DM: make GPU tasks multithreaded again. Big refactor.

DM: make GPU tasks multithreaded again.  Big refactor.

The main meat of things is in SkThreadPool.  We can now give SkThreadPool a
type for each thread to create and destroy on its local stack.  It's TLS
without going through SkTLS.

I've split the DM tasks into CpuTasks that run on threads with no TLS, and
GpuTasks that run on threads with a thread local GrContextFactory.

The old CpuTask and GpuTask have been renamed to CpuGMTask and GpuGMTask.

Upshot: default run of out/Debug/dm goes from ~45 seconds to ~20 seconds.

BUG=skia:

Committed: http://code.google.com/p/skia/source/detail?r=13632

[bsalomon, 2014-02-28 16:40:49]

https://codereview.chromium.org/179233005/diff/40001/dm/DM.cpp
File dm/DM.cpp (right):

https://codereview.chromium.org/179233005/diff/40001/dm/DM.cpp#newcode31
dm/DM.cpp:31: DEFINE_int32(gpuThreads, 4, "Threads for GPU work.");
I wonder if we'll want a different default for Android (1)

https://codereview.chromium.org/179233005/diff/40001/dm/DMTask.h
File dm/DMTask.h (right):

https://codereview.chromium.org/179233005/diff/40001/dm/DMTask.h#newcode54
dm/DMTask.h:54: class GpuTask : public Task, public
SkTRunnable<GrContextFactory> {
AFAIK we have a policy against multiple inheritance.

[mtklein, 2014-02-28 17:18:13]

https://codereview.chromium.org/179233005/diff/40001/dm/DM.cpp
File dm/DM.cpp (right):

https://codereview.chromium.org/179233005/diff/40001/dm/DM.cpp#newcode31
dm/DM.cpp:31: DEFINE_int32(gpuThreads, 4, "Threads for GPU work.");
On 2014/02/28 16:40:49, bsalomon wrote:
> I wonder if we'll want a different default for Android (1)
Good thought.  I'll default this back to 1 for now until we have a better idea.

https://codereview.chromium.org/179233005/diff/40001/dm/DMTask.h
File dm/DMTask.h (right):

https://codereview.chromium.org/179233005/diff/40001/dm/DMTask.h#newcode54
dm/DMTask.h:54: class GpuTask : public Task, public
SkTRunnable<GrContextFactory> {
On 2014/02/28 16:40:49, bsalomon wrote:
> AFAIK we have a policy against multiple inheritance.

If we do, this should be an exemption.  There's one concrete base class here,
and one interface.  As long as you have at most one base class that's got
non-pure-virtual methods and members, you're not going to have any tricky
inheritance problems.

C.f.
http://google-styleguide.googlecode.com/svn/trunk/cppguide.xml#Multiple_Inher...

[bsalomon, 2014-02-28 18:45:37]

https://codereview.chromium.org/179233005/diff/40001/dm/DMTask.h
File dm/DMTask.h (right):

https://codereview.chromium.org/179233005/diff/40001/dm/DMTask.h#newcode54
dm/DMTask.h:54: class GpuTask : public Task, public
SkTRunnable<GrContextFactory> {
On 2014/02/28 17:18:13, mtklein wrote:
> On 2014/02/28 16:40:49, bsalomon wrote:
> > AFAIK we have a policy against multiple inheritance.
> 
> If we do, this should be an exemption.  There's one concrete base class here,
> and one interface.  As long as you have at most one base class that's got
> non-pure-virtual methods and members, you're not going to have any tricky
> inheritance problems.
> 
> C.f.
>
http://google-styleguide.googlecode.com/svn/trunk/cppguide.xml#Multiple_Inher...

Forwarding your appeal to the supreme court for judicial review.

[reed1, 2014-02-28 19:25:40]

I do not feel strongly about the single/multiple inheritance pattern for
tool-only code like this.

[bsalomon, 2014-02-28 19:28:25]

On 2014/02/28 19:25:40, reed1 wrote:
> I do not feel strongly about the single/multiple inheritance pattern for
> tool-only code like this.

lgtm

[mtklein, 2014-02-28 19:44:12]

On 2014/02/28 19:25:40, reed1 wrote:
> I do not feel strongly about the single/multiple inheritance pattern for
> tool-only code like this.

One day I might try to get you on on board with single-class/multiple-interface
in non-tool-only code too.  It's what Java enforces with
single-extends/multiple-implements, and all the cool C++ kids do this too.  :)

[mtklein, 2014-02-28 19:44:53]

The CQ bit was checked by mtklein@google.com

[commit-bot: I haz the power, 2014-02-28 19:45:00]

CQ is trying da patch. Follow status at
https://skia-tree-status.appspot.com/cq/mtklein@chromium.org/179233005/50001

[commit-bot: I haz the power, 2014-02-28 20:31:59]

Change committed as 13632