Heap maps: add VM service message.

Heap maps: add VM service message.

A heap map is serialized as arrays of sizes + class ids:

{"type":"HeapMap","id":"heapmap","pages":[[15,2,10,3,...,100,0],...,]}

R=johnmccutchan@google.com

Committed: https://code.google.com/p/dart/source/detail?r=33608

[Cutch, 2014-03-12 19:02:02]

https://codereview.chromium.org/197663006/diff/40001/runtime/vm/pages.cc
File runtime/vm/pages.cc (right):

https://codereview.chromium.org/197663006/diff/40001/runtime/vm/pages.cc#newc...
runtime/vm/pages.cc:439: array_->AddValue(obj->IsFreeListElement() ? 0 :
obj->GetClassId());
I'm concerned about mapping kFreeListElement to kIllegalCid.

What if kept it as kFreeListElement include the index in the HeapMap message
(see below).

https://codereview.chromium.org/197663006/diff/40001/runtime/vm/pages.cc#newc...
runtime/vm/pages.cc:449: heap_map.AddProperty("id", "heapmap");
heap_map.AddProperty("free_class_id", kFreeListElement);

https://codereview.chromium.org/197663006/diff/40001/runtime/vm/pages.cc#newc...
runtime/vm/pages.cc:451: JSONArray all_pages(&heap_map, "pages");
Maybe add a comment here saying the format of "pages" is:

[page0, page1, ..., pageN]

Format of a page is:

[size, class id, ..., size, class id]

[koda, 2014-03-12 19:08:01]

https://codereview.chromium.org/197663006/diff/40001/runtime/vm/pages.cc
File runtime/vm/pages.cc (right):

https://codereview.chromium.org/197663006/diff/40001/runtime/vm/pages.cc#newc...
runtime/vm/pages.cc:439: array_->AddValue(obj->IsFreeListElement() ? 0 :
obj->GetClassId());
On 2014/03/12 19:02:02, Cutch wrote:
> I'm concerned about mapping kFreeListElement to kIllegalCid.
> 
> What if kept it as kFreeListElement include the index in the HeapMap message
> (see below).

Good idea. Done.

https://codereview.chromium.org/197663006/diff/40001/runtime/vm/pages.cc#newc...
runtime/vm/pages.cc:449: heap_map.AddProperty("id", "heapmap");
On 2014/03/12 19:02:02, Cutch wrote:
> heap_map.AddProperty("free_class_id", kFreeListElement);

Done.

https://codereview.chromium.org/197663006/diff/40001/runtime/vm/pages.cc#newc...
runtime/vm/pages.cc:451: JSONArray all_pages(&heap_map, "pages");
On 2014/03/12 19:02:02, Cutch wrote:
> Maybe add a comment here saying the format of "pages" is:
> 
> [page0, page1, ..., pageN]
> 
> Format of a page is:
> 
> [size, class id, ..., size, class id]

Done.

[Cutch, 2014-03-12 20:37:41]

lgtm w/ comment

https://codereview.chromium.org/197663006/diff/60001/runtime/vm/pages.cc
File runtime/vm/pages.cc (right):

https://codereview.chromium.org/197663006/diff/60001/runtime/vm/pages.cc#newc...
runtime/vm/pages.cc:451: // [size, class id, size, class id, ...] (size unit is
kObjectAlignment).
Should we send the value of kObjectAlignment or is it the same across all
architectures?

[koda, 2014-03-12 20:44:17]

https://codereview.chromium.org/197663006/diff/60001/runtime/vm/pages.cc
File runtime/vm/pages.cc (right):

https://codereview.chromium.org/197663006/diff/60001/runtime/vm/pages.cc#newc...
runtime/vm/pages.cc:451: // [size, class id, size, class id, ...] (size unit is
kObjectAlignment).
On 2014/03/12 20:37:41, Cutch wrote:
> Should we send the value of kObjectAlignment or is it the same across all
> architectures?

Added it as "unit_size_bytes".

(It's two words, hence architecture-dependent.)

[koda, 2014-03-12 20:46:14]

Committed patchset #5 manually as r33608 (presubmit successful).