Update constant pool correctly when resuming a generator

Update constant pool correctly when resuming a generator

This CL depends on the following CLs landing first:
  https://codereview.chromium.org/188063002/
  https://codereview.chromium.org/190793002/

R=ulan@chromium.org

Committed: https://code.google.com/p/v8/source/detail?r=20012

[rmcilroy, 2014-03-07 18:52:35]

Part of the out-of-line constant pool change - PTAL.

[ulan, 2014-03-11 13:07:39]

https://codereview.chromium.org/190823002/diff/1/src/arm/full-codegen-arm.cc
File src/arm/full-codegen-arm.cc (right):

https://codereview.chromium.org/190823002/diff/1/src/arm/full-codegen-arm.cc#...
src/arm/full-codegen-arm.cc:2198: ConstantPoolUnavailableScope
constant_pool_unavailable(masm_);
Shouldn't this scope be
- at the beginning of the "if (resume_mode == JSGeneratorObject::NEXT)" branch?
- or inside the "if (FLAG_enable_ool_constant_pool)" branch?

[rmcilroy, 2014-03-11 16:14:52]

https://codereview.chromium.org/190823002/diff/1/src/arm/full-codegen-arm.cc
File src/arm/full-codegen-arm.cc (right):

https://codereview.chromium.org/190823002/diff/1/src/arm/full-codegen-arm.cc#...
src/arm/full-codegen-arm.cc:2198: ConstantPoolUnavailableScope
constant_pool_unavailable(masm_);
On 2014/03/11 13:07:39, ulan wrote:
> Shouldn't this scope be
> - at the beginning of the "if (resume_mode == JSGeneratorObject::NEXT)"
branch?
> - or inside the "if (FLAG_enable_ool_constant_pool)" branch?
> 

The Constant Pool is unavailable (for this code object) from the ldr(pp, ...) 
until the Jump(r3) instruction (since between these instructions it contains the
pp for the target function).  

I can't put it inside the "if (FLAG_enable_ool_constant_pool) because the scope
would go out of scope before the Jump(r3).  I could put it at the beginning of
the if (resume_mode == JSGeneratorObject::NEXT) scope if you prefer, but I put
it here to make it clear that this was related to the ldr(pp, ..) instruction. 
If you still prefer moving it to the beginning of the scope though I'm happy to
move it. WDYT?

[ulan, 2014-03-11 16:19:57]

https://codereview.chromium.org/190823002/diff/1/src/arm/full-codegen-arm.cc
File src/arm/full-codegen-arm.cc (right):

https://codereview.chromium.org/190823002/diff/1/src/arm/full-codegen-arm.cc#...
src/arm/full-codegen-arm.cc:2198: ConstantPoolUnavailableScope
constant_pool_unavailable(masm_);
I see, thanks for explaining! What about putting instructions from ldr(pp, ...) 
until the Jump(r3) into a new block { }? That would make clear when exactly the
scope is active.

[rmcilroy, 2014-03-11 16:42:22]

https://codereview.chromium.org/190823002/diff/1/src/arm/full-codegen-arm.cc
File src/arm/full-codegen-arm.cc (right):

https://codereview.chromium.org/190823002/diff/1/src/arm/full-codegen-arm.cc#...
src/arm/full-codegen-arm.cc:2198: ConstantPoolUnavailableScope
constant_pool_unavailable(masm_);
On 2014/03/11 16:19:57, ulan wrote:
> I see, thanks for explaining! What about putting instructions from ldr(pp,
...) 
> until the Jump(r3) into a new block { }? That would make clear when exactly
the
> scope is active.

Good call! Done.

[ulan, 2014-03-11 17:27:39]

lgtm

[rmcilroy, 2014-03-17 17:29:11]

Committed patchset #2 manually as r20012 (presubmit successful).