On Sun, Sep 30, 2001 at 11:54:01AM -0400, Brian S. Julin wrote:
> On Sun, 30 Sep 2001, Christoph Egger wrote:
> > The conversion should be as fast as possible, as less overhead as
> > possible. I think, the roughly way to go is this:
> >
> > input sublib -> conversion sublib -> output sublib
> >
> > But I don't want to write a conversion sublib for each possible
> > combination of io-targets as input and output...
> >
> >
> > Any ideas?
>
> Do the above, but make it so that you can do:
>
> input sublib -> generic conversion sublib -> output sublib
>
I am assuming that this goes into an intermediate form. This is OK if
color space characteristics are taken into account...
Each format covers a given color space. Some color spaces are able to represent
more colors. If you go through an intermediate conversion you will always
be limited to that color space at best. This could be a bad thing unless the
intermediate is the most complete space.
> ...in cases where a customized sublib has not been written, like a
> stubs or generic-* renderer in GGI. You may want to consider how to
> layer the generic libraries, e.g. in ggi we have:
>
> Layer 0: stubs color
> Layer 1: generic-*
> Layer 2: fbdev and other accels
>
> I would offer a specific suggestion for how to layer gpf overloads, but
> I have been derilict on actually testing/working with it.
>
> --
> Brian
>
--
Curtis Veit email: [EMAIL PROTECTED]
Lineo Inc. Where Open Meets Smart
390 South 400 West phone: 801.426.5001
Lindon, Utah, USA 84042 fax: 801.852.8071
