Split LInstruction into multiple variant to make our LIR more compact....

Split LInstruction into multiple variant to make our LIR more compact.

This change introduces LTemplateInstruction which is a specialized version
of LInstruction and takes one template parameter to indicate whether the
instruction produces a result operand.

All instruction that do not have a result inherit from LTemplateInstruction<0>.
Instructions that have a result operand from LTemplateInstruction<1>

All the Define* function only operate on instructions with a result.

For this to work I also refactored the places where we do 

  LInstruction* result = new Lxyy

into

  Lxyz* result = new Lxyz 




Committed: http://code.google.com/p/v8/source/detail?r=6245

[fschneider, 2011-01-10 10:53:44]

This is a first step towards a bigger refactoring. Please have a look.

[Kevin Millikin (Chromium), 2011-01-10 11:41:44]

The approach LGTM, and this if fine as a work in progress.

I think the two comments below are worth addressing earlier rather than later.

http://codereview.chromium.org/6219001/diff/1/src/ia32/lithium-ia32.h
File src/ia32/lithium-ia32.h (right):

http://codereview.chromium.org/6219001/diff/1/src/ia32/lithium-ia32.h#newcode312
src/ia32/lithium-ia32.h:312: virtual void set_result(LOperand* operand) = 0;
Maybe we should put these only on the Result=1 specialization as non-virtual
member functions.  We could have a static casting function that checks types in
debug builds but that has no overhead in release builds, rather than incurring a
small overhead for each access.

http://codereview.chromium.org/6219001/diff/1/src/ia32/lithium-ia32.h#newcode337
src/ia32/lithium-ia32.h:337: template <int Result = 0>
I think I'd rather have two separate specializations than the default parameter.

http://codereview.chromium.org/6219001/diff/1/src/ia32/lithium-ia32.h#newcode...
src/ia32/lithium-ia32.h:2103: LInstruction* Define(LTemplateInstruction<1>*
instr, LUnallocated* result);
This is nice.

[fschneider, 2011-01-10 11:55:57]

http://codereview.chromium.org/6219001/diff/1/src/ia32/lithium-ia32.h
File src/ia32/lithium-ia32.h (right):

http://codereview.chromium.org/6219001/diff/1/src/ia32/lithium-ia32.h#newcode312
src/ia32/lithium-ia32.h:312: virtual void set_result(LOperand* operand) = 0;
On 2011/01/10 11:41:44, Kevin Millikin wrote:
> Maybe we should put these only on the Result=1 specialization as non-virtual
> member functions.  We could have a static casting function that checks types
in
> debug builds but that has no overhead in release builds, rather than incurring
a
> small overhead for each access.

Done.

http://codereview.chromium.org/6219001/diff/1/src/ia32/lithium-ia32.h#newcode337
src/ia32/lithium-ia32.h:337: template <int Result = 0>
On 2011/01/10 11:41:44, Kevin Millikin wrote:
> I think I'd rather have two separate specializations than the default
parameter.

Done.