[Kde-graphics-devel] Descriptions of new stuff
dan.duley at verizon.net
Wed Dec 15 18:40:21 CET 2004
On Wednesday 15 December 2004 11:14 am, Lubos Lunak wrote:
> On Wednesday 15 of December 2004 16:21, Mosfet wrote:
> Yes, they're those from resize.c .
> 6461ms Qt
> 275ms none (sampling)
> 4708ms fast (box) - quality roughly equals Qt
> 7106ms normal (triangle) - this is pretty sufficient in most cases
> 12153ms best (mitchell)
> It needed only little help to get faster, and I actually don't think the
> algorithm can get faster than this :(.
Those aren't bad numbers, certainly better than I expected them to be.
Nonetheless if Qt takes around 6,000 the Imlib based code with MMX will take
about 1,000 and without around 3000. If your numbers are right, tho, this
does confirm for me that we should be using your code at least for high
> Are you sure the imlib algorithm actually does smoothscale? Enlarged
> pictures in Kuickshow look just as blocky as with Qt. Since you called that
> section "smoothscaling" I thought it could be used instead of the
> ImageMagick stuff, but if only as good as Qt then I'm afraid it won't help
> in this specific case.
Well, it is smoothscaling. I don't know about Kuickshow but I assume it's
using Imlib smoothscaling. You just don't like the quality of Qt and Imlib
smoothscaling ;-) That's okay, I'm not really arguing with you here. It's not
supposed to be high quality it's supposed to be as fast as possible while
producing okay, (but not great), results. Please read my previous mail.
> PS: Were you so long away from Linux that you forgot how to quote mails :)
As for the code, I can post the MImage effect stuff to the list if you want
it, I'd just like to get confirmation that it's okay to post big C++ files.
If your upset my not quoting emails I dunno how people will feel about big
a*s code attachments ;-) Otherwise I can send it to you personally. I don't
have a site.
More information about the Kde-graphics-devel