new .net floating point types for x64?

views:

127

answers:

+1 Q:

new .net floating point types for x64?

Hello!

Does anybody know if there are any? I would be interested in a .net floating point type with the range of a double (or greater) combined with the precision of a decimal (or greater). Are there any planned? Anybody interested in speculating as to the usefulness or practicality of a new high-precision, high-range floating point type on x64?

+1 A:

x64 still has the same core .NET libraries as x86 (although there may be some internal changes, especially in the CLI itself). So no: x64 doesn't get anything extra.

There is BigInteger in 4.0, though.

Marc Gravell 2009-04-30 09:22:38

Thanks for the answer. What's your opinion though: do you think that a new high-precision, high-range floating point type would be both feasible and useful for those working on the x64 platform on .net?

Shunyata Kharg 2009-04-30 09:28:25

@Shunyata: x86 floating point supports wider types that .NET (or VC++) exposes as it is. More likely (hopefully in .NET 4) is some BigFloat for arbitrary precision.

Richard 2009-04-30 09:38:10

IMO, that is quite a specialized requirement... it would be *nice* if they were there, but we've done OK without them so far. I expect you can pick up some arbitrary precision libraries online.

Marc Gravell 2009-04-30 09:56:41

@Richard: Thanks for the info. I would welcome some sort of BigFloat in x86, I just thought that in x64 the performance overhead would be lower for such a beast.

Shunyata Kharg 2009-04-30 09:57:01

To have better performance, it would need to use new CLI primitives in the implementation... AFAIK, that isn't happening.

Marc Gravell 2009-04-30 10:35:31

@Marc: Ok, I see, and thanks for the info. Am I correct in thinking, then, that such a BigFloat type could perform better on x64 than x86 given the necessary CLI changes?

Shunyata Kharg 2009-04-30 11:10:03

That isn't a simple question... it would depend entirely on the unmanaged part of the system (not my specialist subject). I guess potentially it *could*, **if** there are CPU instructions to handle that data natively.

Marc Gravell 2009-04-30 11:57:46

ansaurus

tags:

views:

answers:

new .net floating point types for x64?

related questions