Add a soft-deopt in keyed element access when current IC is pre-monomorphic and no type feedback wa…

Add a soft-deopt in keyed element access when current IC is pre-monomorphic and no type feedback was collected.

BUG=
R=danno@chromium.org

Committed: https://code.google.com/p/v8/source/detail?r=17335

[danno, 2013-10-21 16:35:58]

here's a few comments, more comming...

https://codereview.chromium.org/32643004/diff/1/src/ic.cc
File src/ic.cc (right):

https://codereview.chromium.org/32643004/diff/1/src/ic.cc#newcode1371
src/ic.cc:1371: return LoadIC::Load(object, Handle<String>::cast(key));
Don't you need to install the generic stub here, too? With a corresponding
trace?

https://codereview.chromium.org/32643004/diff/1/src/ic.cc#newcode1381
src/ic.cc:1381: if (state() == UNINITIALIZED) stub = string_stub();
need else here to TRACE_GENERIC

https://codereview.chromium.org/32643004/diff/1/src/ic.cc#newcode1392
src/ic.cc:1392: }
need else here to TRACE_GENERIC

https://codereview.chromium.org/32643004/diff/1/src/ic.cc#newcode1395
src/ic.cc:1395: TRACE_GENERIC_IC(isolate(), "KeyedLoadIC", "force generic");
Comment is wrong. Should not be "force generic", but rather "access check
needed"

[Hannes Payer (out of office), 2013-10-22 09:29:29]

I added it now all over the place.

https://codereview.chromium.org/32643004/diff/1/src/ic.cc
File src/ic.cc (right):

https://codereview.chromium.org/32643004/diff/1/src/ic.cc#newcode1371
src/ic.cc:1371: return LoadIC::Load(object, Handle<String>::cast(key));
On 2013/10/21 16:35:59, danno wrote:
> Don't you need to install the generic stub here, too? With a corresponding
> trace?

Done.

https://codereview.chromium.org/32643004/diff/1/src/ic.cc#newcode1381
src/ic.cc:1381: if (state() == UNINITIALIZED) stub = string_stub();
On 2013/10/21 16:35:59, danno wrote:
> need else here to TRACE_GENERIC

Done.

https://codereview.chromium.org/32643004/diff/1/src/ic.cc#newcode1392
src/ic.cc:1392: }
On 2013/10/21 16:35:59, danno wrote:
> need else here to TRACE_GENERIC

Done.

https://codereview.chromium.org/32643004/diff/1/src/ic.cc#newcode1395
src/ic.cc:1395: TRACE_GENERIC_IC(isolate(), "KeyedLoadIC", "force generic");
On 2013/10/21 16:35:59, danno wrote:
> Comment is wrong. Should not be "force generic", but rather "access check
> needed"

Done.

[danno, 2013-10-22 14:52:24]

https://codereview.chromium.org/32643004/diff/80001/src/ic.cc
File src/ic.cc (right):

https://codereview.chromium.org/32643004/diff/80001/src/ic.cc#newcode868
src/ic.cc:868: TRACE_GENERIC_IC(isolate(), "KeyedLoadIC", "access check
needed");
change comment appropriately

https://codereview.chromium.org/32643004/diff/80001/src/ic.cc#newcode896
src/ic.cc:896: TRACE_GENERIC_IC(isolate(), "KeyedLoadIC", "access check
needed");
change comment appropriately

https://codereview.chromium.org/32643004/diff/80001/src/ic.cc#newcode1134
src/ic.cc:1134: TRACE_GENERIC_IC(isolate(), "KeyedLoadIC", "access check
needed");
comment should be: not heap object

https://codereview.chromium.org/32643004/diff/80001/src/ic.cc#newcode1405
src/ic.cc:1405: TRACE_GENERIC_IC(isolate(), "KeyedLoadIC", "access check
needed");
Should be: "not otherwise handled"

https://codereview.chromium.org/32643004/diff/80001/src/ic.cc#newcode1408
src/ic.cc:1408: TRACE_GENERIC_IC(isolate(), "KeyedLoadIC", "access check
needed");
should be: "not JSObject"

[danno, 2013-10-23 09:38:06]

lgtm

[Hannes Payer (out of office), 2013-10-23 10:41:31]

Committed patchset #7 manually as r17335 (presubmit successful).