Hi!

For various reasons I need Eigen to handle long long int values. However, support seems to be rather poor at the moment. The NumTraits are defined, but the math functions do not handle these types. As a result it is not possible to do a simple dot-product with Eigen when dealing with Eigen vectors of long long ints. Is there some particular reason why the appropiate implementation in MathFunctions has been left out for long long int? I tried eigen 2.0.5 and the dev-version with no success.

Thank a lot for any help here.

Sebastian

Argh, somehow my post got lost, rewriting...


no reason at all, just that apparently no-one needed it so far.

Is there any chance of SSE vectorization for 64 bit integers? If not, then indeed all what's needed is to add the missing bits to Eigen/src/Core/MathFunctions.h

If yes, perhaps that needs to go in a new module. perhaps join discussion with the current ML thread on short integers.

I dont get this software (forum stuff).

Anyway, a short implementation would be nice as well. However, I have no clue about SSE stuff. Shouldn't a simple copy int code to long int code at apropiate places do the job?

Sebastian

It won't get you vectorization. Getting vectorization to work requires adapting arch/SSE/PacketMath.h and it's not trivial(Again, I don't even know if it's possible for 64 bit integers).