I am not on the devel mailing list, but I took a look over there and saw that complex vectorization support is under way. Am I correct in understanding that this is vectorization of std::complex<> ? The old documentation suggested that there would be a custom eigen complex class (for some technical reason which is beyond my scope). Is this no longer the case?

I also don't understand some issue with copy constructors from real to complex. I have some code (still using an old revision of Eigen ) which tries to initialize a complex vector from a real vector. The underlying types support this - there exists a constructor std::complex<T>(const T& real) where the imaginary part is assumed to be zero. Instead I must write

complexVector c( realVectorExpression.cast<complexType>() );

Is this the right/wrong way to do it?

Cheers
Leo

duetosymmetry wrote:I am not on the devel mailing list, but I took a look over there and saw that complex vectorization support is under way. Am I correct in understanding that this is vectorization of std::complex<> ?

This is vectorizing std::complex<> for Eigen's objects, just like float or double.

The old documentation suggested that there would be a custom eigen complex class (for some technical reason which is beyond my scope). Is this no longer the case?

Indeed, we are not introducing any new complex class anymore.

I also don't understand some issue with copy constructors from real to complex. I have some code (still using an old revision of Eigen ) which tries to initialize a complex vector from a real vector. The underlying types support this - there exists a constructor std::complex<T>(const T& real) where the imaginary part is assumed to be zero. Instead I must write

complexVector c( realVectorExpression.cast<complexType>() );

Is this the right/wrong way to do it?

Yes that's the right way. Allowing to implicitly convert a real matrix to a complex one looks scary.

Sorry, prob. solve, anyone interest could refer to:
viewtopic.php?f=74&t=102168