Add missing type feedback collection to ia32 BinaryOpStubs for bitwise operations

Add missing type feedback collection to ia32 BinaryOpStubs for bitwise operations

Committed: http://code.google.com/p/v8/source/detail?r=13624

[Jakob Kummerow, 2013-02-07 15:28:35]

Fix for the second problem that CryptoJS exposed. Please take a look.

https://codereview.chromium.org/12223039/diff/1/src/ia32/code-stubs-ia32.cc
File src/ia32/code-stubs-ia32.cc (left):

https://codereview.chromium.org/12223039/diff/1/src/ia32/code-stubs-ia32.cc#o...
src/ia32/code-stubs-ia32.cc:1856: // right optimized code, int32 type feedback
is just right.
Yeah, well, turns out it's not quite right after all :-)

https://codereview.chromium.org/12223039/diff/1/src/ia32/code-stubs-ia32.cc#o...
src/ia32/code-stubs-ia32.cc:2802: return;
Who are we kidding?

https://codereview.chromium.org/12223039/diff/1/src/ia32/code-stubs-ia32.cc#o...
src/ia32/code-stubs-ia32.cc:3001: void
FloatingPointHelper::CheckFloatOperandsAreInt32(MacroAssembler* masm,
This seems rather similar in functionality to
FloatingPointHelper::CheckLoadedIntegersWereInt32, maybe the two should be
refactored and combined?

https://codereview.chromium.org/12223039/diff/1/src/ia32/code-stubs-ia32.cc
File src/ia32/code-stubs-ia32.cc (right):

https://codereview.chromium.org/12223039/diff/1/src/ia32/code-stubs-ia32.cc#n...
src/ia32/code-stubs-ia32.cc:920: static void
ConvertHeapNumberToInt32(MacroAssembler* masm,
Basically a drop-in replacement for IntegerConvert() above, except that it jumps
to |conversion_failure| when the heap number did not have an int32 value.
We use SSE2 xmm-to-register conversions all over the place, I don't think it
hurts here performance-wise.

...I'll update the method's comment before landing.

https://codereview.chromium.org/12223039/diff/1/src/ia32/code-stubs-ia32.cc#n...
src/ia32/code-stubs-ia32.cc:2254: BinaryOpIC::GENERIC,
Strictly speaking, this is inconsistent, but it doesn't seem to hurt currently.
Postponing the cleanup to when we hydrogen-generate this stub.

[Yang, 2013-02-07 15:49:45]

On 2013/02/07 15:28:35, Jakob wrote:
> Fix for the second problem that CryptoJS exposed. Please take a look.
> 
> https://codereview.chromium.org/12223039/diff/1/src/ia32/code-stubs-ia32.cc
> File src/ia32/code-stubs-ia32.cc (left):
> 
>
https://codereview.chromium.org/12223039/diff/1/src/ia32/code-stubs-ia32.cc#o...
> src/ia32/code-stubs-ia32.cc:1856: // right optimized code, int32 type feedback
> is just right.
> Yeah, well, turns out it's not quite right after all :-)
> 
>
https://codereview.chromium.org/12223039/diff/1/src/ia32/code-stubs-ia32.cc#o...
> src/ia32/code-stubs-ia32.cc:2802: return;
> Who are we kidding?
> 
>
https://codereview.chromium.org/12223039/diff/1/src/ia32/code-stubs-ia32.cc#o...
> src/ia32/code-stubs-ia32.cc:3001: void
> FloatingPointHelper::CheckFloatOperandsAreInt32(MacroAssembler* masm,
> This seems rather similar in functionality to
> FloatingPointHelper::CheckLoadedIntegersWereInt32, maybe the two should be
> refactored and combined?
> 
> https://codereview.chromium.org/12223039/diff/1/src/ia32/code-stubs-ia32.cc
> File src/ia32/code-stubs-ia32.cc (right):
> 
>
https://codereview.chromium.org/12223039/diff/1/src/ia32/code-stubs-ia32.cc#n...
> src/ia32/code-stubs-ia32.cc:920: static void
> ConvertHeapNumberToInt32(MacroAssembler* masm,
> Basically a drop-in replacement for IntegerConvert() above, except that it
jumps
> to |conversion_failure| when the heap number did not have an int32 value.
> We use SSE2 xmm-to-register conversions all over the place, I don't think it
> hurts here performance-wise.
> 
> ...I'll update the method's comment before landing.
> 
>
https://codereview.chromium.org/12223039/diff/1/src/ia32/code-stubs-ia32.cc#n...
> src/ia32/code-stubs-ia32.cc:2254: BinaryOpIC::GENERIC,
> Strictly speaking, this is inconsistent, but it doesn't seem to hurt
currently.
> Postponing the cleanup to when we hydrogen-generate this stub.

LGTM