Added pretenuring support for fast literal allocation in old data space.

Added pretenuring support for fast literal allocation in old data space.

BUG=
R=mstarzinger@chromium.org

Committed: https://code.google.com/p/v8/source/detail?r=15306

[Michael Starzinger, 2013-06-24 15:04:47]

Looking good, just a few nits.

https://codereview.chromium.org/17580011/diff/1004/src/hydrogen.cc
File src/hydrogen.cc (right):

https://codereview.chromium.org/17580011/diff/1004/src/hydrogen.cc#newcode9964
src/hydrogen.cc:9964: Handle<Map>
free_space_map(isolate()->heap()->free_space_map());
Better use isolate()->factory()->free_space_map() here.

https://codereview.chromium.org/17580011/diff/1004/src/hydrogen.cc#newcode10008
src/hydrogen.cc:10008: // this object and its backing store.
nit: The above comment is out-dated, let's adapt it and maybe even split it up
into to parts (one for the object here and one for the elements below).

https://codereview.chromium.org/17580011/diff/1004/src/hydrogen.cc#newcode10024
src/hydrogen.cc:10024: *offset += object_size;
nit: Let's add an empty new-line after this block to separate it from the next
one.

https://codereview.chromium.org/17580011/diff/1004/test/cctest/test-heap.cc
File test/cctest/test-heap.cc (right):

https://codereview.chromium.org/17580011/diff/1004/test/cctest/test-heap.cc#n...
test/cctest/test-heap.cc:2194: AlwaysAllocateScope always_allocate;
Is the AlwaysAllocateScope really necessary here? I am not sure about the
purpose. If it is needed then a short one-liner comments about it's purpose
would be nice, otherwise let's just drop it.

https://codereview.chromium.org/17580011/diff/1004/test/cctest/test-heap.cc#n...
test/cctest/test-heap.cc:2196: if (full) {
As discussed offline: This will not really trigger the case where the second
allocation in optimized code will fail, because the call to CompileRun() below
will fire up the whole compilation pipeline, so the next GC will most probably
be triggered earlier. Let's just drop this part from the tests.

[Hannes Payer (out of office), 2013-06-24 16:44:22]

https://codereview.chromium.org/17580011/diff/1004/src/hydrogen.cc
File src/hydrogen.cc (right):

https://codereview.chromium.org/17580011/diff/1004/src/hydrogen.cc#newcode9964
src/hydrogen.cc:9964: Handle<Map>
free_space_map(isolate()->heap()->free_space_map());
On 2013/06/24 15:04:47, Michael Starzinger wrote:
> Better use isolate()->factory()->free_space_map() here.

Done.

https://codereview.chromium.org/17580011/diff/1004/src/hydrogen.cc#newcode10008
src/hydrogen.cc:10008: // this object and its backing store.
On 2013/06/24 15:04:47, Michael Starzinger wrote:
> nit: The above comment is out-dated, let's adapt it and maybe even split it up
> into to parts (one for the object here and one for the elements below).

Done.

https://codereview.chromium.org/17580011/diff/1004/src/hydrogen.cc#newcode10024
src/hydrogen.cc:10024: *offset += object_size;
On 2013/06/24 15:04:47, Michael Starzinger wrote:
> nit: Let's add an empty new-line after this block to separate it from the next
> one.

Done.

https://codereview.chromium.org/17580011/diff/1004/test/cctest/test-heap.cc
File test/cctest/test-heap.cc (right):

https://codereview.chromium.org/17580011/diff/1004/test/cctest/test-heap.cc#n...
test/cctest/test-heap.cc:2194: AlwaysAllocateScope always_allocate;
On 2013/06/24 15:04:47, Michael Starzinger wrote:
> Is the AlwaysAllocateScope really necessary here? I am not sure about the
> purpose. If it is needed then a short one-liner comments about it's purpose
> would be nice, otherwise let's just drop it.

Done.

https://codereview.chromium.org/17580011/diff/1004/test/cctest/test-heap.cc#n...
test/cctest/test-heap.cc:2196: if (full) {
On 2013/06/24 15:04:47, Michael Starzinger wrote:
> As discussed offline: This will not really trigger the case where the second
> allocation in optimized code will fail, because the call to CompileRun() below
> will fire up the whole compilation pipeline, so the next GC will most probably
> be triggered earlier. Let's just drop this part from the tests.

Done.

[Michael Starzinger, 2013-06-24 16:45:54]

LGTM.

[Hannes Payer (out of office), 2013-06-25 07:48:01]

Committed patchset #6 manually as r15306 (presubmit successful).