Allocate SkPictureFlat::fScratch lazily.

Allocate SkPictureFlat::fScratch lazily.

Cuts out half to two thirds of allocations of this object.
Since we're already lazily allocating fWriteBuffer, we don't have to add any other code complexity.

BUG=http://skbug.com/1980
R=mtklein@google.com

Committed: http://code.google.com/p/skia/source/detail?r=12992

[tomhudson, 2014-01-09 13:57:13]

PTAL? This was written before your
https://code.google.com/p/skia/issues/detail?id=1980#c3, so is pretty simple,
but it seems like a reasonable first step.

[mtklein, 2014-01-09 14:38:58]

just small stuff

https://codereview.chromium.org/131483005/diff/40001/src/core/SkPictureFlat.h
File src/core/SkPictureFlat.h (right):

https://codereview.chromium.org/131483005/diff/40001/src/core/SkPictureFlat.h...
src/core/SkPictureFlat.h:379: , fScratchSize(kScratchSizeGuess)
For consistency, let's make this fScratchSize(0) then set it to
kScratchSizeGuess just before fScratch = AllocScrach(fScratchSize) in
lazyBufferInit().  It feels spooky to have fScratchSize inconsistent when we
start.

https://codereview.chromium.org/131483005/diff/40001/src/core/SkPictureFlat.h...
src/core/SkPictureFlat.h:382: , fBuffersReady(false) {
Suggestion: fReady and lazyInit() ?

https://codereview.chromium.org/131483005/diff/40001/src/core/SkPictureFlat.h...
src/core/SkPictureFlat.h:518: // be fully set up by the time we get it in the
constructor. We also delay fScratch to
delay -> delay allocating?

https://codereview.chromium.org/131483005/diff/40001/src/core/SkPictureFlat.h...
src/core/SkPictureFlat.h:585: SkASSERT(fScratch);
Suggestion: SkASSERT(fScratch != NULL); as a free reminder it's a pointer?

[tomhudson, 2014-01-09 14:49:31]

https://codereview.chromium.org/131483005/diff/40001/src/core/SkPictureFlat.h
File src/core/SkPictureFlat.h (right):

https://codereview.chromium.org/131483005/diff/40001/src/core/SkPictureFlat.h...
src/core/SkPictureFlat.h:379: , fScratchSize(kScratchSizeGuess)
On 2014/01/09 14:38:58, mtklein wrote:
> For consistency, let's make this fScratchSize(0) then set it to
> kScratchSizeGuess just before fScratch = AllocScrach(fScratchSize) in
> lazyBufferInit().  It feels spooky to have fScratchSize inconsistent when we
> start.

Done.

https://codereview.chromium.org/131483005/diff/40001/src/core/SkPictureFlat.h...
src/core/SkPictureFlat.h:382: , fBuffersReady(false) {
On 2014/01/09 14:38:58, mtklein wrote:
> Suggestion: fReady and lazyInit() ?

Done.

https://codereview.chromium.org/131483005/diff/40001/src/core/SkPictureFlat.h...
src/core/SkPictureFlat.h:518: // be fully set up by the time we get it in the
constructor. We also delay fScratch to
On 2014/01/09 14:38:58, mtklein wrote:
> delay -> delay allocating?

Done.

https://codereview.chromium.org/131483005/diff/40001/src/core/SkPictureFlat.h...
src/core/SkPictureFlat.h:585: SkASSERT(fScratch);
On 2014/01/09 14:38:58, mtklein wrote:
> Suggestion: SkASSERT(fScratch != NULL); as a free reminder it's a pointer?

Done.

[mtklein, 2014-01-09 15:06:08]

lgtm

[commit-bot: I haz the power, 2014-01-09 15:06:17]

CQ is trying da patch. Follow status at
https://skia-tree-status.appspot.com/cq/tomhudson@chromium.org/131483005/80001

[commit-bot: I haz the power, 2014-01-09 16:45:52]

Change committed as 12992