> > The new code is a bit worse but faster. The problem is in the
> > rounding.  I can potentially scale up the outlines temporarily to
> > deal with this problem.
>
> I really wonder whether it makes sense to speed up the exact bbox
> computation if this causes a loss of precision.
>

The original motivation was to kill the last use of FT_SqrtFixed. Then I
read old emails that made perplexing assertion that bisection is too slow
to converge while never questioning iterations in the sqrt algorithm. So I
saw a challenge and produced a fast proof-of-concept code without sqrt. It
is a matter of improving accuracy now. The old algorithm uses upscaling to
achieve its precision and so can I.
_______________________________________________
Freetype-devel mailing list
[email protected]
https://lists.nongnu.org/mailman/listinfo/freetype-devel

Reply via email to