Move the thread local functions to the Thread class in runtime/platform

Move the thread local functions to the Thread class in runtime/platform

This gets rid of all the platform specific isolate_*.h and isolate_*.cc files.

R=asiva@google.com

BUG=
TEST=

Committed: https://code.google.com/p/dart/source/detail?r=3982

[siva, 2012-02-06 19:29:37]

LGTM with some comments.

http://codereview.chromium.org/9328042/diff/4001/runtime/platform/thread.h
File runtime/platform/thread.h (right):

http://codereview.chromium.org/9328042/diff/4001/runtime/platform/thread.h#ne...
runtime/platform/thread.h:32: static ThreadLocalKey kInvalidThreadLocal;
data field is clubbed in with functions, shouldn't this be listed first?

Also should we call the field kUnsetThreadlocalKey?
or if you want to stick with Invalid make it
kInvalidThreadlocalKey so that it is obvious that the key is invalid not the
thread local value.

http://codereview.chromium.org/9328042/diff/4001/runtime/platform/thread_linu...
File runtime/platform/thread_linux.cc (right):

http://codereview.chromium.org/9328042/diff/4001/runtime/platform/thread_linu...
runtime/platform/thread_linux.cc:111: int result = pthread_key_create(&key,
NULL);
ASSERT(key != kInvalidThreadLocal);

http://codereview.chromium.org/9328042/diff/4001/runtime/platform/thread_linu...
runtime/platform/thread_linux.cc:127: }
Can this be defined in the header file so that it gets inlined the way it is
now.

http://codereview.chromium.org/9328042/diff/4001/runtime/platform/thread_maco...
File runtime/platform/thread_macos.cc (right):

http://codereview.chromium.org/9328042/diff/4001/runtime/platform/thread_maco...
runtime/platform/thread_macos.cc:95: int result = pthread_key_create(&key,
NULL);
ASSERT(key != kInvalidThreadLocal);

http://codereview.chromium.org/9328042/diff/4001/runtime/platform/thread_maco...
runtime/platform/thread_macos.cc:111: }
Inline this method in the header so that there is no perf degradation.

http://codereview.chromium.org/9328042/diff/4001/runtime/vm/isolate_linux.cc
File runtime/vm/isolate_linux.cc (left):

http://codereview.chromium.org/9328042/diff/4001/runtime/vm/isolate_linux.cc#...
runtime/vm/isolate_linux.cc:22: // isolate specific information a single pthread
key is sufficient.
This comment has been lost now.

http://codereview.chromium.org/9328042/diff/4001/runtime/vm/isolate_macos.cc
File runtime/vm/isolate_macos.cc (left):

http://codereview.chromium.org/9328042/diff/4001/runtime/vm/isolate_macos.cc#...
runtime/vm/isolate_macos.cc:22: // isolate specific information a single pthread
key is sufficient.
Ditto, comment lost.

[Søren Gjesse, 2012-02-07 09:13:59]

Thanks for the review.

I haved checked the generated code and Thread::GetThreadLocal is inlined now.

http://codereview.chromium.org/9328042/diff/4001/runtime/platform/thread.h
File runtime/platform/thread.h (right):

http://codereview.chromium.org/9328042/diff/4001/runtime/platform/thread.h#ne...
runtime/platform/thread.h:32: static ThreadLocalKey kInvalidThreadLocal;
On 2012/02/06 19:29:37, asiva wrote:
> data field is clubbed in with functions, shouldn't this be listed first?
> 
> Also should we call the field kUnsetThreadlocalKey?
> or if you want to stick with Invalid make it
> kInvalidThreadlocalKey so that it is obvious that the key is invalid not the
> thread local value.

Moved field and changed name to kUnsetThreadlocalKey.

http://codereview.chromium.org/9328042/diff/4001/runtime/platform/thread_linu...
File runtime/platform/thread_linux.cc (right):

http://codereview.chromium.org/9328042/diff/4001/runtime/platform/thread_linu...
runtime/platform/thread_linux.cc:111: int result = pthread_key_create(&key,
NULL);
On 2012/02/06 19:29:37, asiva wrote:
> ASSERT(key != kInvalidThreadLocal);

Done.

http://codereview.chromium.org/9328042/diff/4001/runtime/platform/thread_linu...
runtime/platform/thread_linux.cc:127: }
On 2012/02/06 19:29:37, asiva wrote:
> Can this be defined in the header file so that it gets inlined the way it is
> now.

Added a class ThreadInlineImpl to platform specific .h files with static method
for GetThreadLocal. We could consider just moving all of class Thread to the
platform specific .h files.

http://codereview.chromium.org/9328042/diff/4001/runtime/platform/thread_maco...
File runtime/platform/thread_macos.cc (right):

http://codereview.chromium.org/9328042/diff/4001/runtime/platform/thread_maco...
runtime/platform/thread_macos.cc:95: int result = pthread_key_create(&key,
NULL);
On 2012/02/06 19:29:37, asiva wrote:
> ASSERT(key != kInvalidThreadLocal);

Done.

http://codereview.chromium.org/9328042/diff/4001/runtime/platform/thread_maco...
runtime/platform/thread_macos.cc:111: }
On 2012/02/06 19:29:37, asiva wrote:
> Inline this method in the header so that there is no perf degradation.

Done.

http://codereview.chromium.org/9328042/diff/4001/runtime/vm/isolate_linux.cc
File runtime/vm/isolate_linux.cc (left):

http://codereview.chromium.org/9328042/diff/4001/runtime/vm/isolate_linux.cc#...
runtime/vm/isolate_linux.cc:22: // isolate specific information a single pthread
key is sufficient.
On 2012/02/06 19:29:37, asiva wrote:
> This comment has been lost now.

Copied to isolate.cc where isolate_key is defined. Changed "pthread" to "thread
local"

http://codereview.chromium.org/9328042/diff/4001/runtime/vm/isolate_macos.cc
File runtime/vm/isolate_macos.cc (left):

http://codereview.chromium.org/9328042/diff/4001/runtime/vm/isolate_macos.cc#...
runtime/vm/isolate_macos.cc:22: // isolate specific information a single pthread
key is sufficient.
On 2012/02/06 19:29:37, asiva wrote:
> Ditto, comment lost.

Ditto, now in isolate.cc.