Unify internal and external typed arrays a bit

Unify internal and external typed arrays a bit

Just give internal ones an ArrayBuffer with a NULL backing store. This
simplifies the access checks a lot.

BUG=v8:3996
R=hpayer@chromium.org,verwaest@chromium.org
LOG=n

Committed: https://crrev.com/98140318fac953e675369d0e97dc8f8646324060
Cr-Commit-Position: refs/heads/master@{#28168}

[jochen (gone - plz use gerrit), 2015-04-29 13:38:31]

Hannes, can you look esp. at HGraphBuild::BuildAllocateEmptyArrayBuffer? I'm
adding the MakePrefillWithFiller() call to get a defined value into the internal
fields of the JSArrayBuffer...

Toon, ptal at the rest

[Benedikt Meurer, 2015-04-30 08:16:37]

bmeurer@chromium.org changed reviewers:
+ bmeurer@chromium.org

[Benedikt Meurer, 2015-04-30 08:16:37]

While I can see the point of the optimization, I'm pretty sure the correct fix
would be to not have any kind of "on heap typed array". Each typed array should
have a proper JSArrayBuffer. I'm 99% sure that we're missing the real world
optimization goal once again, and do optimize for useless micro benchmarks with
"on heap typed arrays".

But sanity apart this is LGTM from my side.

[Hannes Payer (out of office), 2015-04-30 09:25:12]

hpayer@chromium.org changed reviewers:
+ kbr@chromium.org

[Hannes Payer (out of office), 2015-04-30 09:25:13]

kbr, could you point us to benchmarks where the on heap typed array provides a
performance improvement?

[jochen (gone - plz use gerrit), 2015-04-30 09:28:22]

On 2015/04/30 at 09:25:13, hpayer wrote:
> kbr, could you point us to benchmarks where the on heap typed array provides a
performance improvement?

I talked a bit with Danno and Toon about this, and we'll try to get an benchmark
internally (and decided to keep onheap typed arrays for now)

[Hannes Payer (out of office), 2015-04-30 10:50:18]

https://codereview.chromium.org/1109353003/diff/1/src/hydrogen.cc
File src/hydrogen.cc (right):

https://codereview.chromium.org/1109353003/diff/1/src/hydrogen.cc#newcode9701
src/hydrogen.cc:9701: result->MakePrefillWithFiller();
Make prefill with filler is not the right thing here. It will preinitialize the
whole space with one word fillers. You are going to overwrite most of them right
after that. Please initialize all fields with sane values.

[jochen (gone - plz use gerrit), 2015-04-30 12:46:32]

done, ptal

[Hannes Payer (out of office), 2015-04-30 13:26:42]

lgtm

[jochen (gone - plz use gerrit), 2015-04-30 13:27:45]

The CQ bit was checked by jochen@chromium.org

[jochen (gone - plz use gerrit), 2015-04-30 13:27:45]

The patchset sent to the CQ was uploaded after l-g-t-m from bmeurer@chromium.org
Link to the patchset: https://codereview.chromium.org/1109353003/#ps20001
(title: "updates")

[commit-bot: I haz the power, 2015-04-30 13:28:11]

CQ is trying da patch. Follow status at
 https://chromium-cq-status.appspot.com/patch-status/1109353003/20001

[commit-bot: I haz the power, 2015-04-30 13:46:34]

Committed patchset #2 (id:20001)

[commit-bot: I haz the power, 2015-04-30 13:46:51]

Patchset 2 (id:??) landed as
https://crrev.com/98140318fac953e675369d0e97dc8f8646324060
Cr-Commit-Position: refs/heads/master@{#28168}