Fixing leaky handling of SkImage in SkDeferredCanvas.

Fixing leaky handling of SkImage in SkDeferredCanvas.

Long lived SkImageHeap objects currently accumulate refs indefinitely.
This leads to massive memory leaks in the gpu-accelerated 2D canvas
code path. This CL does not implement a general fix for SkGPipe, but
it resolves the leak in SkDeferredCanvas (currently the only user
of SkGPipe) by resetting the image heap when the deferral queue is
flushed. This change also fixes the accounting of bytes allocated
by referenced images in order to trigger flushing heuristics
appropriately.

BUG=crbug.com/494148

Committed: https://skia.googlesource.com/skia/+/d26c9fa66c45b5a050580772acfbcc1b5271543e

[Justin Novosad, 2015-06-01 21:51:40]

The CQ bit was checked by junov@chromium.org to run a CQ dry run

[Justin Novosad, 2015-06-01 21:51:59]

junov@chromium.org changed reviewers:
+ reed@google.com

[Justin Novosad, 2015-06-01 21:51:59]

PTAL

[commit-bot: I haz the power, 2015-06-01 21:52:14]

Dry run: CQ is trying da patch. Follow status at
 https://chromium-cq-status.appspot.com/patch-status/1145893007/1

[commit-bot: I haz the power, 2015-06-01 23:54:29]

The CQ bit was unchecked by commit-bot@chromium.org

[commit-bot: I haz the power, 2015-06-01 23:54:31]

Dry run: This issue passed the CQ dry run.

[Justin Novosad, 2015-06-02 01:55:48]

FWIW, the more general solution would be to do what SkBitmapHeap does: maintain
a counter of the number of operations in the pipe that reference the image, and
unref the image when the op count reaches zero.  The simpler solution
implemented here seemed more reasonable considering that this needs to be merged
to the M44 release branch and that SkDeferredCanvas is the only use of SkGPipe
in production.

[reed1, 2015-06-02 14:35:42]

Can we document somewhere (skia bug?) how/why the imageheap and bitmapheaps
differ in how they handle purging/etc.?

https://codereview.chromium.org/1145893007/diff/1/include/pipe/SkGPipe.h
File include/pipe/SkGPipe.h (right):

https://codereview.chromium.org/1145893007/diff/1/include/pipe/SkGPipe.h#newc...
include/pipe/SkGPipe.h:90: void resetImageHeap();
1. lets add some dox
2. can the logically be named resetCaches() or purgeCaches() or something more
general/semantic?

https://codereview.chromium.org/1145893007/diff/1/src/pipe/SkGPipePriv.h
File src/pipe/SkGPipePriv.h (right):

https://codereview.chromium.org/1145893007/diff/1/src/pipe/SkGPipePriv.h#newc...
src/pipe/SkGPipePriv.h:225: size_t bytesAllocated() const { return
fBytesAllocated; }
nit: bytesUsed() or bytesCached()?

[Justin Novosad, 2015-06-02 16:50:34]

created skbug for follow-up:
https://code.google.com/p/skia/issues/detail?id=3884

https://codereview.chromium.org/1145893007/diff/1/include/pipe/SkGPipe.h
File include/pipe/SkGPipe.h (right):

https://codereview.chromium.org/1145893007/diff/1/include/pipe/SkGPipe.h#newc...
include/pipe/SkGPipe.h:90: void resetImageHeap();
On 2015/06/02 14:35:42, reed1 wrote:
> 1. lets add some dox
> 2. can the logically be named resetCaches() or purgeCaches() or something more
> general/semantic?

Done.

https://codereview.chromium.org/1145893007/diff/1/src/pipe/SkGPipePriv.h
File src/pipe/SkGPipePriv.h (right):

https://codereview.chromium.org/1145893007/diff/1/src/pipe/SkGPipePriv.h#newc...
src/pipe/SkGPipePriv.h:225: size_t bytesAllocated() const { return
fBytesAllocated; }
On 2015/06/02 14:35:42, reed1 wrote:
> nit: bytesUsed() or bytesCached()?

Done.

[reed1, 2015-06-02 17:07:52]

lgtm

[Justin Novosad, 2015-06-02 18:41:50]

The CQ bit was checked by junov@chromium.org

[commit-bot: I haz the power, 2015-06-02 18:42:15]

CQ is trying da patch. Follow status at
 https://chromium-cq-status.appspot.com/patch-status/1145893007/20001

[commit-bot: I haz the power, 2015-06-02 18:47:52]

Committed patchset #2 (id:20001) as
https://skia.googlesource.com/skia/+/d26c9fa66c45b5a050580772acfbcc1b5271543e

[reed1, 2015-06-04 13:20:39]

https://codereview.chromium.org/1145893007/diff/20001/src/pipe/SkGPipeWrite.cpp
File src/pipe/SkGPipeWrite.cpp (right):

https://codereview.chromium.org/1145893007/diff/20001/src/pipe/SkGPipeWrite.c...
src/pipe/SkGPipeWrite.cpp:1452: printf("Images reff'ed: %d \n", fArray.count());
EEEEK!

[Justin Novosad, 2015-06-04 15:02:14]

On 2015/06/04 13:20:39, reed1 wrote:
>
https://codereview.chromium.org/1145893007/diff/20001/src/pipe/SkGPipeWrite.cpp
> File src/pipe/SkGPipeWrite.cpp (right):
> 
>
https://codereview.chromium.org/1145893007/diff/20001/src/pipe/SkGPipeWrite.c...
> src/pipe/SkGPipeWrite.cpp:1452: printf("Images reff'ed: %d \n",
fArray.count());
> EEEEK!

Sorry 'bout that