Provide an option in bench_pictures to count pixel ram.

Provide an option in bench_pictures to count pixel ram.

In SkLruImageCache, provide an option to keep all pixels, and document
the new behavior.

In gm/factory.cpp, set the budget for the Lru image cache to 1, to
retain (basically) the old behavior.

BUG=https://code.google.com/p/skia/issues/detail?id=1010

Committed: https://code.google.com/p/skia/source/detail?r=7972

[reed1, 2013-03-04 20:07:27]

I see an API addition, and a mode that emits (lots?) of debugging printfs. Are
these related?

https://codereview.chromium.org/12378075/diff/1/include/lazy/SkLruImageCache.h
File include/lazy/SkLruImageCache.h (right):

https://codereview.chromium.org/12378075/diff/1/include/lazy/SkLruImageCache....
include/lazy/SkLruImageCache.h:30: 
existing API naming we use for font cache. Maybe we share naming styles?

    static size_t GetFontCacheLimit();
    static size_t SetFontCacheLimit(size_t bytes);
    static size_t GetFontCacheUsed();

[scroggo, 2013-03-04 20:49:29]

> I see an API addition, and a mode that emits (lots?) of debugging printfs. Are
> these related?

Yes. The new mode that prints debugging statements calls the new API. (There is
one line of debugging per skp, reporting how much pixel ram is used.)

https://codereview.chromium.org/12378075/diff/1/include/lazy/SkLruImageCache.h
File include/lazy/SkLruImageCache.h (right):

https://codereview.chromium.org/12378075/diff/1/include/lazy/SkLruImageCache....
include/lazy/SkLruImageCache.h:30: 
On 2013/03/04 20:07:27, reed1 wrote:
> existing API naming we use for font cache. Maybe we share naming styles?
> 
>     static size_t GetFontCacheLimit();
>     static size_t SetFontCacheLimit(size_t bytes);
>     static size_t GetFontCacheUsed();

Done. Currently, we do not have a notion like the font cache of having a single
global/single per thread (though I'm happy to revisit that), so I left them
member functions rather than static. They do have similar names though.

[reed1, 2013-03-04 20:55:02]

lgtm

[scroggo, 2013-03-04 21:16:04]

I just realized that it would have been more clear that these aren't debug
statements per se - they're actually the point of the CL, if I used the
BenchLogger.

[scroggo, 2013-03-04 21:35:06]

Committed patchset #3 manually as r7972 (presubmit successful).