Remove RESTORE_CONTEXT flag from ia32 crankshaft codegen.

As part of allowing different contexts for inlined functions, eliminate most explicit reads of the context from the stack frame in ia32 crankshaft codegen.

Eliminates the enum flag RESTORE_CONTEXT and CONTEXT_ADJUSTED, and adds a context HValue and LOperand to many hydrogen and lithium instructions.

Context is still used from the stack from in CallKnownFunction (this seems safe), and in CallRuntimeFromDeferred in lithium-codegen-ia32.cc, which needs to be fixed.


BUG=
TEST=

Committed: http://code.google.com/p/v8/source/detail?r=8529

[William Hesse, 2011-06-28 15:45:00]

OK, this is done except for the LIn instruction, and ready for review.  I'll do
the LIn instruction tomorrow morning.

[Kevin Millikin (Chromium), 2011-06-29 08:36:15]

A round of comments.  Also make sure it lints.

http://codereview.chromium.org/7132002/diff/19001/src/hydrogen-instructions.h
File src/hydrogen-instructions.h (right):

http://codereview.chromium.org/7132002/diff/19001/src/hydrogen-instructions.h...
src/hydrogen-instructions.h:2501: if (index == 0) {
I sort of prefer

return (index == 0) ? Representation::Tagged() : representation();

but I guess you don't?

http://codereview.chromium.org/7132002/diff/19001/src/hydrogen-instructions.h...
src/hydrogen-instructions.h:2644: HValue* left() { return value(); }
It might make sense to define left() before right() here :)

http://codereview.chromium.org/7132002/diff/19001/src/hydrogen-instructions.h...
src/hydrogen-instructions.h:3908: return Representation::Integer32();
Context should be tagged.

http://codereview.chromium.org/7132002/diff/19001/src/ia32/lithium-codegen-ia...
File src/ia32/lithium-codegen-ia32.cc (right):

http://codereview.chromium.org/7132002/diff/19001/src/ia32/lithium-codegen-ia...
src/ia32/lithium-codegen-ia32.cc:3543: __ mov(esi, Operand(ebp,
StandardFrameConstants::kContextOffset));
This needs a comment that we're potentially passing the wrong context (in the
case of an inlined function).

http://codereview.chromium.org/7132002/diff/19001/src/ia32/lithium-codegen-ia...
File src/ia32/lithium-codegen-ia32.h (left):

http://codereview.chromium.org/7132002/diff/19001/src/ia32/lithium-codegen-ia...
src/ia32/lithium-codegen-ia32.h:182: ContextMode context_mode);
Yay!

http://codereview.chromium.org/7132002/diff/19001/src/ia32/lithium-ia32.cc
File src/ia32/lithium-ia32.cc (right):

http://codereview.chromium.org/7132002/diff/19001/src/ia32/lithium-ia32.cc#ne...
src/ia32/lithium-ia32.cc:2167: new LDeleteProperty(UseFixed(instr->context(),
esi),
I like to name these UseXXX subexpressions and linearize them.  I'm not sure if
we still have an evaluation-order dependency, but I'd like the register
allocator to be determinstic wrt. the compiler's evaluation order. 

So feel free to clean them up as you notice them.

[William Hesse, 2011-06-29 10:30:21]

http://codereview.chromium.org/7132002/diff/19001/src/hydrogen-instructions.h
File src/hydrogen-instructions.h (right):

http://codereview.chromium.org/7132002/diff/19001/src/hydrogen-instructions.h...
src/hydrogen-instructions.h:2501: if (index == 0) {
All changed to
return cond
    ? value1
    : value2;

On 2011/06/29 08:36:20, Kevin Millikin wrote:
> I sort of prefer
> 
> return (index == 0) ? Representation::Tagged() : representation();
> 
> but I guess you don't?

http://codereview.chromium.org/7132002/diff/19001/src/hydrogen-instructions.h...
src/hydrogen-instructions.h:2644: HValue* left() { return value(); }
On 2011/06/29 08:36:20, Kevin Millikin wrote:
> It might make sense to define left() before right() here :)

Done.

http://codereview.chromium.org/7132002/diff/19001/src/ia32/lithium-codegen-ia...
File src/ia32/lithium-codegen-ia32.cc (right):

http://codereview.chromium.org/7132002/diff/19001/src/ia32/lithium-codegen-ia...
src/ia32/lithium-codegen-ia32.cc:3543: __ mov(esi, Operand(ebp,
StandardFrameConstants::kContextOffset));
On 2011/06/29 08:36:20, Kevin Millikin wrote:
> This needs a comment that we're potentially passing the wrong context (in the
> case of an inlined function).

Done.

http://codereview.chromium.org/7132002/diff/19001/src/ia32/lithium-ia32.cc
File src/ia32/lithium-ia32.cc (right):

http://codereview.chromium.org/7132002/diff/19001/src/ia32/lithium-ia32.cc#ne...
src/ia32/lithium-ia32.cc:2167: new LDeleteProperty(UseFixed(instr->context(),
esi),
On 2011/06/29 08:36:20, Kevin Millikin wrote:
> I like to name these UseXXX subexpressions and linearize them.  I'm not sure
if
> we still have an evaluation-order dependency, but I'd like the register
> allocator to be determinstic wrt. the compiler's evaluation order. 
> 
> So feel free to clean them up as you notice them.

Done.