012` and `96512512.8671`. This would not be handled properly (float has 7-8 digits of precision). Thanks for gamedev.stackexchange.com recommendation - I'll try there :) — Michał, Jan 10 '14 at 18:26
That's also going to be an interesting physics engine that can handle those differences in degrees of scale. Regardless of the single/double math, Box2D likes all its dynamic objects to be something like 0.1m to 10.m dimensionally or things don't play nicely... — FuzzyBunnySlippers, Jan 10 '14 at 21:01
@FuzzyBunnySlippers The differences will be only in objects' position. Objects' scale is not meant to be unusual — Michał, Jan 10 '14 at 21:19
Do you actually NEED this? I mean, have you actually generated a world so large you've already exceeded the float.MaxValue = 340,282,300,000,000,000,000,000,000,000,000,000,000? Physics engines have been known to have stability issues with the extremes of floating point numbers, you may very well run into other issues doing this. — craftworkgames, Jan 11 '14 at 12:28
@craftworkgames I MAY need this. I am writing open world engine (nearly end) and I should look at the things more broadly. Maybe I won't use world that big in the first game I'll create with it. But what about second, or third? Why should I limit myself? And responding directly to the float.MaxValue - try to store 340,282,299,299,299,299,299,299,299,299,299,299,299 in float. It is less than Max for sure. It is not about size in floating points, it's about precision. — Michał, Jan 11 '14 at 15:29
@Michał My point is that you are creating a LOT of work for yourself for something you don't need yet, and maybe never will. In my experience it's a bad idea to worry about "limiting yourself" because until you hit those limits you don't need to do anything about them. When / if you do hit the limits, you will be able to sort it out then. http://en.wikipedia.org/wiki/You_aren't_gonna_need_it — craftworkgames, Jan 14 '14 at 12:21
@craftworkgames I talked to my programming prof about that and he said I should make offsets instead of doubles (point 3 in my question). Actually farseer implements that. — Michał, Jan 14 '14 at 20:28
@craftworkgames But: my engine is an open world engine and it is MEANT for huge worlds. As for now I write engine, not game. That's why I cannot agree with YAGNI when we talk about engines. Let's assume you're from Epic, making the Unreal Engine. If your first game on UE is going to be only in-doors, it doesn't mean you shouldn't make out-doors - because it will be needed for further projects and it will be hard to add it, when engine is made for in-doors. Some changes need to be thought of much before - it's matter of architecture. — Michał, Jan 14 '14 at 20:30
@Michał Actually, that's exactly what it means. The first iteration of the Unreal engine wasn't meant for huge worlds. It was meant for in-door first person shooters. I'm certain they didn't start by making an open world physics engine. They made an engine so they could make a game. You're goal is to make games. Engine's are there to support game development. http://en.wikipedia.org/wiki/Unreal_Engine — craftworkgames, Jan 15 '14 at 00:24

score 0 · Accepted Answer · answered Jan 10 '14 at 22:13

Ok, I edited the Farseer. It seems to work (on the samples attached). It is not exactly double precision - I didn't edit some internal functions of XNA - it is not, strictly speaking, legal :) But if you want, you can actually find these references by "(float)" query after ctrl + F, and create your own implementations for given methods (shouldn't be difficult - 3 or 4 short methods). I edited engine versions for XNA, Mono and XBox, and edited XNA samples.

Here's the link, if somebody needs it: https://dl.dropboxusercontent.com/u/75904715/farseer%20double%20precision.rar

Farseer and double precision

1 Answers1