Add a skia method to perform an atomic add.

Add a skia method to perform an atomic add.

Complements sk_atomic_inc for when you want to increase by more than one.

This time, use the correct atomic add function on Windows.

[bungeman, 2012-07-16 14:56:57]

Mostly naming, using 'inc' instead of 'value' consistently.

https://codereview.appspot.com/6399050/diff/1/include/core/SkThread_platform.h
File include/core/SkThread_platform.h (right):

https://codereview.appspot.com/6399050/diff/1/include/core/SkThread_platform....
include/core/SkThread_platform.h:62: #define sk_atomic_add(addr, value) 
android_atomic_add(value, addr)
Maybe 'inc' instead of 'value'?

https://codereview.appspot.com/6399050/diff/1/include/core/SkThread_platform....
include/core/SkThread_platform.h:101: specified by the address (in a thread-safe
manner), and returns the
nit: spaces to line up like the others.

https://codereview.appspot.com/6399050/diff/1/include/core/SkThread_platform....
include/core/SkThread_platform.h:106: SK_API int32_t sk_atomic_add(int32_t*
addr, int32_t value);
Like the others, 'inc' or 'addend' instead of 'value'

https://codereview.appspot.com/6399050/diff/1/src/ports/SkThread_win.cpp
File src/ports/SkThread_win.cpp (right):

https://codereview.appspot.com/6399050/diff/1/src/ports/SkThread_win.cpp#newc...
src/ports/SkThread_win.cpp:29: return
_InterlockedExchangeAdd(reinterpret_cast<LONG*>(addr), value);
I suppose value is the name of the formal parameter, but it seems confusing to
call this 'value' since it isn't the value we want to increment (it's the amount
of the increment). I guess 'value' is non-distinguishing. It would probably be
clearer just to have a line break here,

return _InterlockedExchangeAdd(reinterpret_cast<LONG*>(addr),
                               static_cast<LONG>(inc));

[Leon, 2012-07-16 15:07:54]

https://codereview.appspot.com/6399050/diff/1/include/core/SkThread_platform.h
File include/core/SkThread_platform.h (right):

https://codereview.appspot.com/6399050/diff/1/include/core/SkThread_platform....
include/core/SkThread_platform.h:62: #define sk_atomic_add(addr, value) 
android_atomic_add(value, addr)
On 2012/07/16 14:56:58, bungeman wrote:
> Maybe 'inc' instead of 'value'?

Done.

https://codereview.appspot.com/6399050/diff/1/include/core/SkThread_platform....
include/core/SkThread_platform.h:101: specified by the address (in a thread-safe
manner), and returns the
On 2012/07/16 14:56:58, bungeman wrote:
> nit: spaces to line up like the others.

Done.

https://codereview.appspot.com/6399050/diff/1/include/core/SkThread_platform....
include/core/SkThread_platform.h:106: SK_API int32_t sk_atomic_add(int32_t*
addr, int32_t value);
On 2012/07/16 14:56:58, bungeman wrote:
> Like the others, 'inc' or 'addend' instead of 'value'

Done.

https://codereview.appspot.com/6399050/diff/1/src/ports/SkThread_win.cpp
File src/ports/SkThread_win.cpp (right):

https://codereview.appspot.com/6399050/diff/1/src/ports/SkThread_win.cpp#newc...
src/ports/SkThread_win.cpp:29: return
_InterlockedExchangeAdd(reinterpret_cast<LONG*>(addr), value);
On 2012/07/16 14:56:58, bungeman wrote:
> I suppose value is the name of the formal parameter, but it seems confusing to
> call this 'value' since it isn't the value we want to increment (it's the
amount
> of the increment). I guess 'value' is non-distinguishing. It would probably be
> clearer just to have a line break here,
> 
> return _InterlockedExchangeAdd(reinterpret_cast<LONG*>(addr),
>                                static_cast<LONG>(inc));

Done.

[bungeman, 2012-07-16 15:23:38]

LGTM with one last nit.

https://codereview.appspot.com/6399050/diff/2002/src/ports/SkThread_pthread.cpp
File src/ports/SkThread_pthread.cpp (right):

https://codereview.appspot.com/6399050/diff/2002/src/ports/SkThread_pthread.c...
src/ports/SkThread_pthread.cpp:38: int32_t sk_atomic_add(int32_t* addr, int32_t
value)
nit: one last 'inc', just to be consistent.

[Leon, 2012-07-16 15:33:11]

https://codereview.appspot.com/6399050/diff/2002/src/ports/SkThread_pthread.cpp
File src/ports/SkThread_pthread.cpp (right):

https://codereview.appspot.com/6399050/diff/2002/src/ports/SkThread_pthread.c...
src/ports/SkThread_pthread.cpp:38: int32_t sk_atomic_add(int32_t* addr, int32_t
value)
On 2012/07/16 15:23:39, bungeman wrote:
> nit: one last 'inc', just to be consistent.

D'oh! I thought I'd gotten all of them.

Done.

[Leon, 2012-07-16 16:54:31]

Moved to https://codereview.appspot.com/6407048/ and committed at
https://code.google.com/p/skia/source/detail?r=4623