Crankshaft builtins.

Crankshaft builtins.

Enable optimizing compiler for V8 built-ins. Also fixes an issue uncovered in
x64 codegen.

R=danno@chromium.org, mstarzinger@chromium.org, svenpanne@chromium.org

Committed: https://code.google.com/p/v8/source/detail?r=17334

[Dmitry Lomov (no reviews), 2013-10-22 16:40:41]

PTAL

[Michael Starzinger, 2013-10-22 17:01:28]

Drive-by comment.

https://codereview.chromium.org/34503003/diff/40001/src/runtime.cc
File src/runtime.cc (left):

https://codereview.chromium.org/34503003/diff/40001/src/runtime.cc#oldcode2940
src/runtime.cc:2940: target_shared->ReplaceCode(source_shared->code());
I have the feeling that removing the call to SharedFunctionInfo::ReplaceCode
here will break ProxyFix() as the shared function info backing the newly fixed
JSFunction created from that proxy will still point to the Builtins::kIllegal
IIUC.

[Dmitry Lomov (no reviews), 2013-10-22 19:25:02]

https://codereview.chromium.org/34503003/diff/40001/src/runtime.cc
File src/runtime.cc (left):

https://codereview.chromium.org/34503003/diff/40001/src/runtime.cc#oldcode2940
src/runtime.cc:2940: target_shared->ReplaceCode(source_shared->code());
On 2013/10/22 17:01:28, Michael Starzinger wrote:
> I have the feeling that removing the call to SharedFunctionInfo::ReplaceCode
> here will break ProxyFix() as the shared function info backing the newly fixed
> JSFunction created from that proxy will still point to the Builtins::kIllegal
> IIUC.
See below line 2951 where the call to ReplaceCode is repeated.

[Sven Panne, 2013-10-23 05:58:43]

lgtm

https://codereview.chromium.org/34503003/diff/40001/src/x64/lithium-codegen-x...
File src/x64/lithium-codegen-x64.cc (right):

https://codereview.chromium.org/34503003/diff/40001/src/x64/lithium-codegen-x...
src/x64/lithium-codegen-x64.cc:5260: __ movq(temp, Operand(temp,
StandardFrameConstants::kCallerFPOffset));
Hmmm, is this a candidate for backporting?

[Dmitry Lomov (no reviews), 2013-10-23 07:29:54]

On 2013/10/23 05:58:43, Sven Panne wrote:
> lgtm
> 
>
https://codereview.chromium.org/34503003/diff/40001/src/x64/lithium-codegen-x...
> File src/x64/lithium-codegen-x64.cc (right):
> 
>
https://codereview.chromium.org/34503003/diff/40001/src/x64/lithium-codegen-x...
> src/x64/lithium-codegen-x64.cc:5260: __ movq(temp, Operand(temp,
> StandardFrameConstants::kCallerFPOffset));
> Hmmm, is this a candidate for backporting?

No, if we do not optimize the bulitins, we never actually generate any code of
%_IsConstructCall

[Michael Starzinger, 2013-10-23 08:05:19]

Not LGTM.

https://codereview.chromium.org/34503003/diff/40001/src/runtime.cc
File src/runtime.cc (left):

https://codereview.chromium.org/34503003/diff/40001/src/runtime.cc#oldcode2940
src/runtime.cc:2940: target_shared->ReplaceCode(source_shared->code());
On 2013/10/22 19:25:03, Dmitry Lomov (chromium) wrote:
> On 2013/10/22 17:01:28, Michael Starzinger wrote:
> > I have the feeling that removing the call to SharedFunctionInfo::ReplaceCode
> > here will break ProxyFix() as the shared function info backing the newly
fixed
> > JSFunction created from that proxy will still point to the
Builtins::kIllegal
> > IIUC.
> See below line 2951 where the call to ReplaceCode is repeated.

Nope, SharedFunctionInfo::ReplaceCode and JSFunction::RecplaceCode are actually
two different functions.

[Dmitry Lomov (no reviews), 2013-10-23 08:21:19]

https://codereview.chromium.org/34503003/diff/40001/src/runtime.cc
File src/runtime.cc (left):

https://codereview.chromium.org/34503003/diff/40001/src/runtime.cc#oldcode2940
src/runtime.cc:2940: target_shared->ReplaceCode(source_shared->code());
On 2013/10/23 08:05:20, Michael Starzinger wrote:
> On 2013/10/22 19:25:03, Dmitry Lomov (chromium) wrote:
> > On 2013/10/22 17:01:28, Michael Starzinger wrote:
> > > I have the feeling that removing the call to
SharedFunctionInfo::ReplaceCode
> > > here will break ProxyFix() as the shared function info backing the newly
> fixed
> > > JSFunction created from that proxy will still point to the
> Builtins::kIllegal
> > > IIUC.
> > See below line 2951 where the call to ReplaceCode is repeated.
> 
> Nope, SharedFunctionInfo::ReplaceCode and JSFunction::RecplaceCode are
actually
> two different functions.

Rats, you are right.

[Dmitry Lomov (no reviews), 2013-10-23 08:25:46]

Fixed a bug spotted by Michael - thanks!

[Michael Starzinger, 2013-10-23 08:28:09]

LGTM from my end now, didn't fully review the whole thing though.

[danno, 2013-10-23 08:38:21]

lgtm

[Dmitry Lomov (no reviews), 2013-10-23 08:58:02]

Committed patchset #3 manually as r17334 (presubmit successful).