In SkGPipe, set a limit on the amount of flattenables cached.

In SkGPipe, set a limit on the amount of flattenables cached.

Also fix a type in gmmain.

[reed1, 2012-07-03 19:40:37]

hooray for a finite cache! I like the simplicity of the reader's logic.

https://codereview.appspot.com/6351061/diff/1/src/pipe/SkGPipeWrite.cpp
File src/pipe/SkGPipeWrite.cpp (right):

https://codereview.appspot.com/6351061/diff/1/src/pipe/SkGPipeWrite.cpp#newco...
src/pipe/SkGPipeWrite.cpp:352: // we remove one, but do not ensure that we
return to below the limit.
can this be encapsulated in a helper function which handles the policy/heuristic
part of this (i.e. which entry should we overwrite)? That will make it easier to
see how to add new policies (LRU, etc.) w/o changing the make part of the code.

[Leon, 2012-07-03 20:45:01]

https://codereview.appspot.com/6351061/diff/1/src/pipe/SkGPipeWrite.cpp
File src/pipe/SkGPipeWrite.cpp (right):

https://codereview.appspot.com/6351061/diff/1/src/pipe/SkGPipeWrite.cpp#newco...
src/pipe/SkGPipeWrite.cpp:352: // we remove one, but do not ensure that we
return to below the limit.
On 2012/07/03 19:40:37, reed1 wrote:
> can this be encapsulated in a helper function which handles the
policy/heuristic
> part of this (i.e. which entry should we overwrite)? That will make it easier
to
> see how to add new policies (LRU, etc.) w/o changing the make part of the
code.

Done.

[reed1, 2012-07-09 13:50:33]

lgtm

down the road, we should be adding benchmarks that explicitly let us measure
changes to the writer's logic for speed.