[prev in list] [next in list] [prev in thread] [next in thread] 

List:       kde-devel
Subject:    Re: [OT] Any plans to keep up with longhorn?
From:       Kuba Ober <kuba () mareimbrium ! org>
Date:       2003-11-06 20:46:08
[Download RAW message or body]

On Sunday 02 November 2003 06:38 am, Maurizio Colucci wrote:
> Alle 12:05 PM, domenica 02 novembre 2003, Stephan Kulow ha scritto:
> > > full resolution of next generation screens simply because no one would
> > > be able to recognize the current icons and fonts at these high dpi
> > > rates.
> >
> > And of course images on web pages will appear quite useless, but icons
> > are surely the most important aspect :)
>
> Well, browsers could include some "intelligent bitmap stretching" algo,
> which takes resolution into account. It would not be too difficult.

Essentially all it needs is a fast processor and some hefty memory. You 
upsample the image, apply a 2D FIR or IIR filter so that aliasing is removed, 
and you downsample it afterwards. Or a similar algorithm. The trick is that 
your filter should offer a sharp, almost brickboard cutoff. If you do it that 
way, the bitmap scaling is seamless, i.e. your bitmap could be scaled by a 
factor pi:1 and look as good as scaled by a factor 2:1. An equivalent of a 
5GHz PIV CPU and probably at least 512Mb of RAM available w/o swapping is 
what it takes on a typical 1280x1024 desktop, given that the code is in C and 
optimized only by the compiler, not by a human being. You can probably scale 
it down to a 1GHz CPU and 128Mb temporary storage given a capable enough 
mathematician who's proficient in numerical methods. Me comes to mind the guy 
who optimized GIMPS software by a factor of 10 (roughly) by writing the FFT 
code in assembly, with CPU's caches in mind (http://www.mersenne.org).

Cheers, Kuba Ober

 
>> Visit http://mail.kde.org/mailman/listinfo/kde-devel#unsub to unsubscribe <<
[prev in list] [next in list] [prev in thread] [next in thread]