The link you provided says eigen does not support multidimentional fft. For our case, i guess, we need the one.<br><br><div class="gmail_quote">On Mon, Dec 14, 2009 at 10:52 PM, Boudewijn Rempt <span dir="ltr"><<a href="mailto:boud@valdyas.org">boud@valdyas.org</a>></span> wrote:<br>
<blockquote class="gmail_quote" style="border-left: 1px solid rgb(204, 204, 204); margin: 0pt 0pt 0pt 0.8ex; padding-left: 1ex;"><div class="im">On Monday 14 December 2009, <a href="mailto:schumifer@hotmail.com">schumifer@hotmail.com</a> wrote:<br>
<br>
> > Dmitry Kazakov wrote:<br>
> > I might know the reason why it works wrong.<br>
> > Doesn't it look like "border effect"? If so this is surely due to<br>
> > threading of filters. I'm working on this. To prove the fault of<br>
> > threading try limit thread number to 1.<br>
> ><br>
> > Dmitry Kazakov wrote:<br>
> > antiquark, haven't you tried to use fft for that? Does eigen2 support<br>
> > fft? (just an idea)<br>
><br>
> What needs to be done, is port all the convolution code in Krita to an FFT<br>
> system. This would bring speed-ups to a number of filters in Krita. For<br>
> this I was thinking of the FFTW library. However, this requires a<br>
> significant amount of work.<br>
<br>
</div>Let's take this to the mailing list. Eigen is working on fft --<br>
<a href="http://eigen.tuxfamily.org/index.php?title=EigenFFT" target="_blank">http://eigen.tuxfamily.org/index.php?title=EigenFFT</a>, but I won't t block an<br>
additional dependency if it brings speedups, and fftw seems well-maintained.<br>
<br>
Which other filters than convolution would be sped up as well?<br>
<div class="im">--<br>
Boudewijn Rempt | <a href="http://www.valdyas.org" target="_blank">http://www.valdyas.org</a><br>
_______________________________________________<br>
kimageshop mailing list<br>
</div><a href="mailto:kimageshop@kde.org">kimageshop@kde.org</a><br>
<a href="https://mail.kde.org/mailman/listinfo/kimageshop" target="_blank">https://mail.kde.org/mailman/listinfo/kimageshop</a><br>
</blockquote></div><br><br clear="all"><br>-- <br>Dmitry Kazakov<br>