Add support for object pool

Add support for object pool that will be used on ARM and MIPS architectures.
Modify code patching infrastructure to accept code object, which is necessary to
get to the object pool containing patchable target addresses.
Modify assembler test infrastructure to provide associated code object.
Committed: https://code.google.com/p/dart/source/detail?r=18604

[regis, 2013-02-14 23:56:38]

Srdjan, Matthias,

Do not fear, the list is long, but the changes are small.

Zach,

Congrats on getting your first Dart cl to review. No panic. I'll walk you
through it on Friday.

Cheers,
Regis

[hausner, 2013-02-15 00:58:36]

lgtm

[zra, 2013-02-15 18:15:30]

lgtm

[srdjan, 2013-02-15 21:24:27]

LGTM with comments

https://codereview.chromium.org/12260026/diff/17001/runtime/vm/code_generator.cc
File runtime/vm/code_generator.cc (right):

https://codereview.chromium.org/12260026/diff/17001/runtime/vm/code_generator...
runtime/vm/code_generator.cc:766:
CodePatcher::PatchStaticCallAt(caller_frame->pc(), target_code,
Should be caller_code, below as well

https://codereview.chromium.org/12260026/diff/17001/runtime/vm/code_patcher.h
File runtime/vm/code_patcher.h (right):

https://codereview.chromium.org/12260026/diff/17001/runtime/vm/code_patcher.h...
runtime/vm/code_patcher.h:29: const Code& code,
Please modify comment/variable name 'code' to explain what 'code' is for.

https://codereview.chromium.org/12260026/diff/17001/runtime/vm/code_patcher.h...
runtime/vm/code_patcher.h:34: const Code& code,
ditto here and below.

https://codereview.chromium.org/12260026/diff/17001/runtime/vm/code_patcher_i...
File runtime/vm/code_patcher_ia32.cc (right):

https://codereview.chromium.org/12260026/diff/17001/runtime/vm/code_patcher_i...
runtime/vm/code_patcher_ia32.cc:146: const Code& code) {
Here and below, you could assert that return address belongs to code I guess you
are not doing LookupCode(return_address) for performance reasons?

https://codereview.chromium.org/12260026/diff/17001/runtime/vm/debugger.h
File runtime/vm/debugger.h (right):

https://codereview.chromium.org/12260026/diff/17001/runtime/vm/debugger.h#new...
runtime/vm/debugger.h:98: const Code& code_;
Remove function_ use RawCode* instead Code.

[regis, 2013-02-15 22:51:31]

Thanks!

https://codereview.chromium.org/12260026/diff/17001/runtime/vm/code_generator.cc
File runtime/vm/code_generator.cc (right):

https://codereview.chromium.org/12260026/diff/17001/runtime/vm/code_generator...
runtime/vm/code_generator.cc:766:
CodePatcher::PatchStaticCallAt(caller_frame->pc(), target_code,
On 2013/02/15 21:24:27, srdjan wrote:
> Should be caller_code, below as well

Good catch!

https://codereview.chromium.org/12260026/diff/17001/runtime/vm/code_patcher.h
File runtime/vm/code_patcher.h (right):

https://codereview.chromium.org/12260026/diff/17001/runtime/vm/code_patcher.h...
runtime/vm/code_patcher.h:29: const Code& code,
On 2013/02/15 21:24:27, srdjan wrote:
> Please modify comment/variable name 'code' to explain what 'code' is for.

Done.

https://codereview.chromium.org/12260026/diff/17001/runtime/vm/code_patcher.h...
runtime/vm/code_patcher.h:34: const Code& code,
On 2013/02/15 21:24:27, srdjan wrote:
> ditto here and below.

Done.

https://codereview.chromium.org/12260026/diff/17001/runtime/vm/code_patcher_i...
File runtime/vm/code_patcher_ia32.cc (right):

https://codereview.chromium.org/12260026/diff/17001/runtime/vm/code_patcher_i...
runtime/vm/code_patcher_ia32.cc:146: const Code& code) {
On 2013/02/15 21:24:27, srdjan wrote:
> Here and below, you could assert that return address belongs to code I guess
you
> are not doing LookupCode(return_address) for performance reasons? 
LookupCode would be too slow, but I can assert that the return address belongs
to the code with a fast range check.

https://codereview.chromium.org/12260026/diff/17001/runtime/vm/debugger.h
File runtime/vm/debugger.h (right):

https://codereview.chromium.org/12260026/diff/17001/runtime/vm/debugger.h#new...
runtime/vm/debugger.h:98: const Code& code_;
On 2013/02/15 21:24:27, srdjan wrote:
> Remove function_ use RawCode* instead Code.

The purpose of the member code_ was to cache the Code object in order to prevent
repeated allocation of handles.
Since handles cannot be used here, caching is not useful. Field is removed.