Handle non-JSObject heap objects using slow-path IC stub guarded by the map.

Handle non-JSObject heap objects using slow-path IC stub guarded by the map.

BUG=chromium:280632
R=mstarzinger@chromium.org

Committed: https://code.google.com/p/v8/source/detail?r=16689

[Toon Verwaest, 2013-09-12 09:28:55]

PTAL

[Michael Starzinger, 2013-09-12 14:28:48]

LGTM if comments are addressed.

For the record: I still don't like the idea of populating the code cache for
non-JSObject maps, but I don't have a better alternative.

https://codereview.chromium.org/23691056/diff/5001/src/objects.h
File src/objects.h (right):

https://codereview.chromium.org/23691056/diff/5001/src/objects.h#newcode1793
src/objects.h:1793: static void UpdateMapCodeCache(Handle<HeapObject> object,
Please move this up a few lines to above the casting method. Also add a short
one-liner commment describing this method.

https://codereview.chromium.org/23691056/diff/5001/src/objects.h#newcode5495
src/objects.h:5495: inline bool is_jsobject_map();
I think starting to duplicate type checking predicates on Map is not good
(especially with a completely new naming convention). I would rather have this
predicate be inlined into CanInlinePropertyAccess directly.

[Toon Verwaest, 2013-09-12 14:58:23]

Addressed comments

https://chromiumcodereview.appspot.com/23691056/diff/5001/src/objects.h
File src/objects.h (right):

https://chromiumcodereview.appspot.com/23691056/diff/5001/src/objects.h#newco...
src/objects.h:1793: static void UpdateMapCodeCache(Handle<HeapObject> object,
On 2013/09/12 14:28:48, Michael Starzinger wrote:
> Please move this up a few lines to above the casting method. Also add a short
> one-liner commment describing this method.

Done.

https://chromiumcodereview.appspot.com/23691056/diff/5001/src/objects.h#newco...
src/objects.h:5495: inline bool is_jsobject_map();
On 2013/09/12 14:28:48, Michael Starzinger wrote:
> I think starting to duplicate type checking predicates on Map is not good
> (especially with a completely new naming convention). I would rather have this
> predicate be inlined into CanInlinePropertyAccess directly.

Done.

[Toon Verwaest, 2013-09-12 16:13:56]

Committed patchset #2 manually as r16689 (presubmit successful).