Re: [Gnash-commit] gnash ChangeLog server/parser/shape_character_d...

[Sandro Santilli]

On Tue, Aug 28, 2007 at 10:03:19AM +0200, Udo Giacomozzi wrote:

> Edges could be stored in integer TWIPS or perhaps it should be even so
> by definition (ie. align any edge to TWIPS).
> 
> FYI, don't confuse the shape TWIPS scale with the TWIPS scale of the
> stage: a shape may be scaled down much and thus has very precise PIXEL
> coordinates - it's just the *definition* that has a limited accuracy.
> 
> AGG (and most probably tri. renderers as well) have no problems with
> integer edges.
> 
> A different question is if is CPU-expensive to convert integer to
> floats (ie. if it adds additional processor cycles when multiplicating
> an integer with a float)...?

It surely does so, but the same overhead would be in converting int values
to floats when parsing, wouldn't it ? Seems another case of instance vs. defs
rendering (defs are in integer space, instances might be in float, considering
matrix transforms). Also, if no invalidated bounds, no multiple transforms
required..

--strk;