Change static calls in unoptimized code to always call via a stub. Using ICData, the call count of …

Change static calls in unoptimized code to always call via a stub. Using ICData, the call count of static calls is collected as well.
TODO: Use call frequency to guide inlining.

R=asiva@google.com, hausner@google.com, zra@google.com

Committed: https://code.google.com/p/dart/source/detail?r=24307

[zra, 2013-06-21 19:03:35]

arm/mips lgtm

https://codereview.chromium.org/17554003/diff/11001/runtime/vm/stub_code_mips.cc
File runtime/vm/stub_code_mips.cc (right):

https://codereview.chromium.org/17554003/diff/11001/runtime/vm/stub_code_mips...
runtime/vm/stub_code_mips.cc:1812: __ BranchEqual(T0, 0, &ok);
__ beq(T0, ZR, &ok)

https://codereview.chromium.org/17554003/diff/11001/runtime/vm/stub_code_mips...
runtime/vm/stub_code_mips.cc:1847: int num_slots = 3;  // Two preserved
registers, one argument (function).
When the number of slots is fixed, I've been foregoing a variable for the number
of slots like this.

[srdjan, 2013-06-21 20:30:18]

https://codereview.chromium.org/17554003/diff/11001/runtime/vm/stub_code_mips.cc
File runtime/vm/stub_code_mips.cc (right):

https://codereview.chromium.org/17554003/diff/11001/runtime/vm/stub_code_mips...
runtime/vm/stub_code_mips.cc:1812: __ BranchEqual(T0, 0, &ok);
On 2013/06/21 19:03:35, zra wrote:
> __ beq(T0, ZR, &ok)

Done.

https://codereview.chromium.org/17554003/diff/11001/runtime/vm/stub_code_mips...
runtime/vm/stub_code_mips.cc:1847: int num_slots = 3;  // Two preserved
registers, one argument (function).
On 2013/06/21 19:03:35, zra wrote:
> When the number of slots is fixed, I've been foregoing a variable for the
number
> of slots like this.

Done.

[hausner, 2013-06-21 20:52:22]

lgtm

https://codereview.chromium.org/17554003/diff/11001/runtime/vm/object.cc
File runtime/vm/object.cc (right):

https://codereview.chromium.org/17554003/diff/11001/runtime/vm/object.cc#newc...
runtime/vm/object.cc:8537: const intptr_t old_num = NumberOfChecks();
Nit: suggest to switch the above two lines and change to ASSERT(old_num == 0);

[siva, 2013-06-21 23:24:55]

lgtm

https://codereview.chromium.org/17554003/diff/32001/runtime/vm/code_generator.cc
File runtime/vm/code_generator.cc (right):

https://codereview.chromium.org/17554003/diff/32001/runtime/vm/code_generator...
runtime/vm/code_generator.cc:733: // necessary. Only static calls in optimized
code need to be patched.
Why not change the comment to:

// Patches static calls in optimized code with the target's entry point.

https://codereview.chromium.org/17554003/diff/32001/runtime/vm/code_patcher_i...
File runtime/vm/code_patcher_ia32.cc (right):

https://codereview.chromium.org/17554003/diff/32001/runtime/vm/code_patcher_i...
runtime/vm/code_patcher_ia32.cc:52: }
Is it possible to change this to:

RawICData* ic_data() const {
  return *reinterpret_cast<RawICData**>(start_ + 1);
}

That way in the code below you can use:
const ICData& ic_data = ICData::Handle(static_call.ic_data());

[srdjan, 2013-06-21 23:32:20]

https://codereview.chromium.org/17554003/diff/11001/runtime/vm/object.cc
File runtime/vm/object.cc (right):

https://codereview.chromium.org/17554003/diff/11001/runtime/vm/object.cc#newc...
runtime/vm/object.cc:8537: const intptr_t old_num = NumberOfChecks();
On 2013/06/21 20:52:22, hausner wrote:
> Nit: suggest to switch the above two lines and change to ASSERT(old_num == 0);

Done.

https://codereview.chromium.org/17554003/diff/32001/runtime/vm/code_generator.cc
File runtime/vm/code_generator.cc (right):

https://codereview.chromium.org/17554003/diff/32001/runtime/vm/code_generator...
runtime/vm/code_generator.cc:733: // necessary. Only static calls in optimized
code need to be patched.
On 2013/06/21 23:24:55, siva wrote:
> Why not change the comment to:
> 
> // Patches static calls in optimized code with the target's entry point.

Done.

https://codereview.chromium.org/17554003/diff/32001/runtime/vm/code_patcher_i...
File runtime/vm/code_patcher_ia32.cc (right):

https://codereview.chromium.org/17554003/diff/32001/runtime/vm/code_patcher_i...
runtime/vm/code_patcher_ia32.cc:52: }
On 2013/06/21 23:24:55, siva wrote:
> Is it possible to change this to:
> 
> RawICData* ic_data() const {
>   return *reinterpret_cast<RawICData**>(start_ + 1);
> }
> 
> That way in the code below you can use:
> const ICData& ic_data = ICData::Handle(static_call.ic_data());

Discussed off-line, leaving it as is so that we can explicitly check that the
the object is ICData.

[srdjan, 2013-06-21 23:35:22]

Committed patchset #5 manually as r24307 (presubmit successful).