[OpenJDK 2D-Dev] RFC: 7019861 + workaround + minor optimization

[Denis Lila]

Hi Jim.


> line 52: and => or?

No, because our ranges are half open, so if lastCrossing == end,
that means the line ends just before end and hence doesn't cross it.

> line 355: Make that "16" a constant?

Sure.

> Is this blocking anyone until they find the Hotspot problem?

Not to my knowledge, but it has been very annoying when running
some of my own tests.

> line 535: Would it work to use pminY and pmaxY? Will there be any
> active edges to iterate outside of that range?

I'm not sure I understand. I do use pminY and pmaxY - just their
subpixel versions. If we used pminY and pmaxY instead of subpix_pminY
and subpix_pmaxY that would just mean we'd have to compute the subpixel
versions in _endRendering (or ScanlineIterator.<init>).

I've updated the webrev.

Thank you,
Denis.

----- Original Message -----
> Hi Denis,
> 
> Renderer.java:
> 

> 

> 

> 
> Stroker.java:
> 

> 
> ...jim
> 
> On 3/4/2011 1:04 PM, Denis Lila wrote:
> > Hi.
> >
> > In my last push to pisces I introduced a regression.
> > The scanline iterator doesn't iterate over the last
> > scanline. This:
> > http://icedtea.classpath.org/~dlila/webrevs/7019861_skipped_last_scanline/webrev/
> > fixes it.
> >
> > The webrev also includes a workaround for a hotspot
> > bug and a minor optimization in Renderer.java (eliminating
> > a range check on edgeBucketCounts).
> >
> > Is it ok to push?
> >
> > Thank you,
> > Denis.