posix: Avoid static initializers in static/global mutexes

posix: Avoid static initializers in static/global mutexes

This patch removes static initializers related to static and
global mutexes from the final library's machine code when
building on a pthread-capable system.

We use PTHREAD_MUTEX_INITIALIZER to perform POD-style
initialization. You need a line like the following to declare
a global mutex with it:

    SkBaseMutex gMutex = { PTHREAD_MUTEX_INITIALIZER };

We introduce the SK_DECLARE_STATIC_MUTEX and SK_DECLARE_GLOBAL_MUTEX
macros to be able to declare static/global mutexes in the source tree
uniformly.

SkMutex is now defined as a sub-class of SkBaseMutex, with standard
construction/destruction semantics. This is useful if the mutex
object is a member of another C++ class, or allocated dynamically.

We also modify a few places to refer to SkBaseMutex instead of a
SkMutex, where it makes sense. Generally speaking, client code
should hold and use pointers to SkBaseMutex whenever they can
now.

We defined a new built-time macro named SK_USE_POSIX_THREADS
to indicate that we're using a pthread-based SkThread.h
interface. The macro will also be used in future patches
to implement other helper thread synchronization classes.

Finally, we inline the acquire() and release() functions in the
case of Posix to improve performance a bit.

Running: 'bench -repeat 10 -match mutex' on an Android device or
a 2.4GHz Xeon Linux desktop shows the following improvements:

                      Before     After

        Galaxy Nexus    1.64      1.45
        Nexus S         1.47      1.16
        Xoom            1.86      1.66
        Xeon            0.36      0.31

This removes 5 static mutex initializers from the library
Committed: https://code.google.com/p/skia/source/detail?r=3091

[digit, 2011-12-22 17:26:32]

this is the correct version, but previous notes still apply:

1/ Will slightly conflict with http://codereview.appspot.com/5498069/, but no
big deal, can resolve later and re-upload.

2/ Missing Win32 static lazy mutex implementation for now.

[digit, 2012-01-10 10:40:49]

http://codereview.appspot.com/5501066/diff/1/include/core/SkThread_platform.h
File include/core/SkThread_platform.h (right):

http://codereview.appspot.com/5501066/diff/1/include/core/SkThread_platform.h...
include/core/SkThread_platform.h:67: >>>>>>> ab852a0... posix: Avoid static
initializers in static/global mutexes
There is something weird with this patch, I'm going to re-upload a proper
version. Not sure what happened here.

[digit, 2012-01-11 18:02:59]

On 2012/01/10 10:40:49, digit wrote:
> http://codereview.appspot.com/5501066/diff/1/include/core/SkThread_platform.h
> File include/core/SkThread_platform.h (right):
> 
>
http://codereview.appspot.com/5501066/diff/1/include/core/SkThread_platform.h...
> include/core/SkThread_platform.h:67: >>>>>>> ab852a0... posix: Avoid static
> initializers in static/global mutexes
> There is something weird with this patch, I'm going to re-upload a proper
> version. Not sure what happened here.

The second patchset fixes the problem and builds correctly.

A few notes: This defines a new build-time macro SK_USE_POSIX_THREADS to
indicate that we're using the Pthread API. For now, this is only defined when
skia_os != windows.

I have an experimental change that implements lazy-initialized mutexes for
Windows too, but I haven't tested it yet. This will probably appear as a
different patch later.

[digit, 2012-01-19 13:58:10]

ping?

[caryclark1, 2012-01-19 15:18:57]

It looks like the !SK_USE_POSIX_THREADS case has not been compiled or tested.

http://codereview.appspot.com/5501066/diff/3002/include/core/SkThread_platform.h
File include/core/SkThread_platform.h (right):

http://codereview.appspot.com/5501066/diff/3002/include/core/SkThread_platfor...
include/core/SkThread_platform.h:109: class SkMutex : public SkBaseMutex {
This loses the SkNoncopyable base. What about either multiple inheritance or
making SkBaseMutex a member of SkMutex?

http://codereview.appspot.com/5501066/diff/3002/include/core/SkThread_platfor...
include/core/SkThread_platform.h:120: class SkMutex : SkNonCopyable {
This should be SkNoncopyable -- lower case 'c'.

http://codereview.appspot.com/5501066/diff/3002/include/core/SkThread_platfor...
include/core/SkThread_platform.h:134: #define SkBaseMutex  SkMutex
use a typedef instead?

http://codereview.appspot.com/5501066/diff/3002/include/images/SkFlipPixelRef.h
File include/images/SkFlipPixelRef.h (right):

http://codereview.appspot.com/5501066/diff/3002/include/images/SkFlipPixelRef...
include/images/SkFlipPixelRef.h:71: 
remove this formatting change

http://codereview.appspot.com/5501066/diff/3002/src/ports/SkThread_pthread.cpp
File src/ports/SkThread_pthread.cpp (left):

http://codereview.appspot.com/5501066/diff/3002/src/ports/SkThread_pthread.cp...
src/ports/SkThread_pthread.cpp:90: {
why remove this check in the !SK_USE_POSIX_THREADS case?

http://codereview.appspot.com/5501066/diff/3002/src/ports/SkThread_pthread.cp...
src/ports/SkThread_pthread.cpp:120: {
where is this code if SK_USE_POSIX_THREADS is not defined?

[digit, 2012-01-19 15:31:02]

http://codereview.appspot.com/5501066/diff/3002/include/core/SkThread_platform.h
File include/core/SkThread_platform.h (right):

http://codereview.appspot.com/5501066/diff/3002/include/core/SkThread_platfor...
include/core/SkThread_platform.h:109: class SkMutex : public SkBaseMutex {
On 2012/01/19 15:18:57, caryclark1 wrote:
> This loses the SkNoncopyable base. What about either multiple inheritance or
> making SkBaseMutex a member of SkMutex?

Oh right, sorry about that. I will try multiple inheritance, and if this doesn't
work will simply provide a non-implemented private copy constructor.

SkMutex must be a is-a SkBaseMutex because SkAutoMutexAcquire wants to operate
on a SkBaseMutex now.

http://codereview.appspot.com/5501066/diff/3002/include/core/SkThread_platfor...
include/core/SkThread_platform.h:120: class SkMutex : SkNonCopyable {
On 2012/01/19 15:18:57, caryclark1 wrote:
> This should be SkNoncopyable -- lower case 'c'. 

Damn, I need to check the !SK_USE_POSIX_THREADS case.

http://codereview.appspot.com/5501066/diff/3002/include/core/SkThread_platfor...
include/core/SkThread_platform.h:134: #define SkBaseMutex  SkMutex
On 2012/01/19 15:18:57, caryclark1 wrote:
> use a typedef instead?

ok, will do.

http://codereview.appspot.com/5501066/diff/3002/include/images/SkFlipPixelRef.h
File include/images/SkFlipPixelRef.h (right):

http://codereview.appspot.com/5501066/diff/3002/include/images/SkFlipPixelRef...
include/images/SkFlipPixelRef.h:71: 
On 2012/01/19 15:18:57, caryclark1 wrote:
> remove this formatting change

ok

http://codereview.appspot.com/5501066/diff/3002/src/ports/SkThread_pthread.cpp
File src/ports/SkThread_pthread.cpp (left):

http://codereview.appspot.com/5501066/diff/3002/src/ports/SkThread_pthread.cp...
src/ports/SkThread_pthread.cpp:90: {
On 2012/01/19 15:18:57, caryclark1 wrote:
> why remove this check in the !SK_USE_POSIX_THREADS case?

Because this source file is SkThread_pthread.c, i.e. it is only compiled when
SK_USE_POSIX_THREADS is defined. Otherwise, you would be in SkThread_win.cc with
the current build system.

http://codereview.appspot.com/5501066/diff/3002/src/ports/SkThread_pthread.cp...
src/ports/SkThread_pthread.cpp:120: {
On 2012/01/19 15:18:57, caryclark1 wrote:
> where is this code if SK_USE_POSIX_THREADS is not defined?

Again, it is always defined when you build this source.

[digit, 2012-01-19 20:08:40]

The third patch should answer Cary's remarks. Note that I did test the
!SK_USE_POSIX_THREADS case by manually disabling the macro definition in a
custom build to ensure that everything was still building and working properly.

[digit, 2012-01-26 20:47:08]

On 2012/01/19 20:08:40, digit wrote:
> The third patch should answer Cary's remarks. Note that I did test the
> !SK_USE_POSIX_THREADS case by manually disabling the macro definition in a
> custom build to ensure that everything was still building and working
properly.

ping?

[caryclark1, 2012-01-26 20:55:45]

LGTM