Add OptimizedCodeEntry as a new heap object type. An optimized code entry represents an association…

Add OptimizedCodeEntry as a new heap object type. An optimized code entry represents an association between the native context, a function, optimized code, and the literals. Such associations are needed by the deoptimizer and optimized code cache to efficiently find related optimized code and functions for a given context or shared function info.

BUG=
R=mstarzinger@chromium.org

Committed: https://code.google.com/p/v8/source/detail?r=16407

[titzer, 2013-08-28 09:03:30]

On 2013/08/28 08:52:38, titzer wrote:

Note that I did not implement the GC routines to deal with the
weakness--specifically the routine that finds all OptimizedCodeEntries and
updates or clears their weak references. This is because the
OptimizedCodeEntries will be hanging off the native contexts in the next
iteration--currently, no one creates or references them, except in my next set
of changes.

[Michael Starzinger, 2013-08-28 14:59:47]

LGTM with one comment in the allocator.

https://codereview.chromium.org/23691002/diff/1/src/heap.cc
File src/heap.cc (right):

https://codereview.chromium.org/23691002/diff/1/src/heap.cc#newcode3670
src/heap.cc:3670: entry->set_native_context(native_context, SKIP_WRITE_BARRIER);
Skipping WBs for all four of these setters is not safe, as any of these objects
might be on evacuation candidates and the entry itself is not in new-space.

https://codereview.chromium.org/23691002/diff/1/src/objects.h
File src/objects.h (right):

https://codereview.chromium.org/23691002/diff/1/src/objects.h#newcode6857
src/objects.h:6857: // Entries are cleaned up lazily as each list is traversed.
nit: s/cleaned up/removed from the list/ IIUC?

https://codereview.chromium.org/23691002/diff/1/src/objects.h#newcode6904
src/objects.h:6904: // and code.
nit: ... and literals.

[titzer, 2013-08-28 16:05:57]

On 2013/08/28 14:59:47, Michael Starzinger wrote:
> LGTM with one comment in the allocator.
> 
> https://codereview.chromium.org/23691002/diff/1/src/heap.cc
> File src/heap.cc (right):
> 
> https://codereview.chromium.org/23691002/diff/1/src/heap.cc#newcode3670
> src/heap.cc:3670: entry->set_native_context(native_context,
SKIP_WRITE_BARRIER);
> Skipping WBs for all four of these setters is not safe, as any of these
objects
> might be on evacuation candidates and the entry itself is not in new-space.
> 

Good catch. I was originally setting them to Smi 0.

> https://codereview.chromium.org/23691002/diff/1/src/objects.h
> File src/objects.h (right):
> 
> https://codereview.chromium.org/23691002/diff/1/src/objects.h#newcode6857
> src/objects.h:6857: // Entries are cleaned up lazily as each list is
traversed.
> nit: s/cleaned up/removed from the list/ IIUC?

Yep.

> 
> https://codereview.chromium.org/23691002/diff/1/src/objects.h#newcode6904
> src/objects.h:6904: // and code.
> nit: ... and literals.

Done.

[titzer, 2013-08-28 16:48:56]

Committed patchset #1 manually as r16407 (presubmit successful).