AllocationSite::memento_found_count() and friends need Smi access.

AllocationSite::memento_found_count() and friends need Smi access.

Some new AllocationSite fields weren't accessed with the correct
representation.

R=hpayer@chromium.org

Committed: https://code.google.com/p/v8/source/detail?r=18253

[mvstanton, 2013-12-03 17:53:56]

Hi Hannes,
Here is the fix for the bug you were looking at. PTAL, thx,
--Michael

[Hannes Payer (out of office), 2013-12-03 19:31:49]

LGTM

[Benedikt Meurer, 2013-12-04 06:05:07]

https://codereview.chromium.org/101783005/diff/1/src/hydrogen-instructions.cc
File src/hydrogen-instructions.cc (right):

https://codereview.chromium.org/101783005/diff/1/src/hydrogen-instructions.cc...
src/hydrogen-instructions.cc:4272: HObjectAccess
HObjectAccess::ForAllocationSiteOffset(int offset) {
I'd prefer to have one method per allocation site field, just as we do for the
other object types. This would be cleaner IMHO and less likely to break (the
fact that we need a switch here is already an indicator that it'll break at some
point).

[Hannes Payer (out of office), 2013-12-04 07:25:50]

https://codereview.chromium.org/101783005/diff/1/src/hydrogen-instructions.cc
File src/hydrogen-instructions.cc (right):

https://codereview.chromium.org/101783005/diff/1/src/hydrogen-instructions.cc...
src/hydrogen-instructions.cc:4272: HObjectAccess
HObjectAccess::ForAllocationSiteOffset(int offset) {
On 2013/12/04 06:05:07, Benedikt Meurer wrote:
> I'd prefer to have one method per allocation site field, just as we do for the
> other object types. This would be cleaner IMHO and less likely to break (the
> fact that we need a switch here is already an indicator that it'll break at
some
> point).

Breaking is intended. Michael, wdyt?

[mvstanton, 2013-12-04 09:02:20]

Hi guys,
I prefer this style because it's object-oriented. Fields of a single object
deserve an encapsulating structure, and the case statement does that. If each
field access for a single object has to be maintained with three or 4 lines that
only appear "together" by convention in the header file I don't feel much more
secure.

I think it's a little numbing to scroll through pages of those definitions in
hydrogen-instructions.h.

[mvstanton, 2013-12-04 09:54:31]

Committed patchset #1 manually as r18253 (presubmit successful).