Damage as a DIX notion

[Keith Packard]

Michel Dänzer <michel at daenzer.net> writes:

> Right, but I'm questioning if any gains from that vs using and possibly
> tweaking the current damage code (what would the expected gains be?) are
> enough to justify the churn.

The current stack has gotten pretty unwieldy in the normal case, being
largely unchanged since 1990 when I added the whole 'wrapping' notion. I
think we'll see some nice simplifications of code and reductions in the
overhead for small operations.

Now seems like a pretty good time to revisit the 2D rendering
interfaces; we're down to a small handful of implementations that
matter, and we've already identified several places where the current
interface doesn't match what those implementations really want.

> Not really, at least not as currently implemented:
> DamageReportBoundingBox only reports new damage when the extents change,
> but it still records the fully accurate damage region, which can incur
> significant overhead as the region grows rects.

Right. One wonders whether tracking just the damage bounds might not be
a whole lot better idea in all cases.

-- 
-keith
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 800 bytes
Desc: not available
URL: <https://lists.x.org/archives/xorg-devel/attachments/20160926/2b83ee8e/attachment.sig>