Remember all deopt reasons in ic_data, not just the last one.

Remember all deopt reasons in ic_data, not just the last one.
Remember if a JS warning was issued in ic_data.
Save a word in ic_data on 64-bit platforms.

R=iposva@google.com, srdjan@google.com

Committed: https://code.google.com/p/dart/source/detail?r=35457

[srdjan, 2014-04-25 20:54:15]

LGTM

https://codereview.chromium.org/254723003/diff/40001/runtime/vm/code_descript...
File runtime/vm/code_descriptors.h (right):

https://codereview.chromium.org/254723003/diff/40001/runtime/vm/code_descript...
runtime/vm/code_descriptors.h:28: ASSERT(static_cast<uword>(data) <
kDeoptNumReasons);
Remove static_cast

https://codereview.chromium.org/254723003/diff/40001/runtime/vm/code_generator.h
File runtime/vm/code_generator.h (right):

https://codereview.chromium.org/254723003/diff/40001/runtime/vm/code_generato...
runtime/vm/code_generator.h:50: const char* DeoptReasonToText(intptr_t
deopt_reason);
s/intptr_t/DeoptReasonId/

https://codereview.chromium.org/254723003/diff/40001/runtime/vm/object.cc
File runtime/vm/object.cc (right):

https://codereview.chromium.org/254723003/diff/40001/runtime/vm/object.cc#new...
runtime/vm/object.cc:10727: ASSERT(static_cast<uword>(reason.Value()) <
kDeoptNumReasons);
Why the cast?

https://codereview.chromium.org/254723003/diff/40001/runtime/vm/object.h
File runtime/vm/object.h (right):

https://codereview.chromium.org/254723003/diff/40001/runtime/vm/object.h#newc...
runtime/vm/object.h:3665: bool HasDeoptReasons() const { return DeoptReasons() >
0; }
!= 0, since it cannot be negative.

[Ivan Posva, 2014-04-25 20:59:22]

LGTM

-Ivan

https://codereview.chromium.org/254723003/diff/20001/runtime/vm/code_descript...
File runtime/vm/code_descriptors.h (right):

https://codereview.chromium.org/254723003/diff/20001/runtime/vm/code_descript...
runtime/vm/code_descriptors.h:28: ASSERT(static_cast<uword>(data) <
kDeoptNumReasons);
Can we change this into a compile time assert?

https://codereview.chromium.org/254723003/diff/20001/runtime/vm/code_generator.h
File runtime/vm/code_generator.h (right):

https://codereview.chromium.org/254723003/diff/20001/runtime/vm/code_generato...
runtime/vm/code_generator.h:50: const char* DeoptReasonToText(intptr_t
deopt_reason);
DeoptReasonToCString

https://codereview.chromium.org/254723003/diff/40001/runtime/vm/object.h
File runtime/vm/object.h (right):

https://codereview.chromium.org/254723003/diff/40001/runtime/vm/object.h#newc...
runtime/vm/object.h:3183: #define DEOPT_REASONS(V)                              
                        \
I would be more comfortable if this was within a class instead of defined at the
top-level.

https://codereview.chromium.org/254723003/diff/40001/runtime/vm/object.h#newc...
runtime/vm/object.h:3792: kNumArgsTestedBits = 0,
Pos and Size?

[regis, 2014-04-25 23:38:27]

Thanks!

https://codereview.chromium.org/254723003/diff/20001/runtime/vm/code_descript...
File runtime/vm/code_descriptors.h (right):

https://codereview.chromium.org/254723003/diff/20001/runtime/vm/code_descript...
runtime/vm/code_descriptors.h:28: ASSERT(static_cast<uword>(data) <
kDeoptNumReasons);
On 2014/04/25 20:59:22, Ivan Posva wrote:
> Can we change this into a compile time assert?

The value of 'data' is not known at compile time.

https://codereview.chromium.org/254723003/diff/20001/runtime/vm/code_generator.h
File runtime/vm/code_generator.h (right):

https://codereview.chromium.org/254723003/diff/20001/runtime/vm/code_generato...
runtime/vm/code_generator.h:50: const char* DeoptReasonToText(intptr_t
deopt_reason);
On 2014/04/25 20:59:22, Ivan Posva wrote:
> DeoptReasonToCString

Done.

https://codereview.chromium.org/254723003/diff/40001/runtime/vm/code_descript...
File runtime/vm/code_descriptors.h (right):

https://codereview.chromium.org/254723003/diff/40001/runtime/vm/code_descript...
runtime/vm/code_descriptors.h:28: ASSERT(static_cast<uword>(data) <
kDeoptNumReasons);
On 2014/04/25 20:54:16, srdjan wrote:
> Remove static_cast

Done.
Casting to unsigned was to catch negative values. I made the range check
explicit.

https://codereview.chromium.org/254723003/diff/40001/runtime/vm/code_generator.h
File runtime/vm/code_generator.h (right):

https://codereview.chromium.org/254723003/diff/40001/runtime/vm/code_generato...
runtime/vm/code_generator.h:50: const char* DeoptReasonToText(intptr_t
deopt_reason);
On 2014/04/25 20:54:16, srdjan wrote:
> s/intptr_t/DeoptReasonId/

I did not do it, because this required a cast at the call site.
Done now.

https://codereview.chromium.org/254723003/diff/40001/runtime/vm/object.cc
File runtime/vm/object.cc (right):

https://codereview.chromium.org/254723003/diff/40001/runtime/vm/object.cc#new...
runtime/vm/object.cc:10727: ASSERT(static_cast<uword>(reason.Value()) <
kDeoptNumReasons);
On 2014/04/25 20:54:16, srdjan wrote:
> Why the cast?

Done.
Casting to unsigned was to catch negative values. I made the range check
explicit.

https://codereview.chromium.org/254723003/diff/40001/runtime/vm/object.h
File runtime/vm/object.h (right):

https://codereview.chromium.org/254723003/diff/40001/runtime/vm/object.h#newc...
runtime/vm/object.h:3183: #define DEOPT_REASONS(V)                              
                        \
On 2014/04/25 20:59:23, Ivan Posva wrote:
> I would be more comfortable if this was within a class instead of defined at
the
> top-level.

Done.
But I had to move class ICData before class Code, both in object.h and
object.cc.

https://codereview.chromium.org/254723003/diff/40001/runtime/vm/object.h#newc...
runtime/vm/object.h:3665: bool HasDeoptReasons() const { return DeoptReasons() >
0; }
On 2014/04/25 20:54:16, srdjan wrote:
> != 0, since it cannot be negative.

Done.

https://codereview.chromium.org/254723003/diff/40001/runtime/vm/object.h#newc...
runtime/vm/object.h:3792: kNumArgsTestedBits = 0,
On 2014/04/25 20:59:23, Ivan Posva wrote:
> Pos and Size?

Done.

[regis, 2014-04-25 23:45:36]

Committed patchset #4 manually as r35457 (presubmit successful).