Crankshaft is now able to compile top level code even if there is a ScriptContext.

Crankshaft is now able to compile top level code even if there is a ScriptContext.

This CL introduces HPrologue instruction which does the context allocation work and supports deoptimization.

Committed: https://crrev.com/29ebcc32052d486cbc1933ac4738aa5cb68aa851
Cr-Commit-Position: refs/heads/master@{#30496}

[Igor Sheludko, 2015-08-28 19:28:50]

Patchset #2 (id:20001) has been deleted

[Igor Sheludko, 2015-08-28 19:58:32]

Patchset #1 (id:1) has been deleted

[Igor Sheludko, 2015-08-28 19:58:37]

Patchset #1 (id:40001) has been deleted

[Igor Sheludko, 2015-08-31 08:05:31]

ishell@chromium.org changed reviewers:
+ jarin@chromium.org

[Igor Sheludko, 2015-08-31 08:05:32]

PTAL. This CL has some issues on arm ports which don't look related. I'm
investigating now.

[Jarin, 2015-08-31 09:54:53]

lgtm with some non-blocking questions.

https://codereview.chromium.org/1317383002/diff/60001/src/full-codegen/arm/fu...
File src/full-codegen/arm/full-codegen-arm.cc (right):

https://codereview.chromium.org/1317383002/diff/60001/src/full-codegen/arm/fu...
src/full-codegen/arm/full-codegen-arm.cc:247: function_in_register = true;
Why is this here? It looks like function_in_register is not used after this
point...

https://codereview.chromium.org/1317383002/diff/60001/src/full-codegen/arm/fu...
src/full-codegen/arm/full-codegen-arm.cc:293: __ mov(r1,
Operand(Smi::FromInt(rest_index)));
I do not quite understand why this cannot overwrite the function in r1, but I
see that it was similar situation in the original code. Are you sure this is ok?
(Same thing above.)

https://codereview.chromium.org/1317383002/diff/60001/src/hydrogen.h
File src/hydrogen.h (right):

https://codereview.chromium.org/1317383002/diff/60001/src/hydrogen.h#newcode2034
src/hydrogen.h:2034: return HParameter::New(isolate(), zone(), nullptr, index);
Do I understand correctly that you had to add these because you want to pass
nullptr as context?

[Igor Sheludko, 2015-08-31 11:03:05]

Patchset #2 (id:80001) has been deleted

[Igor Sheludko, 2015-08-31 11:07:25]

Thanks a lot! Meet the working version.

https://codereview.chromium.org/1317383002/diff/60001/src/full-codegen/arm/fu...
File src/full-codegen/arm/full-codegen-arm.cc (right):

https://codereview.chromium.org/1317383002/diff/60001/src/full-codegen/arm/fu...
src/full-codegen/arm/full-codegen-arm.cc:247: function_in_register = true;
On 2015/08/31 09:54:53, Jarin wrote:
> Why is this here? It looks like function_in_register is not used after this
> point...

Actually we need this. See the next patch set.

https://codereview.chromium.org/1317383002/diff/60001/src/full-codegen/arm/fu...
src/full-codegen/arm/full-codegen-arm.cc:293: __ mov(r1,
Operand(Smi::FromInt(rest_index)));
On 2015/08/31 09:54:53, Jarin wrote:
> I do not quite understand why this cannot overwrite the function in r1, but I
> see that it was similar situation in the original code. Are you sure this is
ok?
> (Same thing above.)

Thank you for pointing that out! That was the reason of both arm and arm64
crashes.

https://codereview.chromium.org/1317383002/diff/60001/src/hydrogen.h
File src/hydrogen.h (right):

https://codereview.chromium.org/1317383002/diff/60001/src/hydrogen.h#newcode2034
src/hydrogen.h:2034: return HParameter::New(isolate(), zone(), nullptr, index);
On 2015/08/31 09:54:53, Jarin wrote:
> Do I understand correctly that you had to add these because you want to pass
> nullptr as context?

Almost. I had to add this in order to avoid calling GraphBuilder::context() when
it is not available yet (before HContext instruction).

[Igor Sheludko, 2015-08-31 14:41:18]

Patchset #3 (id:120001) has been deleted

[Igor Sheludko, 2015-08-31 15:59:04]

I finished the mips ports and simplified the full-codegen changes.

[Jarin, 2015-09-01 03:57:29]

lgtm.

[Igor Sheludko, 2015-09-01 07:04:43]

The CQ bit was checked by ishell@chromium.org

[commit-bot: I haz the power, 2015-09-01 07:04:57]

CQ is trying da patch. Follow status at
 https://chromium-cq-status.appspot.com/patch-status/1317383002/150001
View timeline at
 https://chromium-cq-status.appspot.com/patch-timeline/1317383002/150001

[commit-bot: I haz the power, 2015-09-01 07:06:57]

Committed patchset #4 (id:150001)

[commit-bot: I haz the power, 2015-09-01 07:07:17]

Patchset 4 (id:??) landed as
https://crrev.com/29ebcc32052d486cbc1933ac4738aa5cb68aa851
Cr-Commit-Position: refs/heads/master@{#30496}