Partially restore small-T optimization for very small (empty) T.

Partially restore small-T optimization for very small (empty) T.

This is particularly helpful for SkRecord::replace<NoOp>, which now doesn't go
off and allocate a pointless byte.

BUG=skia:2378
Committed: http://code.google.com/p/skia/source/detail?r=14622

[bungeman-skia, 2014-05-07 14:35:23]

https://codereview.chromium.org/269543025/diff/1/src/record/SkRecord.h
File src/record/SkRecord.h (right):

https://codereview.chromium.org/269543025/diff/1/src/record/SkRecord.h#newcod...
src/record/SkRecord.h:173: }
I understand what you're doing here, but eck. Can we just make an IsEmpty trait?

SkTLogic.h:
class SkTIsEmpty<T> {
    struct test : public T { char unused; };
public:
    static const bool value = sizeof(test) == sizeof(char);
};

Here:
template <typename T> SK_WHEN<SkTIsEmpty<T>, T*> allocCommand() {
    return NULL;
}
template <typename T> SK_WHEN<!SkTIsEmpty<T>, T*> allocCommand() {
    return this->alloc<T>();
}

You're already writing/using a template DSL here, this seems more in keeping
with the rest of the implementation.

Come to think of it, we should probably put out traits in a namespace or
something. And move SkTIsConst there also.

[mtklein, 2014-05-07 15:27:42]

https://codereview.chromium.org/269543025/diff/1/src/record/SkRecord.h
File src/record/SkRecord.h (right):

https://codereview.chromium.org/269543025/diff/1/src/record/SkRecord.h#newcod...
src/record/SkRecord.h:173: }
Done.  Leaving rearranging / renaming the traits for later.

[bungeman-skia, 2014-05-07 15:33:42]

lgtm with one comment nit.

https://codereview.chromium.org/269543025/diff/20001/src/utils/SkTLogic.h
File src/utils/SkTLogic.h (right):

https://codereview.chromium.org/269543025/diff/20001/src/utils/SkTLogic.h#new...
src/utils/SkTLogic.h:33: /** Is T an empty class? */
Since the other ones also say something like it, 

/** Pre-C++11 version of std::is_empty<T>. */

[mtklein, 2014-05-07 15:36:44]

https://codereview.chromium.org/269543025/diff/20001/src/utils/SkTLogic.h
File src/utils/SkTLogic.h (right):

https://codereview.chromium.org/269543025/diff/20001/src/utils/SkTLogic.h#new...
src/utils/SkTLogic.h:33: /** Is T an empty class? */
On 2014/05/07 15:33:42, bungeman1 wrote:
> Since the other ones also say something like it, 
> 
> /** Pre-C++11 version of std::is_empty<T>. */

Done.

[mtklein, 2014-05-07 17:52:54]

The CQ bit was checked by mtklein@google.com

[commit-bot: I haz the power, 2014-05-07 17:53:31]

CQ is trying da patch. Follow status at
 https://skia-tree-status.appspot.com/cq/mtklein@chromium.org/269543025/60001

[commit-bot: I haz the power, 2014-05-07 18:02:03]

Change committed as 14622