What does 64bit support really mean?

So far, in the betas the transform.position is still limited to a ~7 digit float, so can we access more memory or what?

Are "int" values now 64bit (longs?).

By all means point me to an FAQ with these answers & I'll be off

 

64 bit is referring to how much memory the process can access, and how big the pointers are. It hasn't anything to do with the size of the integers or floats.

http://en.wikipedia.org/wiki/X86-64

If you're wondering what the limit on memory is, the easy answer is "all that you have". The real answer is that it varies by implementation (anywhere from 64GB up to 256TB of RAM).

 

As float is the standard in realtime graphics that goes down even to the hardware, this will very likely never really change. For large worlds (which is the only case where double would give an initial benefit), there's other solutions available out there ;-)

One very nice article about issues with large worlds is:

The Continuous World of Dungeon Siege

Also quite interesting in this context:

About floating point precision and why do we still use it

 

The tldr answer is the 64 bit unity editor can use more than 4gb ram. If your system has more, then you'll be able to take advantage of all your ram.

 

@GregMeach note that you are actually asking a C# question. The size of float and some other built in types is defined in the C# spec. So both float and int are and will always be 32 bit (e.g., see here). Other parts of C# do however change in a 64bit .NET/Mono environment. The probably wont matter to you as long as you dont call unsafe code or perform low-level bit juggling.

 

Thank you all for the excellent replies, I've got some reading to do now. :]

 

But if graphics hardware and the relevant APIs would support it ... all of them ... it would be possible to switch from float to double. As far as I can see, this is never gonna happen but hey, we have eternity, so I might as well almost certainly be wrong ;-)

 

Well, it wouldnt be a "switch" (a change of the public interface of the Transform class) because that would break existing code, more a 64 bit extension of the API. I could see that happen in the not so distant future, 32 bit arithmetic really can be a bottleneck...