Re: [eigen] FFT update

[ Thread Index | Date Index | More lists.tuxfamily.org/eigen Archives ]


2010/1/22 Hauke Heibel <hauke.heibel@xxxxxxxxxxxxxx>:
> On Fri, Jan 22, 2010 at 4:06 PM, Mark Borgerding <mark@xxxxxxxxxxxxxx> wrote:
>> Glad to hear it. Thanks.
>> One slight change that will be pushed soon is that the destination argument
>> for fwd, inv functions is now a pointer.  Benoit suggested this and I think
>> it helps make it more obvious which is which.
>
> Oh, and I thought we were done with this discussion. :)

oooooooops.

This "policy" that destination arguments are pointers, was abandoned 3
months ago, sorry.

Our "policy" is now:
 - try to return by value, optimizing with the ReturnByValue class,
wherever possible;
 - otherwise you can just use references.

Benoit


>
>>> First, you might want to consider taking DenseBase<Derived> as your
>>> input parameter(s) as opposed to MatrixBase<Derived>. This allows us
>>> to use the FFT module out of the box with the Array class.
>>>
>>
>> I will look into this.  I need to familiarize myself more with the Matrix
>> inheritence tree anyway.  I feel a bit "Dense" :)
>
> This should be just a 'search and replace' operation but take your time.
>
>>> Second, it would be a nice addition when we could adapt the
>>> ReturnByValue class and thus making inv(...) and fwd(...) returning
>>> their results.
>>
>> I think this would be a great addition.  I don't think this would be
>> confusing:
>> stdVecDst = fft.fwd( stdVecSrc );      <=>       fft.fwd( &stdVecDst,
>> stdVecSrc );
>> eigMatDst = fft.fwd( eigMatSrc);       <=>       fft.fwd(&eigMatSrc,
>> eigMatDst );
>
> This is what I meant when I said it won't be possible with the
> overloads. For std::vector we cannot use the ReturnByValue. The class
> is relying on internal Eigen assignment tricks...
>
> Once again, thanks for all your effort!
>
> - Hauke
>
>
>



Mail converted by MHonArc 2.6.19+ http://listengine.tuxfamily.org/