Re: [Devel] Problems with bbox code and cubic bezier curves

[Toby J Sargeant]

On Mon, Apr 23, 2001 at 08:20:24AM -0400, Tom Kacvinsky wrote:
> Right  -- there may be multiple divisions by 2 when subdividing a curve.  
> Ugh...
> Hadn't stopped to think about that one!

I like your argument that the bbox calculation should be pixel correct,
though. Maybe it would be useful to have a function for returning a
pixel correct bounding box, in pixel units, and a glyph bounding box
in glyph units that is ``mathematically correct''.

Given some fast isqrt code, computing the solutions to the derivative
polynomial would likely be faster than doing the curve subdivision,
especially in cases where cpu speed exceeds memory bandwidth by a large
amount.

Toby.