add SkClipStack::clipEmpty() as an optimized way to say clipDevRect(empty, intersect)

add SkClipStack::clipEmpty() as an optimized way to say clipDevRect(empty, intersect)
if the caller knows up-front that it wants the clipstact to become empty.
Committed: https://code.google.com/p/skia/source/detail?r=5127

[robertphillips, 2012-08-15 19:11:31]

LGTM

[TomH, 2012-08-15 19:11:58]

Use SkNEW_PLACEMENT_ARGS?

[TomH, 2012-08-15 19:16:13]

https://codereview.appspot.com/6443132/diff/2001/src/core/SkClipStack.cpp
File src/core/SkClipStack.cpp (right):

https://codereview.appspot.com/6443132/diff/2001/src/core/SkClipStack.cpp#new...
src/core/SkClipStack.cpp:651: Rec* rec = (Rec*) iter.prev();
Not constant time accessor?!

[reed1, 2012-08-16 15:59:30]

http://codereview.appspot.com/6443132/diff/2001/src/core/SkClipStack.cpp
File src/core/SkClipStack.cpp (right):

http://codereview.appspot.com/6443132/diff/2001/src/core/SkClipStack.cpp#newc...
src/core/SkClipStack.cpp:651: Rec* rec = (Rec*) iter.prev();
On 2012/08/15 19:16:14, TomH wrote:
> Not constant time accessor?!

This is lifted from clipDevRect. I will leave that question to Robert to address
across the board.

[robertphillips, 2012-08-22 11:08:51]

http://codereview.appspot.com/6443132/diff/2001/src/core/SkClipStack.cpp
File src/core/SkClipStack.cpp (right):

http://codereview.appspot.com/6443132/diff/2001/src/core/SkClipStack.cpp#newc...
src/core/SkClipStack.cpp:651: Rec* rec = (Rec*) iter.prev();
Fixed in r5149. clipEmpty originated as a copy of clipDevRect which did require
the reverse iteration.

[TomH, 2012-09-05 16:52:15]

Right now, clipDevRect() doesn't shortcut to the optimized version. Would that
be too expensive to detect? Are we just going to rely on callers knowing when
they're eligible for the fast path?

[robertphillips, 2012-09-05 17:02:03]

I don't know where Mike is on it but he was working on
http://codereview.appspot.com/6462067/