Fix embedded new-space pointer in LCmpObjectEqAndBranch.

Fix embedded new-space pointer in LCmpObjectEqAndBranch.

R=mvstanton@chromium.org
BUG=chromium:240032
TEST=mjsunit/regress/regress-crbug-240032

Committed: http://code.google.com/p/v8/source/detail?r=14777

[Michael Starzinger, 2013-05-23 12:59:28]

Michael: PTAL; Danno: FYI.

[mvstanton, 2013-05-23 13:09:13]

lgtm. It looks like you can use CmpHeapObject in another place in 32 bit as
well, and it might be good to expose the method in all platforms (though I know
ARM doesn't have some of these register micro-optimizations).

https://codereview.chromium.org/15779004/diff/1/src/x64/lithium-codegen-x64.cc
File src/x64/lithium-codegen-x64.cc (right):

https://codereview.chromium.org/15779004/diff/1/src/x64/lithium-codegen-x64.c...
src/x64/lithium-codegen-x64.cc:4958: __ CmpHeapObject(reg, target);
Nice that you could refactor this code...how about the one in 32bit though? I
see it in ARM too (lithium-codegen-arm.cc)

[Michael Starzinger, 2013-05-23 13:13:51]

Landing ...

https://codereview.chromium.org/15779004/diff/1/src/x64/lithium-codegen-x64.cc
File src/x64/lithium-codegen-x64.cc (right):

https://codereview.chromium.org/15779004/diff/1/src/x64/lithium-codegen-x64.c...
src/x64/lithium-codegen-x64.cc:4958: __ CmpHeapObject(reg, target);
On 2013/05/23 13:09:14, mvstanton wrote:
> Nice that you could refactor this code...how about the one in 32bit though? I
> see it in ARM too (lithium-codegen-arm.cc)

As discussed offline: Yes, I agree, ia32 should use the same pattern, but it
currently decides at LChunkBuilder time. And with parallel compilation the
location might change. This should be cleaned up as well. As for ARM, the bug
doesn't exists there so that's also up for a follow-up CL, this should be kept a
clean bug-fix.

[Michael Starzinger, 2013-05-23 14:06:38]

Committed patchset #1 manually as r14777 (presubmit successful).