Re: [eigen] vectorization of complex

[ Thread Index | Date Index | More Archives ]

On Tue, Jul 20, 2010 at 4:06 PM, Joel Falcou <joel.falcou@xxxxxx> wrote:
> On 20/07/10 16:01, Gael Guennebaud wrote:
>> On Tue, Jul 20, 2010 at 2:38 PM, FMDSPAM<fmdspam@xxxxxxxxx>  wrote:
>>>  Am 20.07.2010 11:47, schrieb Gael Guennebaud:
>>>> [...] but 3 swizzling and one "bitwise or" operation to negate
>>>> some coefficients. This make it far to be optimal. In particular
>>>> swizzling is extremely costly. Moreover all these "complex" operations
>>>> requires many additional registers to store the temporaries. [...]
>>> In another threat matlab was "blamed" to use separate matrices for real
>>> and
>>> imaginary parts instead of one matrix of complex coeff.
>>> How do you think, this would compete?
>> Good point, with such an approach it is straightforward to support
>> complex-complex and mixing types on top of a real-real matrix product
>> with good performance I guess....
> NT2 use 2x array for complex matrix (and for any non scalar type in fact)
> and it significantly simplify the vectorization process. What was the blames
> for doing so ?

Yes it clearly simplify the vectorization of products but for the rest
I'm not sure (but who cares about the rest ?). It is also to be
compatible with BLAS/LAPACK.

> Also note all LAPACK thingy use this format.

hm, BLAS/LAPACK use the fortran COMPLEX type with interleaved real and
imaginary part. Did I miss something?



Mail converted by MHonArc 2.6.19+