New hydrogen instruction to reduce cost of growing an array on keyed stores.

New hydrogen instruction to reduce cost of growing an array on keyed stores.

HMaybeGrowElements moves the situation where you actually have to grow
into deferred code. This means crankshaft doesn't have to spill registers
just to make the bounds comparison to see if it'll need to grow or not.

It makes the growing case a bit more expensive, but reduces the cost of
the general case.

BUG=chromium:484025
LOG=N

Committed: https://crrev.com/3bce9c3afbe792167a54db49b16679289e0aea8f
Cr-Commit-Position: refs/heads/master@{#28359}

[mvstanton, 2015-05-08 10:38:58]

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

[mvstanton, 2015-05-08 10:40:42]

Hi Benedikt,
x64 and ia32 are done, I'll work on ports, thx for the look!
--Michael

[Benedikt Meurer, 2015-05-08 10:46:31]

x64 looking good, except for the LoadSmiConstant optimization, which might need
some thought.

https://codereview.chromium.org/1124443004/diff/80001/src/x64/macro-assembler...
File src/x64/macro-assembler-x64.cc (right):

https://codereview.chromium.org/1124443004/diff/80001/src/x64/macro-assembler...
src/x64/macro-assembler-x64.cc:927: void
MacroAssembler::LoadSmiConstant(Register dst, Smi* source) {
Hm, we tried that before, see https://codereview.chromium.org/1059543004 for the
revert.

[mvstanton, 2015-05-08 13:03:26]

Hi Benedikt, yeah, I got rid of the xorl optimization now, leaving a comment.

arm64 and arm done. MIPS team, can you create a mips/mips64 port?
Thanks,
--Michael

[akos.palfi.imgtec, 2015-05-08 17:48:20]

On 2015/05/08 13:03:26, mvstanton wrote:
> Hi Benedikt, yeah, I got rid of the xorl optimization now, leaving a comment.
> 
> arm64 and arm done. MIPS team, can you create a mips/mips64 port?
> Thanks,
> --Michael

Hi Michael,

I've created the MIPS port and uploaded it here:
https://codereview.chromium.org/1129053003/

[Benedikt Meurer, 2015-05-12 05:13:04]

LGTM with comment.

https://codereview.chromium.org/1124443004/diff/140001/src/x64/lithium-codege...
File src/x64/lithium-codegen-x64.cc (right):

https://codereview.chromium.org/1124443004/diff/140001/src/x64/lithium-codege...
src/x64/lithium-codegen-x64.cc:4648: // Deoptimize if we got a smi back.
Please add a comment here to describe why we deopt in case of a SMI return (i.e.
elements normalized).

[mvstanton, 2015-05-12 07:59:03]

Right on, thanks Benedikt!
--Michael

https://codereview.chromium.org/1124443004/diff/140001/src/x64/lithium-codege...
File src/x64/lithium-codegen-x64.cc (right):

https://codereview.chromium.org/1124443004/diff/140001/src/x64/lithium-codege...
src/x64/lithium-codegen-x64.cc:4648: // Deoptimize if we got a smi back.
On 2015/05/12 05:13:04, Benedikt Meurer wrote:
> Please add a comment here to describe why we deopt in case of a SMI return
(i.e.
> elements normalized).

Good idea, done.

[mvstanton, 2015-05-12 08:02:57]

The CQ bit was checked by mvstanton@chromium.org

[mvstanton, 2015-05-12 08:02:58]

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/1124443004/#ps200001
(title: "REBASE.")

[commit-bot: I haz the power, 2015-05-12 08:03:13]

CQ is trying da patch. Follow status at
 https://chromium-cq-status.appspot.com/patch-status/1124443004/200001

[commit-bot: I haz the power, 2015-05-12 08:47:23]

Committed patchset #11 (id:200001)

[commit-bot: I haz the power, 2015-05-12 08:47:41]

Patchset 11 (id:??) landed as
https://crrev.com/3bce9c3afbe792167a54db49b16679289e0aea8f
Cr-Commit-Position: refs/heads/master@{#28359}