Implement type recording for ToBoolean on x64.

Implement type recording for ToBoolean on x64.

Handle oddballs on ia32 via root indices, similar to other platforms. Added a
special case for Smi types on ia32 to make lithium code generation on both Intel
platforms more similar.
Committed: http://code.google.com/p/v8/source/detail?r=8767

[Kevin Millikin (Chromium), 2011-08-01 11:32:04]

LGTM except that I'd prefer to avoid the #ifdef---it sets a bad precedent and
can get ugly.

http://codereview.chromium.org/7544012/diff/1/src/code-stubs.h
File src/code-stubs.h (right):

http://codereview.chromium.org/7544012/diff/1/src/code-stubs.h#newcode959
src/code-stubs.h:959: Handle<Object> value,
I don't really like using the preprocessor like this.  Couldn't it be avoided by
passing the root list index always and using the roots array in the IA32
implementation?

[danno, 2011-08-01 11:36:13]

please move the platform-specific stuff into the macro assembler then I'll take
another look.

http://codereview.chromium.org/7544012/diff/1/src/code-stubs.h
File src/code-stubs.h (right):

http://codereview.chromium.org/7544012/diff/1/src/code-stubs.h#newcode958
src/code-stubs.h:958: #if V8_TARGET_ARCH_IA32
Why don't you make CheckOddball a method in the MacroAssembler? Then you don't
need the #if def at all

[Kevin Millikin (Chromium), 2011-08-01 11:51:54]

http://codereview.chromium.org/7544012/diff/1/src/code-stubs.h
File src/code-stubs.h (right):

http://codereview.chromium.org/7544012/diff/1/src/code-stubs.h#newcode958
src/code-stubs.h:958: #if V8_TARGET_ARCH_IA32
On 2011/08/01 11:36:13, danno wrote:
> Why don't you make CheckOddball a method in the MacroAssembler? Then you don't
> need the #if def at all

I thought it was because it needs the types_.  They could be passed in as an
argument, but the function is still pretty specific to this stub.

[Sven Panne, 2011-08-01 12:52:36]

Introduced CompareRoot on ia32 to address the review comments.

[Kevin Millikin (Chromium), 2011-08-01 13:02:05]

Seems fine to me.  Are you sure you need BitCast?

http://codereview.chromium.org/7544012/diff/7/src/ia32/code-stubs-ia32.cc
File src/ia32/code-stubs-ia32.cc (right):

http://codereview.chromium.org/7544012/diff/7/src/ia32/code-stubs-ia32.cc#new...
src/ia32/code-stubs-ia32.cc:258: // 'null' -> false.!!!
Don't be so surprised, it's JavaScript :)

http://codereview.chromium.org/7544012/diff/7/src/ia32/macro-assembler-ia32.cc
File src/ia32/macro-assembler-ia32.cc (right):

http://codereview.chromium.org/7544012/diff/7/src/ia32/macro-assembler-ia32.c...
src/ia32/macro-assembler-ia32.cc:268: Handle<Object> value(BitCast<Object**>(
I don't think we need the BitCast here, the roots array elements are Object*.

[Sven Panne, 2011-08-01 13:13:24]

I'll submit a separate tiny cleanup patch fixing these issues.

http://codereview.chromium.org/7544012/diff/7/src/ia32/code-stubs-ia32.cc
File src/ia32/code-stubs-ia32.cc (right):

http://codereview.chromium.org/7544012/diff/7/src/ia32/code-stubs-ia32.cc#new...
src/ia32/code-stubs-ia32.cc:258: // 'null' -> false.!!!
On 2011/08/01 13:02:05, Kevin Millikin wrote:
> Don't be so surprised, it's JavaScript :)

Ooops, a leftover from a previous debugging session... :-}

http://codereview.chromium.org/7544012/diff/7/src/ia32/macro-assembler-ia32.cc
File src/ia32/macro-assembler-ia32.cc (right):

http://codereview.chromium.org/7544012/diff/7/src/ia32/macro-assembler-ia32.c...
src/ia32/macro-assembler-ia32.cc:268: Handle<Object> value(BitCast<Object**>(
On 2011/08/01 13:02:05, Kevin Millikin wrote:
> I don't think we need the BitCast here, the roots array elements are Object*.

Good point, I should have looked more closely at the actual code when using the
time-proven technique of copy-n-paste... :-P