Move most of the implementation of AdjustAmountOfExternalMemory to v8.h

Move most of the implementation of AdjustAmountOfExternalMemory to v8.h

This reduces the overhead of reporting allocations to v8 to an
acceptable level.

BUG=none
LOG=n
R=dcarney@chromium.org, hpayer@chromium.org

Committed: https://code.google.com/p/v8/source/detail?r=21688

[Hannes Payer (out of office), 2014-06-04 11:24:49]

https://codereview.chromium.org/310393003/diff/1/include/v8.h
File include/v8.h (right):

https://codereview.chromium.org/310393003/diff/1/include/v8.h#newcode6614
include/v8.h:6614: 192 * 1024 * 1024) {
192? Please don't use a hard-coded value here.

https://codereview.chromium.org/310393003/diff/1/src/heap-inl.h
File src/heap-inl.h (right):

https://codereview.chromium.org/310393003/diff/1/src/heap-inl.h#newcode544
src/heap-inl.h:544: if (amount_since_last_global_gc > 192 * MB) {
192?

https://codereview.chromium.org/310393003/diff/1/src/heap.h
File src/heap.h (left):

https://codereview.chromium.org/310393003/diff/1/src/heap.h#oldcode1041
src/heap.h:1041: inline int64_t AdjustAmountOfExternalAllocatedMemory(
With that change, can we get rid of this method? The array buffers could use the
new method.

[jochen (gone - plz use gerrit), 2014-06-04 11:41:24]

https://codereview.chromium.org/310393003/diff/1/include/v8.h
File include/v8.h (right):

https://codereview.chromium.org/310393003/diff/1/include/v8.h#newcode6614
include/v8.h:6614: 192 * 1024 * 1024) {
On 2014/06/04 11:24:49, Hannes Payer wrote:
> 192? Please don't use a hard-coded value here.

Done.

https://codereview.chromium.org/310393003/diff/1/src/heap-inl.h
File src/heap-inl.h (right):

https://codereview.chromium.org/310393003/diff/1/src/heap-inl.h#newcode544
src/heap-inl.h:544: if (amount_since_last_global_gc > 192 * MB) {
On 2014/06/04 11:24:49, Hannes Payer wrote:
> 192?

external_allocation_limit_ is hard-coded to 192MB, I moved this to v8.h

https://codereview.chromium.org/310393003/diff/1/src/heap.h
File src/heap.h (left):

https://codereview.chromium.org/310393003/diff/1/src/heap.h#oldcode1041
src/heap.h:1041: inline int64_t AdjustAmountOfExternalAllocatedMemory(
On 2014/06/04 11:24:49, Hannes Payer wrote:
> With that change, can we get rid of this method? The array buffers could use
the
> new method.

The fast path method doesn't handle the case where we need to GC

[jochen (gone - plz use gerrit), 2014-06-04 11:53:07]

https://codereview.chromium.org/310393003/diff/40001/src/heap-inl.h
File src/heap-inl.h (left):

https://codereview.chromium.org/310393003/diff/40001/src/heap-inl.h#oldcode535
src/heap-inl.h:535: // Avoid overflow.
note that signed overflow is undefined, so these checks all don't make sense

[jochen (gone - plz use gerrit), 2014-06-04 11:58:23]

now it even compiles :)

[dcarney, 2014-06-04 12:02:43]

lgtm

[Hannes Payer (out of office), 2014-06-04 12:18:30]

https://codereview.chromium.org/310393003/diff/40001/src/heap-inl.h
File src/heap-inl.h (right):

https://codereview.chromium.org/310393003/diff/40001/src/heap-inl.h#newcode530
src/heap-inl.h:530: int64_t Heap::AdjustAmountOfExternalAllocatedMemory(
Why don't we get rid of this method? It seems like we can completely inline this
method in v8.h except for the call to CollectGarbage which is the slow path.

[jochen (gone - plz use gerrit), 2014-06-04 13:43:23]

On 2014/06/04 12:18:30, Hannes Payer wrote:
> https://codereview.chromium.org/310393003/diff/40001/src/heap-inl.h
> File src/heap-inl.h (right):
> 
> https://codereview.chromium.org/310393003/diff/40001/src/heap-inl.h#newcode530
> src/heap-inl.h:530: int64_t Heap::AdjustAmountOfExternalAllocatedMemory(
> Why don't we get rid of this method? It seems like we can completely inline
this
> method in v8.h except for the call to CollectGarbage which is the slow path.

the FLAG thing won't work in v8.h. is it ok to drop it?

[jochen (gone - plz use gerrit), 2014-06-04 14:37:09]

version without heap-inl.h implementation

[Hannes Payer (out of office), 2014-06-05 08:11:28]

lgtm

[jochen (gone - plz use gerrit), 2014-06-05 08:44:50]

Committed patchset #4 manually as r21688 (presubmit successful).