[gs-code-review] 688177 Eliminate the 64 bit pixel size limit

[Dan Coby]

Ray,

>So how does the 'overprint' compositor work with this?

Overprint does work properly with this scheme.  The overprint
compositor already had logic for 'non separable' devices.  It does
execute slower.  It is possible to add a special overprint compositor
for this device but I am not sure about how large the performance
improvement would be.


>It seems that as pixels are painted over areas that are previously
>painted, the compositing will need to allocate new colors. For example,
>if a pixel was painted with 60%C, 5%M, 30%G and 100%"Varnish" when
>painting with overprint with 100%"Die", then if I understand the
>scheme the new color will be in a different index group with 5
>colorants instead of 4. Is that right?

Yes.  Typical examples of this are the files with a 'Varnish'
overprinted over the entire page.  If the file has more than seven
colorants, then the new Varnish layer creates new component combinations
with the added Varnish component.  This does not happen with seven
or fewer colorants since the component list is started with an entry
that contains all seven colorants.


>Second question, I think we can further optimize bits by having
>indexes for 100% of a colorant that differ from 0 < colorant < 100%.

There are possibilities with this idea.  I did not do it for a couple
of reasons:

1. The current implementation already works well with seven colorants
and I think that this is already a very reasonable upper limit for
the number of colorants that are likely to be in a single pixel.

2. It would make the implementation even more complex.


Dan

-----Original Message-----
From: Ray Johnston [mailto:ray.johnston at artifex.com]
Sent: Friday, September 02, 2005 8:44 AM
To: dan.coby at artifex.com
Cc: Gs-Code-Review
Subject: Re: [gs-code-review] 688177 Eliminate the 64 bit pixel size
limit


Dan,

Thanks for the quick work on implementing this idea!

Sorry for not responding sooner, but I've been sort of swamped with
making sure I had what I needed from Ralph before he vacated.

I had a couple of comments/questions:

So how does the 'overprint' compositor work with this?

It seems that as pixels are painted over areas that are previously
painted, the compositing will need to allocate new colors. For example,
if a pixel was painted with 60%C, 5%M, 30%G and 100%"Varnish" when
painting with overprint with 100%"Die", then if I understand the
scheme the new color will be in a different index group with 5
colorants instead of 4. Is that right?

Second question, I think we can further optimize bits by having
indexes for 100% of a colorant that differ from 0 < colorant < 100%.
This would expand the colorants available since the index group
would define 100% of that colorant, leaving the bits for the
colorants that are no 100%. When painting with a range of a color,
this would result in one index entry for the pixels that had 100%
of that color and a different index entry for pixels that had less
than 100%, but more than 0%, but since we often see 'real world'
layers that are 100%, avoiding the need to reduce the bits per
pixel seems to me to be worthwhile. If we do this, then we may
want to handle CMYK without this since we will usually see each
of these colorants at 100% as well as < 100% levels.

Regards,
Ray
_____________________________________________________________________